免疫评分模型精准预测乳腺癌结局

　　肿瘤进展是一个复杂的过程，需要癌细胞、微环境和免疫系统相互作用，影响肿瘤的启动和进展。免疫细胞对癌症具有重要的辅助功能并影响临床结局，对于免疫细胞浸润丰度较高的患者，治疗效果和临床结局往往较好。近年来，免疫治疗已经成为癌症精准治疗的一大热点，故有必要充分考虑癌症的肿瘤浸润免疫细胞。对此，大连医科大学附属第二医院李曼教授团队开展研究，构建了由肿瘤浸润免疫细胞生物标志组成的免疫评分模型，用于预测和改善乳腺癌的治疗效果和生存结局，并探讨了乳腺癌肿瘤浸润免疫细胞的预后价值和临床意义。

　　该研究分析数据来自美国国家癌症研究所和国家人类基因组研究所癌症基因组图谱TCGA数据库、美国国家生物技术信息中心基因表达数据库GEO、欧洲生物信息研究所功能基因组学数据库Array Express、国际癌症基因组联盟ICGC数据库、国际乳腺癌分子分类学联盟METABRIC数据库的29个数据集，共包括6844个样本，同时还包括临床收集的183个样本。通过CIBERSORT软件评估样本的22种免疫细胞浸润丰度，剔除CIBERSORT分析P值>0.05的样本。通过随机分层将5038例样本以7∶3的比例分为训练集和验证集，并收集中国医院173例乳腺癌组织作为测试集。随后，在训练集中应用包括单因素比例风险回归、多因素比例风险回归、最小绝对收缩选择算子（LASSO）回归等构建由6种肿瘤浸润免疫细胞组成的免疫模型，用于预测患者的化疗效果和总生存结局。最后，对验证集和测试集进一步验证和测试该模型预测结局的准确性和有效性。

　　结果，该研究根据单因素比例风险回归分析，确定了14种免疫细胞与乳腺癌患者的总生存显著相关，进一步采用LASSO分析和多元比例风险回归分析，构建了由6种免疫细胞（静息CD4阳性T淋巴细胞、调节型T淋巴细胞、γδT淋巴细胞、活化自然杀伤细胞、单核细胞、M0型巨噬细胞）组成的免疫评分模型。在训练集中，免疫低风险组和高风险组的总生存差异具有显著意义，其20年总生存率分别为42.6%和26.3%。随后，在验证集和测试集中，也证实了该模型可预测乳腺癌患者总生存，且与乳腺癌的分子分型无关。用免疫评分模型预测乳腺癌患者化疗效果，结果发现无论接受何种化疗方案，低风险组接受化疗的患者都有显著生存优势。改研究还整合了免疫评分及其他临床病理预后因素，包括年龄、肿瘤分级和TNM分期，构建了列线图。列线图预后系统总体上改善了乳腺癌的预后模型。根据校准曲线，列线图对训练组和验证组的5年、10年和20年生存率预测良好，决策曲线表明该列线图比标准TNM分期系统的预测精度更好，充分体现了免疫细胞的临床意义。

　　因此，该研究结果表明，由免疫浸润细胞组成的免疫评分模型可以有效地用于预测乳腺癌患者的化疗效果和生存结局。

Theranostics. 2020 Oct 25;10(26):11938-11949.

An immune cell infiltration-based immune score model predicts prognosis and chemotherapy effects in breast cancer.

Sui S, An X, Xu C, Li Z, Hua Y, Huang G, Sui S, Long Q, Sui Y, Xiong Y, Ntim M, Guo W, Chen M, Deng W, Xiao X, Li M.

The Second Affiliated Hospital of Dalian Medical University; Institute of Cancer Stem Cell, Dalian Medical University, Dalian, China; Sun Yat-Sen University Cancer Center, Guangzhou, China; The First Affiliated Hospital of Dalian Medical University, Dalian, China; Institute (College) of Integrative Medicine, Dalian Medical University, Dalian, China; The First People's Hospital of Yichang (People's Hospital of Three Gorges University), Yichang, Hubei, China; Vagelos College of Physicians and Surgeons, Columbia University, New York, USA.

BACKGROUND: Immune cells have essential auxiliary functions and influence clinical outcomes in cancer, with high immune infiltration being associated with improved clinical outcomes and better response to treatment in breast cancer (BC). However, studies to date have not fully considered the tumor-infiltrating immune cell (TIIC) landscape in tumors. This study investigated potential biomarkers based on TIICs to improve prognosis and treatment effect in BC.

RESULTS: We enrolled 5112 patients for analysis and used cell type identification by estimating relative subsets of RNA transcripts (CIBERSORT), a new computational algorithm, to quantify 22 TIICs in primary BC. From the results of univariate Cox regression, 12 immune cells were determined to be significantly related to the overall survival (OS) of BC patients. Furthermore, least absolute shrinkage and selection operator (LASSO) and multivariate Cox regression analyses were applied to construct an immune prognostic model based on six potential biomarkers. By dividing patients into low- and high-risk groups, a significant distinction in OS was found in the training cohort, with 20-year survival rates of 42.6% and 26.3%, respectively. Applying a similar protocol to validation and test cohorts, we found that OS was significantly shorter in the high-risk group than in the low-risk group, regardless of the molecular subtype of BC. Using the immune score model to predict the effect of BC patients to chemotherapy, the survival advantage for the low-risk group was evident among those who received chemotherapy, regardless of the chemotherapy regimen. In evaluating the predictive value of the nomogram, a decision curve showed better predictive accuracy than the standard tumor-node-metastasis (TNM) staging system.

CONCLUSION: The immune cell infiltration-based immune score model can be effectively and efficiently used to predict the prognosis of BC patients as well as the effect of chemotherapy.

KEYWORDS: CIBERSORT; breast cancer; immune score; prognostic; tumor-infiltrating immune cells

PMID: 33204321

PMCID: PMC7667685

DOI: 10.7150/thno.49451