Abstract Rationale Pulmonary Fibrosis (PF) is a progressive, highly morbid disease with limited therapeutic options. Recent studies have implicated Hypoxia-Inducible Factor 2 (HIF2, encoded by EPAS1) as a key transcriptional regulator whose chronic activation impairs alveolar repair and promotes aberrant “transitional” cell accumulation in the IPF distal lung epithelium. We hypothesized that reversing this HIF2-driven maladaptive phenotype in a human-relevant model could identify novel, epithelial-targeted, disease-modifying therapeutic candidates. To this end, we developed a high-content phenotypic reversion screen using primary human alveolar organoids. Methods Primary human alveolar organoids were generated from CD326+ epithelial cells isolated from declined donor lungs and cultured in Matrigel under serum-free, feeder-free conditions. A disease-relevant phenotype was induced by stimulating a HIF2-biased signaling axis using the prolyl-hydroxylase inhibitor Roxadustat in combination with the HIF2 allosteric activator M1002. Organoids were arrayed in 384-well plates for high-content screening. We utilized a Cell Painting-based approach, staining for nuclei (Hoechst), RNA (SYTO 14), membranes/Golgi (Wheat Germ Agglutinin), cytoskeleton (Phalloidin), and mitochondria (MitoTracker). Imaging was performed on an ImageXpress.AI system, and captured Z-stacks were analyzed using a custom-trained SINAP algorithm in InCarta software for organoid segmentation and feature extraction. A curated library of 362 compounds (metabolic modulators, kinase inhibitors, epigenetic modifiers, and targeted metabolites (including sterols)) was screened at three doses (0.1, 1, and 10 µM). Extracted morphologic features were normalized using ComBAT and analyzed in PCA space to quantify phenotypic distance from untreated controls. Hits were defined as compounds causing a significant phenotypic reversion (Cohen’s d 1.0) and demonstrating a clear dose-response relationship. Results From the 362-compound library, we identified 14 distinct candidates that robustly rescued the HIF2-driven morphologic phenotype in a dose-dependent manner. Neither of the currently standard-of-care PF therapies (pirfenidone and nintedanib) demonstrated capacity to rescue the HIF2-driven phenotypic changes. The validated candidates included several classes of compounds, notably modulators of lipid metabolism (targeting both nuclear receptors and enzymes) and multiple independent hits targeting a single kinase pathway. Conclusion A high-content phenotypic screen identified 14 promising candidates that reverse aberrant epithelial cell states driven by HIF2 activation. These data validate this screening approach and provide a list of novel, epithelial-targeted therapeutic candidates to be explored. These targets represent promising starting points for a new therapeutic campaign aimed at identifying disease-modifying therapies for pulmonary fibrosis. This abstract is funded by: NIH, Pulmonary Fibrosis Foundation, Parker B Francis Foundation
Mccall et al. (Fri,) studied this question.