Poor outcome in hypoxic endometrial carcinoma is related to vascular density

Background Identification of endometrial carcinoma (EC) patients at high risk of recurrence is lacking. In this study, the prognostic role of hypoxia and angiogenesis was investigated in EC patients. Methods Tumour slides from EC patients were stained by immunofluorescence for carbonic anhydrase IX (CAIX) as hypoxic marker and CD34 for assessment of microvessel density (MVD). CAIX expression was determined in epithelial tumour cells, with a cut-off of 1%. MVD was assessed according to the Weidner method. Correlations with disease-specific survival (DSS), disease-free survival (DFS) and distant disease-free survival (DDFS) were calculated using Kaplan–Meier curves and Cox regression analysis. Results Sixty-three (16.4%) of 385 ECs showed positive CAIX expression with high vascular density. These ECs had a reduced DSS compared to tumours with either hypoxia or high vascular density (log-rank p = 0.002). Multivariable analysis showed that hypoxic tumours with high vascular density had a reduced DSS (hazard ratio [HR] 3.71, p = 0.002), DDFS (HR 2.68, p = 0.009) and a trend for reduced DFS (HR 1.87, p = 0.054). Conclusions This study has shown that adverse outcome in hypoxic ECs is seen in the presence of high vascular density, suggesting an important role of angiogenesis in the metastatic process of hypoxic EC. Differential adjuvant treatment might be indicated for these patients.

(CA9), whose encoded protein, carbonic anhydrase IX (CAIX), regulates intracellular pH by converting carbon dioxide to carbonic acid. 14 By adaptation of tumour cells to a hostile microenvironment, tumour proliferation can commence even in hypoxic areas. 15 Also in normoxic conditions, HIF-1 can be activated; however, downstream activation is present in lesser extent. 16,17 In this perspective CAIX expression, one of the key effector proteins of HIF-1, has been shown to be more specifically related to hypoxia and poor outcome. 18 Next to maintenance of intracellular pH, stimulation of angiogenesis is an important response to hypoxia. Vascular endothelial growth factor (VEGF), another downstream target of HIF-1, is also correlated with hypoxia and angiogenesis in several cancer types, including EC. [19][20][21][22] Angiogenesis can be assessed by microvessel density (MVD) and is prognostically associated with deep myometrial invasion (MI), lymphovascular space invasion (LVSI) and poor overall survival in EC. 23 Although earlier studies suggest correlation between hypoxia, angiogenesis and poor outcome, the prognostic value has not yet been studied before. 5,6 Therefore, we have investigated the prognostic value of hypoxia and angiogenesis in EC, assessed with CAIX expression and MVD.

Patients
Data and tumour slides were collected previously for a study analysing the value of L1CAM expression in ECs, which included ECs from 11 collaborating European Network for Individualised Treatment of Endometrial Cancer (ENITEC) centres. 24,25 Only cases diagnosed by an expert gynaecological pathologist, with complete data on treatment and pathological examination and at least 36 months of follow-up, were included. Cases with a nonendometrioid component were categorised as non-endometrioid. The 1199 cases included in the original study were randomly selected using SPSS version 22 (SPSS IBM, New York, NY, USA), resulting in a database of 403 patients for the present study. These cases were not statistically different from the original cases for the most important baseline characteristics.
For analysis of CAIX staining, only membranous expression on epithelial tumour cells was analysed. Areas of necrosis, large vessels and tumour stroma, determined using H&E-stained adjacent tumour slides, were therefore manually excluded from the analysis (i-Vision for Mac; BioVision Technologies, Exton, PA, USA). Next, thresholds for segmentation of the fluorescent signals were interactively set above the background staining for each individual marker and adjusted for each sample in order to optimise the signal to background ratio using ImageJ software (Wayne Rasband, National Institute of Mental Health, National Institutes of Health). An interactively set threshold limits intersample variability by correction for differences in immunofluorescence staining intensity. 26,27 The resulting binary images were used to calculate the fraction of CAIX (fCAIX) relative to the total tumour area. To minimise bias of non-specific staining, only positive signals exceeding 5 pixels were included.
The MVD was measured according to the Weidner method. 28 In short, surrounding epithelial tumour cells three areas with the highest density of vessels were selected by the assessor (M.A.) using a ×200 magnification. To correct for objects that exceed the image borders, only objects exceeding the left and upper border were included. To minimise bias of nonspecific staining, only positive signals exceeding 2 pixels were included.
CAIX expression was considered positive when the fCAIX was above 1%. 29,30 The MVD was dichotomised over the median. A representative example of CAIX and MVD staining is shown in Fig. 1.

Statistical analyses
Clinicopathological differences between subgroups were compared with the χ 2 and Fisher's exact tests for categorical data and the Mann-Whitney U test for continuous variables.
Kaplan-Meier curves were constructed for disease-specific survival (DSS), disease-free survival (DFS) and distant-DFS (DDFS). The association between CAIX and MVD and DSS, DFS and DDFS was determined using Cox regression analysis. DSS was calculated from the date of primary treatment to the date of death caused by the disease or, for surviving patients, to the date of the last followup. DFS and DDFS were defined as the length of follow-up, after completion of the primary treatment, during which women survived without any clinical sign of (distant) disease recurrence. Distant recurrence included metastases in distant organs and para-aortic lymph nodes. Features identified by univariable regression analysis with p < 0.20 were used for multivariable regression analysis. LVSI was coded as negative in case of missing data (n = 108) since only substantial LVSI was recently reported as relevant for prognosis of EC. If LVSI was not reported in the pathological report, it was therefore assumed that LVSI was absent. 31

Patients
After staining for CAIX and CD34, 18 of the 403 patients were excluded due to insufficient tumour tissue (n = 9) and excess of nonspecific background staining (n = 9). Clinicopathological characteristics of the 385 patients included for analysis are shown in Table 1 Figure 2 shows that CAIX expression with high vascular density was associated with a worse DSS compared to CAIX expression with low vascular density and negative CAIX expression (p =  (Fig. 3). In multivariable analysis, high age, CAIX expression with high vascular density and tumour grade 3 remained significantly associated with reduced DSS, with CAIX and MVD as the most significant parameter (hazard ratio [HR] 3.71, 95% confidence interval (CI) 1.59-8.63, p = 0.002). Multivariable analysis showed that age, FIGO stage and LVSI were significantly associated with DFS. CAIX expression with high vascular density was nearly significant (HR 1.87, 95% CI 0.99-3.55, p = 0.054, Fig. 4). Multivariable analysis for DDFS showed that LVSI and CAIX expression with high vascular density was significantly associated with a reduced DDFS (CAIX and MVD: HR 2.68, 95% CI 1.27-5.65, p = 0.009, Fig. 5).
Individual contribution of CAIX and MVD Positive CAIX expression was associated with high tumour grade, non-endometrioid histology, higher median MVD and treatment with radiotherapy. In multivariable analysis, CAIX and grade were significantly associated with DSS (HR 2.45, 95% CI 1.05-5.73, p = 0.039) ( Supplementary Tables 1 and 2). High MVD was correlated with deep MI, but not with other clinicopathological factors. In multivariable analysis, high MVD remained an independent predictor of reduced DSS (HR 2.92, 95% CI 1.13-7.54, p = 0.027) (Supplementary Tables 3 and 4). Continuous scoring of CAIX expression showed a significant correlation with DSS as well (data not shown).

DISCUSSION
In the present study, we have investigated the prognostic value of angiogenesis and hypoxia, assessed with MVD and CAIX expression. We hypothesised that angiogenesis would facilitate haematogenous spread of hypoxic tumour cells with subsequent poor clinical outcome. Additionally, we assumed that this would specifically be facilitated in hypoxic ECs, because of activation of intracellular pathways that induce an aggressive and metastatic phenotype. We have shown that CAIX expression with high vascular density is associated with reduced DSS and DDFS. Interestingly, CAIX-positive ECs with low vascular density had a similar outcome as CAIX-negative ECs. Finally, multivariable analyses for CAIX expression and vascular density showed that both were independent prognostic markers as well.
This is the largest study to date studying CAIX in EC. In contrast to previous studies in EC, we did find significant correlations between CAIX expression and poor outcome, especially in case of high vascular density. Seeber et al. 29 included 93 patients and found CAIX expression in 76% of ECs. 29 In this study, no correlation between CAIX expression and outcome was found; however, small sample size and different cut-off value (all degrees of positive staining were regarded as positive) could explain why no correlation was found. Similarly, Pijnenborg et al. 22 investigated CAIX expression in 59 ECs and did not find a correlation. Again, possibly this study was underpowered due to a limited sample size and low number of distant recurrences. Also, differences in study design (case-control study) hamper valid comparison with our results. In other cancer types, including breast carcinoma, hepatocellular carcinoma, cervical carcinoma and renal cell carcinoma, CAIX expression is associated with poor prognosis. [33][34][35][36][37] More specifically, increased distant failure was seen in several solid tumour types with positive CAIX expression. 37,38 The metastatic process is a complex step-wise process, including acquisition of an aggressive phenotype, invasion in surrounding tissues and blood vessels, survival in the circulation with subsequent extravasation and colonisation in new organs. 39 Hypoxia and subsequent neoangiogenesis will intervene with several steps of this process, including promoting tumour cell survival by acquisition of a malignant phenotype and increased invasion in blood vessels. 9 A recent meta-analysis has shown that high MVD was associated with several poor prognostic variables, including deep MI, positive LVSI and poor outcome in EC, although heterogeneity due to differences in used antibodies and cut-off values hampers interpretation of these results. 23 Biologically, intratumoural neoangiogenesis in response to hypoxia will promote the formation of vasculature with high degrees of permeability and potential for rapid growth. 40 Our hypothesis that CAIX expression with high degrees of vascular density would be associated with unfavourable prognostic features and poor outcome was based both on the facilitation of haematogenous spread in areas with high angiogenesis and on the aggressive biological behaviour of tumour cells after hypoxia. 41,42 HIF-1α is stabilised and accumulates under hypoxia, and activates transcription of numerous genes involved in angiogenesis, proliferation and pH regulation (VEGF, CAIX, glucose transporter-1). 9 Our hypothesis was supported by the fact that ECs with positive CAIX expression and high vascular density had a decreased DSS compared to ECs with only one or none of both features. This observation supports the complex interplay underlying the metastatic processes. The observation that CAIX-positive ECs with high vascular density did not have more lymph node metastasis or local recurrences, but instead have more distant recurrences, could support the role of angiogenesis in the haematogenous rather than the lymphogenic metastatic process.
The obvious strengths of this study are the inclusion of a large and representative cohort of EC patients within the ENITEC network and the objective and reproducible measurement of CAIX and MVD using digital imaging analyses. However, there are some limitations that need to be addressed. Due to the retrospective nature of the study, there were missing values, specifically for LVSI and lymph node metastasis. Substantial LVSI is a stronger predictor for prognosis of EC compared to moderate LVSI. Also, LVSI is not routinely reported in the pathologic report at all centres. Therefore, we assumed that if substantial LVSI was present, it was reported, and if LVSI was not reported, no substantial LVSI was present. 31 Missing cases were therefore coded as negative for LVSI. Separate analyses of patients with available LVSI status did not alter the results of the primary outcome (data not shown). Another general limitation in interpretation of CAIX and MVD is the lack of standardised criteria in the current literature, which hampers comparison of previous studies and this study. 29,30 However, the applied digital With the integration of digital pathology into clinical practice, comparison of future studies with our results might be easier. 43,44 Although widely used to quantify MVD, CD34 is known to also identify lymph vessels and stem cell populations, which theoretically could have led to an overestimation of our results. On the other hand, other antibodies, for example, CD31, also carry the risk of aspecific staining. Compared to CD31, CD34 staining is known to have stronger reactivity with endothelial cells, resulting in a lower risk of staining failure. 45 Finally, generalisability to non-endometrioid subtypes can be questioned, as they comprise only 3.4% of the entire cohort. More research focused on this specific subgroup could help to strengthen these results. This study identifies a group of patients with a poor DSS and DDFS based on CAIX and MVD. Given the increased risk of distant metastases, differential adjuvant treatment for these ECs could be explored either in the form of chemotherapy or, in the future, targeted therapies directed against angiogenesis. Because of the focal character of CAIX expression in the tumour tissue, performing the analysis on preoperative biopsies might be challenging, but visualisation of hypoxia and angiogenesis on FDG-PET/CT (fluorodeoxyglucose-positron emission tomography/computed tomography) scan and magnetic resonance imaging could be an alternative, as Berg et al. 5 showed recently.
In summary, we have found that CAIX expression and high vascular density are prognostic markers for decreased survival in EC. Combining these two markers revealed that ECs with positive CAIX expression and high vascular density have an impaired outcome compared to ECs that have only one or none of both features. These patients experienced more distant recurrences, and therefore differential adjuvant treatment for these tumours should be explored.