Figure 3.

Gene set enrichment and data integration reveals a profibrotic phenotype of COVID-19-associated macrophages

(A) Gene set module score of “IPF-expanded macrophages” (IPFe-Mφ) and alveolar FABP4+Mφ (Ayaub et al., 2021), calculated based on single transcriptomes. Projected onto the UMAP embedding (top) and plotted as violin plots (bottom) across the monocyte/macrophage clusters (annotated in Figure 2B). Dot color indicates signature module score. Violin colors show cluster identity, numbers indicate -log10 transformed adjusted p values (one-sided wilcoxon test compared to average), and lines in violins indicate median scores per cluster.

(B) Heatmap representing -log10 transformed adjusted p values (one-sided Fisher’s exact test) assessing the overlap of gene sets from monocyte/macrophage clusters identified in Figure 2B (y axis) and published transcriptional signatures of monocyte/macrophage clusters derived from the indicated IPF datasets (cluster names and reference studies indicated on the x axis; Table S3).

(C) Schematic depicting monocyte/macrophage data integration from present study and Bharat et al. (2020) with two human lung fibrosis reference datasets (Adams et al., 2020; Morse et al., 2019) via scVI. COVID-19 macrophages were mapped to IPF or control macrophages based on a kNN (k-nearest neighbor)-proximity mapping.

(D) UMAP of 138,341 cells derived from all four datasets based on integrated scVI embedding.

(E) UMAP as in (D) highlighting COVID-19-associated macrophage clusters annotated in Figure 2B. Cells from reference datasets shaded in gray.

(F) Proximity analysis of macrophage clusters annotated in Figure 2B and macrophages identified in IPF and healthy controls, respectively. Circle size shows cell fraction, color codes indicate the -log10 transformed adjusted p values, and bold black circle indicates statistical significance (adjusted p < 0.0001, Fisher’s exact test, one-tailed with Benjamini-Hochberg correction).