Transcriptomic Profiling of Canine Atrial Fibrillation Models After One Week of Sustained Arrhythmia

Abstract

[Figure: see text].

Keywords: atrial fibrillation; atrial remodeling; glutamate; microRNAs; transcriptomics.

Figure 1.

Deconvolution of canine atria cell composition using bulk RNA-sequencing.A, We inferred cell fractions with CIBERSORTx and an atrial-specific gene signature matrix obtained using orthologous murine genes. We present cell fractions for each dog sample that we analyzed in this study. B, When we group animals per treatment arm, we observed a significantly higher fraction of fibroblasts in the atrial fibrillation dog models (AF and AF+AVB) than in the control animals (AFvsCTL Wilcoxon test, P=0.0087 and AF+AVBvsCTL, P=0.015). Principal component analysis of the top 1000 most variable genes expressed in canine atria before (C) and after (D) correction for fibroblast fraction show treatment-dependent clustering after correction for cell composition. AF indicates Atrial fibrillation; AF+AVB, AF with atrioventricular block; and CTL, control.

Figure 2.

Validation of highly expressed RNA by proteomics.A, In 18 atrial samples, 755 genes (NS=619, RNA.sign=122, both.sign=14) found in both data sets are highly correlated at the protein (x axis) and RNA (y axis) levels (Pearson, r=0.49; P=1.57×10⁻⁴⁶). For reference, we annotated 15 genes that are differentially expressed in the RNA-seq experiment and have high protein expression levels. NS, not differentially expressed in the RNA-seq or proteomic experiment; RNA.sign, genes that are differentially expressed in the RNA-seq assay only; Both.sign, differentially expressed genes in both the RNA-seq and proteomic experiments. DE genes in the RNAseq data set have an false discovery rate (FDR) <0.01 (likelihood ratio test) and proteomics data set an FDR <0.05 (F test). The gray area around the line corresponds to the 95% CI. B, Relative expression level of all transcripts measured in the RNA-seq experiment. The histogram shows that genes that are present in both the RNA-seq and proteomic experiment are highly expressed (Common, dark gray) in comparison to the expression levels of all transcripts measured (all_RNA, light gray).

Figure 3.

Analyses of differentially expressed atrial genes identify many biological pathways that are dysregulated in atrial fibrillation (AF) dog models.A, Volcano plots of all transcripts that we analyzed in this study. Transcripts in red have a false discovery rate (FDR) <0.01. We found 434, 5971, and 7867 genes that were differential expression (DE) in the AFvsCTL, AF+AVBvsCTL, and AFvsAF+AVB analyses, respectively. The full DE results are available in Table II in the Data Supplement. B, Upset plot showing the intersection of upregulated and downregulated DE genes (FDR <0.01) in each analysis. C, The 5 most significant biological pathways identified using gene set enrichment analyses (GSEA) for each set of DE genes (FDR <0.01). Full results are available in Table IV in the Data Supplement. AF+AVBvsCTL indicates atrial fibrillation+AV block versus control samples; AFvsCTL, atrial fibrillation versus control samples; and AVB, atrioventricular block.

Figure 4.

Eleven differentially expressed microRNAs (miRNAs) map to a canine chromosome 8 region that is syntenic to human DLK1-DIO3.A, Volcano plots of all miRNA that we measured in our experiments. We identified 31, 19, and 20 miRNA that are differentially expressed (false discovery rate [FDR] <0.01) in the AFvsCTL, AF+AVBvsCTL and AFvsAF+AVB analyses, respectively. B, Miami plots of miRNA and their corresponding statistical significance (y axis) for the AF+AVBvsCTL (top) and ATvsCTL (bottom) analyses. An arrow indicates the miRNA cluster located on the canine chromosome 8 region that is syntenic to human DLK1-DIO3. The odd and even chromosomes FDR values are in blue and red, respectively. C, Upset plot showing the DE miRNA targets located in the syntenic DLK1-DIO3locus and their corresponding number of potential target RNA. We identified potential targets with the MiRNAtap package (predicted by ≥3 databases) from DE miRNA (FDR <0.01) and DE mRNA (FDR <0.01). D, Gene set enrichment analyses (GSEA) with the potential gene targets (x axis) of the DE miRNA located at the syntenic DLK1-DIO3 locus. We only present the top 5 pathways enriched in this analysis. A red square in the heatmap indicates membership of a given target gene to the biological pathways located on the left (empty columns were removed for clarity). GSEA FDR and AF+AVBvsCTL DE FDR are on the right and top of the heatmap, respectively. AF indicates atrial fibrillation; AF+AVBvsCTL, atrial fibrillation+AV block versus control samples; AFvsCTL, atrial fibrillation versus control samples.

Figure 5.

Overlaps in genes differentially expressed in canine AF models and human AF patients. We compared differentially expressed genes in our canine AF models with annotated human orthologues that are DE in human AF left atrial appendages. Homo Sapiens; hsa, hsa_AF; AF in AF rhythm, hsa_AF.SR; AF in sinus rhythm, hsa_CTL; no AF.