A total of 154 eligible surgically resected, formalin-fixed, paraffin-embedded (FFPE) oligodendroglioma specimens were collected from the Department of Pathology, the First Affiliated Hospital of USTC, Division of Life Sciences and Medicine, University of Science and Technology of China from 2010 to 2018. Diagnosis of oligodendroglioma was performed according to the 2016 revision of the WHO Classification of Tumors of CNS, with demonstration of IDH1 immune-positive and 1p/19q co-deletion concurrently. Clinical data of each case was summarized in Table 1. The follow-up period was from 12 to 110 months with a median time of 48 months. Progression-free survival (PFS) was defined as the time span from the date of surgery to the first relapse, secondary tumor occurrence, death, or last follow-up. Overall survival (OS) was evaluated from the date of surgery to death caused by tumor or last follow-up. 18 cases were excluded from the study because of incomplete clinical information. Finally, 136 PFS and OS data were collected in which 23 of patients had relapsed or secondary tumor and 18 died. This study was conducted with approval by the Ethics Committee of the First Affiliated Hospital of USTC and in accordance with the Declaration of Helsinki. The written informed consent was obtained from each patient.

Briefly, 3 µm thick slices of FFPE specimens were deparaffinized and hydrated through a series of xylenes and alcohols, followed by antigen retrieval and blocking. The slices were then incubated with primary antibodies including pATM (Abcam, ab81292, 1:1,600), pATR (Abcam, ab178407, 1:4,000) and Hypoxia-inducible factor-1α (HIF1α, ZSGB-BIO, ZA-0552) for 2 h at room temperature. Next, the slices were washed with PBS and incubated with avidin-biotin-peroxidase (MXB Biotechnologies, KIT-5220) at room temperature for 15 min followed by reaction with diaminobenzidine (MXB Biotechnologies, KIT-5220) for 1–2 min. Finally, all stained slices were washed with running tap water for 15 min and then counterstained with hematoxylin. pATM and pATR levels were analyzed in a semiquantitative manner via composite score system by assessing both the percentage and intensity of stained tumor cells. The percentage of positive cells was evaluated per high-power field (HPF) by ×400 magnification. The slices with 5% or less positive cells score 0 point. In the same way, each slice scores 1 (5–25%), 2 (26–50%), 3 (51–75%) or 4 (76–100%). In addition, stained signals in each slice score 0, 1, 2 or 3, corresponding to negative, weak, moderate or high intensity, respectively. Scores for the percentage and intensity were multiplied as the overall score of the staining. Tumor specimens were classified into low expression and high expression groups according to overall score: low expression indicates 0 to 3 points and high expression indicates 4 to 12 points. All slices were evaluated independently by two pathologists without knowledge of the identity and clinical outcome of patients.

Both CD31-PAS and CD34-PAS dual staining are widely and well accepted for detecting VM vessels. In our preliminary experiment, no significant difference was detected in vessles stained by these two method (data not shown). Considering the relatively restrictive expression of CD31 in endothelium and extensive spectrum expanding from immature to mature vascular compared with a occasionally nonspecific expression of CD34 in oligodendrogliomas, we chose CD31 as the major marker accompany with PAS for proving VM existence in this study. After IHC staining with the anti-CD31 monoclonal antibody (Cell Signaling Technology, #3528, 1:800), the slices were exposed to sodium periodate (BaSO, BA4080B) for 10 min, rinsed with distilled water for 5 min, and then incubated with Schiff solution (BaSO, BA4080B) for 5 min. Slices were counterstained with hematoxylin, dehydrated, and mounted. Stained sections were subjected to microscopy at a magnification of ×400 and analyzed by two pathologists without knowledge of demographic or outcome data. PAS-positive tubes lined with CD31-positive cells (PAS+/CD31+) were defined as endothelium-dependent vessels. VM was defined as PAS-positive tubes enclosed by tumor cells, with red blood cells (RBCs) in the lumens but without ECs lining. The absence of ECs in VM was confirmed by hematoxylin-eosin (H&E) staining and IHC staining with anti-CD31 antibodies (PAS+/CD31−). Each section was thoroughly checked to detect VM formation. VM ratio was counted which was expressed as a percentage of total vessels (VM vessels/total vessels per HPF). In addition, we performed GFAP/PAS dual-staining (anti-GFAP monoclonal antibody, ZSGB-BIO, ZA-0529), similar to the procedure mentioned above, to further confirm VM vessles which exhibited as PAS-positive micro-channels surrounded by tumor cells expressed GFAP strongly.

The ROC curve was utilized to determine the best cutoff value of VM ratio. The chi-square test or Fisher’s exact and t-test were used for categorical and continuous variables respectively. PFS and OS were evaluated by Kaplan-Meier analysis, and the difference between groups was determined by Log-rank test. p values less than 0.05 were considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001). Spearman coefficient was applied to determine the correlation between pATM/pATR levels and VM formation. Cox proportional hazards regression model was applied to perform univariate and multivariate analyses. Variables with a significance of p < 0.05 in univariate analysis were included in the subsequent multivariate analysis using stepwise backward elimination procedures. All statistical analyses were performed using SPSS Statistics 22.0.

We used CD31/PAS and GFAP/PAS dual-staining to distinguish CD31−/PAS+ and GFAP+/PAS + VM vessels from CD31+/PAS + endothelium-dependent vessels (Figure 1). Among the 136 oligodendroglioma specimens inspected, VM vessels was detected in 30 (22.1%) cases with 0–2 VM vessels/10HPF in each section. VM ratio ranged from 0 to 3.85% (mean 1.58%, media 1.43%). Over a range of VM ratios, receiver operating characteristic (ROC) analyze showed an area under the curve (AUC) of 0.743 for PFS and 0.582 for OS. A “cutoff” of 0.87% was yielded according to the greatest Youden index for PFS ROC (Supplementary Figure S1). Thus, we defined VM-positive as VM ratio greater than 0.87%. As a result, 28 cases were defined as VM-positive in this study. Chi-square test analysis suggested that VM-positive cases was significantly associated with higher WHO grade, lower Karnofsky performance status (KPS) score and recurrent tumor (p < 0.001, p = 0.040 and p = 0.020, respectively) (Table 1). In particular, 14 (50%) patients in the VM-positive group and 9 (8.3%) patients in the VM-negative group had tumor relapse (p < 0.001) (Table 1). Glomeruloid microvascular proliferation (GMP) that was commonly seen in high-grade gliomas was detected in 28 AOs. However, no obvious connection was identified between GMP and VM formation through Chi-square test (p = 0.199). Moreover, Kaplan-Meier analyses demonstrated that the VM-positive group had a shorter PFS relative to the VM-negative group (p < 0.001), although no significant difference was detected in OS between these two groups (Figure 1). These data indicate the potential value of VM formation in predicting prognosis of oligodendroglioma patients.

We next investigated the impact of VM formation on prognosis in oligodendroglioma patients with different tumor grades as well as in primary and recurrent oligodendrogliomas. VM was correlated with PFS but not OS in WHO grade Ⅱ oligodendroglioma patients (p = 0.039 and p = 0.521). Likewise, in AOs, patients with VM formation had a significantly shorter PFS compared to those without VM formation (p = 0.001) (Figure 2). Stratified multivariate analyses were utilized to study the prognostic value of KPS, WHO grade, extent of surgical resection, tumor localization (frontal lobe or not), status of tumor (primary or recurrent) and VM formation [29]. The results showed an independent predictive impact of VM on PFS in AO patients (HR = 7.058, 95% CI: 2.148–23.189, p = 0.001). Meanwhile, VM formation was correlated with PFS but not OS both in primary and recurrent oligodendroglioma patients (p < 0.001 and p = 0.005) (Figure 2). Considering the potential bias due to the small number of recurrent oligodendroglioma patients in the statistical analysis, we performed stratified multivariate analyses only in primary oligodendroglioma patients. The results revealed an independent predictive impact of VM for PFS (HR = 10.778, 95% CI: 2.409–48.225, p = 0.002) (Figure 2). These results suggest that VM formation may affect the progression of oligodendroglioma but not the overall tumor growth, resulting in a shorter PFS but no impact on OS.

The activation of ATM and/or ATR contributes to treatment resistance and pathological neoangiogenesis in cancers. IHC of pATM and pATR was performed on oligodendroglioma specimens to determine the activation of ATM/ATR. pATM and pATR signals mainly localized in the nuclei of tumor cells (Figure 3). Low pATM was detected in 105 (77.2%), and high pATM was detected in 31 (22.8%) out of the 136 specimens. Meanwhile, low pATR was detected in 102 (75%), and high pATR was detected in 34 (25%) patients (Table 1). The correlation between the levels of pATM/pATR and clinicopathological features was investigated and summarized in Table 1. Higher pATM was significantly correlated with higher WHO grade (p < 0.001) and recurrent tumor (p = 0.01). pATR level demonstrated the same trend in terms of tumor grade (p = 0.002) and recurrence (p = 0.025), and higher level of it was associated with lower KPS scores (p = 0.002) (Table 1). No significant difference was detected between pATM/pATR levels and other clinical parameters including age, gender, predominant side or lobe of tumor location and the extent of surgical resection (Table 1). The results remind us the correlation between VM formation and tumor grade and recurrence, suggesting a probable link between VM formation and ATM/ATR activation.

Next, we sought to explore the potential involvement of ATM/ATR in VM formation in oligodendrogliomas. We determined the IHC scores in oligodendroglioma sections with or without VM formation. The results showed that VM-positive specimens exhibited higher pATM and pATR scores relative to VM-negative specimens (p < 0.001 for pATM and p < 0.001 for pATR). Moreover, there was a positive correlation between pATM/pATR levels and VM formation according to Spearman correlation analysis (r_s = 0.435, p < 0.001 and r_s = 0.317, p < 0.001) (Figure 4). As the classical downstream effector molecule in the ATM/ATR signaling pathway, HIF1α was examined by immumohistochemical staining as well. The results showed that only 14 cases expressed HIF1α with weak-moderate staining (Supplementary Figure S2). HIF1α expression was positively correlated with higher WHO grade and non-frontal lobe (p = 0.010, p = 0.029), but not correlated with VM formation (p = 0.537). In summary, activation of ATM/ATR but not HIF1α shows a positive correlation with VM formation in oligodendroglioma patients.

Because VM formation is associated with tumor progression and pATM/pATR correlates with VM formation, we investigated the association between pATM/pATR and tumor progression in oligodendroglioma patients. Oligodendrogliomas with high pATM/pATR had an obvious tendency to relapse compared to those with low pATM/pATR (p < 0.001 for pATM and p < 0.001 for pATR) (Table 1). Patients with high pATM/pATR levels had a shorter PFS relative to those with low pATM/pATR levels (p < 0.001 for pATM and p < 0.001 for pATR), but no significant difference in OS was observed (p = 0.151 for pATM and p = 0.054 for pATR) (Figure 5). In patients who received chemotherapy or radiation-therapy, high pATM/pATR levels were also associated with shorter PFS (Supplementary Figure S3). Taken together, the levels of pATM/pATR in oligodendrogliomas may be associated with the progression but not overall growth of tumors, which recapitulates the situation of VM formation.

We then assess the prognostic and predictive values of VM, pATM/pATR and other indicators by using Cox proportional hazards regression model. Univariate analysis showed that higher WHO grade (p = 0.002), recurrent tumor (p = 0.040), high pATM (p < 0.001), high pATR (p = 0.002) and VM formation (p < 0.001) were unfavorable prognostic factors for PFS in oligodendroglioma patients. Multivariate analysis containing these five factors revealed that VM formation is an independent unfavorable predictor for tumor progression in oligodendroglioma patients (p = 0.007, HR = 4.534, 95% CI: 1.504–13.675) (Table 2).