Detection of IDH1-R132H mutations and ATRX expression
In this study, we detected IDH1-132H mutations and ATRX nuclear protein by immunohistochemistry in 104 glioma cases (Fig 1a-1d). We found that IDH1-132H positive mainly occurred in astrotytoma (20/33, 60.6%) and oligodendroglioma (11/14, 78.6%) while glioblastoma (5/57, 8.8% ) only at a low frequency (Table 1, P<0.0001). Although the level of ATRX loss in astrotytoma (16/33, 48.5%) was higher than oligodendroglioma (4/14, 28.6%) and glioblastoma (17/57, 29.8%), there was no significant difference among them (Table 1, P=0.310).
Choice of cutoff score for ZMYND8 expression
ZMYND8 expression in glioma tissues was shown in Fig 1. (1e-1h). To single out a suitable score of ZMYND8 to categorize into low expression group and high expression group, median and ROC curve (Fig 2) were used to determine the expression level of ZMYND8. The median score is 2.1 and the cutoff score is 2.175. ROC-curve analysis can determine the optimal cutoff value which reveals maximum sensitivity and specificity. Therefore, the ROC curve was used to determine the expression level of ZMYND8, which was classified into low expression group (total score≦2.175) and high expression group (total score>2.175). The low expression group was identified in 53/104 (50.9%) of adult glioma patients, and 51/104 (49.0%) in the high expression group.
Expression levels of ZMYND8 between normal brain tissue and glioma
We detected the expression level of ZMYND8 between normal brain tissue and glioma. We found that ZMYND8 was rarely expressed in normal brain tissue (median score 0.33) and expressed in glioma including LGG (median score = 1.49) and HGG (median score = 2.09) with different degrees. In LGG and HGG, ZMYND8 levels were significantly higher than normal brain tissue (Fig 3) (P<0.0001).
Expression levels of ZMYND8 correlate with clinicopathological variables of glioma patients
The expression levels of ZMYND8 in glioma patients to several clinicopathological variables are displayed in Table 2. The significant difference in ZMYND8 levels was observed between LGG and HGG (P<0.001). We also found that high expression of ZMYND8 was significantly correlated with Glioblastoma (P<0.05). In addition, ATRX status was observed to correlated with ZMYND8 levels (P<0.05), which demonstrated that they had a relationship in some kind. Furthermore, no significant association was found between ZMYND8 levels and other clinicopathological variables, such as age, gender, tumor sizes, KPS, IDH1-R132H status (P>0.05).
Association between IDH1-R132H, ATRX, ZMYND8 and clinical outcome in glioma patients
Univariate analysis revealed that ZMYND8 expression, age, pathological grade, IDH1-R132H status, AYRX status were correlated with glioma patients survival rates (Table 3). High expression of ZMYND8 was closely associated with worse overall survival in glioma patients (P<0.05, Fig 4a, Mean OS of ZMYND8 low expression = 38.35 months, Mean OS of ZMYND8 high expression = 25.38 months). We observed that IDH1-R132H wildtype had a shorter OS than those with IDH1-R132H mutation (P<0.001, Fig 4b, Mean OS of IDH1-R132H wildtype=22.00 months; Mean OS of IDH1-R132H mutation=50.86 months). ATRX expression was significantly correlated with poor overall survival (P<0.001, Fig 4c, Mean OS of ATRX expression=25.25 months, Mean OS of ATRX loss=44.20 months). Notably, low ZMYND8 expression had a better PFS (Fig 4d).
Classification of gliomas defined by IDH1-R132H, ATRX and ZMYND8
Considering the above results, we classified gliomas into three groups, A: IDH1-R132Hmut and ATRX loss; B: IDH1-R132Hmut and ATRX expression; IDH1wt-ZMYND8 low expression; C: IDH1wt-ZMYND8 high expression. We observed that OS and PFS of patients were significantly different (Fig 4e-4f, P<0.0001). Group C was correlated with poor outcome (Table2, Mean OS = 15.90 months). Conversely, Group A was associated with better clinical survival (Mean OS =54.65 months). The mean OS of Group A is lower than Group B and C. Multivariate cox regression analysis confirmed that ZMYND8 and IDH1-R132H were independent prognostic factors of PFS and ATRX and IDH1-R132H were independent prognostic factors of OS (Table 4). Furthermore, we found that the distribution of patients’ age in the four groups was significantly different (Fig 4g, P<0.0001). It showed that Group A’s mean age (45) is younger than group C (57), and the distribution of three groups was significantly different (P<0.0001).