In recent decades, our understanding of periodontitis has evolved from that based on a gross/histologic level to one on a cellular/molecular level. Previous landscape studies have explored molecular subtyping, diagnosis, and gingival tissue cell decomposition in periodontitis, and meaningful results have been obtained at a transcriptomic level. However, current periodontitis transcriptomic studies lack a finer dissection of the intercommunication between immune cells and the biological processes of specific immune cell subtypes. In this study, we classified 15 immune cell types in periodontitis at a single-cell level and conducted a cell communication analysis based on a multicenter integrated single-cell transcriptome profile, in which plasma cell–generated macrophage migration inhibitory factor can communicate with most other immune cells in periodontitis. A pseudotime analysis focusing on B/plasma cell infiltration in periodontitis revealed 2 distinct cell fates (CFs) for B/plasma cells. In addition, at a bulk tissue level, a single-sample gene set enrichment analysis showed a similar immune cell infiltration trend, and a weighted gene coexpression network analysis identified an immune-related gene module. Combined with the above findings, we used machine learning methods to further narrow down potential gene candidates for developing and validating molecular diagnostic models of periodontitis. Multivariable logistic regression of a large public cohort (68 healthy vs. 235 periodontitis) and an independent validation cohort (12 healthy vs. 7 periodontitis) showed the CF1 signature provides a good discrimination and calibration performance with clinical benefits at a proper threshold probability. Furthermore, quantitative real-time polymerase chain reaction validation of the gene candidates was performed in both snap-frozen gingival tissues and gingival crevicular fluids. Our transcriptomic landscape analysis at both single-cell and bulk tissue resolutions thereby illustrates the B/plasma cell infiltration process in periodontitis and reveals a gene signature that may assist in molecular diagnosis of the disease.

近几十年来，我们对牙周炎的理解已经从基于大体/组织学水平的理解发展到基于细胞/分子水平的理解。以前的景观研究已经探索了牙周炎中的分子亚型、诊断和牙龈组织细胞分解，并在转录组水平上获得了有意义的结果。然而，目前的牙周炎转录组学研究缺乏对免疫细胞之间的相互交流和特定免疫细胞亚型的生物学过程的更精细剖析。在这项研究中，我们在单细胞水平上对牙周炎中的 15 种免疫细胞类型进行了分类，并基于多中心整合的单细胞转录组谱进行了细胞通讯分析，其中浆细胞产生的巨噬细胞迁移抑制因子可以与牙周炎中的大多数其他免疫细胞进行交流。一项针对牙周炎中 B/浆细胞浸润的伪时间分析揭示了 B/浆细胞有 2 种不同的细胞命运 (CF)。此外，在大块组织水平上，单样本基因集富集分析显示出类似的免疫细胞浸润趋势，加权基因共表达网络分析确定了免疫相关基因模块。结合上述发现，我们使用机器学习方法进一步缩小潜在基因候选范围，以开发和验证牙周炎的分子诊断模型。大型公共队列（68 名健康对 235 名牙周炎）和一个独立验证队列（12 名健康对 12 名牙周炎）的多变量逻辑回归 7 牙周炎）表明 CF1 特征在适当的阈值概率下提供了良好的辨别和校准性能，具有临床益处。此外，在速冻牙龈组织和龈沟液中对候选基因进行了定量实时聚合酶链反应验证。因此，我们在单细胞和大块组织分辨率下的转录组景观分析说明了牙周炎中的 B/浆细胞浸润过程，并揭示了可能有助于对该疾病进行分子诊断的基因特征。在速冻牙龈组织和龈沟液中对候选基因进行了定量实时聚合酶链反应验证。因此，我们在单细胞和大块组织分辨率下的转录组景观分析说明了牙周炎中的 B/浆细胞浸润过程，并揭示了可能有助于对该疾病进行分子诊断的基因特征。在速冻牙龈组织和龈沟液中对候选基因进行了定量实时聚合酶链反应验证。因此，我们在单细胞和大块组织分辨率下的转录组景观分析说明了牙周炎中的 B/浆细胞浸润过程，并揭示了可能有助于对该疾病进行分子诊断的基因特征。