Distinct lung cell signatures define the temporal evolution of diffuse alveolar damage in fatal COVID-19

Abstract

Background: Lung damage in severe COVID-19 is highly heterogeneous however studies with dedicated spatial distinction of discrete temporal phases of diffuse alveolar damage (DAD) and alternate lung injury patterns are lacking. Existing studies have also not accounted for progressive airspace obliteration in cellularity estimates. We used an imaging mass cytometry (IMC) analysis with an airspace correction step to more accurately identify the cellular immune response that underpins the heterogeneity of severe COVID-19 lung disease.

Methods: Lung tissue was obtained at post-mortem from severe COVID-19 deaths. Pathologist-selected regions of interest (ROIs) were chosen by light microscopy representing the patho-evolutionary spectrum of DAD and alternate disease phenotypes were selected for comparison. Architecturally normal SARS-CoV-2-positive lung tissue and tissue from SARS-CoV-2-negative donors served as controls. ROIs were stained for 40 cellular protein markers and ablated using IMC before segmented cells were classified. Cell populations corrected by ROI airspace and their spatial relationships were compared across lung injury patterns.

Findings: Forty patients (32M:8F, age: 22-98), 345 ROIs and >900k single cells were analysed. DAD progression was marked by airspace obliteration and significant increases in mononuclear phagocytes (MnPs), T and B lymphocytes and significant decreases in alveolar epithelial and endothelial cells. Neutrophil populations proved stable overall although several interferon-responding subsets demonstrated expansion. Spatial analysis revealed immune cell interactions occur prior to microscopically appreciable tissue injury.

Interpretation: The immunopathogenesis of severe DAD in COVID-19 lung disease is characterised by sustained increases in MnPs and lymphocytes with key interactions occurring even prior to lung injury is established.

Funding: UK Research and Innovation/Medical Research Council through the UK Coronavirus Immunology Consortium, Barbour Foundation, General Sir John Monash Foundation, Newcastle University, JGW Patterson Foundation, Wellcome Trust.

Keywords: COVID-19; Diffuse alveolar damage; Imaging mass cytometry; Immunopathology; Post-mortem lung tissue.

Fig. 1

Overview of the cohort demographics and histological model. a) A graphical summary of the cohort composition and key clinical metadata, including comorbidities and exposure to medication. b) A graphical and histological summary of the different pathology states present in the cohort as identified by expert pathologist input (H&Es at ×100 magnification except for IPM which is at ×200 magnification). Note that the conventional progression stages are shown with red directional arrows and that the pathoevolutionary divergent stages are linked by grey lines to denote that progression and origin are unknown with respect to conventional stages of pathology. Histology images are derived from H&E stained FFPE serial tissue sections adjacent to those used for IF and IMC analysis. Please also note that “PRES” pathology falls in to two classes; derived from SARS-CoV2 infected and uninfected tissue. Scale bar = 100 μm. Some figure components were created with Biorender.com.

Fig. 2

OPTIMAL analysis of single cells in COVID19 PM lung tissue reveals a progressive loss of lung air space leading to elevated cellularity due to tissue obliteration rather than a de novo cellular influx. a) PM lung tissue slides were stained with a panel of 40 metal tagged antibodies alongside an additional control TMA slide (1). The pathologist-marked ROIs were then set using an OMERO reference image and ablated using a Hyperion IMC system (2) to produce a set of 41 multispectral images (3) that were segmented to single cell data (4), corrected for spill-over and other factors that could affect clarity (5) and converted to FCS file format with additional key metadata added (6). Batch effect was determined and corrected for using a z-score normalisation approach (7). b) A PacMap dimensionality reduction plot for all 901,602 single cells representing ∼195 mm² of COVID19 PM lung tissue. c) Cell counts per mm² of the ablated ROI area for each of the 8 pathology classes. d) A graph showing the % of air space within each ROI as a function of pathology class. e) A graph of the cell counts per mm² of actual lung tissue in each ROI. Differences between pathology classes were considered statistically significant where P ≤ 0.05 (Kruskal–Wallis test). ∗P ≤ 0.05, ∗∗P ≤ 0.01, ∗∗∗P ≤ 0.001, ∗∗∗∗P ≤ 0.0001. The whiskers of all box-and-whisker plots represent the range of data for all similar graphs in this Figure and ongoing.

Fig. 3

Analysis of Tier 1 cell type clusters reveals key immune and structural cell signatures defining temporal stages of DAD and alternate pathology classes. a) A heat map showing the median Z-score normalised values for all 38 phenotypic and functional markers (SARS-CoV2 Spike and Capsid were deemed to be negative so not used) for tier 1 level clusters. The coloured bars denote the frequency of each cluster across the entire single cell data set for all 8 pathology classes. b) A PacMap of the single cell level data coloured by tier 1 cluster as shown in the associated legend. c) Upper panels show pseudo-coloured overlaid key immune cell markers for representative ROIs from all 8 pathology classes. Middle panels show pseudo-coloured overlaid key structural cell markers for the same 8 representative ROIs. Lower panels show cluster maps for the same 8 representative ROIs with all 10 tier 1 clusters. Each column corresponds to duplicate ROIs. d) Graphs showing the cell counts per mm² of lung tissue per ROI for key tier 1 immune cell types. From left to right; neutrophils, mononuclear phagocytes, T cells and B/Plasma Cells. e) Analogous graphs as shown in D but for major tier 1 structural cell types. From left to right; AT1, AT2, vascular endothelium and lymphatic endothelium. Differences between pathology classes were considered statistically significant where P ≤ 0.05 (Kruskal–Wallis test). ∗P ≤ 0.05, ∗∗P ≤ 0.01, ∗∗∗P ≤ 0.001, ∗∗∗∗P ≤ 0.0001.

Fig. 4

Tier 2 cluster analysis reveals a unique set of cell signatures linked to DAD progression. a) A heat map showing the median Z-score normalised values for all 38 phenotypic and functional markers (SARS-CoV2 Spike and Capsid were deemed to be negative so not used) for tier 2 level clusters. The coloured bars denote the frequency of each cluster across the entire single cell data set for all 8 pathology classes. The cluster ID is given by the number below the column (cluster) and denoted in the legend. b) Graphs showing the cell counts per mm² of lung tissue per ROI for key tier 2 lymphocytic cell types. From left to right; memory CD4 T cells, memory CD8 T cells, CD4 T cells with active complement and tissue resident features, B cells and epithelial-associated plasma cells. c) Graphs showing the cell counts per mm² of lung tissue per ROI for key tier 2 non-lymphocytic immune cell clusters. From left to right; neutrophils with signatures of interferon signalling and MHC-class 1 presentation, neutrophils with interferon signalling and NET-associated signatures, mononuclear phagocytes and mononuclear phagocytes with a repair-promoting signature. d) As in B and C but showing key tier 2 structural cell clusters. From left to right; AT2 cells with a transitioning and platelet associated signature, AT 2 cells, AT1 cells with active complement, vascular endothelium 1 (CD61+ CD31+ cells mapping to capillaries) and vascular endothelium 2 (CD61+ CD31+ cells mapping to vascular structure outlines of greater size than capillaries). Differences between pathology classes were considered statistically significant where P ≤ 0.05 (Kruskal–Wallis test). ∗P ≤ 0.05, ∗∗P ≤ 0.01, ∗∗∗P ≤ 0.001, ∗∗∗∗P ≤ 0.0001.

Fig. 5

Spatial neighbourhood analysis of tier 1 clusters reveals key interactions and avoidances that correlate with DAD initiation and progression. a) Heat maps for each of the 5 major pathology classes involved in classical disease progression showing the significance of interaction, avoidance or indifference for all 10 tier 1 clusters as per the legend. b) Representative cluster maps with coloured boxes denoting areas of focus demonstrating highly biologically relevant interactions. Magnified duplications of coloured boxes are provided in the lower part of the panel. A section of the panel component representing EDAD in Fig. 5 is a duplicate of the panel component representing the EDAD cluster map in Fig. 3.