An integrated computational-experimental approach identifies Malate Synthase G as therapeutic target to disrupt carbon metabolism in chronic Pseudomonas aeruginosa infection.

Pseudomonas aeruginosa is a leading cause of nosocomial infections, particularly in individuals with a compromised immune system. Due to its strong adaptive ability, P. aeruginosa tends to develop antibiotic resistance and establish chronic infection, making its eradication through traditional antibiotics challenging. Thus, the development of novel therapeutic targets and corresponding inhibitors is urgently required. We conducted bioinformatics analyses of gene-chip datasets (GSE10362, GSE21966) of P. aeruginosa from patients with cystic fibrosis (CF) in the Gene Expression Omnibus (GEO) database, ranging from early to late stages of infection. Differentially expressed gene (DEG) analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment, and Weighted Gene Co-expression Network Analysis (WGCNA) were performed to identify key functional DEGs. Among these genes, candidate genes were refined using three machine learning algorithms: LASSO regression, random forest (RF), and support vector machine (SVM). Using this integrated approach and published reports, target gene was identified consequently. With reported inhibitors, a structure-based virtual screening model against the target protein was established to screen inhibitors from an FDA-approved drug library. Screened hits were experimentally validated through in vitro MIC and checkerboard broth microdilution assays to determine their antibacterial activity and the synergistic effect of tobramycin (TOB) and amikacin (AMK). Enzyme activity was measured to evaluate target protein inhibition. Transcriptome sequencing analysis was performed to explain the effect of target inhibition and reveal the possible mechanism of inhibition. We identified 210 upregulated genes in the late stage of infection. KEGG functional enrichment demonstrated that the up-regulated genes were mainly enriched in metabolic pathways. Eight metabolism-related key DEGs were obtained by the intersection of genes from metabolic pathways and genes from the key module of WGCNA. Machine learning algorithms (LASSO, RF, and SVM) and literature investigation results identified glcB, which encodes malate synthase G (MS), as a therapeutic target for developing inhibitors. In virtual screening, we formulated four rules for a screening model based on the docking results, and finally identified three FDA-approved drugs prednisolone, dienogest and carbinoxamine maleate salt (CAR) as candidates. In vitro MIC assay and checkerboard assay suggested that CAR had an antibacterial effect with a concentration-dependent trend and could enhance the effect of tobramycin and amikacin. Consistent with MIC assay, the inhibition rate in MS activity assay showed a concentration-dependent trend. Transcriptome sequencing analysis showed that carbon metabolism was remodeled when MS was inhibited, thus impairing the growth of P. aeruginosa, and the synergistic effect with TOB and AMK may be explained by the impaired ability to combat oxidative stress. This study identified and validated drug target glcB (encodes MS) and potential inhibitor CAR in treating chronic P. aeruginosa infection based on an integrated method of bioinformatics, virtual screening, and in vitro experiments.