基于芯片和生物信息学分析结直肠癌肝转移中差异表达的微小RNA及其意义

Analysis of differentially expressed microRNAs in colorectal cancer with liver metastases and their signifi-cance based on microarray and bioinformatics

目的分析有肝转移与无肝转移的结直肠癌微小RNA(microRNA,miRNA)表达的差异及其与结直肠癌肝转移(colorectal liver metastases,CRLM)发生的关系。方法收集2011年2月1日至2018年9月10日浙江大学医学院附属第二医院60例有肝转移(n=29)和无肝转移(n=31)的结直肠癌新鲜标本,采用Agilent芯片检测miRNA表达,利用Gene Spring GX v11.5.1软件筛选出差异表达的miRNA。进一步收集2003年3月1日至2012年12月31日浙江大学医学院附属第二医院62例有肝转移(n=31)和无肝转移(n=31)的结直肠癌4%甲醛固定的石蜡包埋(formalin-fixed and paraffinembedded,FFPE)标本,对差异miRNA表达进行验证。利用Gene Ontology(GO)功能数据库及Kyoto Encyclopedia of Genes and Genomes(KEGG)数据库进行靶基因的信号通路的富集,推测CRLM相关通路。TCGA数据库分析CRLM组织与正常肝脏组织中miR494、miR19a、miR223和miR20a的表达。通过反向传播(back propagation,BP)神经网络模型构建CRLM的预测模型。结果对新鲜结直肠癌样本肝转移组和无肝转移组进行miRNA芯片分析,筛选出差异表达的miRNA,包括miR19a、miR20a、miR223和miR494。实时定量PCR(quantitative real-time PCR,qRT-PCR)验证显示,FFPE肝转移组和无肝转移组miR19a、miR20a、miR223和miR494的表达水平比较,差异均具有统计学意义(均P<0.05)。Kaplan-Meier法分析显示,FFPE肝转移组miR494、miR19a和miR223高表达和低表达患者总生存期比较,差异均具有统计学意义(均P<0.05)。TCGA数据库分析显示,miR494、miR19a、miR223和miR20a在正常肝脏组织和CRLM组织中的表达比较,差异均具有统计学意义(均P<0.05)。通过BP神经网络模型构建miR494、miR19a、miR223和miR20a联合预测CRLM发生的准确度高,曲线下面积(area under the curve,AUC)为1.000。结论miR494、miR19a、miR223和miR20a在CRLM患者原发灶中差异表达,与患者生存相关。miR494、miR19a、miR223和miR20a共同预测CRLM发生的准确度高。

Objective To analyze the differentially expressed microRNAs(miRNAs)in colorectal cancer with and without liver metas-tases and their correlation with colorectal liver metastases(CRLM).Methods Fresh specimen of 60 colorectal cancer patients with liver metastases(n=29)and without liver metastases(n=31)were collected from the Second Affiliated Hospital of Zhejiang University School of Medicine from February 1st,2011,to September 10th,2018,and miRNA expression was detected by Agilent microarray.Gene Spring GX v11.5.1 software was used to screen out differentially expressed miRNAs.Sixty-two 4%formalin-fixed and paraffin-embedded(FFPE)specimen of colorectal cancer with(n=31)and without(n=31)liver metastases were collected from the Second Affiliated Hospital of Zheji-ang University School of Medicine,from March 1st,2003 to December 31st,2012,to validate the differential miRNA expression.The Gene Ontology(GO)functional database and Kyoto Encyclopedia of Genes and Genomes(KEGG)database were used to enrich the signaling pathways of the target genes and infer the relevant pathways in CRLM.The expressions of miR494,miR19a,miR223 and miR20a in CRLM and normal liver tissues were analyzed in TCGA database.The prediction model for CRLM was constructed by back propagation(BP)neural network model.Results Using the fresh specimen of colorectal cancer patients with and without liver metastases,differentially expressed miRNAs,including miR19a,miR20a,miR223 and miR494,were screened out by miRNA microarray.For the FFPE specimen,quantitative real-time PCR(qRT-PCR)validated the differential expressions of miR19a,miR20a,miR223 and miR494 between the colorectal cancer tissues with and without liver metastases(all P<0.05),and Kaplan-Meier method showed the statistically significant differences in overall survival between patients with high and low expressions of miR19a,miR223 and miR494(all P<0.05).The expressions of miR494,miR19a,miR223,and miR20a were significantly different between CRLM tissues and normal liver tissues in the TCGA database(all P<0.05).The prediction model for CRLM using a combination of miR494,miR19a,miR223 and miR20a constructed by BP neural network model had high accuracy,and the area under the curve(AUC)was 1.000.Conclusions miR494,miR19a,miR223 and miR20a were differentially expressed in the primary foci of CRLM patients and related to patients'survival.The combination of miR494,miR19a,miR223,and mi-R20a predicts CRLM occurrence with high accuracy.