The cardiac blood transcriptome predicts de novo onset of atrial fibrillation in heart failure

Abstract

Heart failure (HF) increases the risk of developing atrial fibrillation (AF), leading to increased morbidity and mortality. Therefore, better prediction of this risk may improve treatment strategies. Although several predictors based on clinical data have been developed, the establishment of a transcriptome-based predictor of AF incidence in HF has proven to be more problematic. We hypothesized that the transcriptome profile of coronary sinus blood samples of HF patients is associated with AF incidence. We therefore enrolled 192 HF patients who were selected for biventricular cardioverter defibrillator implantation. Both coronary sinus and peripheral blood samples were obtained during the procedure. Patients were followed-up during two years and AF occurrence was based on interrogation of the defibrillator. A total of 96 patients stayed in sinus rhythm (SR) during follow-up, 13 patients developed AF within 1 year and 10 patients developed AF during the second year of follow up. Gene expression profiling of coronary sinus samples led to the identification of 321 AF predictor genes based on their differential expression between patients developing AF within 1 year of blood sampling and patients remaining in SR. The expression levels of these genes were combined to obtain a molecular atrial fibrillation prediction score for each patient which was significantly different between both patient groups (Mann-Whitney, p = 0.00018). We conclude that the cardiac blood transcriptome of HF patients should be further investigated as a potential AF risk prediction tool.

Keywords: Atrial fibrillation; Coronary sinus blood; Heart failure; Prognostic biomarkers; Transcriptomic profiling.

Graphical abstract

Fig. 1

Flow-chart and set-up of the study. A. Flow-chart of the study. B. Comparison between 1-year and 2-year AF groups to the SR group C. Comparison between the 1-year AF group and the matched SR group. AF - atrial fibrillation, SR - sinus rhythm.

Fig. 2

Visualization of differential gene expression between 1-year AF and SR patients. A. Heatmap of two-way hierarchical clustering of 140 genes differentially expressed between thirteen 1-year AF patients (black boxes in the patient tree) and 96 SR patients (grey boxes in the patient tree) (global comparison). B. Violin plot of the MAPS (Molecular Atrial fibrillation Prediction Score) of each patient. AF - atrial fibrillation, SR - sinus rhythm. C. ROC curve based on the 140 differentially expressed genes in the global comparison.

Fig. 3

Visualization of differential gene expression between 1-year AF and matched SR patients. A. Heatmap of two-way hierarchical clustering of 321 genes differentially expressed between thirteen 1-year AF patients (black boxes in the patient tree) and thirteen matched SR patients (grey boxes in the patient tree). B. Violin plot of the MAPS (Molecular Atrial fibrillation Prediction Score) of each patient from the case-control comparison. C. ROC curve based on the 321 differentially expressed genes in the case-control comparison. D. Left; comparison between the log2(FC) values obtained by the global and the case-control comparisons. Right; Venn diagram showing the overlap between the number of genes differentially expressed using either the global or the case-control comparison. AF - atrial fibrillation, CaseCtrl - case-control, SR - sinus rhythm, FC - Fold change, AUC - area under the curve.

Fig. 4

Digital cytometry results of the case-control patient samples. Visualization of the relative percentage of 22 immune cell types in the 26 blood samples of the case-control patients, based on digital cytometry. B. Heatmap of two-way hierarchical clustering of 60 genes from the AF predictor in 22 immune cell types.

Fig. 5

Weighted gene co-expression network analysis (WGCNA). A. Dendrogram depicting hierarchical clustering to detect outlier samples. B. Heatmap of correlations between sample traits and WGCNA module eigengenes. The selected modules are indicated by black boxes C. Left side: Donut chart showing the distribution of the 10,000 input genes over the different modules. Right side: Donut chart showing the distribution of the AF predictor genes over the different modules. Arrows indicate the selected modules MEturquoise, MEdarkred and MEblack. BMI - body mass index, HR – heart rate, DABP - diastolic arterial blood pressure, DEGs - differentially expressed genes, IHD - ischemic heart disease, LBBB - left bundle branch block, LVEDD - left ventricular end diastolic diameter, LVEF - left ventricular ejection fraction, MR - mitral regurgitation >2/4, NYHA - New York Heart Association class, SABP - systolic arterial blood pressure, VT - ventricular tachycardia.

Fig. 6

Comparison of the AF predictor genes to the genes from the AF-prediction-associated WGCNA modules MEturquoise, MEdarkred and MEblack. A. Venn diagram showing the overlap between the AF predictor genes (DEGs) and genes from the MEturquoise, MEdarkred and MEblack modules. B—D. Scatterplots displaying the relationship between the Gene Significance for AF prediction and the module membership of the gene for the three selected modules (MEturquoise, MEblack and MEdarkred resp.). The overlapping AF predictor genes (DEGs) are indicated by black (B and D) or grey (C) dots. Genes verified by quantitative PCR are identified in (B). DEGs - differentially expressed genes.

Fig. 7

Concept network of Gene Ontology Biological Processes and Disease Ontology enrichment analysis. A-D. Gene Ontology Biological Processes enrichment analyses of (A) the overlap between the AF predictor genes and either the MEturquoise or the MEdarkred or the MEblack genes; (B) the turquoise module; (C) the black module and (D) the darkred module. E. Disease Ontology enrichment analysis of the turquoise module. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 1

ROC Leave one out. ROC curve based on the MAPS values of the test samples obtained from the leave-one-out strategy.

Fig. 2

Analysis of AF predictor genes in the global analysis. A. Violin plot of the MAPS values of each patient from the global comparison, based on the AF predictor genes. B. ROC curve based on the 321 AF predictor genes and patients from the global analysis.

Fig. 3

AF prediction by genes identified by LASSO regression analysis.

Fig. 4

Concept network of Gene Ontology Biological Processes analysis of AF predictor genes.

Fig. 5

GSEA enrichment plots and corresponding heatmaps of genes from the core enrichment.

Fig. 6

Quantitative PCR results.