目的：探讨年轻乳腺癌独特的DNA甲基化特征，筛选与预后相关的甲基化位点，构建并验证预后预测模型，并通过多组学整合及关键蛋白验证评估其临床价值。方法：收集癌症基因组图谱（TCGA）数据库中的乳腺癌样本DNA甲基化及临床数据，筛选年轻乳腺癌特有的甲基化位点。采用随机森林、LASSO回归及单因素和多因素Cox比例风险回归分析构建预后模型。使用受试者工作特征（ROC）曲线、校准曲线和一致性指数（C-index）评估模型性能，并比较高、低风险组间的多组学特征差异，在独立队列中通过免疫组化法检测模型关键基因LAMA5的蛋白表达，分析其与预后的关系。结果：年轻乳腺癌相较于老年乳腺癌，在启动子区域（尤其是CpG岛）呈现显著低甲基化特征。通过多步骤差异分析获得3 489个年轻乳腺癌特有的甲基化位点，经Cox比例风险回归筛选出12个与总生存期（OS）独立相关的位点，其中启动子区、基因体区和基因间区位点各4个。基于筛选出的12个甲基化位点构建的预测模型，在TCGA训练集中显示出良好的区分度，高风险组总生存期显著更差（P = 0.002 5），5、8和10年OS预测曲线下面积（AUC）达0.88～0.94。多组学分析显示，高风险组丝裂原活化蛋白激酶（MAPK）信号通路激活，TP53基因突变频率更高，雌激素受体α蛋白表达较低。将TP53突变状态及雌激素受体1（ESR1）mRNA表达水平等关键多组学指标纳入模型，构建综合列线图，其预测效能进一步提升。组织芯片验证显示，模型关键基因LAMA5蛋白表达阴性患者的OS更差（P =0.046），且富集于三阴性乳腺癌，与模型高风险组的临床病理特征一致。结论：整合启动子和非启动子区域信息的基于12个甲基化位点的模型可用于年轻乳腺癌患者的预后分层。联合年龄、TP53突变及ESR1 mRNA表达水平构建的综合列线图显示出较单一甲基化模型更高的预测效能。

Objective: This study aims to systematically characterize the unique DNA methylation landscape of breast cancer in young women (BCYW), identify prognosis-associated methylation loci, develop and externally validate a robust prognostic prediction model, and further elucidate its clinical utility via multi-omics integration analysis and protein-level validation of a pivotal candidate gene. Methods: DNA methylation and clinical data were obtained from The Cancer Genome Atlas (TCGA). BCYW?specific methylation loci were identified. A prognostic prediction model was developed using random forest, LASSO regression, and univariate and multivariate Cox regression analyses. The predictive performance of the constructed model was comprehensively evaluated using receiver operating characteristic (ROC) curves, calibration plots, and the concordance index (C-index). Multi?omics characteristics were compared between high?risk and low?risk subgroups. In an independent cohort, immunohistochemistry was performed to assess the association between LAMA5 protein expression and clinical outcomes. Results: Compared with their older counterparts with breast cancer, young patients exhibited prominent hypomethylation in promoter regions, which was particularly pronounced within CpG islands. Through a multi-step differential methylation analysis, a total of 3 489 BCYW?specific methylation loci were identified. Subsequent univariate and multivariate Cox regression analyses further screened out 12 loci that were independently correlated with overall survival (OS), among which 4 were located in promoter regions, 4 in gene body regions, and the remaining 4 in intergenic regions. The prognostic prediction model constructed based on these 12 methylation loci exhibited satisfactory discriminative performance in the TCGA training cohort, where patients assigned to the high-risk subgroup presented significantly worse overall survival outcomes (P = 0.002 5). The areas under the curve (AUCs) for 5?, 8?, and 10?year OS prediction ranged from 0.88 to 0.94. Multi?omics analysis revealed that the high?risk subgroup exhibited the activation of the mitogen-activated protein kinase (MAPK) signaling pathway, higher frequency of TP53 mutations, and lower expression of estrogen receptor alpha protein. Incorporating key multi-omics indicators, including TP53 mutation status and estrogen receptor 1 (ESR1) mRNA expression level, into the comprehensive nomogram further significantly improved the model's predictive performance. Tissue microarray-based immunohistochemical validation demonstrated that young breast cancer patients with negative LAMA5 protein expression exhibited significantly shorter OS (P = 0.046), and this phenotype was enriched in triple-negative breast cancer, consistent with the clinicopathological features of the high-risk group defined by the methylation model. Conclusion: The 12-methylation-site model integrating promoter and non-promoter region information shows potential for prognostic stratification in young patients with breast cancer. The multi-omics nomogram constructed by combining age, TP53 mutations and ESR1 mRNA expression level demonstrated superior predictive performance over the methylation-only model.

天津市医学重点学科建设项目（TJYXZDXK-3-003A）