AUTHOR=Koh Hiromi W.L. , Pilbrow Anna P. , Tan Sock Hwee , Zhao Qing , Benke Peter I. , Burla Bo , Torta Federico , Pickering John W. , Troughton Richard , Pemberton Christopher , Soo Wern-Miin , Ling Lieng Hsi , Doughty Robert N. , Choi Hyungwon , Wenk Markus R. , Richards A. Mark , Chan Mark Y. 

TITLE=An integrated signature of extracellular matrix proteins and a diastolic function imaging parameter predicts post-MI long-term outcomes

JOURNAL=Frontiers in Cardiovascular Medicine

VOLUME=10

YEAR=2023

URL=https://www.frontiersin.org/journals/cardiovascular-medicine/articles/10.3389/fcvm.2023.1123682

DOI=10.3389/fcvm.2023.1123682

ISSN=2297-055X

ABSTRACT=<sec><title>Background</title><p>Patients suffering from acute myocardial infarction (AMI) are at risk of secondary outcomes including major adverse cardiovascular events (MACE) and heart failure (HF). Comprehensive molecular phenotyping and cardiac imaging during the post-discharge time window may provide cues for risk stratification for the outcomes.</p></sec><sec><title>Materials and methods</title><p>In a prospective AMI cohort in New Zealand (<italic>N</italic> = 464), we measured plasma proteins and lipids 30 days after hospital discharge and inferred a unified partial correlation network with echocardiographic variables and established clinical biomarkers (creatinine, c-reactive protein, cardiac troponin I and natriuretic peptides). Using a network-based data integration approach (iOmicsPASS+), we identified predictive signatures of long-term secondary outcomes based on plasma protein, lipid, imaging markers and clinical biomarkers and assessed the prognostic potential in an independent cohort from Singapore (<italic>N</italic> = 190).</p></sec><sec><title>Results</title><p>The post-discharge levels of plasma proteins and lipids showed strong correlations within each molecular type, reflecting concerted homeostatic regulation after primary MI events. However, the two molecular types were largely independent with distinct correlation structures with established prognostic imaging parameters and clinical biomarkers. To deal with massively correlated predictive features, we used iOmicsPASS + to identify subnetwork signatures of 211 and 189 data features (nodes) predictive of MACE and HF events, respectively (160 overlapping). The predictive features were primarily imaging parameters, including left ventricular and atrial parameters, tissue Doppler parameters, and proteins involved in extracellular matrix (ECM) organization, cell differentiation, chemotaxis, and inflammation. The network signatures contained plasma protein pairs with area-under-the-curve (AUC) values up to 0.74 for HF prediction in the validation cohort, but the pair of NT-proBNP and fibulin-3 (<italic>EFEMP1</italic>) was the best predictor (AUC = 0.80). This suggests that there were a handful of plasma proteins with mechanistic and functional roles in predisposing patients to the secondary outcomes, although they may be weaker prognostic markers than natriuretic peptides individually. Among those, the diastolic function parameter (E/e' - an indicator of left ventricular filling pressure) and two ECM proteins, <italic>EFEMP1</italic> and follistatin-like 3 (<italic>FSTL3</italic>) showed comparable performance to NT-proBNP and outperformed left ventricular measures as benchmark prognostic factors for post-MI HF.</p></sec><sec><title>Conclusion</title><p>Post-discharge levels of E/e', <italic>EFEMP1</italic> and <italic>FSTL3</italic> are promising complementary markers of secondary adverse outcomes in AMI patients.</p></sec>