目的：肿瘤免疫微环境对肿瘤患者的临床结局起着至关重要的作用，肿瘤免疫治疗的成功依赖于免疫效应机制的诱导，免疫效应机制与免疫微环境中肿瘤特异性细胞毒性T淋巴细胞（cytotoxic T lymphocytes，CTL）的产生有关。CD8阳性T淋巴细胞是免疫微环境的主要效应细胞之一，在肺腺癌(lung adenocarcinoma，LUAD)的发生发展中起重要作用。本研究旨在探索与LUAD中CD8阳性T淋巴细胞浸润相关的关键基因，并以此为基础建立一种新的预后模型。

方法：本研究选取了肿瘤基因组图谱(The Cancer Genome Atlas，TCGA)数据库中LUAD相关数据进行了分析，对基于这些癌症样本的差异表达基因(differently expressed genes，DEG)进行加权基因共表达网络分析(Weighted Gene Co-Expression Network Analysis，WGCNA)，结合CIBERSORT算法，选取WGCNA中与CD8阳性T淋巴细胞相关性最显著的基因模块进行后续分析。然后通过共表达网络、蛋白质相互作用网络（protein -protein interaction network，PPI network）、cox单因素回归分析和lasso降维方法确定关键基因。基于这些关键基因构建了风险评估模型，并利用基因表达综合数据库(Gene Expression Omnibus，GEO)中的数据集以及肺腺癌组织芯片的多色免疫荧光原位杂交实验进行了验证。

结果：根据TCGA数据库中529例LUAD相关数据 (癌症样本478例，癌旁样本51例)调取出8807个基于癌症的差异表达基因，其中包括2172个上调基因和6635个下调基因；然后基于DEGs进行加权共表达网络分析(WGCNA)，共得到11个分类模块，利用CIBERSORT算法对每个样本进行免疫细胞浸润分析，得到各T cell亚类的比例，将分类模块与细胞比例进行关联分析，得到与CD8+T cell浸润相关的模块，其内包含275个基因；进一步筛选模块中的关键基因（hub genes），得到93个具有高节点度(degree)的基因，利用模块中的基因构建PPI network，从中筛选到46个具有高degree的蛋白质编码基因，合并两种筛选的结果，最终得到117个免疫细胞浸润相关的hub genes；利用批量cox单因素回归进行筛选，得到34个预后相关基因，对这些基因进行lasso降维，最终确定了CD8阳性T淋巴细胞浸润相关的5个关键基因(MZT2A、ALG3、ATIC、GPI、GAPDH)，同时在此基础上建立了风险评估模型，即：风险评分（Risk score）= MZT2A * 0.035+ ALG3 * 0.084+ ATIC * 0.104+ GPI * 0.125+ GAPDH * 0.134。为了验证模型效能，计算每个样本的风险得分，然后以中位数为节点划分高低风险组，并绘制Kaplan-Meier曲线，曲线中高低风险组的差异显著，同时绘制风险模型的受试者工作特征曲线(Receiver Operating Characteristic Curve, ROC)，计算曲线下面积（Area Under Curve，AUC），得到的AUC值大于0.6，表明风险评分可以很好地预测LUAD的预后，风险评分与CD8阳性T淋巴细胞浸润程度呈负相关，与肿瘤分期呈负相关。GEO数据库和组织芯片结果与TCGA数据库结果一致。此外，风险评分在肿瘤组织中显著上调，且与肿瘤分期相关。

结论：本研究筛选出了与CD8阳性T淋巴细胞浸润相关的关键基因，利用这些关键基因建立了肺腺癌的预后风险评估模型，此可能为LUAD预后的预测提供一种新的方法，即根据评分确定肺腺癌的预后情况，评分同时与CD8阳性T淋巴细胞浸润相关，其为探讨LUAD中与CD8阳性T淋巴细胞浸润相关的肿瘤免疫微环境机制提供新的前景。

Objective：The tumor immune microenvironment plays a vital role in the clinical outcome of tumor patients. The success of tumor immunotherapy depends on the induction of immune effect mechanisms, immune effect mechanisms and tumor-specific cytotoxic T lymphocytes (CTL). CD8⁺ T cells are one of the main effector cells of the immune microenvironment and play an important role in the occurrence and development of lung adenocarcinoma (LUAD). This study aims to explore the key genes related to CD8+ T cell infiltration in LUAD and establish a new prognostic model based on this.

Method：This study selected the LUAD-related data from the tumor genome atlas (TCGA) database to analyze, and performed a weighted gene co-expression network analysis (WGCNA) on the differentially expressed genes (DEGs) based on these cancer samples. Combined with the CIBERSORT algorithm, select the most significant gene modules in WGCNA with CD8+ T cells for subsequent analysis. Then the key genes were identified by co-expression network, protein-protein interaction network (PPI network), Lasso-penalized Cox regression analysis. Based on these key genes, a risk assessment model was constructed and verified using the data set in Gene Expression Omnibus (GEO) and multiple fluorescence in-situ hybridization experiment of lung adenocarcinoma tissue microarray.

Result：According to the data of 529 cases of LUAD in the TCGA database (478 cases of cancer samples, 51 cases of adjacent cancer samples), 8807 differently expressed genes (DEGs) based on cancer were selected, including 2,172 up-regulated genes and 6,635 down-regulated genes. Then weighted co-expression network analysis (WGCNA) based on DEGs was performed, a total of 11 classification modules are obtained, and use the CIBERSORT algorithm to analyze the immune cell infiltration of each sample to obtain the proportion of each T cell subtype, Analyze the correlation between the classification module and the cell ratio, and obtain the module related to CD8⁺T cell infiltration, which contains 275 genes. The key genes in the module were further screened hub genes, and 93 genes with high degree were obtained. The genes in the module were used to construct a PPI network, and 46 protein-coding genes with high degree were screened out. Combine the results of the two screenings, and finally get 117 hub genes related to immune cell infiltration; Screening by batch cox single factor regression, 34 prognostic-related genes were obtained, and then these genes were lasso-reduced, and finally 5 key genes related to CD8⁺ T cell infiltration were identified (MZT2A, ALG3, ATIC, GPI, GAPDH), At the same time, a risk assessment model was established on this basis, Risk Score = MZT2A * 0.035+ ALG3 * 0.084+ ATIC * 0.104+ GPI * 0.125+ GAPDH * 0.134. In order to verify the effectiveness of the model, we calculate the risk score of each sample, and then divide the high and low risk groups with the median as the node, and draw the KM curve. The difference between the high and low risk groups in the curve is significant. At the same time, the ROC curve of the risk model is drawn to obtain the AUC. A value greater than 0.6 indicates that the risk score can predict the prognosis of LUAD well. The risk score is negatively correlated with the degree of CD8⁺ T cell infiltration, and negatively correlated with tumor staging. The results of GEO database and tissue microarray are consistent with the results of TCGA database. In addition, the risk score is significantly up-regulated in tumor tissues and is related to tumor staging.

Conclusion: This study screened out the key genes related to CD8⁺ T cell infiltration, and used these key genes to establish a prognostic risk assessment model for lung adenocarcinoma. This may provide a new method for predicting the prognosis of LUAD, that is, determine the prognosis of lung adenocarcinoma based on the score, the score is also related to CD8⁺T cell infiltration, which provides new prospects for exploring the tumor immune microenvironment mechanism related to CD8⁺T cell infiltration in LUAD.