HMOX2-driven crosstalk between vascular aging and heart failure: A multimodal bioinformatics and explainable machine learning approach with experimental validation.
HMOX2 was identified as the top predictor (AUC=0.978) of vascular aging-heart failure comorbidity through integrative bioinformatics and machine learning, with experimental validation confirming upregulation of HMOX2, S1PR3, and SERPINA3 in cellular and mouse models.
Key Findings
Results
Analysis of the GSE57338 dataset identified differentially expressed genes in heart failure that, when intersected with WGCNA hub genes and vascular aging-related targets, yielded 272 consensus genes.
GSE57338 dataset comprised 136 controls and 177 HF samples
Vascular aging-related targets were retrieved from GeneCards (n=16,243)
Consensus genes (CGs) were derived by intersecting DEGs, WGCNA hub genes, and VA-related targets
The 272 CGs were enriched in cGMP-PKG signaling, cytoskeletal regulation, and PPAR pathways
Results
Machine learning prioritization identified 12 core genes, with HMOX2 as the top predictor of heart failure-vascular aging comorbidity.
Three machine learning methods were applied: LASSO regression, Random Forest, and SHAP-XGBoost
HMOX2 achieved an AUC of 0.978, making it the highest-ranked predictor
A total of 12 core genes were prioritized across the combined machine learning approaches
Network analysis using GeneMANIA was also applied to identify potential regulators
Results
HMOX2, S1PR3, and SERPINA3 were upregulated in doxorubicin-treated rat primary vascular smooth muscle cells compared to controls.
Validation was performed using qPCR in doxorubicin-induced rat primary vascular smooth muscle cells
Upregulation of all three genes was statistically significant (P < 0.05)
The doxorubicin treatment was used as a model of vascular aging in primary cells
Results
HMOX2, S1PR3, and SERPINA3 were upregulated in the human VSMC cell line under doxorubicin treatment.
Validation was performed in a human VSMC cell line in addition to rat primary cells
Gene expression was assessed by qPCR
Upregulation compared to controls was statistically significant (P < 0.05)
Results
HMOX2, S1PR3, and SERPINA3 were upregulated in doxorubicin-treated H9C2 cardiomyoblasts, representing the heart failure cellular model.
H9C2 cardiomyoblast injury model was used as the in vitro HF model
Gene expression was confirmed by qPCR
Upregulation compared to controls was statistically significant (P < 0.05)
This finding paralleled expression changes seen in the vascular aging cell models
Results
Doxorubicin administration in mice induced significant cardiac dysfunction and myocardial fibrosis.
A mouse model of doxorubicin-induced heart failure was used for in vivo validation
Cardiac function was assessed by echocardiography
Myocardial fibrosis was evaluated by Masson's trichrome staining
Both cardiac dysfunction and fibrosis were significant findings in treated animals (P < 0.05)
Results
Doxorubicin-treated mice showed elevated expression of vascular senescence markers P16 and P21 in vascular tissues.
P16 and P21 are established markers of cellular senescence and vascular aging
Expression was assessed by qPCR in vascular tissues from the mouse model
Elevation of both markers was statistically significant (P < 0.05)
These findings link the cardiac injury model to concurrent vascular aging phenotypes
Results
The 272 consensus genes were functionally enriched in cGMP-PKG signaling, cytoskeletal regulation, and PPAR pathways.
Functional enrichment analysis was applied to the 272 consensus genes
Three major pathway categories were identified: cGMP-PKG signaling, cytoskeletal regulation, and PPAR pathways
These pathways were identified as potentially involved in VA-HF synergy
What This Means
This research suggests that vascular aging (the deterioration of blood vessels with age) and heart failure share common molecular mechanisms, and that a gene called HMOX2 may play a central role in both conditions. The researchers used a large gene expression dataset from heart failure patients combined with known vascular aging genes to identify 272 genes active in both diseases. They then applied multiple machine learning techniques to narrow this list down to 12 key genes, with HMOX2 being the strongest predictor of the combined condition (achieving a near-perfect accuracy score of 0.978 out of 1.0).
To confirm these computational findings in the laboratory, the researchers treated cells — both rat and human blood vessel muscle cells (to model vascular aging) and heart muscle cells (to model heart failure) — with doxorubicin, a chemotherapy drug known to cause heart and vascular damage. In all cell types, three genes — HMOX2, S1PR3, and SERPINA3 — were consistently turned up compared to untreated cells. They also treated mice with doxorubicin and confirmed that these animals developed heart dysfunction, heart scarring (fibrosis), and increased expression of aging markers (P16 and P21) in their blood vessels, closely mirroring both heart failure and vascular aging simultaneously.
This research suggests that HMOX2, along with S1PR3 and SERPINA3, may serve as important molecular links between aging blood vessels and failing hearts. The findings could point toward new targets for treatments aimed at patients who suffer from both conditions at the same time, which is a common and difficult clinical challenge. The pathways identified — particularly cGMP-PKG signaling and PPAR pathways — are known to be involved in cardiovascular regulation and may represent areas where future therapies could be developed.
Li J, Zhou G, Wang Q, Song Z, Liu J, Shen C, et al.. (2026). HMOX2-driven crosstalk between vascular aging and heart failure: A multimodal bioinformatics and explainable machine learning approach with experimental validation.. PloS one. https://doi.org/10.1371/journal.pone.0357886