Identification and validation of circulating microRNAs as biomarkers for heart failure with preserved ejection fraction

Abstract Background Heart failure (HF) is a syndrome due to cardiac insufficiency, manifested by excessive volume overload, increased sympathetic activity, and circulatory redistribution [1]. Heart failure with preserved ejection fraction (HFpEF) accounts for approximately 40%-50% of the total HF events, and its development is often accompanied by various complications, including hypertension, obesity, and metabolic syndrome. Biomarker associations in heart failure with reduced ejection fraction were mostly cardiac stretch based, while many biomarkers of HFpEF are also associated with inflammation [2]. To date, treatments that effectively improve the clinical prognosis of HFpEF are relatively limited, thus HFpEF remains one of the challenges that needs to be further explored. A full understanding of the molecular mechanisms of HFpEF is essential to improve the early diagnosis, treatment and prognosis of these patients. Purpose The study aims to explore key circulating miRNAs and related molecular pathways in HFpEF patients through bioinformatics analysis and machine learning, and to further validate them using a double-hit HFpEF mouse model. Methods We downloaded HF datasets including circulating miRNA expression profiles from the GEO database. LASSO regression, SVM-RFE, Random Forest and XGboost algorithm were conducted to identify key miRNAs for diagnostic model construction. We used R software, GO, KEGG and Cytoscape to better understand the functions of key miRNA targets. Besides, the mRNA-miRNA interaction networks were predicted by starbase. We also constructed a HFpEF mouse model subjected to high fat diet and L-NAME (0.5 g/L) in drinking water to discover the correlation between the screened circulating miRNA and HFpEF. Results After differential expression analysis of the combined HFpEF dataset, we identified 37 differentially circulating miRNAs, of which 15 were down-regulated and 22 were up-regulated. Subsequently, we screened by machine learning and identified 4 key miRNAs that may be involved in the progression of HFpEF, including hsa-miR-645, hsa-miR-296-5p, hsa-miR-199b-5p, and hsa-miR-509-3-5p. We validated these miRNAs based on their expression, and the ROC plots suggest that these miRNAs may have diagnostic capability in HFpEF. The results of GO and KEGG enrichment analyses showed that the target genes of the differentiated miRNAs were mainly enriched in the MAPK signaling pathway, mTOR signaling pathway, protein processing in endoplasmic reticulum, hippo signaling pathway, cell cycle, Foxo signaling pathway, hypertrophic cardiomyopathy, dilated cardiomyopathy, and circadian rhythm. The mRNA-miRNA interaction networks suggested that TMEM135 was regulated by both has-miR-199b-5p and hsa-miR-296-5p. The RT-PCR results also suggested that miR-199b-5p was upregulated in the HFpEF mice. Conclusion We validated 4 circulating miRNAs with strong correlation to HFpEF, which provide a more reliable basis for pre-symptomatic diagnosis.