MicroRNA-centered theranostics for pulmoprotection in critical COVID-19

Abstract

Elucidating the pathobiological mechanisms underlying post-acute pulmonary sequelae following SARS-CoV-2 infection is essential for early interventions and patient stratification. Here, we investigated the potential of microRNAs (miRNAs) as theranostic agents for pulmoprotection in critical illness survivors. Multicenter study including 172 ICU survivors. Diffusion impairment was defined as a lung-diffusing capacity for carbon monoxide (D_LCO) <80% within 12 months postdischarge. A disease-associated 16-miRNA panel was quantified in plasma samples collected at ICU admission. Bioinformatic analyses were conducted using KEGG, Reactome, GTEx, and Drug-Gene Interaction databases. The results were validated using an external RNA-seq dataset. A 3-miRNA signature linked to diffusion impairment (miR-27a-3p, miR-93-5p, and miR-199a-5p) was identified using random forest. Levels of miR-93-5p and miR-199a-5p were independently associated with the outcome, improving patient classification provided by the electronic health record. The experimentally validated targets of these miRNAs exhibited enrichment across diverse pathways, with telomere length quantification in an additional set of samples (n = 83) supporting the role of cell senescence in sequelae. Analysis of an external dataset refined the pathobiological fingerprint of pulmonary sequelae. Gene-drug interaction analysis revealed four FDA-approved drugs. Overall, this study advances our understanding of lung recovery in postacute respiratory infections, highlighting the potential of miRNAs and their targets for pulmoprotection.

Keywords: MT: Novel therapeutic targets and biomarker development Special Issue; long COVID; microRNA; post-COVID syndrome; pulmoprotection; telomere; telomere length; theranostics.

Graphical abstract

Figure 1

miRNAs as theranostic agents for pulmoprotection in postacute sequelae in survivors of critical illness (A) Random forest feature selection model. Top, the importance of the contribution of each miRNA to the model. Bottom, out-of-bag (OOB) error rate when including accumulated miRNAs in the model. The optimal model selection process does not include more miRNAs if they do not represent a significant improvement in the model prediction. The optimal model included 3 miRNAs. (B) Integration of selected miRNA into a clinical model for the prediction of lung-diffusion impairment (D_LCO <80%) previously constructed by our group. miRNA levels were dichotomized for the optimal cutoff. The final model presented was constructed using stepwise logistic regression. The odds ratio represents the risk change per 1 SD in continuous predictors. (C) Receiver operating characteristic (ROC) curves comparing the clinical model (black curve) with the clinical model incorporating the miRNA signature (red curve). The discriminative performance of both models is quantified by the area under the ROC curve (AUC). Reclassification analyses, the Integrated Discrimination Improvement (IDI), and continuous Net Reclassification Improvement (NRI) indexes were implemented to quantify the added value of miRNAs. (D) Venn diagram including the experimentally validated targetome of the selected miRNAs (TarBase version 8 database). (E) STRING protein–protein interaction network. The analysis included 1,133 genes (targets of at least 2 miRNAs). Edges indicate both physical and functional associations (interaction score cutoff is set at 0.95). (F) Top 25 terms (ranking based on q value and plotted according to the Rich Factor) of Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome of the 1,133 target genes. The Rich Factor quantifies pathway enrichment, whereas the q value denotes significance. (G) Telomere length between D_LCO <80% and D_LCO ≥80% groups. (H) Volcano plot representing differential gene expression in an external whole-blood RNA-seq dataset (GSE228320) between D_LCO <80% and D_LCO ≥80% groups. Differential expression criteria were set at p < 0.05 and fold change ≥ 1.5. (I) Enrichment analysis of lung cell types based on single-cell RNA-seq data from the GTEx Project database. Each column shows a cell type and each row shows a gene. Point size indicates the number of cells where the gene was detected, and color represents expression level. (J) Top 25 enriched terms (ranked by q value) from KEGG and Reactome among the differentially expressed target genes. Functional enrichment analysis was performed using the clusterProfiler package (version 4.8.3) of the Bioconductor software (version 3.17). EGFR, estimated glomerular filtration rate; NIMV, noninvasive mechanic ventilation.