Comprehensive T1-relaxation mapping in 35,160 individuals identified leptin (p=1.13×10⁻⁷³) and novel genetic loci near IGF1R, ZNF596, and SYNPO2L as key markers of myocardial fibrosis.
Observational (n=35,160)
Deep learning-based comprehensive T1-relaxation mapping of cardiac MRI identified novel proteomic (leptin) and genomic (IGF1R, ZNF596, SYNPO2L, FRK) markers of myocardial fibrosis.
Abstract Background Heart failure remains a leading global cause of mortality, with myocardial fibrosis playing a pivotal role in its pathogenesis1,2. While current therapies can slow disease progression, they do not directly target fibrotic remodeling. Novel strategies to measure and characterize myocardial fibrosis could thus inform more precise therapeutic approaches3. Purpose We hypothesized that comprehensive T1-relaxation mapping of the entire myocardium, through advanced machine learning models, would identify novel protein and genetic determinants of myocardial fibrosis. Methods Using the UK Biobank cardiac MRI dataset, we developed an open-source U-Net–based software to automatically segment the myocardium from SHMOLLI T1-relaxation images (Dice=0.85)4,5. Unlike prior studies focusing only on mean septal T1 values, we extracted multiple T1-derived features—mean, SD, and percentile cutoffs—from the entire myocardial region6. Additionally, we utilized a variational autoencoder (VAE) to capture latent representations of T1-relaxation maps (MSE=0.0008, PSNR=30.8, SSIM=0.99)7,8,9. All phenotypes were adjusted for age, sex, BMI, and principal genetic components to minimize confounding. We then performed a protein-wide association study (PWAS) using Olink proteomics data, and a genome-wide association study (GWAS) on 35,160 individuals to find genetic loci linked to fibrosis10. Results Strong negative correlations were observed with liver function markers (e.g., gamma-glutamyltransferase r up to -0.952) and hematocrit (r up to -0.908), the latter aligning with existing literature on blood T1 relaxation11,12. HDL cholesterol showed a positive correlation, reproducing results in other studies13. Elevated T1 times were negatively correlated with stroke volume, indexed ventricular stroke volumes, and systolic blood pressure (consistent with prior findings of r=-0.28 between native T1 and LVEF, and potential confounding from concentric hypertrophy)14,15. PWAS highlighted leptin (LEP, p=1.13×10⁻⁷³) as the top protein significantly associated with T1 metrics, alongside IGFBP1/2, FABP4, and IL1RN. GWAS identified variants near IGF1R (rs99281833, p=6.16×10⁻⁹), ZNF596 (rs85351332, p=6.47×10⁻¹⁰), and SYNPO2L (rs75415677, p=2.94×10⁻⁸) as key genetic contributors (Figure 1). Additionally, iron homeostasis genes (HFE/TMPRSS6) emerged, supporting their role in fibrotic pathways. VAE analysis identified the FRK locus (Src kinase family), whose inhibition reduces tissue fibrosis (Figure 2)16. Conclusions This comprehensive T1-relaxation mapping approach identified specific proteomic and genomic markers of myocardial fibrosis. The leptin pathway, iron regulation genes, and novel loci near IGF1R, ZNF596, SYNPO2L, and FRK reveal new targets for potential therapeutic intervention. Future work will validate these findings in independent cohorts and further explore the interplay between myocardial structure, fibrotic remodeling, and clinical heart failure outcomes.Figure 1.T1-mapping percentiles GWAS Figure 2.T1-mapping latent space GWAS
Reddy et al. (Sat,) conducted a observational in Myocardial fibrosis (n=35,160). Comprehensive T1-relaxation mapping in 35,160 individuals identified leptin (p=1.13×10⁻⁷³) and novel genetic loci near IGF1R, ZNF596, and SYNPO2L as key markers of myocardial fibrosis.