Ang-2 is a potential molecular marker for lymphatic metastasis and better response to bevacizumab therapy in ovarian cancer

Samples from 205 patients with epithelial ovarian cancer treated at the University Medical Center Hamburg-Eppendorf were available. All patients underwent primary debulking surgery including lymphadenectomy between 1997 and 2019 and gave written informed consent to gain biomaterial and view their clinical records. This was approved by the local ethics committee “Ethik-Kommission der Ärztekammer Hamburg” (no. 04.05.2004). All experiments were performed in accordance with the relevant guidelines and regulations. Patient data were tracked from date of first diagnosis until October of 2020.

Patient cohorts for Western blot, immunohistochemistry (IHC) and enzyme-linked Immunosorbent assay (ELISA) analyses were classified with regard to tumor spread pattern as seen in Table 1.

All tissue samples were gained during surgery and were immediately stored in liquid nitrogen as fresh frozen samples, or fixed in 4% buffered formalin and subsequently embedded in paraffin as described before (Kuerti et al. 2017; Milde-Langosch et al. 2005).

Only tissue samples consisting of at least 70% tumor cells, as quantified by H&E staining, were used for protein extraction. Protein extraction was performed as described previously (Trillsch et al. 2016).

For each sample 20 μg protein lysate was loaded per well. Equal amount of loading was verified by immunoblotting with β-actin antibody (SantaCruz, Dallas, Texas, USA, sc-47778, Lot E0919, 1:4000). Protein lysate from human pulmonary microvascular endothelial cells (HPMEC) served as positive control and reference sample. To separate the proteins, a 10% sodium dodecyl sulfate (SDS) gel was used. Electrophoresis was performed for approximately 18 h at 55 V. Next, the proteins were blotted for 2 h at 1.00 A to polyvinylidene difluoride membranes. Membranes were blocked with 5% non-fat dry milk for 1 h at room temperature in Tris-buffered saline (TBS) + 20% Tween (TBST) and subsequently incubated with a monoclonal Ang-2 antibody (SantaCruz, sc-74403, Lot B0916, 1:1000) overnight at 4 °C. Afterward, membranes were incubated with the corresponding secondary antibody (goat anti mouse, southern biotech, Birmingham, Alabama, USA, 1030-05, Lot K3515T566C, 1:8000) for 1 h at room temperature. In between incubation steps, membranes were washed with TBST. All antibodies were diluted in 1.5% non-fat dry milk in TBST. The detection was performed with chemiluminescence reagent (Westar Nova 2.0 CYANAGEN, Bologna, Italy, XLS071) to visualize the protein expression on FUJI-super RX medical X-ray films. Band intensities were quantified by densitometry (GS-700 Imaging Densitometer, Bio-Rad, Munich, Germany) and calculated as percentage of the HPMEC control sample.

Paraffin sections of tumor samples were dehydrated in descending concentrations of ethanol and xylene. In between incubation steps, slides were washed twice with TBST and once with TBS. All antibodies were diluted in Antibody Diluent (Agilent, Carpinteria, California, USA, DAKO S0809).

For Ang-2 staining, slides were heated in antigen retrieval buffer (Agilent, DAKO S1699) in a pressure cooker for 10 min at 121 °C. Afterward, the tissue was blocked with 10% normal horse serum (Vector Laboratories, Newark, California, USA, S-2000) for 30 min at room temperature. The primary antibody for Ang-2 (SantaCruz, sc-74403, Lot A0813, 1:50) incubated over night at 4 °C. Peroxidase blocking was performed with 1% H₂O₂ in methanol for 30 min at room temperature.

For CD31 staining, the slides were pretreated in Target-Retrieval-Buffer pH9 (Agilent, DAKO S2367) for 20 min at 85 °C in a steamer. The primary antibody (Agilent, DAKO M0823, Lot 00054859, 1:60) was incubated over night at 4 °C. Peroxidase blocking was performed with 0.5% H₂O₂ in methanol for 30 min at room temperature.

Biotinylated secondary horse anti-mouse IgG (Vector Laboratories, BA-2000) was incubated 1 h at room temperature. For signal amplification, avidin–biotin complex (Vector Laboratories, PK-6100) was applied. A DAB staining kit (Vector Laboratories, SK-4100) was used, and slides were counterstained with hematoxylin. The tissue was rehydrated using ascending concentrations of ethanol and xylene. Placenta tissue served as positive control, while isotype control antibody (mouse IgG1, Agilent, DAKO X0931) served as negative control.

Slides were independently reviewed by two researchers. The Ang-2 staining intensity was scored on a scale from 0 (no staining), 1 (weak staining), 2 (moderate staining) to 3 (strong staining) and multiplied by the amount of tumor tissue that was stained from 1 (< 25%), 2 (25–50%), 3 (50–75%) to 4 (> 75%). Immunoreactive scores for Ang-2 tissue and vascular staining were created separately. For CD31 staining only the density of vessels was considered and assessed on a scale of 0 to 3.

The patients’ blood was collected prior to therapy in serum-gel monovettes (Sarstedt, Germany) and was prepared as previously described (Ding et al. 2022). All serum samples were stored at − 80 °C and thawed on ice right before measurement. Serum Ang-2 was measured with the Human Angiopoietin-2 Quantikine ELISA set (R&D Systems, Minneapolis, Minnesota, USA, Catalog No. DANG-20). Serum samples were diluted 1:5. The procedure was performed according to protocol provided by the manufacturers.

Initially, 175 patients were analyzed in Western blot. For statistical evaluation patients who received neoadjuvant or no chemotherapy at all (n = 15) as well as patients with a histology subtype other than serous/serous papillary (n = 22) were excluded from analyses. A total of 138 patients were left to be included.

Twenty-five (18.1%) patients in tumor stage pT1a–pT3b, pN1, suffering from lymph node metastases only and without significant intraperitoneal tumor growth, were included in the retroperitoneal group. In contrast, 31 (22.5%) patients in tumor stage pT3c, pN0, only suffering from intraperitoneal tumor growth without lymph node metastases, were assigned to the intraperitoneal group. The mixed group consists of 58 (42.0%) patients with both lymph node metastases as well as significant intraperitoneal tumor growth in tumor stage pT3c, pN1. A small group of 9 (6.5%) patients was diagnosed at an early tumor stage pT1a–pT3a, pN0.

Also, there were 15 patients with tumor spread patterns that could not be assigned to any of these four groups (Fig. 1).

As seen in Fig. 2A, an Ang-2 specific band could be detected at approximately 75 kDa. The endothelial cell line HPMEC served as positive control. Corresponding bands of β-actin were detected at 45 kDa (Fig. 2A). There was a heterogeneous Ang-2 expression within the patient cohort. Defining the expression level of HPMEC at 100%, the expression level of analyzed tumors ranged from 7 to 15,962%. Two patients did not express Ang-2. The median expression level of the overall cohort was 1876%.

As protein lysates used in Western blot analyses did not solely contain carcinoma cells and as Ang-2 is also expressed by vascular endothelium, we performed IHC with paraffin sections of 58 tumors to visualize the distribution and Ang-2 expression intensity in the different tumor tissue compartments. Four slides were not analyzable because of insufficient quality and were therefore excluded from the analysis. Representative images of variously strong stained tumors are shown in Fig. 2B. Ang-2 expression was detected at different intensities in tumor cells, as well as in endothelial cells of tumor associated vessels (arrows, Fig. 2B). IHC scores ranged from 0 to 12. Weak staining was defined from 1 to 4, moderate staining from 5 to 8 and strong staining from 9 to 12. Regarding the staining of tumor tissue, 20 (37.04%) tumors were stained weakly, 22 (40.74%) moderately and 9 (16.67%) strongly. Regarding vascular staining, 20 (37.04%) slides were stained weakly, 11 (20.37%) moderately and 18 (33.33%) strongly.

Additionally, CD31 staining was performed to assess the vascular density on the corresponding tumors. Scores for CD31 ranged from 1 to 3, with 4 (7.02%) tumors with low vessel density, 38 (66.67%) with intermediate and 15 (26.32%) with high density.

Ang-2 staining of tumor cells did not correlate with vascular density, determined by CD31 staining (p = 0.986). Vascular density and endothelial Ang-2 expression did not correlate either (p = 0.461). However, Ang-2 tumor cell staining correlated with Ang-2 vascular staining (p < 0.001). Interestingly, when comparing Western blot and IHC expression levels, Ang-2 levels in protein lysates only correlated with Ang-2 vascular staining (p = 0.001), whereas no correlation between Ang-2 tissue staining and Western blot expression levels was noted (p = 0.625).

Serum samples of 38 patients with ovarian cancer at tumor stage pT1a-pT2c were analyzed. A total of 15 patients (39.5%) whose tumors had already spread to lymph nodes (pT1a–pT2c, pN1) were assigned to the retroperitoneal group, while 23 patients (60.5%) were diagnosed at an early stage without lymph node metastases (pT1a–pT2c, pN0) (Table 1). Serum Ang-2 levels ranged from 860.3 to 6114, with a median expression level of 1902.5. Later on, one patient from the early stage group with unusual high levels of serum Ang-2 was excluded. Clinical records revealed that the patient concerned suffered from ulcerative colitis. In 2006 Koutroubakis et al. revealed that Ang-2 serum levels are elevated in patients with inflammatory bowel disease, leading us to exclude said patient (Koutroubakis et al. 2006).

Ang-2 expression in Western blot varied strongly depending on the tumor spread pattern observed in the patient. Defining high Ang-2 levels above the median, 20 (80%) patients of the retroperitoneal group expressed high levels. In comparison, 16 (51.61%) of the intraperitoneal group, 27 (46.55%) of the mixed group and 3 (33.33%) of the early stage group expressed high Ang-2 levels.

The statistical analyses revealed a significantly higher Ang-2 expression in the retroperitoneal group compared to the intraperitoneal group (mean 55.85 vs. 33.91, p = 0.039, Fig. 3A). Also, Ang-2 expression in tumors with a retroperitoneal spread pattern was significantly higher compared to tumors with mixed type of metastasis (mean 55.85 vs. 34.10, p = 0.022, Fig. 3A). A similar trend could be detected between the retroperitoneal group and tumors diagnosed at an early stage. Mean expression levels of these tumors were lower compared to those of the retroperitoneal group (mean 55.85 vs. 24.94). However, these findings did not reach statistical significance (p = 0.053, Fig. 3A).

Between all the other groups, no significant differences in expression levels were found.

In line with these data, a significantly higher Ang-2 expression in serum samples was noted in the retroperitoneal group compared to the tumors diagnosed at an early stage without lymph node metastases (mean 2638 vs. 1913, p = 0.029, Fig. 3B).

Patients were divided into two groups according to the median Ang-2 expression level, defining high (> median) and low (< median) levels. The postoperative residual tumor, which was categorized into “no postoperative tumor residual”, “tumor residual < 1 cm” and “tumor residual > 1 cm”, correlates significantly with Ang-2 expression in the Western blot cohort. Here, lower Ang-2 levels were detected in patients with more residual tumor after surgery (p = 0.022, Supplementary Fig. 1A). Further analyses revealed no significant differences of Ang-2 expression levels in tumor stage (pT, pM) nor histological grading (Supplementary Fig. 1B–E). However, dividing the patient cohort according to quartiles, we observed high Ang-2 expression levels (= quartile 4) in tumors with positive lymph nodes (p = 0.060, Supplementary Fig. 1E). As most of our patients were diagnosed at an advanced tumor stage (FIGO IIIC-IV; n = 116) and due to a small number of patients diagnosed at an early stage (FIGO I-IIA; n = 2), no correlation analyses were performed in regard to FIGO stage.

In IHC as well as in ELISA there were no significant correlations between Ang-2 level and clinical or pathological characteristics. No correlation with CA-125 expression levels could be detected (data not shown).

No significant association between Ang-2 expression and progression free survival (PFS; p = 0.134, Supplementary Fig. 2A–C) was found, whereas patients with high Ang-2 expression levels showed significantly longer overall survival rates (OAS; p = 0.017, Fig. 4A). Interestingly, stratified survival analyses revealed that the prognostic effect of Ang-2 was restricted to the subgroup of patients treated with bevacizumab as maintenance therapy.

Here, high Ang-2 levels correlated significantly with longer OAS (p = 0.008, Fig. 4B), while no prognostic difference was observed in the cohort of patients receiving platinum-based therapy alone with respect to Ang-2 expression levels (p = 0.527, Fig. 4C).

Indeed, patients showing high Ang-2 expression levels did benefit from antiangiogenic therapy. As shown in Fig. 4D, patients treated with platin-based therapy alone had significantly shorter OAS than those who subsequently received bevacizumab maintenance therapy (p = 0.013, Fig. 4D). In contrast, no significant difference in OAS was observed between the two treatment arms in patients with low Ang-2 expression levels (p = 0.737; data not shown). Similar trends were found in PFS analyses, where high Ang-2 expression levels are associated with longer PFS in the cohort of patients treated with bevacizumab. However, this correlation did not reach significance (p = 0.069, Supplementary Fig. 2C).

In multivariate Cox regression analysis, including Ang-2 expression levels and clinical prognostic parameters as FIGO stage and nodal status, postoperative residual tumor is the only independent and significant prognostic indicator (p = 0.001). Ang-2 expression did not turn out to be an independent prognostic indicator (p = 0.157, data not shown).

In addition, survival analyses in the IHC cohort were performed. As stated before, only vascular staining scores in IHC correlated with Western blot expression levels. As vascular Ang-2 expression was therefore assumed to be responsible for the biggest share of expression levels, only scores of Ang-2 vascular staining were used for analyses. Patients were divided into two groups. Tumors with weak to moderate staining (score 1–8; n = 36; 66.7%) were compared to tumors with strong staining (score > 8; n = 18; 33.3%). In line with the Western blot results, strong Ang-2 staining correlated with significantly longer OAS (p = 0.024, Fig. 5A). A similar trend could be shown in patients treated with platin-based-chemotherapy and bevacizumab (p = 0.086, Fig. 5B). Again, no statistical significance was reached in regard to Ang-2 staining in PFS (p = 0.336, Supplementary Fig. 3A).

In ELISA analyses, Ang-2 serum levels did not correlate with OAS or PFS (Supplementary Fig. 3B, C).