Integrated Analysis of RNA-Binding Proteins in Glioma6.162
Cancers (Basel) . 2020 Apr 7;12(4):892. doi: 10.3390/cancers12040892.
Abstract
RNA-binding proteins (RBPs) play important roles in many cancer types. However, RBPs have not been thoroughly and systematically studied in gliomas. Global analysis of the functional impact of RBPs will provide a better understanding of gliomagenesis and new insights into glioma therapy. In this study, we integrated a list of the human RBPs from six sources-Gerstberger, SONAR, Gene Ontology project, Poly(A) binding protein, CARIC, and XRNAX-which covered 4127 proteins with RNA-binding activity. The RNA sequencing data were downloaded from The Cancer Genome Atlas (TCGA) (n = 699) and Chinese Glioma Genome Atlas (CGGA) (n = 325 + 693). We examined the differentially expressed genes (DEGs) using the R package DESeq2, and constructed a weighted gene co-expression network analysis (WGCNA) of RBPs. Furthermore, survival analysis was also performed based on the univariate and multivariate Cox proportional hazards regression models. In the WGCNA analysis, we identified a key module involved in the overall survival (OS) of glioblastomas. Survival analysis revealed eight RBPs (PTRF, FNDC3B, SLC25A43, ZC3H12A, LRRFIP1, HSP90B1, HSPA5, and BNC2) are significantly associated with the survival of glioblastoma patients. Another 693 patients within the CGGA database were used to validate the findings. Additionally, 3564 RBPs were classified into canonical and non-canonical RBPs depending on the domains that they contain, and non-canonical RBPs account for the majority (72.95%). The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis showed that some non-canonical RBPs may have functions in glioma. Finally, we found that the knockdown of non-canonical RBPs, PTRF, or FNDC3B can alone significantly inhibit the proliferation of LN229 and U251 cells. Simultaneously, RNA Immunoprecipitation (RIP) analysis indicated that PTRF may regulate cell growth and death- related pathways to maintain tumor cell growth. In conclusion, our findings presented an integrated view to assess the potential death risks of glioblastoma at a molecular level, based on the expression of RBPs. More importantly, we identified non-canonical RNA-binding proteins PTRF and FNDC3B, showing them to be potential prognostic biomarkers for glioblastoma.
Keywords: RIP-seq; RNA-binding protein; WGCNAs; glioma; non-canonical RBP; prognosis; survival.
数据和方法
作者根据RBPs的结构域是否得到证实将它们分为标准(canonical)RBPs和非标准(non-canonical)RBPs两组,结构域信息来源是Pfam数据库;用DAVID做了GO和KEGG的富集分析以及可视化等分析。
作者从TCGA获得了包含3564个RBPs的表达谱,然后用R包WGCNA构建了一个共表达网络。首先,基于RBPs表达谱,将胶质瘤样本按照不同的临床特征进行层次聚类;然后,根据RBPs间的皮尔森相关系数筛选软阈值β,来确保网络的无标度性,这里β为6;根据TOM这一相似性量度,研究通过AL层次聚类法(组间距离等于两组对象之间的平均距离)将表达相似的RBPs聚到了同一模块。接下来,文中定义了两个参数MEs和GS, MEs代表第一个主成分相关的模块,它的值代表模块中的所有基因,GS表示基因和临床性状之间的相关性,用于量化单个基因与感兴趣的性状的关联关系。
生存分析
单因素COX回归分析用来评估RBPs的预后价值,并筛选出与生存时间相关的RBPs作为基因特征对TCGA和CGGA数据集进行风险分类。然后,通过多元Cox比例风险回归分析构建了风险特征,计算了每个患者的风险评分,最后根据中值将患者分成高低风险组。
有钱真好系列
该研究从ATCC购买了人类GBM细胞系LN229和U251,然后培养、切片、转染等。此外,作者还做了siRNA转染、细胞增殖、qPCR和Western Blot分析,以及RNA免疫沉淀和高通量测序。
统计分析
作者评估了大约3564个RBPs的表达水平。归一化后,用和作为单个患者中RBPs和TFs的表达丰度。DEGs使用R包DEseq2筛选的,绘制箱式图和对表达差异的统计分析用了R包ggplot2、ggpubr和ggsignif。单变量和多变量Cox比例风险回归分析用的都是survival包。生存曲线采用的是Kaplan-Meier法。
结果部分
图1. RBP在胶质瘤中的表达明显高于转录因子
RPBs表达与临床特征
图2. RBPs表达与胶质瘤临床特征的关系
RBPs的功能富集分析
图3. 基于RNA结合域的功能富集分析
RBPs共表达网络模块
图4. WGCNAs识别的共表达网络模块
绿色模块中RBPs的生存分析
图5. 不同数据集的生存分析
非标准RBPs参与维持细胞生长
图6. MTS法测定的LN229细胞增殖
PTRF靶标的RIP分析
图7. RIP-seq分析
转自生信人