基于公共数据库构建肺腺癌肿瘤干性评分模型预测免疫治疗疗效

Construction of a stemness-based scoring model predicting the efficacy of immunotherapy in lung adenocarcinoma based on public databases

目的鉴定肺腺癌肿瘤干细胞相关基因亚型,构建肿瘤干性评分模型以预测肺腺癌免疫检查点抑制治疗疗效。方法从TCGA数据库下载肺腺癌RNA测序数据,使用"limma"包分析肺腺癌(535例)与癌旁组织(59例)中329个肿瘤干细胞相关基因的差异表达(FDR<0.05,|log_(2)Fold Change|>2),利用差异基因鉴定肺腺癌肿瘤干细胞相关亚型,通过单因素Cox回归分析进一步筛选出肿瘤干细胞相关亚型之间对预后有意义的共同差异基因。基于主成分分析(PCA)算法,利用123个预后有意义的共同差异基因对TCGA与GEO合并后的630例肺腺癌患者进行肿瘤干性评分,利用Kaplan-Meier曲线分析确定最佳截断值,将肺腺癌患者分成高、低肿瘤干性评分组(截断值为-1.91)。探究不同肺腺癌肿瘤干细胞相关亚型和肿瘤干性评分组在肿瘤微环境、免疫治疗方面的差异。结果鉴定出了36个差异表达基因和3个预后有统计学意义的肿瘤干细胞相关亚型(CSC-A、CSC-B、CSC-C)(P=0.033),其在免疫细胞浸润方面差异有统计学意义并与抗原递呈、细胞毒性作用等多条免疫通路相关。单因素Cox回归分析筛选出123个对预后有意义的共同差异基因,构建了肿瘤干性评分模型。低肿瘤干性评分组各类免疫细胞浸润程度普遍上升,PD1、PD-L1、CTLA4表达显著升高。无论是单独的抗CTLA4或抗PD1治疗,亦或是二者联合治疗,低肿瘤干性评分组的疗效都优于高肿瘤干性评分组,无免疫检查点抑制治疗时,高、低肿瘤干性评分组的疗效差异无统计学意义(P=0.060)。在抗PD-L1和抗PD1的两个独立免疫治疗队列中,低肿瘤干性评分组的反应率均高于高肿瘤干性评分组(抗PD-L1治疗队列反应率:50%vs 20%;抗PD1治疗队列反应率:23%vs 0%)。结论肿瘤干性评分模型在预测肺腺癌患者免疫检查点抑制治疗疗效方面具有潜在价值,有望为肺腺癌患者免疫检查点抑制治疗提供理论依据。

Objective To predict immune checkpoint blockade(ICB)response in lung adenocarcinoma(LUAD)by identifying LUAD subtypes related to cancer stem cells and constructing a stemness-based scoring model.Methods LUAD RNA-seq data were obtained from TCGA database.By“limma”package,329 differentially expressed genes(DEGs)related to cancer stem cells between LUAD(535 cases)and adjacent tissues(59 cases)were identified to classify LUAD into different subtypes(FDR<0.05,|log_(2)Fold Change|>2).By univariate Cox regression analysis,the common prognostic DEGs among different subtypes were further screened out.Using principal component analysis(PCA)and the 123 common prognostic DEGs,a stemness-based scoring model was established for 630 LUAD patients from TCGA and GEO.The cutoff value,determined by Kaplan-Meier curves analysis,was used to stratify LUAD patients into high-and low-score groups(cutoff value=-1.91).Furthermore,difference of distinct subtypes and stemness-based scores on tumor microenvironment(TME)and ICB therapy were analyzed.Results Thirty-six differentially expressed genes and three LUAD subtypes related to cancer stem cells(CSC-A,CSC-B,and CSC-C)were identified,overall survival rates of which were statistically different(P=0.033).The three subtypes greatly affected immune infiltration levels and were associated with multiple immune pathways,such as antigen presentation and cytotoxicity.A total of 123 common prognostic genes(P<0.05)were screened out to construct stemness-based scoring model by univariate Cox regression.In the low-score group,the infiltration of various immune cells and mRNA expressions of PD1,PD-L1 and CTLA4 were up-regulated.No matter anti-CTLA4 or anti-PD1 treatment alone,or combination of them,efficacy of the low-score group was better than that of the high-score group,and there was no significant difference in the efficacy of the two groups without ICB(P=0.060).In the anti-PD-L1 and anti-PD1 immunotherapy cohorts,the response rates of the low-score group were higher than that of the high-score group(the response rate of the anti-PD-L1 treatment cohorts:50%vs 20%;the response rate of anti-PD1 treatment cohorts:23%vs 0%).Conclusion Stemness-based scoring model has a potential to predict the efficacy of ICB therapy in LUAD patients,which is expected to provide a theoretical basis for ICB therapy in LUAD patients.