利用新筛选的血清分子标志物构建神经管畸形产前诊断模型及其机制研究

Neural tube defects (NTDs) are the most serious congenital malformation harmful to human health and current therapeutics for NTDs are unsatisfactory. The study on non-invasive early prenatal diagnosis of NTDs can provide clues to guide clinical interventions before the formation of severe structural deformity. Using various high-quality screening technology, we have found 21 new molecule biomarkers (including miRNAs and proteins) in serum of pregnant women with neural tube defects. In the present project, these diversiform biomarkers data will be analyzed in artificial neural network (ANN), a powerful and innovative mathematical method that tracks/categorizes numerical inter-relationships between the nonlinear data, to establish a new prenatal diagnosis model for NTDs. And we will take advantages of the cooperation with a number of domestic birth defects research centers for multicenter and large sample validation. To identify the origin of these molecular markers from the maternal or fetal intrinsic proteins, we will detect the expressive disciplinarian of proteins in the serum from pregnant rats, serum from fetal rats, amniotic fluid and placenta using proteomic technology. At the same time, more novel NTDs specific molecular markers can be discovered, and be used to improve the ANN diagnosis model. Moreover, based on the novel discovery of a large of candidate biomarkers obtained in maternal serum belong to the cytoskeleton family, we will generate a LMNA/GFP transgenic mouse model, and transfer the zygote from LMNA/GFP transgenic mouse to oviducts of a normal mouse recipient. To explore the pathway of cytoskeletal proteins enter into maternal serum, the spatio-temporal expression of LMNA\GFP fusion protein will be detected in the serum of maternal receptor, serum of fetal rats, amniotic fluid, placenta, fetal heart and spinal cord on different development stages. This study not only could promote the application of complex biomarkers in clinical screening of NTDs, and also might offer a further avenue for non-invasive prenatal diagnosis for other congenital malformations disease.

研究神经管畸形（NTDs）的早期无创诊断方法，使其能够在神经不可逆损伤出现之前获得诊断，对于早期干预有重要意义。本课题将利用网络科学中可以对非线性数据分析与处理的人工神经网络（ANN）技术，将前期研究利用不同组学技术筛选出的21个NTDs 分子标志物（包括miRNA和蛋白质）构建新的NTDs诊断模型，并进行多中心、大样本的验证。同时利用蛋白质组学技术分析NTDs孕鼠血清、胎鼠血清、羊水和胎盘的蛋白质谱变化规律，从母体免疫、应激反应和胎儿本身固有蛋白质两方面综合分析，明确这些分子标志物的来源，同时也可筛选更多新的NTDs特异性分子标志物，不断完善ANN诊断模型。另外，在发现分子标志物中存在大量细胞骨架蛋白这一新现象的基础上，本课题将以表达LMNA和GFP融合蛋白的转基因小鼠为供体，以正常小鼠为受体母鼠进行受精卵移植，系统观察LMNA和GFP融合蛋白进入母鼠血清的时间点和途径，明确其具体机制。

神经管畸形等常见先天畸形是由环境因素和遗传因素共同作用的复杂多基因疾病，发病机制不清，治疗上也无良策，寻找新的产前筛查与诊断的方法对于早期干预治疗、减少致残率具有重要意义。本课题利用出生队列收集的孕妇血清标本，完成了常见严重先天畸形的胚胎早期无创诊断的分子标志物筛查，并利用计算机机器学习的方法构建畸形早期诊断与预警模型，同时也对神经管畸形等常见畸形的相关致病分子机制进行深入探讨。本课题招募了4117名孕妇加入出生队列, 采集6646例孕妇血样，其中包括先天性心脏病192例、唇腭裂82例和神经管缺陷100例。利用蛋白组质组学方法筛选、生物信息学分析和扩大样本量验证，发现补体系统蛋白质可以作为神经管畸形胎儿诊断的分子标志物。利用人工智能机器学习的方法进行分类和计算机诊断模型构建，结果显示补体（C1QA、C1S和C3）和AFP联合构建神经管畸形的诊断模型的准确率、敏感度和特异度均提高到75%。利用神经管畸形动物模型和临床孕妇外周血的外泌体进行蛋白质组学筛选和验证，发现CORO1A和DNM2是神经管畸形早期诊断分子标志物。利用miRNA芯片进行筛选和验证，发现三种miRNA是先心病诊断分子标志物。利用全转录组测序分析对唇腭裂、神经管畸形和先心病孕妇外周血的血浆外泌体进行筛选和验证，发现三种piRNA这三种畸形诊断的分子标志物。利用全转录组学测序筛选和验证，首次证实EN2/LIN28A/has-let-7a-3p/HHIP/GLI2调控轴可能参与唇腭裂胚胎发生过程。应用人工智能计算机卷积神经网络技术建立流产的早期预警计算机模型，结果显示其准确性为72.68％。利用全基因组甲基化芯片、全转录组测序和验证，发现TRIM4甲基化异常、miR-322调控NOX4异常、氧化应激与线粒体代谢相关分子表达异常与神经管畸形发生有关。本课题的研究已发表标注课题资助的论文22篇（其中SCI收录论文17篇），申请发明专利5项，其中1项于2020年获得成果转化，转化经费100万元。本课题培养5名博士研究生和4名硕士研究生。