Introduction

Almost 80% of ovarian cancer deaths result from high-grade serous ovarian cancer (HGSOC), a distinct subtype that has shown little improvement in overall survival during the past 30 years (Holmes 2015; Bowtell et al. 2015; Seidman et al. 2004). The development of treatment resistance has been identified as a crucial factor for patients’ poor prognosis. Eighty percent of the patients will benefit from initial cytoreduction and respond to platinum-based chemotherapy. However, almost all of them will experience disease recurrence and ultimately develop platinum resistance (Holmes 2015). In this context, identification of biomarkers to characterize the long-term survivors after completion of platinum-based chemotherapy is urgently needed. Furthermore, including long-term survivors into standard treatment regimens may result in patients’ overtreatment, whereas into clinical trials may lead to selection biases and subsequently unreliable results. A recent large-scale clinical analysis demonstrated that long-term survival was “not unusual” in patients with advanced epithelial ovarian cancer (EOC). Among 3582 women with EOC who survived longer than 10 years, nearly one-third was initially diagnosed at advanced stages (FIGO-International Federation of Gynecology and Obstetrics-stage III-IV), including high-grade serous cancer (Cress et al. 2015). It is still unclear which factors may contribute to long platinum-free intervals for HGSOC patients. Thus, investigating such biomarkers is important to set up more tailored approaches for cancer surveillance, to improve the poor prognosis.

Angiogenesis has been highlighted as a central event in tumor growth and progression of ovarian cancer (Dvorak 2002; Monk et al. 2013; Eskander and Tewari 2014). This process is considered to be mainly regulated through the vascular endothelial growth factor (VEGF) A and VEGF receptor 2 (VEGFA/VEGFR2) signaling pathway (Ferrara 2005). However, in ovarian cancer, the prognostic value of VEGFA/VEGFR2 pathway still remains controversial (Williams et al. 2012; Siddiqui et al. 2010, 2011; Yamamoto et al. 1997; Sallinen et al. 2014; Hartenbach et al. 1997; Engels et al. 2009; Raspollini et al. 2004; Nishida et al. 2004; O’Toole et al. 2007; Wimberger et al. 2014; Klasa-Mazurkiewicz et al. 2011). VEGFA or VEGFR2 often loses prognostic significance when established prognostic factors are included into multivariate analyses, such as age, tumor stage, grade, histologic subtypes and tumor residual mass after surgery (Yamamoto et al. 1997; Engels et al. 2009; Raspollini et al. 2004; Nishida et al. 2004). Besides, the follow-up time in these studies were mostly shorter than 5 years, whereas patients with a PFS ≥ 5 years were rarely included. Herein, better patient selection and long-term follow-up are imperatively needed, to clarify the role of VEGFA/VEGFR2 pathway in HGSOC patients’ prognosis.

This study aims to investigate whether intratumoral VEGFA/VEGFR2 expressions correlate with long-term PFS, analyzing the homogeneous samples from 124 advanced primary HGSOC patients.

Methods and materials

Sample collection

We retrospectively included 124 matched tumor tissue samples collected from primary advanced HGSOC patients, who were treated from 1985 to 2013 and were included in the Tumor Bank Ovarian Cancer Consortium (TOC Consortium, http://www.toc-network.de). Patients underwent surgery in five European high-volume Gynecologic Oncology Centers: Charité Medical University of Berlin, Germany; University-Medical-Center Hamburg-Eppendorf, Germany; Medical University of Innsbruck, Austria; University Hospital Leuven, Belgium; University of Medicine and Pharmacy Iuliu Hatieganu Cluj-Napoca, Romania. Patients’ inclusion criteria were: (1) primary HGSOC diagnosis; (2) FIGO stage III–IV; (3) no relapse within 6 months after completion of primary platinum-based chemotherapy; (4) no anti-angiogenic therapy was performed from the date of first diagnosis till the first relapse or the last follow-up; (5) availability of patients’ follow-up and clinico-pathological data; (6) availability of formalin-fixed, paraffin-embedded ovarian cancer tissue samples. Exclusion criterion was no chemotherapy naïve tumor tissue available for immunohistochemistry. The 124 samples include 62 “long-term survivors” and 62 controls, which were matched 1:1 by age at first diagnosis (older or younger within 5 years) and macroscopic tumor residual status after primary surgery (no residual vs. with residual). “Long-term survivors” were defined as primary HGSOC patients without recurrence 5 years after primary chemotherapy (PFS ≥ 5 years). Controls were defined as primary HGSOC patients experiencing recurrence within 6 months to 3 years after the completion of adjuvant chemotherapy (PFS = 6 months-3 years). Samples from the long-term group were collected from all five centers of TOC Consortium, whereas control group samples were retrieved only from Charité Medical University of Berlin.

All patients underwent cytoreduction and platinum-based chemotherapy following the standard procedure as described previously (Braicu et al. 2012). Ovarian cancer tissue samples were prospectively collected during primary cytoreduction within TOC consortium. All samples underwent central histopathological review to confirm the diagnosis of HGSOC and to evaluate tissue quality and tumor content. Tumor grading was re-evaluated according to World Health Organization (WHO) pathological classification (2014). Staging was performed and defined in accordance with the FIGO criteria for ovarian cancer (22). Clinical data and follow-up information were also collected prospectively in TOC Consortium Centers.

PFS was defined as time interval from the end of first chemotherapy to first recurrence of disease. Overall survival (OS) was defined as time interval between diagnosis and patients’ death or loss of follow up. Response to treatment and diagnosis of recurrence was determined according to the “Response Evaluation Criteria in Solid Tumors” (RECIST) criteria (Eisenhauer et al. 2009).

The study protocol was approved by each local Ethics Committees (Charité 2004-000034, Innsbruck AN2015-0237 354/4.7, Hamburg EK200313, Leuven MML1022, Cluj 39). Prior to surgery and sample collection, all patients signed written informed consent for using their bio-specimens and clinico-pathological data for research purposes.

Immunohistochemical staining

For tissue microarrays (TMAs), representative tumor areas of chemotherapy naïve ovarian carcinoma tissue were selected in hematoxylin and eosin-stained sections by an experienced pathologist (SD). Consecutively tissue microarrays (TMAs) were constructed as previously described (Buckendahl et al. 2011). For detection of VEGFA and VEGFR2 on tissue samples, two rabbit polyclonal antibodies directed against the VEGFA or VEGFR2 (Abcam, Cambridge, United Kingdom) were used. Slides were first deparaffinated and rehydrated in a series of descending alcoholic concentration. For antigen retrieval, slides were boiled for 5 min in 0.01 M sodium citrate buffer at pH 9.0 in a pressure cooker and afterwards put in TBS buffer for the same time. After blocking the endogenous peroxidase, slides were incubated with the primary antibody, diluted 1:250 for VEGFA and 1:500 for VEGFR2 in antibody diluents solution (Zytomed Systems, Berlin, Germany) for 30 min at room temperature. For visualization, Dako Real Detection System (Dako, Glostrup, Denmark) was applied according to a standard protocol, as provided by the manufacturer using DAB + Chromogen. Counterstaining was carried out with Haemalaun (Dr Hollborn, Leipzig, Germany). Afterwards, the tissue was dehydrated and cover-slipped with Vitroclud (Medizintechnik Langenbrinck, Emmerdingen, Germany).

The immunohistochemical expression of VEGFA and VEGFR2 were independently evaluated by two co-authors (JG and ETT), who were blinded to the patient characteristics and outcome. In case of disagreement, results were discussed in a multi-headed microscope till a consensus was reached. The immunoreactivity of VEGFA and VEGFR2 was calculated applying a semi-quantitative immunoreactivity score (IRS), as published before (Yamamoto et al. 1997; Raspollini et al. 2004; Nishida et al. 2004; Horikawa et al. 2017). The scores of staining proportion (0, no cells stained; 1, < 10% of cells stained; 2, 11–50% of cells stained; 3, 51–80% of cells stained; 4, > 80% of cells stained) and intensity (0, negative; 1, weak; 2, moderate; 3, strong) were multiplied to obtain the final IRS (range 0–12). As no established cutoffs for VEGFA/VEGFR2 expression exist, we performed a logistic regression model with different cutoffs to conclude IRS of 0–6 as “low expression” and IRS of 7–12 as “high expression”, to dichotomize the expression of the two biomarkers.

Statistical analyses

Statistical analysis was performed by IBM SPSS Statistics 22.0 (SPSS, Inc., Chicago, IL) The Shapiro–Wilk test was used to ascertain whether continuous variables had a normal distribution. Continuous variables were summarized by means and standard deviations, or median and inter quartile range (IQR) where appropriate. Categorical variables were presented as frequency with percentage. The intra-group differences of continuous variables were investigated by Student t test when normally distributed, or Mann–Whitney U test when non-normally distributed. Chi square test or Fisher’s exact test was used to analyze the difference between categorical variables. Correlation between expressions of two biomarkers was analyzed by correlation test (Spearman coefficient, two-tailed). Multivariable logistic regression model was performed to evaluate the independent value of biomarkers. A two-tailed p value of < 0.05 was considered as statistically significant.

Results

The clinico-pathological characteristics of matched long-term and control groups are summarized in Table 1. Between two groups, no difference in age, tumor residual status and FIGO stage was observed (Table 1; Fig. 1).

Table 1 Characteristics of patients in the long-term and control groups
Full size table
Fig. 1
figure 1

Age, tumor residual and FIGO stage of patients in long-term and control groups

Full size image

VEGFA and VEGFR2 expression

Immunohistochemical staining showed that VEGFA and VEGFR2 proteins were localized in the cytoplasm (Fig. 2a, b, e and f). The IRS distributions of VEGFA and VEGFR2 in long-term and control groups have been presented in Fig. 2c, d, g and h. A high expression of VEGFA (VEGFAhigh) was found in 79.0% (49/62) and 64.5% (40/62) of long-term and control samples, respectively (p = 0.073, Chi square test, Fig. 3a; Table 2). A high expression of VEGFR2 (VEGFR2high), however, was significantly more frequent in the long-term (77.4%, 48/62) compared with the control group (51.6%, 30/62, p = 0.001, Chi square test, Fig. 3b; Table 2). Significant correlation between VEGFA and VEGFR2 expression was observed (p > 0.0001, Spearman coefficient 0.347). After adjusting for FIGO stage and VEGFA expression in multivariate analysis, VEGFR2 expression still remained significantly different between long-term survivors and controls (p = 0.005, logistic regression model, Table 2).

Fig. 2
figure 2

VEGFA and VEGFR2 expression by immunohistochemistry: weak expressions of VEGFA (a) and VEGFR2 (e) in tumor cells; strong expressions of VEGFA (b) and VEGFR2 (f) in tumor cells with weak background staining in stroma cells; and IRS distribution of VEGFA (c, d) and VEGFR2 (g, h) expression in long-term and control groups

Full size image
Fig. 3
figure 3

Difference in VEGFA and VEGFR2 expressions between long-term and the control groups: VEGFA (a) and VEGFR2 (b) expression and their co-expression (c) between two groups in all patients (n = 124); difference in VEGFA (d) and VEGFR2 (e) expressions between two groups in no-residual patients (n = 96); and difference in VEGFA (f) and VEGFR2 (g) expressions between two groups in with-residual patients (n = 28)

Full size image
Table 2 Differences in VEGFA and VEGFR2 expressions between long-term and control groups
Full size table

Co-expression of VEGFA and VEGFR2

A combined expression of VEGFAhigh and VEGFR2high was more common in the long term (62.9% 39/62) compared with the control group (42.0%, 26/62, p = 0.003, Chi square test, Fig. 3c).

VEGFA and VEGFR2 expression in subgroups (with or without residual)

We investigated VEGFA and VEGFR2 expression in HGSOC patients according to their residual tumor status (no residual, n = 96, and with residual, n = 28). Among no-residual patients, VEGFR2high was more frequently observed in the long-term group (75.0%, 36/48) than in controls (54.2%, 26/48, p = 0.033, Fig. 3d, e). This difference was even more pronounced among with-residual patients, as 85.7% of (12/14) long-term survivors showed VEGFR2high compared with 28.6% (4/14) of controls (p = 0.006, Fisher’s exact test, Fig. 3f, g). On the other hand, VEGFAhigh was also more frequent among long-term survivors in both no-residual (79.2%, 38/48) and with-residual subgroups (78.6%, 11/14) than in control (66.7%, 32/48, and 57.1%, 8/14, Fig. 3a, b). Nevertheless, these differences were not statistically significant (P = 0.168, and P = 0.221).

VEGFA and VEGFR2 expression according to PFS duration

To investigate whether the expression of VEGFA or VEGFR2 may change in relation to PFS, we finally analyzed the two biomarkers in four subgroups (Fig. 4): PFS ≥ 10 years (120 months, n = 23), PFS = 5–10 years (60–119 months, n = 39), PFS = 1–3 years (13–36 months, n = 40), and PFS = 6–12 months (n = 22). All subgroups were well balanced regarding age, FIGO stage and tumor residual. VEGFA expression was not associated with PFS (p = 0.298, Chi square test, Fig. 4c) at the median IRS of 8 in all subgroups (Fig. 4a). On the contrary, the IRS of VEGFR2 was the highest in patients with PFS ≥ 10 years (median IRS = 9, VEGFR2high 91.3%, 21/23), whereas the lowest in patients with PFS = 1–3 years (median IRS = 6, VEGFR2high 48.5%, 17/40, p = 0.001, Chi square test). Interestingly, VEGFR2 expression increased with PFS duration among the three subgroups: PFS = 1–3 years, PFS = 5–10 years and PFS ≥ 10 years (p < 0.001, Chi square test). However, between subgroups of PFS = 6–12 months and PFS = 1–3 years, VEGFR2 expression remained stable (P = 0.211, Chi square test).

Fig. 4
figure 4

Correlation between VEGFA/VEGFR2 expressions and PFS: IRS of VEGFA (a) and VEGFR2 (b), and high expression of VEGFA (c) and VEGFR2 (d) in patients with subgroups according to PFS duration

Full size image

Discussion

In the present study, we investigated whether the VEGFA/VEGFR2 pathway plays a role in long-term survival for primary HGSOC patients. VEGFR2high was found to be significantly associated with long-term PFS, independent of age, FIGO stage and residual disease. VEGFA expression was not significantly correlated with long-term survival, although it was positively associated with VEGFR2 expression. This result indicated that other mechanisms might be also involved in VEGFR2 activation.

Although anti-VEGFA and anti-VEGFR treatment have been implemented into first-line treatment of ovarian cancer (Wimberger et al. 2010, 2014; Burger et al. 2011a, b; du Bois et al. 2009), their effects on long-term survival still remain unclear. A median improvement of 4–6 months in PFS has been shown by three phase 3 randomized trials using anti-VEGFA (bevacizumab) and anti-VEGFR treatments (pazopanib) in primary ovarian cancer (Burger et al. 2011a, b; du Bois et al. 2014). Nevertheless, no overall survival differences were found between patients receiving bevacizumab plus chemotherapy compared to chemotherapy alone in GOG 0218 (ClinicalTrials.gov Identifier: NCT00262847). Since grade 3–4 adverse events were significantly higher in patients using bevacizumab and pazopanib compared with placebo control (Burger et al. 2011a, b; du Bois et al. 2014), it is imperative to precisely identify patients who may or may not benefit from anti-VEGFA/VEGFR treatment. However, patients with PFS ≥ 5 years were hardly included in these trials, where median PFS was around 15 months and the follow-up duration was less than 3 years. Thus, our findings are clinically important showing that HGSOC patients with VEGFR2high may be more likely to achieve long-term survival. The anti-VEGFR therapy for these patients might need further consideration.

Our study shows for the first time that VEGFR2 might contribute to long-term PFS of patients with advanced primary HGSOC. Previous findings remained controversial regarding prognostic values of VEGFA/VEGFR2 in ovarian cancer. High expression of VEGFA/VEGFR2 have been found to promote tumor metastases and thus correlate with poor outcome in ovarian cancer (Williams et al. 2012; Siddiqui et al. 2011; Yamamoto et al. 1997; Raspollini et al. 2004; Nishida et al. 2004; Wimberger et al. 2014; Horikawa et al. 2017; Avril et al. 2017). An inverse association has also been reported (Sallinen et al. 2014; Engels et al. 2009; Raspollini et al. 2004; O’Toole et al. 2007). In particular, Nishida N et al. analyzed tumor specimens of 80 patients with ovarian cancer and showed that a high VEGFA expression independently predicted longer disease-free survival in multivariate analyses (Nishida et al. 2004). However, the follow-up time in these studies was mostly around 3 years. Whether the VEGFA/VEGFR2 pathway contributes to long-term PFS in this context has not been clearly elucidated. The association between VEGFR2 overexpression and long-term survival of HGSOC patients was for the first time reported in this study.

The strength of our study is the homogeneous patient selection. On one hand, earlier analyses included patients with all tumor stages, histological subtypes and tumor residual status (Engels et al. 2009; Raspollini et al. 2004; Klasa-Mazurkiewicz et al. 2011), which are the most important prognostic factors for ovarian cancer (du Bois et al. 2009; Wimberger et al. 2007, 2010). These factors were also strongly associated with VEGFA/VEGFR2 expression (Williams et al. 2012; Sallinen et al. 2014; Raspollini et al. 2004; Wimberger et al. 2014; Klasa-Mazurkiewicz et al. 2011). Including established prognostic factors into multivariate analyses, VEGFA and VEGFR2 often lost their prognostic significance (Yamamoto et al. 1997; Engels et al. 2009; Raspollini et al. 2004; Nishida et al. 2004). On the other hand, previous studies mainly recruited both platinum-resistant and platinum-sensitive patients (Williams et al. 2012; Siddiqui et al. 2011; Yamamoto et al. 1997; Raspollini et al. 2004; Nishida et al. 2004; Wimberger et al. 2014; Horikawa et al. 2017; Avril et al. 2017). Since PFS is much shorter in platinum-resistant patients, it might conceal the independent impact of angiogenesis markers on long-term survival. Considering these factors above, we only included platinum-sensitive patients with advanced HGSOC and matched them according to established prognostic factors (age and residual disease). Based on the well-selected samples, we revealed an independent correlation between VEGFR2 overexpression and good prognosis in HGSOC patients.

Defining “long-term survivors” as patients with PFS ≥ 5 years is also a strength of our patient selection. Since the follow-up period in earlier studies was mostly around 3 years, the long-term impacts of VEGFA/VEGFR2 expression in HGSOC remained unclear (Siddiqui et al. 2011; Hartenbach et al. 1997; Raspollini et al. 2004; Wimberger et al. 2014). Furthermore, extended follow-up duration is highly important as the prognostic values of VEGF might change over time. One study showed that VEGFA expression was associated with treatment response within 6 months but not within 12 months after chemotherapy (O’Toole et al. 2007). A similar trend was observed where VEGFA inversely impacted on PFS according to the length of tamoxifen treatment (Sanchez et al. 2010). In this randomized clinical trial, including 402 breast cancer patients with a median follow-up of 9.8 years, VEGFA overexpression predicted shorter PFS in the group of 2-year tamoxifen treatment but correlated with longer PFS in patients treated for 5 years. In our study, VEGFR2 expression remained stable in the group of PFS = 6 months–3 years, whereas it increased significantly among patients with PFS ≥ 5 years. Thus, long-term follow-up and appropriate patient selection should be considered when assessing prognostic values of VEGFA/VEGFR2.

A high expression of VEGFR2 was found more frequently among long-term survivors in our study. Similarly, Zhang et al. found VEGFR2 protein levels were higher in control tissue compared with human squamous cell carcinomas (Zhang et al. 2010). Using anti-VEGFA monoclonal antibody bevacizumab (Avastin) can upregulate VEGFR2 expression in tumor-associated endothelial cells, through the JNK/c-Jun pathway and activity of the ubiquitin–proteasome system (Zhang et al. 2010). In this scenario, improved PFS by bevacizumab in clinical trials might partially attribute to upregulated VEGFR2 induced by the blockage of VEGFA. This may help to explain the association between VEGFR2high and long-term PFS in our study. Anti-VEGFR or anti-VEGFR2 treatment for HGSOC patients may need reconsideration.

The VEGFR2 overexpression in our “long-term” cohort might also be due to the patient selection, as the long-term group might include more patients with BRCA1/2 mutations or other homologous recombination repair (HRR) defects that are strongly related to improved prognosis (Yang et al. 2011; Bolton et al. 2012; Norquist et al. 2018). Our previous study of HGSOC found that VEGF overexpression was most common among patients with somatic BRCA1/2 mutations (Ruscito et al. 2018). Confirming what was already observed in a study on breast cancer, which reported that BRCA mutation carriers were found with higher levels of VEGF mRNA (p = 0.04) compared to patients without BRCA mutations (Danza et al. 2013). Another study on ovarian cancer revealed, that a VEGF-dependent gene signature (VDGs) was overexpressed in BRCA mutation carriers (Yin et al. 2016). The proposed mechanism of correlation between BRCA1 mutation and VEGF overexpression involves Caveolin-1 expression, which down-regulates VEGF and regulates endothelial function; in HGSOC, dysfunctional BRCA1 leads to Caveolin-1 expression inhibition, thus resulting in increased VEGF levels (Desai et al. 2015). However, information on BRCA1/2 mutation was not available for the patients in our study. Further investigations should consider BRCA1/2 mutation status when accessing effect of VEGF expression on long-term survival of HGSOC.

As the major blood vessel endothelial cell receptor for VEGFA, VEGFR2 can also be activated by VEGFC and VEGFD to collect lymphatic vessels and capillaries undergoing active lymph-angiogenesis (Saharinen et al. 2004), VEGFR2 signaling affects not only primary tumor growth but also lymphatic vessel functionality and tumor cell metastatic spread (Joukov et al. 1997; Alam et al. 2004). In our study, the correlation between VEGFR2 expression and long-term PFS was independent from VEGFA expression, although VEGFA and VEGFR2 expression were significantly correlated. Our findings suggest that VEGFR2 might be activated by other factors such as VEGFC and VEGFD to affect development of HGSOC. Nevertheless, in our study other members of the VEGF family have not been evaluated, which is our major limitation. Further investigation and validation of this signaling pathway is therefore needed.

Another strength of this study, besides the homogeneous population, is the large sample size of matched long-term survivors and controls. As very few patients with advanced HGSOC could achieve 5-year PFS, our 124 matched HGSOC patients is the largest homogeneous population regarding VEGFR2 in long-term survivors. In addition, centers in our tumor bank consortium are all high-volume centers of ovarian cancer, with long-lasting experience in surgical treatment and standardized protocols of sample collection. Most centers have also been approved and allowed to participate in the LION (ClinicalTrials.gov Identifier: NCT00712218), DESKTOP III (Clinical-Trials.gov Identifier: NCT01166737) and TRUST (ClinicalTrials.gov Identifier: NCT02828618) studies, based on the high quality of the tumor debulking.

In conclusion, we found VEGFR2high was strongly and independently associated with long-term PFS in primary advanced HGSOC. VEGFR2high patients may be less likely to experience disease recurrence within 5 years after primary chemotherapy. Our findings may provide a new insight into understanding tumor development and a relevant cornerstone for further studies on HGSOC patients’ prognosis.