【作者】 常江；

【导师】 林东昕；

【作者基本信息】 北京协和医学院 ， 肿瘤学， 2014， 博士

【摘要】 第一部分：我国人群食管癌预测模型研究我们最近的全基因组关联研究(genome-wide association study, GWAS)已发现25个食管癌易感基因或位点。是否和如何能将这些遗传易感标志应用于食管癌风险预测是后GWAS的重要课题。本研究运用GWAS发现的食管癌遗传易感标志即单核苷酸多态(SNP),结合传统的食管癌风险因素即性别、年龄及吸烟和饮酒暴露资料,建立中国人群的食管癌风险预测模型。建模在9805个食管癌患者及10493个正常对照中进行。我们发现在25个食管癌易感位点中,17个具有显著的边际效应(marginal effect),8个具有显著的基因-饮酒交互作用。我们将这些SNP进行遗传风险评分(genetic risk score, wGRS)和权重遗传风险评分(weighted genetic risk score, wGRS)后,用非条件性logistic回归进行关联分析。结果表明17个呈边际效应的SNP显著增加食管癌发病风险约4倍(P=1.49×10-164),8个呈基因-饮酒交互作用的SNP则仅在饮酒者中增加食管癌发病风险约4倍(P=8.76×10-41)。通过绘制ROC曲线(receiver operating characteristic cuver)和计算赤池信息量准则(Akaike’s Information Criterion, AIC)评估不同模型的拟合资料优良性,我们发现25个SNP以独立方式建立的食管癌风险预测模型AIC值最低,即具有最优拟合性。因此,我们以独立方式将这25个易感SNP及其与饮酒的交互作用,联合性别、年龄、吸烟及饮酒,建立预测模型。该模型的ROC曲线下面积(area under the curve, AUC)为70.9%,表明该模型可识别70.9%的食管癌病人。加入遗传易感标志比只用性别、年龄、吸烟及饮酒建立的模型预测能力提高了7%(P<0.0001)。这些结果表明整合遗传变异、生活方式因素及其交互作用于食管癌风险预测模型有助于鉴别食管癌患者。但是,还需要进一步研究,发现更多的遗传变异包括罕见变异,以更新该风险预测模型。第二部分：食管癌全基因组关联研究数据挖掘研究显示,有些染色体易感区段是多种不同类型肿瘤共有的易感区段。我们以往的GWAS发现13q22.1是重要的胰腺癌易感区段。通过在食管癌GWAS数据中对该区段进行基因型填补(imputation)和非条件性logistic回归分析,我们发现,在来自北京、河北和武汉的6177个食管癌患者和6179个正常对照中,位于该区段KLF5基因上游的rs1924966(OR=0.84,95%CI:0.80-0.89, P=1.37×10-10)和KLF5基因内含子区的rs115797771(OR=0.69,95%CI:0.62-0.78, P=2.32R10-10)与食管癌的易感性显著相关。我们还分析了这两个SNP与贲门癌(病例=1894,对照=1912)、胃体癌(病例=1007,对照=2243)以及结直肠癌(病例=1111,对照=1138)的关系。结果表明,它们与贲门癌的易感性相关(rs1924966, OR=0.84,95%CI:0.77-0.93, P=0.0003;rs115797771, OR=0.73,95%CI:0.60-0.89, P=0.0018),而与胃体癌(rs1924966, OR=1.04,95%CI:0.93-1.16, P=0.5001; rs115797771, OR=0.82,95%CI:0.65-1.05,P=0.1123)和结直肠癌(rsl924966, OR=0.89,95%CI:0.79-1.00, P=0.0527; rs115797771, OR=1.16,95%CI:0.91-1.48, P=0.2265)的易感性无关。rs115797771A/C变异还与IV期食管癌生存期相关；与AA基因型比较,AC和CC基因型生存时间长(MST=8vs MST=24;HR=0.41,95%CI:0.24-0.73, P=0.0021)。我们发现位于KLF5启动子区有一单碱基缺失变异(rs58090485),该变异与rs115797771高度连锁(r2=0.94)。生化实验表明该变异影响转录抑制因子结合,缺失等位基因的KLF5RNA表达水平显著高于插入等位基因。已知KLF5可通过激活NOTCH1表达而在食管癌中发挥抑癌基因的作用,因此rs58090485对KLF5转录活性的影响可能是其与食管癌易感性相关背后的机制。本研究表明13q22.1区段的遗传变异可能与多种消化道肿瘤发病风险相关,同时也提示KLF5可能在消化道肿瘤的发生发展中有重要作用。更多还原

【Abstract】 Part Ⅰ:Study on risk prediction of esophageal cancer in Chinese populationGenome-wide association studies (GWAS) have identified multiple genetic variants associated with risk of esophageal squamous-cell carcinoma (ESCC) in Chinese populations. In this study, we examined whether these genetic factors, along with non-genetic factors, could contribute to ESCC risk prediction. We examined25single nucleotide polymorphisms (SNPs) and4non-genetic factors (sex, age, smoking and drinking) associated with ESCC risk in9,805cases and10,493controls from Chinese populations. Weighted genetic risk score (wGRS) and unweighted genetic risk score (GRS) were calculated and logistic regression was used to analyze the association between wGRS or GRS and ESCC risk. We calculated the area under the curve (AUC) using receiver-operation-characteristic curve analysis to measure the discrimination after adding genetic variants to the model with only non-genetic factors. Net reclassification improvement (NRI) was used to quantify the degree of correct reclassification using different models. wGRS of the combined17SNPs with significant marginal effect (G SNPs) increased～4-fold ESCC risk (P=1.49×10-164) and the associations were significant in both drinkers and non-drinkers. However, wGRS of the8SNPs with significant effect in gene×drinking interaction (GE SNPs) increased～4-fold ESCC risk only in drinkers (Pinteration=8.76×10-41). The AUC for a risk model with4non-genetic factors,17G SNPs,8GE SNPs and their interactions with drinking was70.9%, with the significant improvement of7.0%compared with the model with only non-genetic factors (P<0.0001). Our results indicate that incorporating genetic variants, life-style factors and their interactions in ESCC risk models can be useful for identifying patients with ESCC. However, additional risk variants including rare risk variants are still need to be discovered by further studies to update this risk prediction model.Part Ⅱ:Data mining of esophageal cancer genome-wide association studyDifferent types of human cancer may share the same genetic susceptibility region. To test this hypothesis, we conducted association studies between variants at chromosome13q22.1, which has previously been identified as a susceptibility locus for pancreatic cancer and various other cancers in digestive system. By using imputation analysis and logistic regression analysis, we found two variants, rs1924966and rs115797771located respectively in the upstream and intronic region of KLF5, associated with risk of ESCC in a case-control cohort comprising6,177cases and6,179controls recruited from Beijing, Hebei province and Hubei province (rs1924966, OR=0.84,95%CI:0.80-0.89, P=1.37×10∷10; rs115797771, OR=0.69,95%CI:0.62-0.78,2.32×10-10). These two SNPs were also proved to be associated with cardiac cancer risk (rs1924966, OR=0.84,95%CI:0.77-0.93, P=0.0003; rs115797771, OR=0.73,95%CI:0.60-0.89, P=0.0018) but not gastric cancer (rs1924966, OR=1.04,95%CI:0.93-1.16, P=0.5001; rs115797771, OR=0.82,95%CI:0.65-1.05, P=0.1123) and colorectal cancer (rs1924966, OR=0.89,95%CI:0.79-1.00, P=0.0527; rs115797771, OR=1.16,95%CI:0.91-1.48, P=0.2265). Additionally, we found that rs11579771is associated with length of survival of patients with stage IV ESCC. Compared with the AA genotype, the AC and CC genotypes had longer survival time (8months vs24months, P=0.0021). Fine-mapping of this genetic region revealed a single base (A/-) Indel variant (rs58090485) that was also associated with risk of ESCC. This Indel variant is located in the promoter region of KLF5and is in strong linkage disequilibrium with rs115797771(r2=0.94). Biochemical assays showed that the A deletion disturbs a transcriptional repressor binding site and results in increased expression of KLF5compared with the A insertion. These results suggest that genetic variants at13q22.1are associated with risk of multiple types of cancer in the digestive system, which might point out an important role of KLF5in the carcinogenesis.更多还原