Background: 5-Methylcytidine (m5C) is the most common RNA modification and plays an important role in multiple tumors including cervical cancer (CC). We aimed to develop a novel gene signature by identifying m5C modification subtypes of CC to better predict the prognosis of patients.

Methods: We obtained the expression of 13 m5C regulatory factors from The Cancer Genome Atlas (TCGA all set, 257 patients) to determine m5C modification subtypes by the “nonnegative matrix factorization” (NMF). Then the “limma” package was used to identify differentially expressed genes (DEGs) between different subtypes. According to these DEGs, we performed Cox regression and Kaplan-Meier (KM) survival analysis to establish a novel gene signature in TCGA training set (128 patients). We also verified the risk prediction effect of gene signature in TCGA test set (129 patients), TCGA all set (257 patients) and GSE44001 (300 patients). Furthermore, a nomogram including this gene signature and clinicopathological parameters was established to predict the individual survival rate. Finally, the expression and function of these signature genes were explored by qRT-PCR, immunohistochemistry (IHC) and proliferation, colony formation, migration and invasion assays.

Results: Based on consistent clustering of 13 m5C-modified genes, CC was divided into two subtypes (C1 and C2) and the C1 subtype had a worse prognosis. The 4-gene signature comprising FNDC3A, VEGFA, OPN3 and CPE was constructed. In TCGA training set and three validation sets, we found the prognosis of patients in the low-risk group was much better than that in the high-risk group. A nomogram incorporating the gene signature and T stage was constructed, and the calibration plot suggested that it could accurately predict the survival rate. The expression levels of FNDC3A, VEGFA, OPN3 and CPE were all high in cervical cancer tissues. Downregulation of FNDC3A, VEGFA or CPE expression suppressed the proliferation, migration and invasion of SiHa cells.

Conclusions: Two m5C modification subtypes of CC were identified and then a 4-gene signature was established, which provide new feasible methods for clinical risk assessment and targeted therapies for CC.

背景：5-甲基胞苷（m5C）是最常见的RNA修饰，在包括宫颈癌（CC）在内的多种肿瘤中发挥着重要作用。我们的目标是通过鉴定 CC 的 m5C 修饰亚型来开发新的基因特征，以更好地预测患者的预后。

方法：我们从癌症基因组图谱（TCGA 全部集，257 例患者）中获取了 13 个 m5C 调节因子的表达，通过“非负矩阵分解”（NMF）确定 m5C 修饰亚型。然后使用“limma”包来识别不同亚型之间的差异表达基因（DEG）。根据这些 DEG，我们进行了 Cox 回归和 Kaplan-Meier (KM) 生存分析，以在 TCGA 训练集（128 名患者）中建立新的基因特征。我们还在TCGA测试集（129名患者）、TCGA所有集（257名患者）和GSE44001（300名患者）中验证了基因签名的风险预测效果。此外，建立了包括该基因特征和临床病理学参数的列线图来预测个体存活率。最后，通过 qRT-PCR、免疫组织化学 (IHC) 以及增殖、集落形成、迁移和侵袭测定来探索这些特征基因的表达和功能。

结果：基于13个m5C修饰基因的一致聚类，CC分为两个亚型（C1和C2），C1亚型预后较差。构建了包含 FND​​C3A、VEGFA、OPN3 和 CPE 的 4 基因特征。在TCGA训练集和三个验证集中，我们发现低风险组患者的预后明显好于高风险组。构建了包含基因特征和 T 阶段的列线图，校准图表明它可以准确预测存活率。 FNDC3A、VEGFA、OPN3、CPE在宫颈癌组织中均呈高表达。 FNDC3A、VEGFA或CPE表达的下调抑制了SiHa细胞的增殖、迁移和侵袭。

结论：鉴定了CC的两种m5C修饰亚型，并建立了4基因标签，为CC的临床风险评估和靶向治疗提供了新的可行方法。