Single-cell analysis reveals cell communication triggered by macrophages associated with the reduction and exhaustion of CD8+ T cells in COVID-19

Abstract

Background: The coronavirus disease 2019 (COVID-19) outbreak caused by severe acute respiratory syndrome coronavirus 2 (SARS-Cov-2) has become an ongoing pandemic. Understanding the respiratory immune microenvironment which is composed of multiple cell types, together with cell communication based on ligand-receptor interactions is important for developing vaccines, probing COVID-19 pathogenesis, and improving pandemic control measures.

Methods: A total of 102 consecutive hospitalized patients with confirmed COVID-19 were enrolled in this study. Clinical information, routine laboratory tests, and flow cytometry analysis data with different conditions were collected and assessed for predictive value in COVID-19 patients. Next, we analyzed public single-cell RNA-sequencing (scRNA-seq) data from bronchoalveolar lavage fluid, which offers the closest available view of immune cell heterogeneity as encountered in patients with varying severity of COVID-19. A weighting algorithm was used to calculate ligand-receptor interactions, revealing the communication potentially associated with outcomes across cell types. Finally, serum cytokines including IL6, IL1β, IL10, CXCL10, TNFα, GALECTIN-1, and IGF1 derived from patients were measured.

Results: Of the 102 COVID-19 patients, 42 cases (41.2%) were categorized as severe. Multivariate logistic regression analysis demonstrated that AST, D-dimer, BUN, and WBC were considered as independent risk factors for the severity of COVID-19. T cell numbers including total T cells, CD4⁺ and CD8⁺ T cells in the severe disease group were significantly lower than those in the moderate disease group. The risk model containing the above mentioned inflammatory damage parameters, and the counts of T cells, with AUROCs ranged from 0.78 to 0.87. To investigate the molecular mechanism at the cellular level, we analyzed the published scRNA-seq data and found that macrophages displayed specific functional diversity after SARS-Cov-2 infection, and the metabolic pathway activities in the identified macrophage subtypes were influenced by hypoxia status. Importantly, we described ligand-receptor interactions that are related to COVID-19 serverity involving macrophages and T cell subsets by communication analysis.

Conclusions: Our study showed that macrophages driving ligand-receptor crosstalk contributed to the reduction and exhaustion of CD8⁺ T cells. The identified crucial cytokine panel, including IL6, IL1β, IL10, CXCL10, IGF1, and GALECTIN-1, may offer the selective targets to improve the efficacy of COVID-19 therapy.

Trial registration: This is a retrospective observational study without a trial registration number. Video Abstract.

Keywords: COVID-19; Macrophage; SARS-CoV-2; Single cell RNA-sequencing; T cell.

Fig. 1

Logistic regression model performance and clinical usefulness of selected laboratory characteristics nomogram. a Forest plot showing the odds ratio of clinical parameters analyzed by univariate logistic regression in patients with moderate or severe COVID-19. The length of the horizontal line represents the 95% confidence interval for each indicator. The vertical dotted line represents the odds ratio (OR) = 1, OR > 1.0 implies a positive relationship. (i.e. odds ratio for gender had been above 1, this means that being male would correspond with higher odds of being the severe outcome.) b The nomogram was built by the multivariate logistic regression model, with the laboratory characteristics including WBC, AST, BUN, and D-dimer. c Calibration curve with Hosmer–Lemeshow test of the nomogram. The calibration curve depicts the calibration of the fitted model in terms of the agreement between the predicted risk of severe COVID-19 and real observed outcomes

Fig. 2

The phenotypes and counts of T cell subtypes. a, b Fluorescence-activated cell sorting (FACS) dot plot examples, gated on total CD45⁺ cells (left), the expression of CD4 and/ or CD3 on CD45⁺ cells (middle), and the expression of CD8 and/ or CD3 on CD45⁺ cells (right). Case A indicates one of the moderate patients with COVID-19 (a). Case B, one representative case of severe patients (b). c, d The violin graph showing the counts of CD4⁺ T cells (c) and CD8⁺ T cells (d) in patients with different groups. *** p < 0.001 (Student’s t-test). e ROCs are created by plotting the true positive rate (TPR) against the false positive rate (FPR) at various threshold settings with corresponding AUCs labeled around the curves

Fig. 3

Single-cell RNA-seq of total bronchoalveolar lavage fluid immune (BAI) cells population. a t-SNE plots showing detected lineages and sub-populations in BAI cells across conditions (e.g. Control means the cells derived from healthy donors). b t-SNE plot of aggregate BAI cells with identified sub-populations. c Cell population percentages across conditions determined to be significantly modulated according to Differential proportion Analysis (DPA). *p < 0.05, **p < 0.01 (Fisher’s exact test). Abbreviation: C, M, S indicate Control, Moderate, and Severe, respectively; Mac, Macrophage; Mons /DC, Monocytes/Dendritic cells; PMC, Plasma cells; EPC, Epithelial cells; NK, natural kill cells. d Expression of select marker genes across BAI cells as visualized on t-SNE plots. e Left: bubble plot showing representative GO terms according to cell types. Right: heatmap characterizing the expression signatures of top 50 specifically expressed genes in each cell type; the value for each gene is row-scaled Z score

Fig. 4

BAI macrophage subpopulations. a t-SNE plot showing macrophage subpopulations. b Percentage of the subpopulation of macrophages derived from patients with COVID-19 across different conditions. c Expression of marker genes across macrophage subtypes as visualized on t-SNE (upper) and violin plots. d The difference of cytokine expression levels between moderate and severe groups. e Hypoxia scores of macrophages derived from moderate and severe samples. f Hypoxia score enriched in each of macrophage subpopulation, the score of subpopulations were respectively compared to M0. * p < 0.05, ** p < 0.01, *** p < 0.001 (Student’s t test)

Fig. 5

Functional heterogeneity of macrophage subpopulations after SARS-CoV-2 infection. a–c Differences in pathway activities scored per macrophages by GSVA with Moderate vs. Control (a), Severe vs. Control (b), Severe vs. Moderate (c). d Heatmap of GSVA score of top 25 signal pathway for each macrophage across conditions. the value for each GSVA score is row-scaled. e, f Signal pathway activity of glycolysis (e) and oxidative phosphorylation (f) under hypoxia status across macrophage subtypes. * p < 0.05, ** p < 0.01 (Student’s t test)

Fig. 6

Reconstruction of a trajectory with Macrophage Subpopulation. a The single-cell trajectory reconstructed by Monocle contains five main branches and two decision points. Cells are colored based on pseudotime (upper) and sample types. Abbreviation: Con, MO, SE indicate Control, Moderate, and Severe, respectively. b Dot plot showing the variability of gene expressions, such as C1QA (upper), CXCL2 (middle), and TREM2, following pseudotime based on cell states. A natural spline was used to model gene expression as a smooth, nonlinear function over pseudotime. c Dot plot showing the variability of pathway activity, such as hypoxia (upper), glycolysis (middle), and senescence, following pseudotime in the path that contains cells of states 1, 2, 3, 4, and 5. d, e Each heatmap presents genes differentially expressed between two branch comparisons, and each row represents the expression level of a gene along the branch trajectory. Enriched pathways are summarized for each gene cluster. From root to state 1 and state 2 branches (d), from root to state 2, and combined state 3 with 4 branches (e)

Fig. 7

Cell–cell ligand–receptor network analysis. a Comparison of total incoming path weights vs total outgoing path weights across BAI cell populations. b Circle network diagram of significant cell–cell interaction pathways. Arrows and edge color indicate direction (ligand: receptor) and edge thickness indicates the sum of weighted paths between populations. c–e Heatmap showing the interaction weights calculated as the product of the average ligand expression from the source cell type including monocyte/ DC (c), epithelial cell (d) macrophages (e), to the average receptor expression of the target cell types. Grey boxes indicate interactions that are not significantly present across all cell types (one-sided Wilcoxon rank-sum test and Benjamini Hochberg false discovery rate [FDR] > 0.05). f Heatmap showing the global five communication patterns calculated by the key signals for subpopulations between macrophage and T cells. g Hierarchical network diagram of significant cell-cell communication patterns. Edge thickness indicates the sum of weight key signals between populations (from outgoing to incoming)

Fig. 8

The key communications between macrophage (M) and T cell subpopulation. a The dot plot showing the key ligands in the outgoing signaling pattern of subpopulation as secreting cells. b Circle plot showing the inferred intercellular communication network for CXCL signaling. Arrows and edge color indicate direction ((source: target). Edge thickness indicates the sum of weight key signals between populations. c Hierarchical network diagram of significantly inferred intercellular communication network for IGF signaling based on above-mentioned cell-cell communication pattern (from macrophage to T cell). d Heatmap shows the relative importance of each cell subtype based on the computed four network centrality measures of CXCL signaling (upper) and relative contribution of each ligand–receptor pair to the overall communication network of CXCL signaling. e Heatmap shows the relative importance of each cell subtype based on the computed four network centrality measures of IGF signaling and the relative contribution of each ligand–receptor pair to the overall communication network of IGF signaling. f Significant ligand–receptor pairs involving CXCL and IGF pathway sending signals from TREM2^high M to four T cell states. g FACS dot plots showing the expression of PD-1 on CD8⁺ T cells. h The percentages of PD-1⁺ in CD8⁺ T cells (the first panel). The serum levels of IL6, IL1β, IL-10, TNF-α, CXCL10, GALECTIN-1, and IGF1 in different groups are shown in boxplot (mean ± SD). Statistical analysis by Student’s t-test