Identify Mesothelial Cell States Across 12 Organs

A mesothelial cell compendium across healthy mouse organs

To obtain a detailed panoramic view of mesothelial cell states across mouse internal organs, we compiled and re-analyzed 32 single-cell RNA-seq datasets from public repositories across twelve different tissues/organs including lung, heart, liver, spleen, mesentery, kidney, pancreas, adipose tissue, uterus, vasculature, prostate and omentum—generating a compendium of 23,134 mesothelial cells. When available, we used pre-processed data and extracted mesothelial cell datasets based on the authors’ annotation. When annotation was unavailable, we used raw data, which we preprocessed, filtered, and annotated, using restricted criteria of defined mesothelial markers. We also excluded immune, endothelial, and epithelial cell types. We then corrected our compiled dataset for any batch effects using BatchBench25 before merging, filtering, normalizing (Fig. 1a and Supplementary Fig. 1a).

We then performed dimension reduction UMAP analysis, followed by Leiden analysis (resolution = 0.5), which revealed eight distinct mesothelial cell clusters. We annotated these according to their most distinguishing gene’s expression: Cadherin 12 (Cdh12), Sodium Leak Channel Non-Selective (Nalcn), Prothymosin Alpha (Ptma), Delta Like Non-Canonical Notch Ligand 1 (Dlk1), Dermatopontin (Dpt), Mitochondrially Encoded Cytochrome C Oxidase I (Cox1), Thymosin Beta 10 (Tmsb10) and Actin Gamma 1 (Actg1) (Fig. 1b).

Each of these eight mesothelial cell types, expressed unique transcript profiles with over 200 differentially expressed genes. When comparing the tissues origins of the groups we found they had distinct organ-distributions with organ-specific expression pattern. Some of the cell types were found in a unique organ, whereas others were globally present across different organs. For example, adipose tissue and lung cells, contain organ-specific mesothelial subsets (Supplementary Fig. 1b).

Lung mesothelium was more complex, including two different mesothelial cell types, distinguished by expression of Ptma and Cox1. Indeed, immunostaining on healthy mouse lung samples confirmed both subtypes were present (Supplementary Fig. 1c). Interestingly, the Ptma group of cells also overexpress infectious disease and innate immune response genes and those involved in cellular responses to stimuli. As with lung adipose tissue was composed of two different cell types expressing either Cdh12 or Nalcn. Gene ontology (GO) enrichment indicated Cdh12 cells are involved in regulation of macromolecule biosynthesis process, cell projection organization, cellular response to stress and cytoskeleton organization, whereas Nalcn cells are involved in immune response and cellular responses to stimuli (Supplementary Fig. 1d).

Our analysis further identified two multi-organ mesothelial cell types. The first expressed Tmsb10 and was present across many different organs such as prostate, uterus, spleen, heart, and vascular. Tmsb10 plays an important role in cytoskeletal organization by binding to actin monomers, suggesting a role for this mesothelial cell type in organ surface rigidity. The second multi-organ mesothelial cell type overexpressed Actg1 and was specific to abdominal organs including liver, kidney and omentum. Actg1 is a non-muscular cytoskeletal actin isoform that regulates cell proliferation, apoptosis and tumor cell migration.

We then employed “pseudotime” analysis to determine the hierarchical “distance” between the mesothelial cells from different tissues. Our analysis indicated mesothelial cells within individual body cavities (either pericardial, pleural or peritoneal) share gene expression patterns, have a closer hierarchical distance, and have a more recent common cellular ancestor than mesothelial cells from adjacent cavities (Fig. 1c and Supplementary Fig. 1e, f).

A mesothelial cell compendium across mouse diseased organs

Next, we extended our analysis to diseased tissues and organs. For a deep dive into the different mesothelial cell states and their potential functions in disease, we collected mesothelial cell scRNA-seq data from twenty-five mouse datasets representing seven different fibrotic-related diseases, including lung and liver fibrosis, heart and kidney ischemia, heart aortic constriction, pancreatic adenocarcinoma, peritoneum adhesion, vasculature high fat diet. This revealed three new disease-specific cell states (Fig. 1d).

We determined the most enriched signaling pathways in each state, using reactome pathway analysis, and annotated these three mesothelial cell states as (i) a “fibrogenically-active” cell state expressing pathways for collagen biosynthesis and its modifying enzymes, (ii) a “metabolically-active” state expressing metabolic pathways of amino acids and derivatives, and (iii) a “proteolytically-active” state related to connective tissue dismantling, including matrix metalloproteinases and collagen degradation enzymes (Supplementary Fig. 1g–j). These three cell states were universal, appearing across many diseases, animal models and organs (Fig. 1e). We further identified distinguishing markers for each cell state, including Matrix Gla protein (Mgp) and Secreted Protein Acidic and Cysteine Rich (Sparc) for “fibrogenically active”, interferon alpha-inducible protein 27 like 2A (Ifi27l2a) and Cysteine Rich Protein 1 (Crip1) for “metabolically-active” cells and Decorin (Dcn) and Placenta Associated 8 (Plac8) for “proteolytically-active” cells (Fig. 1f).

Lung fibrosis is driven by mesothelial cell state differentiation

To determine the interrelationships between the three universal mesothelial cell states across a disease progression timeline, we analyzed scRNA-seq of mesothelial cells from a Bleomycin-instilled lung fibrosis model, using eighteen different time points spanning 54 days (Fig. 2a). This lung fibrosis progression analysis revealed the same three cell states with the same markers as before: (i) fibrogenic, (ii) metabolic, and (iii) proteolytic with the addition of a fourth (iv) “immune modulatory” state with Serum Amyloid A3 (Saa3) as marker (Fig. 2b, c). This fourth cell state was enriched for interleukin and cytokine signaling and other immune system pathways (Supplementary Fig. 2a–g).

The earliest disease-specific mesothelial state was predicted to be metabolically active cells expressing Ifi27l2a, and Crip1. Metabolically active cells were predicted to differentiate into proteolytic-active cells overexpressing metalloprotease and endopeptidases, including Matrix Metallopeptidase 8 and 14 (Mmp8, Mmp14) Cathepsin B (Ctsb), Dcn and Plac8. Proteolytic cells transiently were predicted to differentiate into immune-active cells expressing chemoattractant Saa3. Finally, immune modulatory cells were predicted to differentiate into fibrogenically active cells, overexpressing Mgp, Sparc and collagens 1 and 3 (Fig. 2d and Supplementary Fig. 2g). This unprecedented differentiation pathway is likely key to the fibrotic process.

The predictive differentiation steps were mirrored by sequential enrichment of the mesothelial cell states over the time-course. For example, mesothelial cells adopted a metabolically active phenotype in the initial pre-fibrotic period from day 2 until day 6 of the Bleomycin model. On day 7 after Bleomycin installation these metabolically active cells differentiated into proteolytic cells, and the two cell types co-existed within the lung pleura until day 10 when some proteolytic cells adopted immune modulatory phenotypes. This immune phenotype emerges around day 5 and gradually increases, reaching its peak at day 10. Then, these immune modulatory cells transition progressively into fibrogenic cells, coinciding with the onset of lung fibrosis at day 11. Starting from day 21 fibrogenic cell numbers plummet, reverting to metabolically active cells, which coincides with fibrosis resolution (Fig. 2e). We confirmed the temporal appearance of these four cell types on sections from fibrotic mouse lungs across day 0, 5, 10, and day 14 (Supplementary Fig. 3).

Next, to identify the genetic regulation of this cell state differentiation, we estimated fate probability and analyzed driver genes across all mesothelial cell states, using the CellRank package26,27. We identified the same drive genes between the cells states in disease as we had observed in our cross-tissue analysis above including the metabolic driver Ifi27l2a, Dcn and Plac8 controlling proteolytic differentiation, Saa3 driving Immune modulatory differentiation, and Mgp, and Sparc are at the apex of fibrogenic differentiation (Fig. 2f).

Mesothelial differentiation triggers pathological matrix internalization

To functionally explore the role of these drivers, ex vivo, we overexpressed individual cell state drivers on pleural lung surfaces. Transfection was made possible with a specialized AAV capsid containing an RGD motif on its viral coat, enabling immediate binding and internalization of the AAV into pleural lining cells. Our AAV strategy contained the genetic driver under the control of a CMV promoter, to force a strong and constitutive expression on pleural surfaces, as well as an mCherry reporter tag, to visualize and quantify transfection/transduction efficiency. To directly assess proteolysis and surface material transfer, we tagged pleural ECM with Fluorescein Isothiocianate (NHS-ester), and 300 µM thick slices were incubated with medium for 5 days then immunolabeled and visualized with confocal microscopy (Fig. 3a).

Overexpression of the above drivers induced a robust differentiation of lining mesothelium ex vivo. Overexpression of Ifi27l2a increased metabolism and mitotic activity, evidenced by increased Ki67 and PCNA staining (Supplementary Fig. 4a). Overexpressing Dcn and Plac8 induced further differentiation into proteolytic cells, resulting in massive proteolysis of pleural ECM and inward transfer of digested macromolecules (Fig. 3b, c). High resolution multi-photon images showed a reorganization of ECM occurred at pleural surfaces, consistent with ECM proteolytic disassembly, accompanied by interstitial accrual of macromolecules (Fig. 3d). Proteolytic activity was confirmed by increased Ctsb expression on mesothelial cells (Supplementary Fig. 4b). Finally, overexpression of the fibrogenic driver Mgp, induced mesothelial-to-mesenchymal transition, cell migration and α-SMA protein expression, consistent with a fibrogenic cell fate (Supplementary Fig. 4c).

Forced mesothelial differentiation results in lung fibrosis

To directly investigate the role of mesothelial cell state drivers in disease progression, we transfected AAV into the pleural space in animals, then injected NHS-FITC intrapleurally to label the ECM there, followed by Bleomycin instillment through the trachea to induce lung fibrosis (Fig. 4a). ECM labeling protocol adopted from ref. 28 was used to track ECM movement and structure in the pleural surface.