Despite intensive research since the emergence of SARS-CoV-2, it has remained unclear precisely which components of the early immune response protect against the development of severe COVID-19. To address this issue, we performed a comprehensive immunogenetic and virologic analysis of nasopharyngeal and peripheral blood samples obtained during the acute phase of infection with SARS-CoV-2. We found that soluble and transcriptional markers of systemic inflammation peaked during the first week after symptom onset and correlated directly with the upper airways viral loads (UA-VLs), whereas the contemporaneous frequencies of circulating viral nucleocapsid (NC)-specific CD4+ and CD8+ T cells correlated inversely with various inflammatory markers and UA-VLs. In addition, we observed high frequencies of activated CD4+ and CD8+ T cells in acutely infected nasopharyngeal tissue, many of which expressed genes encoding various effector molecules, such as cytotoxic proteins and IFN-γ. The presence of functionally active T cells in the infected epithelium was further linked with common patterns of gene expression among virus-susceptible target cells and better local control of SARS-CoV-2. Collectively, these results identified an immune correlate of protection against SARS-CoV-2, which could inform the development of more effective vaccines to combat the acute and chronic illnesses attributable to COVID-19.

Federated Learning approaches are becoming increasingly relevant in various fields. These approaches promise to facilitate an integrative data analysis without sharing the data, which is highly beneficial for applications with sensitive data such as healthcare. Yet, the risk of data leakage caused by malicious attacks needs to be assessed carefully. In this study, we consider a new attack route and present an algorithm that depends on being able to compute sample means, sample covariances, and construct known linearly independent vectors on the data owner side. We show that these basic functionalities, available in several established Federated Learning frameworks, suffice to reconstruct privacy-protected data. Moreover, the attack algorithm is robust to defence strategies that build on random noise. We demonstrate this limitation of existing frameworks and discuss possible defence strategies. The novel insights will facilitate the improvement of Federated Learning frameworks.