Kusejko Katharina, Arefian Mohammad, Duroux Diane, Zeeb Marius, Dollé Cédric, Hoffmann Matthias, Labhardt Niklaus, Wandeler Gilles, Cavassini Matthias, Haller Sabine, Bernasconi Enos, Russenberger Doris, Kouyos Roger D, Günthard Huldrych F, Collins Ben C, Nemeth Johannes
Department of Infectious Diseases and Hospital Epidemiology, University Hospital Zurich and University of Zurich, Zürich, Switzerland.
Institute of Medical Virology, University of Zurich, Zürich, Switzerland.
mBio. 2025 Oct 8;16(10):e0158525. doi: 10.1128/mbio.01585-25. Epub 2025 Aug 28.
Improved biomarkers for predicting progression to active tuberculosis (TB) are urgently needed, especially in people with HIV, who are at elevated risk. We used high-throughput plasma proteomics and machine learning to identify signatures associated with TB progression in this population. From the Swiss HIV Cohort Study, we analyzed plasma samples collected at least 6 months before TB diagnosis from 91 participants who later developed TB. We selected 293 controls matched for demographic and clinical parameters who remained TB-free to achieve a risk score specific to active TB. In total, 583 samples were analyzed, with 613-1,283 proteins quantified per sample. A random forest classifier predicted a significantly higher median probability of TB progression for cases (33%) than for controls (16%; < 0.001). In this matched population, the score achieved an area under the receiver-operating characteristic curve of 0.77, an area under the precision-recall curve (AUPRC) of 0.60 (as compared to an expected AUPRC of 0.29), as well as a specificity of 87.3% and a sensitivity of 58.6% using the optimal threshold of 0.311. The plasma proteome of individuals who progressed to active TB showed a distinct shift toward systemic inflammation, B cell activation, and immunoglobulin production. Independent of progression to active TB, the proteome score correlated with broader indicators of immune suppression, including lower CD4 counts and unsuppressed HIV RNA. This suggests that integrating proteomic and clinical data could enhance the overall predictive power of the score.IMPORTANCEWe still lack reliable tools to predict who will develop tuberculosis (TB) among people with HIV. Moreover, the underlying biological events driving progression remain poorly understood. Our study reveals early immune changes that include unexpected alterations in B cell activation and antibody responses. These findings suggest that humoral immunity may play a more important role in TB pathogenesis than previously recognized and offer promising new directions for biomarker discovery and targeted prevention.
迫切需要改进用于预测进展为活动性结核病（TB）的生物标志物，尤其是在风险较高的HIV感染者中。我们使用高通量血浆蛋白质组学和机器学习来识别该人群中与结核病进展相关的特征。在瑞士HIV队列研究中，我们分析了91名后来发展为结核病的参与者在结核病诊断前至少6个月采集的血浆样本。我们选择了293名在人口统计学和临床参数上匹配且未患结核病的对照，以获得针对活动性结核病的风险评分。总共分析了583个样本，每个样本定量了613 - 1283种蛋白质。随机森林分类器预测病例结核病进展的中位概率（33%）显著高于对照（16%；<0.001）。在这个匹配人群中，该评分在受试者工作特征曲线下的面积为0.77，精确召回率曲线下的面积（AUPRC）为0.60（与预期的AUPRC 0.29相比），使用最佳阈值0.311时特异性为87.3%，敏感性为58.6%。进展为活动性结核病的个体的血浆蛋白质组显示出明显向全身炎症、B细胞活化和免疫球蛋白产生的转变。独立于进展为活动性结核病，蛋白质组评分与免疫抑制的更广泛指标相关，包括较低的CD4计数和未受抑制的HIV RNA。这表明整合蛋白质组学和临床数据可以提高评分的整体预测能力。重要性我们仍然缺乏可靠的工具来预测HIV感染者中谁会发展为结核病。此外，驱动疾病进展的潜在生物学事件仍知之甚少。我们的研究揭示了早期免疫变化，包括B细胞活化和抗体反应的意外改变。这些发现表明体液免疫在结核病发病机制中可能比以前认识到的发挥更重要的作用，并为生物标志物发现和靶向预防提供了有希望的新方向。
