Long COVID: Deep single-cell immunophenotyping and machine learning reveal a general signature for fatigue

BACKGROUND: The post COVID-19 condition, commonly referred to as “Long COVID” (LC), is a constellation of long-lasting and debilitating symptoms following acute SARS-CoV-2 infection, which closely resembles other post-infective fatigue states. The underlying immunological disturbances of LC are poorly understood. Multiple explanatory mechanisms, such as persisting SARS-CoV-2 reservoirs, reactivation of latent viruses, endothelial dysfunction, auto-antibodies, tissue damage, and immune abnormalities, have been proposed but remain incompletely characterized, particularly in younger populations. METHODS: The present study included 12 to 25-year-old females with and without mild SARS-CoV-2 infection, who were prospectively followed for six months after infection and assessed according to the WHO definition of post COVID-19 condition, resulting in four groups (“Long COVID” (LC), recovered convalescents (RC), fatigued controls (FC), healthy controls (HC)). Peripheral blood mononuclear cells were either stimulated with Phorbol-12-myristate-13-acetate/ionomycin or left unstimulated, and analyzed using a 41-antibody CyTOF panel, unsupervised clustering with FlowSOM, dimensionality reduction via UMAP, polyfunctionality assessment with COMPASS, and machine learning for identifying correlates of LC and fatigue. All participants were female with a median age of 18.5 years. RESULTS: Higher frequencies of Terminal NK cells were associated with LC and FC, showing enhanced polyfunctional responses upon stimulation. Increased CD4 + T cell activation and exhaustion were observed in both LC and FC, with elevated effector memory and effector T cells expressing PD-1, alongside decreased marginal zone B cells and transitional B cells. Machine learning analysis revealed terminal NK cells as the most important feature for predicting fatigue. Importantly, no unique LC-specific immune changes were detected. CONCLUSION: Our results point to a shared underlying pathophysiology in LC and other forms of fatigue. The absence of unique LC-related immune changes supports the notion that LC represents a specific example of post-infective fatigue syndrome triggered by SARS-CoV-2 in susceptible individuals. Innate and adaptive immune dysregulation, particularly higher levels of hyperresponsive terminal NK cells, is most striking and could serve as a general fatigue marker rather than an LC-specific biomarker. These findings highlight the need for further studies to correlate immune perturbations with symptom clusters and inform diagnostic and treatment strategies for fatigue states.