a Uniform Manifold Approximation and Projection (UMAP) UMAP representation of CD8+ T cell landscape. b Heat map representing different clusters identified by FlowSOM, with relative identity and percentages in controls and patients. The color in the heat map represents the median of the arcsinh, 0–1 transformed marker expression calculated over cells from all the samples, varying from blue for lower expression to red for higher expression. The dendrogram on the left represents the hierarchical similarity between the metaclusters (metric: Euclidean distance; linkage: average). Each cluster has a unique color assigned (bar on the left). Barplots along the rows (clusters) and values on the right indicate the relative sizes of clusters. c Differential analysis between CTR (bar color: salmon; n = 13) and COVID-19 (emerald; n = 19). The heat represents arcsine-square-root transformed cell frequencies that were subsequently normalized per cluster (rows) to mean of zero and standard deviation of one. The color of the heat varies from blue indicating relative under-representation to orange indicating relative over-representation. Bar and numbers at the right indicate significant differentially abundant clusters (green) and adjusted p values. Clusters are sorted according to adjusted p values, so that the cluster at the top shows the most significant abundance changes between the two conditions. d Representative dot plots related to the expression of different chemokine receptors and lineage-specifying transcription factors in gated CD8+ T from a control donor (upper) and a patient (lower panel). Two experiments (one for the control group, one for patients) out of 13 are shown. Numbers indicate the percentage in each quadrant. The gating strategy for the identification of CD8+ T cells is reported in Supplementary Fig. 1. e Percentages of different CD8+ T cell subpopulations in controls (n = 6) and patients (n = 7), obtained by manual gating. Data represent individual values (dots), mean (centre bar) ± SEM (upper and lower bars). Statistical analysis is performed by two-sided Mann–Whitney nonparametric test; if not indicated, p value is not significant. Source data are provided as a Source Data file.