Inflammation-Associated Stromal Reprogramming Can Initiate Metaplasia and Facilitate Dysplastic Progression via ECM

SUMMARY
In lungs, chronic exposure to tobacco smoke triggers chronic inflammation, fostering a fibrotic microenvironment and promoting the focal transition of columnar human bronchial epithelial cells (hBECs) to squamous metaplasia. These cells that change identity, metaplasias, can progress to squamous dysplasia and, ultimately, lead to squamous cell carcinoma. The cellular stress associated with chronic inflammation (e.g., oxidative stress, DNA damage) enhances TGF-β signaling and upregulates HSP47 expression in fibroblasts which results in increased collagen fiber alignment, elevated tissue stiffness, and subsequent activation of YAP-dependent mechanotransduction in hBECs. In air-liquid interface cultures, stressed fibroblasts alone were sufficient to induce epithelial metaplasia in hBECs. Furthermore, when tumor suppressor function was compromised in hBECs, stressed fibroblasts induced dysplastic phenotypes both in vitro and in vivo. Mechanistically, fibroblasts modulated epithelial cell identity via extracellular matrix (ECM)-dependent mechanotransduction. Notably, a stiff matrix alone could induce squamous metaplasia and dysplasia. Inhibition of HSP47-dependent collagen processing effectively prevented fibroblast-induced squamous metaplasia and dysplasia. Moreover, this intervention was capable of reversing fibroblast-induced metaplasia, restoring bronchial epithelial identity.
The attached h5ad file represent the processed single cell RNA-seq data used in analysis.

Data dictionary for the .h5ad data file.
The count matrix: AnnData.X
=========
X.Shape: (199271 cells, 35606 genes)
X contains raw counts data (non-logged).
X.__class__ == scipy.sparse._csr.csr_matrix
X.dtype == 'float32'

AnnData.obs
===========
index - cell barcodes + sample_diagnosis
samplename - coded sample ID
n_genes - number of measured genes in the cell
n_molecules - number of molecules sequenced
doublet_score - whether the droplet contained two cells (scrublet)
percent_mito - percent of genes measured that are mitochondrial
leiden - cluster labels from leiden algorithm
louvain - cluster labels from the louvain algorithm
nobatch_leiden - non-batch corrected leiden cluster labels
nobatch_louvain - non-batch corrected louvain cluster labels
diagnosis - tissue diagnosis, N normal, M metaplasia, D dysplasia, T tumor
phase - cell cycle phase
sample_diagnosis - sample ID + tissue diagnosis
patient - patient ID
treatment - whether the patient recieved any treatment
procedure - how the sample was aquired
hcl_refined - human cell landscape refined cell type name
hcl_celltype - human cell landscape cell type best match
hcl_score - human cell landscape matching score
CLid - cell ontology ID
CL_name - cell ontology cell type name

AnnData.var
===========
index - gene symbols
gene_ids - ensembl gene IDs
feature_types - type of the feature
genome - genome build
is_mito - whether the gene is mitochondrial
is_ribo - whether the gene is ribosomal

AnnData_embeddings:
========================
PCA (obsm.X_pca)
UMAP (obsm.X_umap)
PCA_nobatch (obsm.X_pca_original)
UMAP_nobatch (obsm.X_umap_nobatch)
neighbors (AnnData.uns)

Marker Genes:
=============
AnnData.uns['rank_genes_groups_filtered'].keys()
names - one list per leiden cluster
logfoldchanges - one cluster vs all others
scores - wilcoxon statistic
pvals - wilcoxon p-value
pvals_adj - BH adjusted p-values
params = {'corr_method': 'benjamini-hochberg',
'groupby': 'leiden',
'method': 'wilcoxon',
'reference': 'rest',
'use_raw': True}