1 Introduction

1.1 Atrial Fibrillation and the Pulmonary Vein

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, affecting more than 7 million Americans and 33 million people worldwide [1]. AF is characterized by an irregular pattern of atrial depolarization, resulting in rapid and disorganized atrial conduction and lack of effective atrial chamber contraction. The rhythm abnormality in AF manifests with circulatory deficits and systemic thromboembolism that greatly increase morbidity and mortality. Patients with AF have an increased risk of developing major complications including heart failure and stroke. In addition, the prevalence of AF is expected to rise significantly as the population ages. AF has become a major clinical and economic burden, owing to the limitations and side effects associated with current AF therapies. Although AF most often manifests in the context of pre-existing cardiac pathologies, idiopathic or lone AF forms indicated a heritable component [2]. Over the last decade, genome-wide association studies (GWAS) have identified over one hundred AF-associated loci [3, 4]. These advances in understanding the genetic basis of AF portend an understanding of the complex molecular pathways provoking AF.

The current paradigm of AF requires two arrhythmogenic components: (1) ectopic (triggered) atrial myocardial activity such as early or delayed afterdepolarizations and (2) a fibrillogenic substrate that propagates these triggers throughout the atrial chambers [5]. The most common source of triggered activity is the pulmonary veins, which connect the lungs to the left atria [6, 7]. It has been shown that the pulmonary vein (PV) and peri-pulmonary vein (PPV) myocardium can demonstrate automaticity which can trigger atrial depolarization [8]. A pacemaker current has been observed in the PV-left atrial junction in atrial fibrillation, and gene expression contributing to myocardial automaticity, such as the HCN4 channel protein, is observed in the pulmonary veins [9]. A common treatment for AF is to isolate the PV myocytes from the atrial myocardium by catheter ablation, severing the electrical connectivity between these two populations and preventing inappropriate PV depolarizations from initiating global atrial depolarization and atrial arrhythmias [6, 7].

An important advance in the understanding of the pathophysiology of AF has been the demonstration that the pulmonary vein plays an important role in the initiation of AF [6, 10]. Other initiating foci include the superior vena cava and the left atrium; however, the pulmonary vein remains the most common source of focal activity. This observation has stimulated consideration of the distinctions between the myocardium of the pulmonary/peri-pulmonary vein region and other regions of the atrium. However, the gene expression profile promoting focal activity around the PVs in humans has been difficult to elucidate, owing to the limitations of tissue samples from patients. The molecular identity of the PV myocardium, including specific ion channels and transcription factors expressed in this tissue, has only recently been described. Analysis of a whole-lung transcriptome data set uncovered a set of 24 transcripts, including sarcomeric structural proteins, genes regulating sarcomere assembly, ion transport proteins and hormone signaling in the pulmonary cardiomyocyte gene network [11]. Some of these genes have been linked to AF in humans, suggesting that perturbation of this transcriptional network might lead to altered calcium handling, altered myocardium contractile force and electrophysiological properties which may contribute to the initiation of atrial arrhythmias.

1.2 TBX5 and Pulmonary Vein Development

Holt-Oram syndrome is an autosomal dominant disorder caused by mutations in the T-box family transcription factor TBX5 [12,13,14], occurring in approximately 1 in 100,000 live births [15, 16]. The clinical diagnosis of Holt-Oram syndrome includes completely penetrant upper-limb malformations and variable congenital heart defects, most commonly septal and cardiac conduction defects [17]. Specifically, the conduction defects can manifest as long PR interval, AV and bundle branch block, bradycardia, sick sinus syndrome and atrial fibrillation with or without overt structural defects [18, 19]. The observation that conduction abnormalities can occur in the absence of structural heart defects was an early indication that TBX5 may play a direct role in controlling gene expression essential for normal cardiac conduction. Recently, GWAS have implicated TBX5 in altered cardiac conduction speed (PR and QRS intervals) and AF in structurally normal hearts. Although strict clinical diagnosis of Holt-Oram Syndrome does not involve defects of the pulmonary veins [14, 17], pulmonary vein abnormalities are often associated with Holt-Oram syndrome and TBX5 mutations [20,21,22]. Furthermore, homozygous mutation of Tbx5 in the mouse germline [23] results in complete morphological failure of the cardiopulmonary progenitors, which would normally generate portions of the atria, the atrial septum, pulmonary vein myocardium and smooth muscle cells and airways smooth muscle cells [24, 25]. These observations suggested that TBX5 may drive gene expression important for the development of the pulmonary vasculature in the embryo and for the suppression of atrial arrhythmias in the pulmonary veins or peri-pulmonary vein atrial myocardium in the adult.

Here, we describe regional differences between the transcriptome of the left atrial appendage and the peri-pulmonary vein of the left atrium from adult mice. RNA sequencing revealed previously known and novel regional differences, including several genes implicated in atrial fibrillation. We further compared Tbx5-depedent transcripts between the left atrial appendage and the peri-pulmonary vein of the left atrium from adult mice. Important distinctions were observed between the two regions including Hcn4, which demonstrated differential Tbx5-dependence. Our data suggests that underlying regional differences in the transcriptome of the atrial myocardium may enlighten susceptibility to atrial fibrillation.

2 Results

2.1 Comparison of Peri-Pulmonary Vein and Atrial Appendage Transcriptomes

We investigated transcriptional differences between regions of the left atrium in 7-week-old mice. The left atrium was microdissected into two parts: the left atrial appendage (LAA) and the peri-pulmonary vein (PPV) (Fig. 51.1a). Transcriptional profiling was performed by RNA sequencing (RNA-seq; n = 4 in each group) as previously described [26]. Principal component analysis (PCA) demonstrated that atrial region was responsible for the first principal component, accounting for 32% of the variation between samples (Fig. 51.1b). Differential expression testing between the PPV and LAA samples identified 747 significantly different genes (|log2FC| > 1; FDR < 0.01) with 582 expressed more strongly in the PPV samples and 165 expressed more strongly in the LAA samples (Fig. 51.1c, full table available at GEO accession GSE128870). Gene ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were examined using metascape (http://metascape.org) for both the LAA and PPV genes [27]. GO terms associated with the PPV indicated a preponderance of genes associated with immune cells. We observed GO terms that were unambiguously related to the immune system and only containing genes associated with the PPV, including phagocytosis (GO:0006909), inflammatory response (GO:0006954), humoral immune response (GO:0006959), lymphocyte differentiation (GO:0030098) and cytokine-cytokine receptor interaction (mmu04060). Although these 141 genes are of potential interest, they were separated from the remaining 441 PPV genes to focus on those genes previously verified as myocardially expressed or those with unknown roles or expression domains (Fig. 51.1c).

Fig. 51.1
figure 1

Transcriptional comparison of the left atrial appendage and peri-pulmonary vein atrium. (a) The left atrial appendage (LAA, blue) and the peri-pulmonary vein region of the left atrium (PPV, red) were microdissected from 7-week-old adult mice. (b) Principal component analysis comparing the LAA (blue) and PPV (red) transcriptomes. (c) Volcano plot showing log2(fold-change PPV/LAA). One hundred and sixty five were associated with LAA and 582 were associated with the PPV (|log2FC|>1 and FDR < 0.01). Of the PPV genes, 141 genes were associated with GO terms related to the immune system (red) and removed from downstream analysis. (d) Overlap of genes associated with atrial fibrillation and either the LAA (7 genes) or PPV (12 genes). P-values derived from Fisher’s exact test

Full size image

Because of the relationship between the left atria, pulmonary vein and AF, we examined region-specific gene expression against a list of 235 AF associated genes [3, 28,29,30]. Of the 606 genes that demonstrated regionalized expression in our RNA-seq, 19 have been associated with AF (Fig. 51.1d). Specifically, Adrb1, Nkx2-5, Tmem182, Nppa, Nppb, Gja5 and Kcnh7 are more strongly expressed in the LAA while Ntrk2, Tubb4a, Cacna1d, Tbx3, Scn3b, Shox2, Kcne4, Hcn4, Adra1a, Adrb3, Tubb3 and Cacna2d2 have higher expression in the PPV.

2.2 Tbx5-Dependent Transcriptome of Peri-Pulmonary Vein Atrium Identifies Genes Involved with Cardiac Conduction

We hypothesized that Tbx5-dependent gene expression may vary between different regions in the left atria. Previous work has demonstrated that Tbx5 regulates a gene regulatory network for atrial identity and its removal from the adult mouse results in atrial fibrillation [26]. Tbx5 is robustly expressed in both the LAA and PPV tissues with an average TPM of 79.7 ± 9.9 SEM and 97.1 ± 18.4 SEM, respectively putting Tbx5 within the top 8% of all genes expressed in either area of the left atrium (TPM > 1). To investigate differential Tbx5-dependent regional gene expression in the left atrium of the adult mouse, we performed Tbx5-dependent transcriptional profiling of the PPV tissue and compared it to previously described Tbx5-dependent transcriptional profiling of the left atrial appendage [26]. We performed RNA-seq on the PPV of an adult-specific conditional knockout of Tbx5 [26]. Control (R26 creERT2/creERT2; Tbx5+/+) and conditional mutant (R26creERT2/creERT2; Tbx5flox/flox) animals were treated with tamoxifen at 6-weeks old and tissue was harvested one week later prior to the onset or atrial arrhythmias (Fig. 51.1a) [26]. PCA analysis demonstrated clustering along PC1 and PC2 of samples (cumulative variation > 50%). PC1 demonstrated clustering associated with genotype (control versus conditional Tbx5 mutant), and PC2 demonstrated clustering associated with LA region (LAA versus PPV), indicating that genotype differences drove the majority of variance followed by regional atrial differences (Fig. 51.2a).

Fig. 51.2
figure 2

Tbx5 conditional knockout from the peri-pulmonary vein reveals Tbx5-dependence of multiple unique features. (a) Principal component analysis comparing the left atrial appendage (LAA, blue) and peri-pulmonary vein (PPV, red) transcriptomes from control (R26 creERT2/creERT2; Tbx5+/+, “Tbx5Ctrl”, circles) and Tbx5 conditional knockout (R26creERT2/creERT2; Tbx5flox/flox, “Tbx5CKO”, squares) adult mice. (b) Volcano plot comparing the PPV Tbx5 CKO and control samples. One hundred and forty nine genes were significantly up-regulated and 189 were significantly down-regulated (|log2FC|>1 and FDR< 0.01). (c) Summary GO terms representing all GO terms associated with up- and down-regulated genes (red and blue, respectively) with greater than five associated genes and −log10(P-value) greater than 5. (d) Full GO term list represented by “multicellular organismal signaling” with percent of genes identified (bar) and corresponding −log10(P-value) (dot). (e) Correlation of fold changes between the LAA and PPV samples. Colors and shapes represent whether the genes residuals were in the outermost one-thousandth quantiles (purple square, red circle, blue triangle, dark grey circle) or not (light grey circle). Furthermore, colors and shapes represent whether those outliers were significantly different in both datasets (purple square), the PPV only (red circle), the LAA only (blue triangle) or neither (dark grey circle)

Full size image

We identified 338 significantly dysregulated genes (|log2FC| > 1; FDR < 0.01) comparing the PPV between Tbx5 conditional mutant and controls (Fig. 51.2b, GEO accession GSE128870). Of these, 149 were up-regulated and 189 were down-regulated in the Tbx5 mutants (Fig. 51.2b). Utilizing metascape [27], we identified seven representative GO terms associated with the up-regulated genes and two representative GO terms associated with the down-regulated genes (Fig. 51.2c). The down-regulated GO terms are reminiscent of the LAA previously reported [26]. In particular, examining the full set of GO terms under “multicellular organismal signaling,” we identified multiple GO terms associated with the regulation of cardiac conduction, including ion channels, action potential and transmembrane signaling (Fig. 51.2d).

We identified 1492 significantly dysregulated genes in the LAA conditional mutant and control, with 770 up-regulated and 722 down-regulated, similar to our prior analysis of this dataset [26]. Although the numbers of significantly dysregulated genes were different between the PPV and LAA, the dysregulation between the LAA and PPV experiments were significantly correlated (p-value < 2.2E-16, Fig. 51.2e), suggesting that the removal of Tbx5 had broadly similar effects in the LAA and PPV regions. In order to identify genes demonstrating distinct Tbx5-dependent effects between the atrial sub-regions, we correlated the Tbx5-dependent changes between the LAA and PPV samples. We identified the 28 genes with residuals in the outermost one-thousandth quantiles (Table 51.1). Amongst those genes, we identified 11 genes that were significantly different in both experiments with 10/11 demonstrating changes in the same direction, 4 genes that were significantly increased in only the LAA samples (Cemip, Col9a2, Itgb7, Lars2), 7 genes that were significantly different in only the PPV samples (increased: Gm42417, Lrtm1; decreased: Alox8, Cntn2, Hcn4, Scarna10, Vsnl1) and 6 genes that were not significantly different in either sample. Interestingly, among the genes that were in the outermost one-thousandth quantiles, we identified Hcn4, which encodes the potassium channel essential for the depolarizing “funny current” of the sinoatrial node. Hcn4 was significantly down-regulated only in the PPV and was also significantly more expressed in the PPV samples than the LAA (Fig. 51.1d).

Table 51.1 Genes displaying distinct Tbx5-dependent expression patterns
Full size table

3 Discussion

3.1 The Genetics of Atrial Fibrillation Implicates Developmental Transcription Factors

Most genetic loci implicated in AF encode ion channels that affect arrhythmia trigger or substrate. Initial transcriptional studies of AF have focused on a restricted set of candidates, associating remodeling with changes in ion channels and components of cellular signaling cascades [31, 32]. Early transcriptomic or proteomic investigation of AF has been limited to animal models and/or chronic AF [33,34,35,36]. For example, animal models based on disruption of specific ion channels do not exhibit spontaneous AF in the absence of concomitant cardiac pathophysiology [37]. However, a recent transcriptomic study using the left atrial appendage of patients with AF revealed that AF susceptibility was associated with decreased expression of transcription factors involved in inflammation, oxidation and the cellular stress response [38]. This study further suggested that remodeling of ion channel expression occurs later in the onset of the disease. Accumulating evidence suggests that transcription factors are important contributors to the pathogenesis of AF [39].

GWAS of AF have identified numerous transcription factor loci that play fundamental roles during cardiac development, including TBX5, NKX2-5, GATA4 and PITX2. The implication of transcription factor genes essential for cardiac development in in AF raises the fundamental question of whether AF is a function or a result of a developmental defect, for example, of the pulmonary veins, or a later manifestation of altered transcriptional control of cardiac rhythm genes in the adult. The first GWAS of AF identified a susceptibility locus on chromosome 4q25 adjacent to PITX2 [40]. There is evidence that Pitx2c affects adult gene expression in the atrium. PITX2 is expressed both in the left atrium and the pulmonary myocardial sleeve and that its levels are decreased in patients with AF [41, 42]. Atrial-specific deletion of Pitx2c leads to hallmarks of AF, which is also observed in Pitx2c heterozygous mice or adult-specific deletion of this gene [41, 43, 44]. Furthermore, microarray analysis of Pitx2c heterozygous mice revealed impaired gap and tight junction gene expression, consistent with the hypothesis that structural genes are remodeled later in the onset of AF [44].

We previously demonstrated that adult-specific deletion of Tbx5 causes spontaneous and sustained AF with atrial gene regulatory network dysfunction [26]. TBX5 drives the atrial expression of Pitx2, and TBX5 and PITX2 co-modulate the expression of cardiac rhythm effector genes, including Ryr2 and Atp2a2. These findings indicated that interactions between TBX5 and PITX2 provide tight control of an atrial rhythm gene regulatory network and that perturbation of this network triggered AF susceptibility. This example suggested that cardiac TFs implicated in AF by genetic association may co-regulate a gene regulatory network for atrial rhythm homeostasis in the adult atrium.

3.2 Transcription Factors and AF: A Pulmonary Vein Developmental Relationship?

The requirement of developmental TFs in the adult for normal cardiac rhythm control does not rule out the possibility that developmental defects caused by their deficiency in the embryo also contributes to AF susceptibility. During embryonic development, the cardiopulmonary progenitors give rise to parts of the atria, pulmonary vein and lungs [25]. Distinct functional characteristics between the PPV and LAA regions may reflect distinct developmental ontogeny. The myocardium of the pulmonary veins has a distinct developmental origin from that of the atrial free walls [45]. Subsequently, the myocardium surrounding the pulmonary veins is distinct from that of other regions in the atrium [46]. These developmental distinctions may set the stage for regional gene expression differences within the mature atrium of the adult that normally maintains atrial rhythm but which make the PPV region susceptible to trigger formation when faced with environmental or genetic insult.

Mutations in NKX2-5, PITX2C and TBX5 increase susceptibility to AF, and all of these transcription factors (TF) play critical roles during cardiovascular development [26, 47,48,49,50]. Each of these TFs are expressed in the pulmonary veins, suggesting that perturbations of the transcription program in the PV myocardium may underlie the pathogenesis of AF. The caudal myocardium, including the developing pulmonary vein myocardium, expresses Tbx5, Pitx2c, Nkx2-5 and Gja5 (Cx40) at high levels [51, 52] consistent with a fast-conducting atrial myocardial phenotype. In a Nkx2-5 haploinsufficient mouse model, Gja5 levels are decreased whereas Hcn4 levels are increased in the pulmonary vein myocardium. Nkx2-5 and Pitx2c are both required for the development and maintain identity of the pulmonary myocardium [45, 53]. The changes in gene expression observed in Nkx2-5 haploinsufficient mice could be reflective of a loss of fast conduction and acquisition of pacemaking conduction phenotype. This type of transition provided a conceivable model for how genetic changes could result in automaticity in the PPV region that resulted in arrhythmia initiation.

Absence of the lungs and pulmonary vasculature in Tbx5 knockout mice suggests that it may play an essential developmental role specifically within the pulmonary veins as well [24]. We observed that a decrement in Tbx5 in the PPV region decreased Hcn4 expression, which is not overtly consistent with this transformation model. On the other hand, we observed that many genes with important roles in myocardial automaticity, including Tbx3, Shox2, Cacna1d, Cacna2d2 and Hcn4 were more highly expressed in the PPV than the LAA region, and remained more highly expressed in the context of adult-specific Tbx5 conditional deletion. Furthermore, other Tbx5-dependent genes in the PPV region are classical fast conduction genes, including Scn5a, Gja5 and Kcnk3. Reduction in the expression of these fast conduction genes in the context of retained high expression of automaticity genes such as Hcn4 could diminish fast conduction physiology and allow the emergence of myocardial automaticity, providing a nidus for the initiation of atrial arrhythmias including AF.

4 Methods

4.1 Animal Experiments and Ethics Statement

All murine experiments were performed under University of Chicago Institutional Animal Care and Use Committee (IACUC) protocol number 71737. Mice harboring the ROSA26 tamoxifen-inducible cre recombinase transgene, Gt(ROSA)26Sor tm1(cre/ERT2)Tyj (R26 creERT2), were bred with mice harboring the Tbx5 tm1Jse allele (Tbx5 flox) to generate Tbx5 conditional mutants, R26 creERT2/creERT2; Tbx5flox/flox, and controls, R26 creERT2/creERT2; Tbx5+/+ [23, 54]. Tamoxifen was administered by intraperitoneal injection for three consecutive days at 6 weeks of age and tissue samples were harvested 1 week after treatment as previously reported [26].

4.2 Transcriptional Profiling and Analysis

Transcriptional profiling was performed on the peri-pulmonary vein region of the left atrium (Fig. 51.1a). Total RNA was extracted from four controls and six conditional mutants. Following rRNA removal (Ribo-Zero rRNA Removal Kit, Illumina), libraries were prepared using the TruSeq RNA Sample prep kit v2 (Illumina). Samples were pooled in equimolar ratios and sequenced on the Illumina HiSeq4000 platform. Library preparation and sequencing were performed by the Genomics Core Facility at the University of Chicago.

The peri-pulmonary vein samples were compared with the previously published corresponding left atrial appendage samples [26]. Both the peri-pulmonary vein and left atrial appendage samples were analyzed in parallel as described below.

Between 20 and 22 million reads were generated for each replicate (14–19 million reads for left atrial appendage) and aligned to the GRCm38/mm10 build of the Mus musculus genome (retrieved from UCSC May 23, 2012) using TopHat v2.1.1 [55, 56]. Reads were filtered to remove poorly aligned (MAPQ<10), duplicated and unmapped reads using BamTools v2.4.1 [57]. Post-alignment, post-filtered reads were assigned to transcripts using StringTie v1.3.3 [58, 59].

Downstream analysis was performed using R v3.4.0 [60]. Differential expression testing was performed using edgeR v3.18.1 [61] and limma v3.32.10 [62]. Low level genes were removed within each condition using median log-transformed counts per gene per million mapped reads (cpm) cutoff of 1. A generalized linear model (GLM) framework was used to test for differential expressed genes between Tbx5 fl/fl;R26 cre-ERT2 conditional mutants and Tbx5 +/+;R26 cre-ERT2 controls for each tissue and between controls of the left atrial appendage and peri-pulmonary vein tissues. Data was plotted using the ggplot2 v2.3.0 [63] and scales v0.5.0 [64] packages. GO term analysis was performed using metascape (http://metascape.org) [27]. The raw and processed data as well as the full outputs were deposited in GEO at accession number GSE128870.