Integrated histopathology, spatial and single cell transcriptomics resolve cellular drivers of early and late alveolar damage in COVID-19

Abstract

The most common cause of death due to COVID-19 remains respiratory failure. Yet, our understanding of the precise cellular and molecular changes underlying lung alveolar damage is limited. Here, we integrate single cell transcriptomic data of COVID-19 and donor lung tissue with spatial transcriptomic data stratifying histopathological stages of diffuse alveolar damage. We identify changes in cellular composition across progressive damage, including waves of molecularly distinct macrophages and depletion of epithelial and endothelial populations. Predicted markers of pathological states identify immunoregulatory signatures, including IFN-alpha and metallothionein signatures in early damage, and fibrosis-related collagens in late damage. Furthermore, we predict a fibrinolytic shutdown via endothelial upregulation of SERPINE1/PAI-1. Cell-cell interaction analysis revealed macrophage-derived SPP1/osteopontin signalling as a key regulator during early steps of alveolar damage. These results provide a comprehensive, spatially resolved atlas of alveolar damage progression in COVID-19, highlighting the cellular mechanisms underlying pro-inflammatory and pro-fibrotic pathways in severe disease.

Fig. 1. Single cell transcriptomic atlas of the healthy and COVID-19 lung.

A Schematic overview. A multi-study sc/snRNA-seq dataset was integrated with histopathology driven spatial whole transcriptome analysis. The cell2location tool was used to map cell types/states to spatial transcriptomic data, with DGE, cell-colocalization, abundance analysis, and cell-cell interaction interrogation performed downstream. Created in BioRender. Barnett, S. (2025) https://BioRender.com/n08s131. B Number of COVID-19 patients / donors contributing to the integrated sc/snRNA-seq dataset. Created in BioRender. Barnett, S. (2025) https://BioRender.com/h64x272. C Percentage contribution of sc/snRNA-seq datasets from organ donors and COVID-19 patients. D UMAP representation of integrated COVID-19 (red) and healthy control (blue) datasets contributing to the final sc/snRNA-seq object. E UMAP representation of the global object with broad cell type (dotted) and mid-level annotation. F Heatmap representation of markers used for mid-level annotation. G Sankey plot visualisation of cell state level annotations derived from subclustering of the broad cell type compartments (Supplementary Fig. 1C, D and Supplementary Data 3, 4).

Fig. 2. Transcriptome-wide spatial profiling of alveolar damage progression in COVID-19.

A Number of COVID-19 patients contributing to the spatial WTA dataset (left) and histopathological regions of interest (ROI). Created in BioRender. Barnett, S. (2025) https://BioRender.com/m70f077. B Representative images from H&E stained tissue sections illustrating preserved (PRES) tissue structure and ROIs including different DAD histopathological features. Upper: Indication of multiple ROIs taken from the same tissue section. Lower: Tissue morphology of PRES (n = 21), EDAD (n = 108), MDAD (n = 24) and ODAD (n = 52) states. C Number of histopathological ROIs for each COVID-19 patient used in the study. D Schematic representing random forest classifier approach for predicting pathology-associated gene signatures. E Dotplot representation of histopathological state-associated gene signatures obtained from random forest classifier analysis. Dot colour = scaled mean gene expression within a column. Dot size = percentage expressed in group. Importance score = ratio of correct pathology classification in all ROIs.

Fig. 3. Transcriptional progression of alveolar damage progression.

A Volcano plot representing genes upregulated in EDAD (yellow) and ODAD (purple). Differentially expressed genes were calculated using Pseudobulk EdgeR. B Hallmark MSigDB gene set enrichment of genes upregulated in EDAD (top) vs. ODAD (bottom). C Dot plot illustrates DGE between COVID-19 and controls related to fibrinolysis and coagulation in the sc/snRNA-seq dataset calculated using Pseudobulk EdgeR with QLF test. Red = upregulated (logFC ≥ 1) in COVID-19, blue = downregulated (logFC ≤ -1) in COVID-19. All data points shown are significant (FDR < 0.05). For a complete list of DGE, see Supplementary Data 5. D Schematic representation of the fibrinolysis pathway. Genes upregulated (red) and downregulated (blue) in COVID-19 EC compared to healthy controls are shown. Grey arrows indicate upstream and downstream factors that are not visualised. E Healthy control and COVID-19 lung parenchyma samples stained by smFISH for SERPINE1 (magenta), and CDH5 (VE-Cadherin) protein (green) counterstained with DAPI (blue). F Boxplot shows log2fc area of SERPINE1 staining per number of cells in COVID-19 lung parenchyma (n = 4 patients x 1 region) compared to healthy controls (n = 2 donors x 1 region). Boxplot elements: centre line, median; box limits, upper and lower quartiles; whiskers, 1.5 x interquartile range.

Fig. 4. Evolving cellular composition across alveolar damage progression.

A Schematic representing the integration of sc/snRNA-seq data with ST data for cell state deconvolution. B, C Cell state abundances in WTA and sc/snRNA-seq data. B Healthy and COVID-19 cell state gene expression signatures were mapped to healthy and COVID-19 ROIs using cell2location, respectively. The scatter plot shows the percentage of estimated cell abundance enriched in ROIs of COVID-19 (red) or healthy (blue) samples, black spot indicating the median of abundance. A two-sided Welch’s t test with the Benjamini-Hochberg adjustment for multiple comparisons was used. C Beeswarm plots illustrate the enrichment (red) or decrease (blue) of neighbourhoods in COVID-19 for each indicated cell type calculated using MiloR (FDR < 0.05). D, E The distribution of selected healthy (D) and COVID-19 (E) cell state signatures across DAD pathologies. Dot plots show cell abundance values that were z-score normalised per cell type across rows/pathologies (colour) and the fraction of ROIs (percent) with cell abundance above the average value of all ROIs in a given pathology (size).

Fig. 5. Cellular niches and cell-cell signalling across alveolar damage stages.

A Diagram of analysis approach for mapping cell-cell communication across early versus late DAD. Created in BioRender. Barnett, S. (2025) https://BioRender.com/n08s131. B Waterfall plot visualisation of global pathway analysis between healthy (blue) and COVID-19 (red) sc/snRNA-seq, with significant pathways highlighted in blue and red respectively. C NMF analysis across histopathological states. D SPP1 signalling within the COVID-19 sc/snRNA-seq compartment, mapped to EDAD niche 2. Arrows indicate the directionality of ligand (SPP1) signalling to receptors between cell states. E Dotplot visualisation of SPP1 signalling to specific receptors across EDAD niche 2 cell states. P-values are computed from a one-sided permutation test. F Representative confocal images of endothelial cells treated with 1 µg/mL rhOPN or vehicle (0 µg/mL) for 24 h. Scale bar = 10 µm. G Box-whiskers plot of PAI-1 average fluorescence intensity. Values normalised to vehicle control (0 µg/mL rhOPN). n = 3 independent experiments, performed in triplicate. One-way ANOVA; ** adjusted p < 0.01. 0 vs. 0.5 adjusted p-value = 0.0093. 0 vs. 1 adjusted p-value = 0.0086. Boxplot elements: centre line, median; box limits, upper and lower quartiles; whiskers, 1.5 x interquartile range. Arb. units = arbitrary units. H Summary of Macrophage subtypes contributing to pro-thrombotic and anti-fibrinolytic states in early DAD through SPP1 signalling. Created in BioRender. Barnett, S. (2025) https://BioRender.com/f13w044.