广西医学

目的基于癌症相关成纤维细胞（CAFs）构建骨肉瘤预后风险评分预测模型，并分析其与免疫微环境的关系。方法基于GSE162454数据集中6名骨肉瘤患者的单细胞RNA测序数据，鉴定CAFs亚群并筛选CAFs亚群的差异表达基因（DEGs）。针对CAFs的DEGs，使用加权基因共表达网络分析方法进行基因聚类，并利用cor函数筛选得出 CAFs 的免疫基因（IG_CAFs）。采用单因素和多因素COX回归模型、最小绝对收缩与选择算子回归模型分析与骨肉瘤患者预后独立相关的基因，用于构建风险评分预测模型，并评估其预测性能。使用xCell、MCPcounter和CIBERSORT算法评估风险评分预测模型与免疫细胞浸润水平的相关性。采用qRT⁃PCR分析IG_CAFs在人类成骨细胞（hFOB1.19）、骨肉瘤细胞系（143B、SA0S2、HOS 和 MG63）中的表达情况。结果共筛选出912个CAFs的DEGs，6个IG_CAFs（BCAR1、BMP1、IFITM3、RGS5、SERPINE2和TUSC1）是与骨肉瘤患者预后独立相关的基因，风险评分预测模型=（-0.120×BCAR1）+（0.039×BMP1）+（-0.005×IFITM3）+（-0.119×RGS5）+（0.013×SERPINE2）+（0.112×TUSC1）。该风险评分预测模型在测试数据集和验证数据集中的预测性能良好。在3个免疫浸润算法中，风险评分预测模型与单核巨噬细胞系统之间存在较强的相关性。与hFOB1.19细胞相比，BCAR1、TUSC1、RGS5、BMP1、IFITM3和SERPINE2在4种骨肉瘤细胞系中存在差异表达。结论基于SERPINE2、BMP1、TUSC1、RGS5、IFITM3和BCAR1构建的风险评分预测模型，对骨肉瘤患者预后的预测性能较好，这6个IG_CAFs可能通过调节单核巨噬细胞来影响免疫微环境和患者预后。

Objective To establish the risk score prediction model of osteosarcoma prognosis based on cancer⁃associated fibroblasts (CAFs), and to analyze its relation with immune microenvironment. Methods Based on single cell RNA sequencing data of 6 patients with osteosarcoma in GSE162454 dataset, CAFs subsets were identified and differentially expressed genes (DEGs) in CAFs subsets were screened. For DEGs of CAFs, weighted gene co⁃expression network analysis was used for gene clustering, and cor function was used to screen immune genes of CAFs (IG_CAFs). The univariate and multivariate COX regression models and least absolute shrinkage and selection operator regression model were used to analyze genes independently correlated with the prognosis of patients with osteosarcoma to establish a risk score prediction model and evaluate its predictive performance. The xCell, MCPcounter and CIBERSORT algorithms were used to evaluate the correlation of risk score prediction model with level of immune cell infiltration. The qRT⁃PCR was used to analyze the expression of IG⁃CAFs in human osteoblasts (hFOB1.19) and osteosarcoma cell lines (143B, SA0S2, HOS and MG63). Results A total of 912 DEGs of CAFs were screened, and 6 IG_CAFs (BCAR1, BMP1, IFITM3, RGS5, SERPINE2 and TUSC1) were independently correlated genes with the prognosis of patients with osteosarcoma. The risk score prediction model was equal to (-0.120×BCAR1)+(0.039×BMP1)+(-0.005×IFITM3)+(-0.119×RGS5)+(0.013×SERPINE2)+(0.112×TUSC1). The prediction performance of risk score prediction model in test dataset and validation dataset was favorable. Among the three immune infiltration algorithms, there was a relatively strong correlation between the risk score prediction model and monocyte⁃macrophage system. Compared with hFOB1.19 cells, BCAR1, TUSC1, RGS5, BMP1, IFITM3 and SERPINE2 were differentially expressed in four categories of osteosarcoma cell lines. Conclusion The risk score prediction model established based on SERPINE2, BMP1, TUSC1, RGS5, IFITM3 and BCAR1 has a good prediction performance for the prognosis of patients with osteosarcoma. These 6 IG_CAFs may affect the immune microenvironment and the prognosis of patients by regulating monocyte⁃macrophages.