The COVID-19 pandemic, caused by SARS-CoV-2, has had a significant global impact. While vaccination rates have improved outcomes, the emergence of new variants necessitates the development of outpatient treatments for severe COVID-19. COVID-19 exhibits remarkable symptomatic heterogeneity, ranging from mild to severe cases involving pneumonia, acute respiratory distress syndrome, and multiple organ failure. Preclinical in vitro models, derived from human and other species' cell lines, have been instrumental in COVID-19 research, aiding in virus isolation, pathogenesis studies, and drug identification. Current drug repositioning strategies and in vitro testing of antiviral compounds have yielded limited success, possibly due to the complexity of the disease and inconsistencies between theoretical and practical models. Existing antiviral drugs like Remdesivir and Paxlovid show promise but face the risk of resistance development. Targeting virus-induced host gene expression alterations offers an alternative strategy to overcome resistance and potentially treat infections by other coronaviruses. This study aims to analyze COVID-19's molecular signature in clinical samples and cell models to identify master regulators and drug repositioning candidates. A holistic approach that considers the complete molecular context of the disease could enhance drug repositioning strategies and increase the success rate of clinical trials.

The introduction extensively reviews the existing literature on COVID-19 pathogenesis, treatment strategies, and the limitations of current approaches. It highlights the need for new therapeutic strategies that target the host response to infection, rather than solely focusing on viral components. The use of preclinical models, while valuable, requires careful evaluation to ensure their suitability for clinical translation. The review emphasizes the importance of a holistic approach to better understand the complex molecular landscape of COVID-19 to improve drug discovery efforts and clinical trial success rates.

The study used several publicly available gene expression datasets from the Gene Expression Omnibus (GEO) database. These datasets included transcriptional data from lung autopsies of COVID-19 patients and healthy controls (GSE155241), nasopharyngeal swabs from SARS-CoV-2 positive and negative patients (GSE152075), A549 and NHBE cells infected with SARS-CoV-2 (GSE147507), and Vero cells infected with SARS-CoV-2 (GSE159316). Data quality was verified using the fastqcr package and transcription quantification was conducted using the salmon package (except for GSE152075, where pre-calculated counts were used). Differential expression analysis was performed using DESeq2, identifying differentially expressed genes (DEGs) with a q-value ≤ 0.05. Gene Ontology (GO) analysis was done using ClueGO in Cytoscape to identify overrepresented biological processes (q-value ≤ 0.05). Master regulators (MRs) were inferred using a transcriptional network centered on transcription factors from healthy lung tissue (GSE23546) and the RTN package. Only regulatory units with ≥100 gene hits and a hypergeometric test q-value ≤ 0.05 were considered significant. The Connectivity Map (CMap) approach using the PharmacoGx package identified potential drug candidates that could counteract the disease signature (q-value ≤ 0.05 by permutation test). Jaccard indexes and Fisher's exact tests were used to compare similarities between datasets. The R environment (version 4.1.0) and Cytoscape (version 3.8.2) were used for the analyses.

Analysis of lung autopsy data revealed 299 DEGs in severe COVID-19 compared to healthy controls. Top upregulated genes included HLA-DQB1, TNFSF12/TNFSF13, and FAP, while top downregulated genes included DENND11 and EIF3CL. GO analysis associated these genes with immune response, extracellular matrix remodeling, and signaling pathways. Six MRs were identified: TAL1, TEAD4, EPAS1, ATOH8, ERG, and ARNTL2. These MRs are involved in inflammatory response and cell morphogenesis. CMap analysis suggested 52 drugs negatively related to the severe COVID-19 signature, including corticosteroids, antibiotics, and psychoanaleptics. Analysis of preclinical cell models showed a varying number of DEGs across different cell lines (A549, NHBE, Vero). Only four genes (MAFF, CSRNP1, NFKBIA, DUSP1) were consistently DEGs across all models, including lung autopsies. GO analysis revealed associations with immune system regulation, cell death, and stress response across all models, but specific pathways varied. The number of MRs also varied widely across cell models. CMap analysis revealed various drug candidates for each model. Jaccard index analysis revealed higher similarity between NHBE cells and severe COVID-19 samples in terms of GO terms, MRs, and drug candidates than other cell lines. Nasopharyngeal swabs showed greater similarity with A549 MOI 2 cells in terms of GO terms, MRs and drug candidates compared to other cell models. Several consensus MRs were identified across different models.

The identified MRs (TAL1, ERG, TEAD4, ARNTL2, EPAS1, ATOH8) are known to be involved in inflammatory response and cell morphogenesis, aligning with the clinical features of severe COVID-19 (hyperinflammation and cellular changes). The drug repositioning candidates (corticosteroids, antibiotics, psychoanaleptics) support the clinical observation that immunomodulatory drugs are beneficial in severe COVID-19. The analysis of preclinical cell models revealed that NHBE cells show greater similarity to clinical lung samples than other cell lines, suggesting their suitability for studying lung tissue gene expression in the context of COVID-19. A549 cells better represent upper respiratory tract epithelium. None of the cell models provided a statistically reliable representation of clinical lung data for drug repositioning purposes. The study highlights the complex interplay between viral infection and the host response and emphasizes the need for more robust preclinical models for accurate translation to the clinic.

This study identifies key master regulators and potential drug candidates for severe COVID-19 using a holistic approach incorporating both clinical and preclinical data. Immunomodulatory drugs, along with antibiotics, psychoanaleptics, and cardiovascular drugs, emerge as potential therapeutic candidates warranting further investigation. The study also provides insights into the translational potential of different COVID-19 cell models, suggesting that NHBE cells provide better representation of lower respiratory tract and A549 cells may be better for upper respiratory tract. Further research should focus on improving preclinical models to better reflect the complex molecular landscape of COVID-19 for improved drug repositioning strategies and clinical trial success.

The study's limitations include the limited number of samples in the lung autopsy dataset, which might affect the generalizability of findings. The use of publicly available datasets introduces the possibility of variations in experimental protocols. The in silico nature of the drug repositioning approach necessitates further validation through in vitro and in vivo experiments.