【作者】 张朋军；

【导师】 田亚平；

【作者基本信息】 南开大学 ， 生物化学与分子生物学， 2012， 博士

【摘要】 目的：本研究旨在证明Illumina450K芯片可以作为一种有效的血清甲基化谱筛选工具，同时以期能够得到原发性肝细胞癌相关甲基化位点。探索基于外周血的原发性肝癌诊断模型和基于血清多参数联合诊断用于鉴别健康对照组和疾病组，良性疾病组和恶性疾病组的诊断价值，为临床诊断提供一种辅助诊断方法。方法：利用Illumina450K甲基化芯片对原发性肝癌和正常对照组血清甲基化水平进行检测，亚硫酸盐甲基化测序验证DBX2和THY1甲基化位点；利用Affymetrix基因芯片对原发性肝癌和正常对照组外周血mRNA表达进行检测，然后从中筛选肝癌相关基因建立GeXP检测系统，通过多参数联合分析，以期能够建议一套基于外周血mRNA、GeXP检测平台和多参数联合分析的标准化诊断模型，用于原发性肝癌的临床辅助诊断。生化项目在日立7600全自动生化分析仪检测和罗氏Modular全自动生化分析仪上检测，免疫项目在罗氏E170EE全自动免疫分析仪和雅培i2000全自动免疫发光分析仪上检测，经标准品和质控品校准仪器后，检测试剂盒检测。细胞因子的10项指标在Luminex200液态芯片检测仪上检测。受试者工作曲线用来评价指标的诊断价值。多参数联合分析使用二元Logistic回归分析、判别分析、分类树分析和人工神经网络分析。结果：在本研究中，我们发现原发性肝癌组血清的全基因组甲基化水平显著低于正常对照组。与健康对照组相比，原发性肝细胞癌组的差异甲基化位点为7333个，约占450K芯片检测位点的1.5%。在这7333个差异甲基化位点当中，与健康对照组相比，甲基化水平降低的甲基化位点为6953个，甲基化水平升高的甲基化位点为380个。差异甲基化位点在每条染色体上的分布及比率不一致，可能导致染色体不稳定从而导致肿瘤的发生。差异甲基化位点主要位于启动子区域和CpG岛，尤其是甲基化水平升高的差异甲基化位点。我们将β值大于0.5定义为超甲基化，而β值小于0.2定义为超低甲基化。我们筛选出453个健康对照组和37个原发性肝细胞癌组的超甲基化位点，分别对这些超甲基化位点进行基因本体轮和功能富集分析后，在原发性肝细胞癌组中，我们筛选出3个基因功能富集的基因，在健康对照组中，我们筛选出28个基因功能富集的基因。基因相互关系分析后，我们选择DBX2超甲基化位点和THY1超甲基化位点进行验证。DBX2对于健康对照组和原发性肝细胞癌组的诊断灵敏性和特异性分别为89%和87%，而THY1分别为81%和85%。本研究通过对原发性肝癌和正常对照组外周血mRNA表达进行检测，基因芯片质量检测合格后，从中筛选肝癌相关的上调和下调各40个基因，然后从中选择15个相关基因，基于GeXP检测系统，经过基因特异性检测和引物浓度调节后，建立GeXP检测方法，建立最佳诊断模型(CALR、PFN1、SPAG9、ANXA1、HGF、FOS、GPC3和HPSA1B基因)区分正常对照组、乙肝组、肝硬化组、肝癌组和其他组的诊断准确率分别为80.57%，78.17%，84.48%，73.24%和85.85%，然后对模型进行验证，其区分正常对照组、乙肝组、肝硬化组、肝癌组和其他组的平均准确率分别是83.33%，73.33%，100%，75.00%和95.24%。通过多参数联合分析，能够建议一套基于外周血mRNA、GeXP检测平台和多参数联合分析的标准化诊断模型，用于原发性肝癌的临床辅助诊断。本研究通过检测61项指标在多种肿瘤中的含量，在区分健康对照组和疾病组时，发现受试者工作曲线下面积大于0.9以上的有CRP和IL-8两项指标，当CRP的检测阈值为0.29mg/L时，其检测灵敏性和特异性分别为89.90%和97.00%。当IL-8的检测阈值为25.38pg/mL时，检测灵敏性和特异性分别为83.90%和85.50%。CRP用于区分健康对照组和疾病组的诊断价值最好，联合诊断分析用于区分健康对照组和疾病组的诊断价值优于单项指标检测，但由于单项指标CRP的诊断灵敏性和特异性已经较高，因此，联合检测的诊断价值优势不明显。在区分良性疾病组和恶性疾病组时，受试者工作曲线下面积最大的指标CA724，其AUC值仅为0.589，当其诊断阈值为1.685时，其检测灵敏性和特异性分别为56.80%和59.00%。诊断价值非常有限。人工神经网络的诊断价值最好，将样品的70%用于训练，30%用于测试。进行ROC曲线分析，我们发现人工神经网络分析良性疾病组和肿瘤组的受试者工作曲线下面积为0.941。验证集的正确率为81.30%，肿瘤组在训练集中的正确率为88.70%，总体正确率为85.40%。结论：本研究证实450K芯片可以作为一种非常有前景的血清甲基化差异位点筛选工具，并且证明原发性肝细胞癌相关差异甲基化位点可能会作为一种临床原发性肝细胞癌的辅助诊断手段。基于外周血、GeXP技术和多参数联合分析技术，可以有效提高原发性肝癌的诊断价值。当CRP可有效用于区分健康对照组和疾病组时。当CA724用于区分良性疾病组和肿瘤组时，诊断价值有限，人工神经网络用于区分良性疾病组和肿瘤组时诊断价值最好。多参数联合分析结合了多个参数，可有效提高灵敏性和特异性，是一种非常有前景的临床辅助诊断方法。更多还原

【Abstract】 Objective: We aimed to demonstrate Illumina450K chips can beuseful and effective serum methylation profile screening tool and getsome methylation sites which was related to primary hepatocellularcarcinoma. We aimed to explore the primary hepatocellular carcinomadiagnostic model based on peripheral blood and multi-parameter jointdiagnosis. Based on the serum indicators, we aimed to identify thediagnostic value of healthy controls and disease groups, benigndisease and malignant disease groups which may provide acomplementary diagnostic method for clinical diagnosis.Methods: Illumina450K methylation chip was used to detect theserum methylation levels of hepatocellular carcinoma and normalcontrol group. Bisulfite methylation sequencing was used to validatethe methylation levels of DBX2and THY1sites. Affymetrix GeneChip was used to detect the peripheral blood mRNA expressionin hepatocellular carcinoma and normal control group, and then thescreened genes which were related to hepatocellular carcinoma wereused to establish GeXP detection system. Based on multi-parameterjoint diagnosis, peripheral blood mRNA and GeXP detection platform,a standardized diagnostic model which were used for clinicalcomplementary diagnosis was built. Biochemical indicators weredetected by Hitachi7600automatic biochemical analyzer and RocheModular automatic biochemical analyzer detection. Immunizationindicators were detected by Roche E170EE automatic immunoassayanalyzer and the Abbott i2000automatic immune luminescenceanalyzer.10cytokines indicators were detected by Luminex200liquid chip detector. The receiver operating curve was used to evaluatethe diagnostic value of the index. Multi-parameter joint analysis wasevaluated by the binary Logistic regression analysis, discriminant analysis, classification tree analysis and artificial neural networkanalysis.Results: In our study, we found that the serum level of wholegenome-wide methylation in primary liver cancer group wassignificantly lower than the healthy control group. Compared withhealthy control group, the number of differentially methylation siteswas7333in hepatocellular carcinoma group, accounting for1.5%of450K microarray. In the7333differentially methylated sites, thenumber of reduced methylation sites was6953, the number ofincreased methylation sites was380. Distribution and percentage ofdifferentially methylated loci on each chromosome was inconsistent, itmay be related to chromosome instability which leaded tohepatocellular carcinoma. Differentially methylated sites locatedmainly in the promoter region and CpG islands. Beta value greaterthan0.5was defined as hypermethylation. Beta value which was less than0.2is defined as hypomethylation. We had screened453hypermethylation sites in healthy controls group and37hypermethylation sites in hepatocellular carcinoma group. After geneontology and functional enrichment analysis, we screened28genefunction enrichment of genes in the healthy control group and3genefunction enrichment of genes in the hepatocellular carcinomagroup. After gene interaction analysis, we selected DBX2and THY1hypermethylation sites for validation. The diagnostic sensitivity andspecificity of DBX2for differentiating healthy control group and thethe hepatocellular carcinoma group were89%and87%, while THY1were81%and85%respectively. In our study, we detected the mRNAexpression in peripheral blood of primary hepatocellular carcinomaand healthy normal control group. After the quality of gene chip wasevaluated. We selected40up-regulated genes and40down-regulatedgenes in hepatocellular carcinoma group, and then15genes were selected and detected by GeXP detection system, after the adjustmentof the gene-specific detection and primer concentration. We built theGeXP detection methods for hepatocellular carcinoma diagnosis. Thebest diagnostic model was CALR, PFN1, SPAG9, ANXA1, HGF,FOS, GPC3and HPSA1B gene. The diagnostic accuracy rate ofdistinguishing the normal control group, hepatitis B group group, livercirrhosis, liver cancer group and other groups were80.57%,78.17%,84.48%,73.24%and85.85%, and then the model was validated, thediagnostic accuracy rate of distinguishing the normal control group,hepatitis B group group, liver cirrhosis, liver cancer group and othergroups were83.33%,73.33%,100%,75.00%and95.24%,respectively. We had built a standardized diagnostic model based onperipheral blood mRNA, GeXP detection platform andmulti-parameter for the clinical diagnosis of primary liver cancer. Inour study, we detected61indicators in serum of tumors to distinguish between healthy control group and disease group. We found the areaunder the receiver operating curve of CRP and IL-8were greater than0.9. When the threshold value of CRP detection was0.29mg/L, thediagnostic sensitivity and specificity were89.90%and97.00%. Whenthe threshold of IL-8detection was25.38pg/mL, the sensitivity andspecificity were83.90%and85.50%. CRP was the best indicator todistinguish the healthy controls and disease groups when the indicatorwere detected alone. The joint diagnostic analysis to distinguishbetween healthy controls and disease groups can increase diagnosticvalue when compared to the single indicator detection, however, thediagnostic sensitivity and specificity of CRP were high enough,therefore, the diagnostic value of joint detection showed no obviousadvantages. The largest area under the receiver operating curve fordistinguishing between benign disease and malignant disease groupwas CA724, but the value of area under curve was only0.589. When the diagnostic threshold was1.685, the diagnostic sensitivity andspecificity were56.80%and59.00%. The diagnostic value was verylimited. The artificial neural network diagnostic method divided70%of the samples used for training,30%used for testing. After areaunder curve analysis, we found that the area under the receiveroperating curve analyzed by artificial neural network analysis fordifferentiating the benign disease group and tumor group was0.941. The overall correct rate of the validation set was81.30%,88.70%in the tumor training group and85.40%in the healthy controlgroup.Conclusion: Our study demonstrated that the450K chip was a verypromising serum methylation screening tool. The methylated siteswhich were related to primary hepatocellular carcinoma may serve asa complementary diagnostic method for clinical primaryhepatocellular carcinoma diagnosis. When we combined the peripheral blood, GeXP technology and multi-parameter analysistechniques together, it can effectively improve the diagnostic value ofprimary hepatocellular carcinoma. CRP was an effectively indicatorfor distinguishing between healthy controls and diseasegroups. When CA724was used for distinguishing between thebenign disease group and tumor group, the diagnostic value waslimited. The artificial neural network which was used fordistinguishing the benign disease group and tumor group had the bestdiagnostic value. Multi-parameter joint analysis which combinedmultiple parameters can effectively improve the diagnostic sensitivityand specificity, it may be a very promising clinical complementarydiagnosis.更多还原