Objective: To establish a recurrence risk prediction model for meningioma based on HOXA9 DNA methylation.
Methods: Meningioma-related datasets were downloaded from GEO database for screening homeobox genes (HOXs) with prognostic values using differential methylation and ROC curve analysis and Cox regression analysis. The differentially methylated CpG sites with high predictive efficacy were selected to establish the risk prediction model using Lasso-Cox regression analysis, based on which the patients were divided into high- and low-risk groups by the cutoff value. The methylation levels of CpG sites were verified at the cell and tissue levels using methylation-specific PCR (MS-PCR). Clinical meningioma tissue samples were used to validate the predictive efficacy of the model.
Results: HOXA9 methylation level was significantly up-regulated in meningiomas (P< 0.001) and showed a high diagnostic efficiency (AUC=0.884) as an independent risk factor for overall survival (P< 0.01) positively correlated with the degree of malignancy and poor prognosis of meningioma (P< 0.05). Risk stratification by HOXA9 methylation was more accurate than WHO grading for predicting recurrence and patient survival time. The AUCs of the sites cg03217995 and cg21001184 were both above 0.8 for meningioma diagnosis and above 0.6 for predicting recurrence. The patients' clinical characteristics differed significantly between the high- and low-risk groups (P< 0.001), and the prediction score of the model was an independent prognostic factor for meningioma (P< 0.05). MSPCR results showed that the methylation levels of the two sites increased significantly in meningioma cells. In clinical samples, the combined model showed a high prediction efficiency (AUC=0.857), and the predicted risk of progression was highly consistent with the patients' actual condition.
Conclusion: High HOXA9 methylation level is a predictor for poor prognosis of meningiomas, and the combined prediction model based on its CpG sites provides a new approach to early screening of meningioma patients at risk of progression.
目的: 探讨同源异形盒基因（HOXs）在脑膜瘤中的甲基化模式，筛选对脑膜瘤复发风险分层有临床指导意义的HOXs基因，建立预测模型并评估其预测效能。
方法: 利用GEO数据库下载脑膜瘤相关数据集。通过甲基化差异分析和ROC曲线分析筛选有预后评估价值的HOXs基因，再通过Cox回归分析、分子特征分析对特征基因的临床应用价值进行验证。进一步筛选差异CpG位点并评估其预测效能，通过Lasso-cox回归分析建立预测模型，根据cut off值将患者分成高、低风险组并进行分析。最后通过甲基化特异性PCR（MS-PCR）在细胞和组织水平验证差异CpG位点的甲基化水平，并纳入脑膜瘤组织样本验证该模型的预测效能。
结果: HOXA9甲基化水平在脑膜瘤中显著上调（P<0.001），且具有较高诊断效能（AUC=0.884）。验证分析表明HOXA9甲基化是影响脑膜瘤患者总生存期的独立危险因素（P<0.01），与脑膜瘤恶性程度和不良预后正相关（P<0.05），且基于HOXA9甲基化水平的分组方法在预测患者复发和生存时间时比WHO分级精度更高。筛选出的CpG位点cg03217995和cg21001184对脑膜瘤诊断的AUC均大于0.8，预测脑膜瘤患者复发的AUC均大于0.6。构建的两位点联合预测模型cut off值为1.226，以此分组的患者临床特征均有显著性差异（P<0.001），并且该模型预测评分是脑膜瘤的独立预后因素（P<0.05）。MS-PCR结果显示，位点cg03217995和cg21001184甲基化水平在脑膜瘤细胞中升高（P<0.0001），在不同WHO分级患者间无统计学差异。临床样本分析表明联合模型有较高预测效能（AUC=0.857），预测状态与患者真实临床进展结果高度一致。
结论: HOXA9甲基化是脑膜瘤预后不良的有效预测指标，基于其CpG位点的联合预测模型有望成为恶性进展风险病例早筛的新方法。
Keywords: CpG sites; DNA methylation; HOXA9; diagnosis; meningioma; prognosis.