Common SYNE2 Genetic Variant Associated with Atrial Fibrillation Lowers Expression of Nesprin-2α1 with Downstream Effects on Nuclear and Electrophysiological Traits

Abstract

Background.

Atrial fibrillation (AF) genome-wide association studies (GWAS) identified significant associations for rs1152591 and linked variants in the SYNE2 gene encoding Nesprin-2, which connects the nuclear membrane with the cytoskeleton.

Methods.

Reporter gene vector transfection and CRISPR-Cas9 editing were used to identify the causal variant regulating expression of SYNE2α1. After SYNE2 knockdown (KD) or SYNE2α1 overexpression (OE) in human stem cell-derived cardiomyocytes (iCMs), nuclear phenotypes were assessed by imaging and atomic force microscopy. Gene expression was assessed by RNAseq and gene set enrichment analysis. Fura-2 AM staining assessed calcium transients. Optical mapping assessed action potential duration (APD) and conduction velocity (CV).

Results.

The risk allele of rs1152591 had lower promoter and enhancer activity and was significantly associated with lower expression of the short SYNE2α1 isoform in iCMs, without an effect on the expression of the full-length SYNE2 mRNA. SYNE2α1 OE had dominant negative effects on the nucleus with its OE or SYNE2 KD leading to increased nuclear area and decreased nuclear stiffness. Gene expression results from SYNE2α1 OE demonstrated both concordant and non-concordant effects with SYNE2 KD. SYNE2α1 OE had a gain of function on electrophysiology, leading to significantly faster calcium reuptake and decreased APD, while SYNE2 KD showed both shortened APD and decreased CV.

Conclusions.

rs1152591 was identified as a causal AF variant, with the risk allele decreasing SYNE2α1 expression. Downstream effects of SYNE2α1 OE include changes of nuclear stiffness and electrophysiology, which may contribute to the mechanism for the risk allele association with AF.

Introduction

Atrial fibrillation (AF), the most common clinically significant arrhythmia, is characterized by rapid and irregular activation in the atria without discrete P waves on surface electrocardiograms (ECG). It is estimated that there are 5.2 million AF patients in the United States, and the number is expected to increase to 12.1 million over the next 2 decades¹. This disorder is a major cause of morbidity, mortality, and health care expenditure. Two AF genome-wide association studies (GWAS) in 2012 and 2017 identified the single nucleotide polymorphisms (SNP) rs1152591, located in an intron of SYNE2 on chromosome 14q23, to be significantly associated with AF ^{2, 3}. Our prior RNA sequencing of human left atrial appendages (LAA) from a biracial cohort of 265 subjects found that rs1152591 is associated with SYNE2 gene expression, thus acting as a cis-expression quantitative trait locus (cis-eQTL) for SYNE2, with the risk allele associated with lower SYNE2 expression ⁴.

The SYNE1 and SYNE2 genes encode Nesprin-1 and −2, respectively, which are two related large outer nuclear membrane-spanning proteins. Alternative transcription start sites (TSS), RNA splicing, and polyadenylation sites give rise to multiple isoforms of Nesprin-1/2. The longest isoform of Nesprin-2, referred to as the giant isoform, contains three different functional domains: the C-terminal transmembrane KASH (Klarsicht-ANC-Syne-homology) domain, which anchors within the nuclear membrane, multiple spectrin-repeat rod domains, and the N terminal paired calponin-homology domain connecting with cytoskeletal actin. Nesprin-2 giant isoform KASH domain binds to SUN proteins (Sad1p/UNC-84) between the outer and inner nuclear membranes. The SUN proteins span the inner membrane to bind lamin A/C underneath the inner nuclear membrane, and together, these proteins comprise the Linker of Nucleoskeleton and Cytoskeleton (LINC) complex, which mechanically links the nucleoskeleton to the cytoskeleton ^5–7. The LINC complex has diverse functions, including maintenance of nuclear position and morphology and the mediation of various signal transduction pathways between the cell surface and the nucleus^{8, 9}. Nesprin-2 short isoforms using alternative TSS lack the N-terminal calponin-homology domain, thus they can only bind to the nuclear membrane but not to cytoskeletal actin.

Here we investigate the regulatory activity of two linked AF associated SNPs on the expression of SYNE2 mRNA and protein isoforms in human LAA. We found that the risk allele of rs1152591 decreased expression of the short Nesprin-2α1 isoform, which may work as a dominant-negative by competing with the Nesprin-2 giant isoform. We determined that the influence of either KD of all SYNE2 isoforms or OE of SYNE2α1 on nuclear size and stiffness, gene expression, calcium transients, and electrophysiology. Together, our findings may explain the mechanisms for decreased Nesprin-2α1 expression to be associated with increased AF susceptibility.

Methods

Data Availability

The human LAA RNAseq data is available at Gene Expression Omnibus (www.ncbi.nlm.nih.gov/a) accession # GSE69890. The left atrial eQTL summary data is available at https://afeqtls.lerner.ccf.org/. All other data that support the findings of this study are available from the corresponding author upon reasonable request.

Human Left Atrial Tissue

Human LAA tissues were obtained from patients undergoing elective surgery to treat AF, valve disease, or other cardiac disorders, and from nonfailing donor hearts not used for transplant. The living donors all provided informed consent, and consent was obtained from family members for the heart transplant donors, under protocols approved by the Cleveland Clinic IRB.

Full methods are available in Supplemental Material.

Results

AF associated SNPs are eQTLs for SYNE2α1, a short mRNA isoform

LAA tissues were obtained from cardiac surgery or from nonfailing donor hearts of 235 European ancestry subjects and 30 self-reported African Americans; the clinical characteristics were reported in our previous study⁴. In that study, we performed a gene-level transcriptome analysis and we found that the AF GWAS SNP rs1152591 was a strong cis-eQTL for SYNE2 gene expression (q=6.7E-17) ⁴. Here we performed a transcript isoform-specific cis-eQTL analysis for SYNE2 expression. The SYNE2 giant isoform (ENST00000357395.7, called long isoform here) has 116 exons, and the position of the GWAS SNPs and nearby exons are shown in Figure 1A. We identified the most abundant LAA SYNE2 mRNA transcript as a short isoform (ENST00000458046.6, referred to hereafter as SYNE2α1 isoform) that initiates transcription at exon 108 of the full-length transcript, leading to an N-terminal deleted 62 kDa protein called Nesprin-2α1 (Figure 1A). The AF GWAS SNP rs1152591 is located only 10 bp upstream of the SYNE2α1 isoform TSS in the intron between exons 107 and 108 (Figure 1A). Our LAA RNAseq demonstrated abundant expression of SYNE2α1 with its expression levels inversely correlated with the number of rs1152591 risk alleles (Figure 1B). qPCR of LAA RNA samples was also performed to verify the expression levels of SYNE2α1 (primers in exon 108 and 109, with the upper primer specific for the unique transcribed sequence in exon 108) and another short isoform SYNE2β2 (primers spanning exons 101 and 102). LAA samples with two vs. zero risk alleles of rs1152591 had significantly decreased levels of the SYNE2α1 isoform (Figure 1C), but not SYNE2β2 (Figure 1D) isoform. Although the full-length SYNE2 mRNA isoform may be underrepresented in RNAseq based on oligo dT priming, there was no association of rs1152591 genotype with the SYNE2 long isoform as assessed by qPCR (Figure1E). The eQTL locus zoom plot shows that the GWAS SNP, rs1152591, was a significant eQTL for SYNE2α1 expression (Figure 1F). However, the strongest cis-eQTL was identified as rs1152595, located 6,128 bp upstream of the SYNE2α1 TSS between exons 102 and 103, which is in linkage disequilibrium (LD) with the GWAS SNP in the CEU population (r²= 0.78. D’=1, Figure 1F). Consistent with the RNAseq and qPCR results, LAA samples with one or two risk alleles of rs1152591 had lower expression levels of Nesprin-2α1 in Western blot (Figure 1G, H). We evaluated LAA RNAseq Sashimi plots for selected subjects homozygous for two vs. zero risk alleles of rs1152591, by counting the ratio of the slice junction reads over the splice acceptor reads. This revealed significantly lower ratio for the risk allele subjects, demonstrating lower SYNE2α1 TSS usage in exon 108 (the first exon of SYNE2α1) while there no difference in this ratio for exon 109 (second exon of SYNE2a1, Supplemental Figure 1 A–C).

Figure 1.

AF risk alleles of rs1152591 and rs1152595 are associated with decreased expression of SYNE2α1 in LAA. A. Schematic diagram of a portion of the SYNE2 long and SYNE2α1 transcript isoforms (not to scale). B. Inverse gene dose relationship between SYNE2α1 expression, assessed by LAA RNAseq, and the number of risk allele of rs1152591 carried by each subject (N=235, box and whiskers show median, interquartile range, 10^th to 90^th percentiles). C. LAA RNA qPCR confirmation of rs1152591 risk allele associated with SYNE2α1 expression (N=10 per genotype). D, E. LAA RNA qPCR shows no effect of rs1152591 risk allele on expression of SYNE2β2, an alternate short isoform (D) or SYNE2 long isoform(E) (N=10 per genotype). F. Locus zoom plot of the eQTL –log10 p-values for the association of each cis-SNP with the expression of SYNE2α1 mRNA isoform showing rs1152595 as the top eQTL SNP, and the GWAS SNP rs1152591 as the second top SNP. The color bar shows linkage disequilibrium with rs1152595. The position of SYNE2α1 exon is shown below. G. Nesprin-2 western blot of LAA lysates showing three short isoforms and GAPDH as a loading control. H. Quantification of Nesprin-2α1 (62 kDa) protein level normalized by Nesprin-2β2 (76 kDa) isoform (n=18 subjects), showing subjects with 1 or 2 AF risk alleles of rs1152591 have significantly decreased expression of Nesprin-2α1 isoform. (TPM, transcripts per million).

Risk allele of rs1152591 decreases SYNE2a1 expression in iCMs

The GWAS SNP, rs1152591, and to a lesser extent the top eQTL SNP, rs1152595, were found centered over DNAase I hypersensitivity sites in fetal human heart (Figure 2A, B). As rs1152591 is 10 bp upstream the SYNE2α1 mRNA TSS, we investigated the effect of this SNP on the strength of the SYNE2α1 promoter. Reporter gene transfections were performed using 1196 bp promoters of the rs1152591 risk and protective alleles driving firefly luciferase. Compared to the promoterless plasmid, the risk allele increased reporter gene expression by 7.5-fold, while the protective allele increased expression by 25.8-fold (Figure 2C, all values different from each other, p<0.001). To exam just the proximal region of rs1152591 for enhancer activity, we inserted a 60 bp double stranded oligos of the two alleles of into a firefly luciferase plasmid driven by a minimal viral promoter. Compared to the enhancerless plasmid, the risk allele of rs1152591 had 2.0-fold enhancer activity, while the protective allele had 4.0-fold enhancer activity (Figure 2D, all values different from each other, p<0.001). The top eQTL SNP, rs1152595 was also tested similarly for enhancer activity, and compared to the enhancerless plasmid, the risk allele had no significant enhancer activity, while the protective allele had 2.7-fold enhancer activity (Figure 2E, protective allele different from other two, p<0.001). Thus, both linked SNPs were regulatory variants that may alter SYNE2α1 expression.

Figure 2.

Functional tests of rs1152591 and rs1152595. A, B. Browser view of fetal heart DNAase I hypersensitivity of rs1152591 and rs1152595 from ENCODE. C. Firefly luciferase reporter gene transfection study in iCMs with a promoterless plasmid containing no insert or 1.2 kb promoter region of the SYNE2α1 isoform with the protective (G) or risk (A) allele of rs1152591, normalized for transfection efficiency by co-transfection with renilla luciferase. D, E. Firefly luciferase reporter gene transfection study in iCMs with a minimal promoter plasmid containing no insert or 60-mer oligos containing protective or risk alleles of rs1152591 or rs1152595, respectively (C-F, p-values determined by ANOVA with Tukey’s posttest). F. Endogenous SYNE2α1 mRNA expression normalized to TNNT2 expression assesed by qPCR in iCMs from an unedited subject homozgyous for the protective alles of rs1152591 and rs1152595 (rs91/95 GG/TT), and from iCMs dervied from the same iPS line after CRISPR Cas9 editing to homozygosity for their respective AF-risk alleles (p-values determined by ANOVA with Dunnett’s multiple comparison test vs. the control unedited iCM).

To further investigate the regulatory impact of the rs1152591 and 1152595, we used CRISPR-Cas9-mediated homology directed repair in a human iPS cell line to convert the protective alleles to their respective risk alleles. iPS colonies were screened by Sanger sequencing, and those that were homozygous for the risk alleles were differentiated into atrial like iCMs. Compared with the unedited iCMs homozygous for the protective allele of both SNPs, SYNE2α1 mRNA expression decreased ∼60% in the iCMs harboring the rs1152591 risk allele (Figure 2F, p=0.047), while there was no significant difference in SYNE2α1 expression in iCMs homozygous for the risk allele of rs1152595. Thus, the rs1152591 risk allele yielded decreased expression of the endogenous SYNE2α1 gene in otherwise isogenic iCMs, in agreement with the results of the reporter gene transfection studies (Figure 2C, D), as well as the RNAseq, qPCR and WB results (Figure 1B, C, G). Although rs1152595 is in high LD with rs1152591, it is about 6 kb upstream of the SYNE2α1 TSS, and it may influence the expression of other SYNE2 isoforms.

Nuclear phenotypes of SYNE2 knockdown and SYNE2α1 OE

To investigate the effects of altered Nesprin-2α1 expression on cardiomyocyte nuclei, we designed an αMHC-driven GFP-SYNE2α1 fusion protein expression vector. 72 h after transfection into iCMs, we measured the nuclear area in GFP positive (overexpressing the Nesprin-2α1 fusion protein) and GFP negative (non-transfected) iCMs (Figure 3A). Separately, scramble siRNAs and SYNE2 siRNAs (targeting all major isoforms) were transfected into iCMs (Figure 3B). Compared with GFP negative control cardiomyocytes, those overexpress GFP-SYNE2α1 had 12.5% larger nuclei (p<0.001, Figure 3C). Similarly, SYNE2 KD cells had 12.6% larger nuclear size (p< 0.001, Figure 3D).

Figure 3.

SYNE2α1 OE or KD of all SYNE2 isoforms in iCMs altered nuclear area and stiffness. A. Transient transfection of GFP-SYNE2α1 plasmid into iCMs yielded GFP expression in the nuclear envelope and diffusely in the cytosol (left side GFP fluourescence, right side GFP and dapi stained nuclei; scale bar = 20 µm). B. Dapi stained nuclei in scramble (Scr, left) siRNA and SYNE2 KD (right) iCMs (scale bar = 10 µm). C. GFP positive cells overexpressing Nesprin-2α1 isoform had a 12.5% increase in median nuclear area vs. adjacent GFP negative cells (N=473 control and 174 OE). D. All SYNE2 isoforms KD in iCMs led to a 12.6% increase in median nuclear area vs. scr siRNA (N=411 scr and 368 KD). E, F. Young modulus nuclear siffness values for control and GFP-SYNE2α1 cells, and scr siRNA vs. SYNE2 siRNA KD cells 72 h after transfection (N=60 control, 60 GFP-SYNE2α1, 40 Scr siRNA, and 60 SYNE2 siRNA). For B-D, median values denoted, p-values determined by two-tailed Mann Whitney t-test. G. Immunostaining of Lamin A/C in scr siRNA (left) and SYNE2 KD (right) iCMs. H, I. Yap luciferase activity in control and GFP-SYNE2α1 OE cells, scr siRNA and SYNE2 siRNA KD cells 48 h after Yap/renilla transfection (all data normally distrubuted, mean ± SD, p valued determined by one-tailed t-test, N=6 control, 6 GFP-SYNE2α1, 6 scr siRNA, and 5 SYNE2 siRNA).

To determine the effects on nuclear mechanics, atomic force microscopy was performed to assess the nuclear stiffness, using Young’s modulus as a measure of nuclear stiffness. We compared GFP-SYNE2α1 and GFP negative cells within the same dish and found that OE of the GFP-Nesprin-2α1 fusion protein led to a 33% decrease in nuclear stiffness (from 3.17 ± 0.42 kPa to 2.11 ± 0.13 kPa, p<0.0001, Figure 3E). Similarly, KD of all SYNE2 isoforms led to a 57% decrease in nuclear stiffness (from 3.09 ± 0.43 kPa to 1.31 ± 0.19 kPa, p< 0.0001, Figure 3F). Meanwhile, Lamin A/C immunofluorescent staining was performed to determine if SYNE2 KD altered the Lamin A/C structure or localization in the inner nuclear membrane; no differences were observed (Figure 3G). Thus, SYNE2 effects on nuclear stiffness were not due to changes in the lamin A/C nuclear lamina, and increased SYNE2α1 expression behaves like a dominant-negative and has the same effect on the nuclear area and stiffness as KD of all SYNE2 isoforms.

LINC complex disruption is related to changes on nuclear membrane shape, nuclear stiffness as well as mechanotransduction¹⁰, such that decreased nuclear stiffness leads to decreased mechanotransduction¹¹. To study the influences of SYNE2α1 OE or SYNE2 KD on Yap-TAZ signaling mechanotransduction, we performed a YAP activity reporter transfection assay using a synthetic TEAD responsive promoter driving firefly luciferase plasmid along with a Renilla luciferase plasmid to control for transfection efficiency. Using this dual luciferase assay, SYNE2α1 OE or SYNE2 KD in iCMs decreased Yap activity by 23% or 20%, respectively (p<0.05, Figure 3H, I), indicating the dominant-negative effect of SYNE2α1 on mechanotransduction, in agreement with the other nuclear phenotypes assessed.

Gene expression changes due to SYNE2α1 OE or SYNE2 KD

To explore the underlying genomic and transcriptomic mechanism of all SYNE2 isoforms KD in iCMs, we performed RNA sequencing to distinguish the differentially expressed genes between SYNE2 KD vs. scramble control iCMs. A total of 1975 genes were identified using the cutoff of 0.5 log2-fold change and q- value <0.05 with 1236 upregulated and 739 downregulated. We also performed RNA sequencing on the GFP-SYNE2α1 OE (OE) vs. GFP control iCMs. A total of 5719 genes were identified using the cutoff of 0.5 log2-fold change and q- value <0.05 with 2454 upregulated and 3265 downregulated (Supplemental Table 1). Furthermore, we also found decreased expression of selected YAP/TEAD targeted genes such as MFN1/2, TBX5, DNM1L, FOSB in both SYNE2α1 OE and SYNE2 KD iCMs (Supplemental Table 2).

To further explore the effects of SYNE2 KD and SYNE2α1 OE, we merged their iCM differential gene expression profiles revealing genes whose expression were concordant (positively correlated, in red) and discordant (negatively correlated, in blue) between these two groups (Figure 4A). The concordantly expressed genes could be downstream of the dominant-negative effects of SYNE2α1 OE, as we observed for the nuclear phenotypes, which may also explain the Gene Set Enrichment Analysis (GSEA) similarities (see below) after either SYNE2 KD or SYNE2α1 OE. The discordantly expressed genes could be downstream of the total amount of all SYNE2 isoforms (regulated in the opposite directions in SYNE2 KD and in SYNE2α1 OE), or due to a gain of function of the short SYNE2a1 isoform.

Figure 4.

Transcriptomic differences due to SYNE2a1 OE and SYNE2 KD. A. Concordant (red) and discordant (green) differential gene expression determined by RNAseq, for genes with |logFC| > 0.5 and Q−value < 0.01; larger dot size indicates more highly significant difference for OE vs. KD. B,C. Molecular signature reactome gene set Venn diagram overlaps among the gene sets positively (B, pos) or negatively (C, neg) associated with SYNE2α1 OE, SYNE2 KD, and LAA RNAseq coexpression with SYNE2, respectively. D. SYNE2α1 colocalization with SERCA2 in nuclear membrane and SR; iCM SERCA2 immunoflourescent staining (left), GFP-SYNE2α1 flourescence (center), and merged including nuclear DAPI stain (right, 60x objective via confocal microscopy).

To validate our SYNE2 KD and SYNE2α1 OE in iCMs, we compared this data to the human LAA RNAseq results, where every expressed gene was analyzed for coexpression with SYNE2 in 235 European descent LAA samples (Supplemental Table 3)¹². The LAA SYNE2 coexpression correlation coefficients (r) and theSYNE2α1 KD and OE mediated log2 fold changes in gene expression were analyzed as three separate ranked files with GSEA using the molecular signature reactome gene set (c2. reactome. v2023). For each of these analyses we divided the outcomes by whether the pathways were positively or inversely associated with SYNE2 coexpression, SYNE2α1 KD, and SYNE2α1 OE. We found high overlap among the LAA SYNE2 coexpression positively correlated molecular signatures lists with the molecular signatures for genes upregulated after either SYNE2 KD or SYNE2α1 OE, including mitochondrial protein import, potassium channels, nuclear events kinase, and transcription factor activation (Figure 4B, Supplemental Table 3). There were also overlaps between the LAA SYNE2 coexpression negatively correlated molecular signatures lists with the with the molecular signatures for genes down regulated after either SYNE2 KD or SYNE2α1 OE, including mitotic spindle checkpoint, cytosolic tRNA aminoacylation, and metabolism of folate/pterines (Figure 4C, Supplemental Table 3). These results may seem incongruous, as there were similar gene set effects after either SYNE2 KD or SYNE2α1 OE. However, as we saw in the nuclear phenotypes, SYNE2α1 OE behaves as a dominant negative similar to KD of all SYNE2 isoforms, and the substantial overlap of GSEA molecular signatures emphasizes the concordant changes in gene expression after both treatments. When we looked in more detail at the localization of the GFP-Nesprin-2α1 fusion protein, we found it was expressed not only on the nuclear membrane, but also in the sarcoplasmic reticulum (SR) as determined by co-localization with the SR marker protein SERCA2 (Figure 4D).

Calcium cycling changes caused by SYNE2α1 OE or SYNE2 KD in iCMs

RNAseq after SYNE2 KD or SYNE2α1 OE showed varying effects on calcium handling genes (Table1); thus, we stained the cells with Fura 2-AM and investigated calcium cycling in iCMs paced at 0.5 Hz. GFP-SYNE2α1 OE vs. GFP OE control iCMs showed no difference for the calcium peak amplitude, but significantly faster calcium release (median values 0.1314 s vs, 0.2154 s, p<0.001) and calcium reuptake (median values, 1.1215 s vs. 1.237 s, p=0.046) (Figure 5A–D). In addition, SYNE2 KD led to significantly faster calcium release (mean values 0.2158 s vs. 0.2480 s, p<0.0001), but did not change the peak calcium amplitude and calcium reuptake (Figures 5E–H).

Table 1.

Calcium related genes in SYNE2α1 OE and SYNE2 KD iCMs, ranked by Q-value of OE vs. KD

Gene Symbol	Log2 FC (GFP-SYNE2α1 OE vs SYNE2 siRNA)	Q Value (GFP-SYNE2α1 OE vs SYNE2 siRNA)	Log2 FC (GFP-SYNE2α1 OE vs GFP Cntl)	Q Value (GFP-SYNE2α1 OE vs GFP Cntl)	Log2 FC (SYNE2 siRNA vs Scr siRNA)	Q Value (SYNE2 siRNA vs Scr siRNA)
ATP2A2	−1.37	5.37E-07	−0.51	1.93E-03	0.86	1.53E-04
CASQ2	−1.54	5.92E-06	−0.64	3.77E-03	0.90	6.03E-04
PLN	−1.36	2.11E-05	−1.34	2.91E-04	0.01	4.21E-01
FKBP9	1.66	2.52E-05	−0.02	3.24E-01	−1.69	4.70E-05
ATP2B2	−2.32	8.46E-05	0.82	1.58E-02	3.14	6.77E-05
CAMK2N1	2.32	1.60E-04	2.64	2.92E-04	0.31	2.27E-01
FKBP2	−1.14	2.26E-04	−0.73	1.55E-03	0.41	5.23E-02
ITPR2	2.00	2.30E-04	0.82	8.20E-03	−1.17	1.07E-02
FKBP3	−0.88	2.68E-04	−0.59	2.46E-03	0.29	5.22E-02
ATP2A1	1.91	4.66E-04	2.26	2.92E-04	0.36	2.00E-01
CALM1	0.81	6.04E-04	0.41	1.80E-02	−0.40	2.41E-03
CAMK2D	−0.68	1.88E-03	−0.75	7.76E-04	−0.06	3.46E-01
HAX1	−0.61	3.02E-03	−0.99	3.80E-04	−0.38	1.49E-02
CACNA2D2	−0.81	3.22E-03	−0.44	1.48E-02	0.37	6.97E-02
HCN2	−0.85	3.23E-03	0.31	5.71E-02	1.16	8.21E-04
ATP2B1	0.63	1.11E-02	0.85	1.42E-03	0.22	1.41E-01
CAMK2B	−0.67	1.45E-02	−0.29	5.37E-02	0.38	8.95E-02
FKBP1B	0.81	1.55E-02	1.20	1.31E-03	0.39	9.54E-02
FKBP11	−0.67	2.31E-02	−0.50	3.29E-02	0.17	2.29E-01
SLC8B1	−0.79	2.68E-02	0.21	1.65E-01	0.99	1.07E-02
CALM2	−0.34	2.91E-02	−0.41	9.87E-03	−0.06	2.74E-01
ATP2A3	−0.77	3.25E-02	−0.28	1.51E-01	0.49	6.26E-02
STK11IP	−0.50	4.39E-02	−0.37	4.32E-02	0.13	2.86E-01
FKBP10	−0.35	4.59E-02	−0.73	1.80E-03	−0.38	6.33E-03
CAMK2A	0.45	5.80E-02	0.74	1.78E-03	0.30	1.37E-01
HACE1	−0.42	6.16E-02	−0.31	3.40E-02	0.11	3.30E-01
CACNA1D	0.85	6.59E-02	0.99	7.41E-03	0.15	3.75E-01
TRPC3	0.57	9.85E-02	1.60	7.58E-04	1.03	1.51E-02
RYR3	1.70	9.98E-02	1.95	1.04E-02	0.24	4.03E-01
SLC8A1	−0.22	1.11E-01	−0.61	1.28E-03	−0.39	1.62E-02
SLC8A2	0.87	1.40E-01	1.88	7.20E-03	1.01	7.64E-02
RYR2	−0.14	1.79E-01	−0.44	7.17E-03	−0.30	6.77E-03
CACNA1C	−0.13	1.99E-01	−0.46	5.52E-03	−0.34	3.98E-03
SLC9A9	−0.53	2.28E-01	−0.91	1.84E-02	−0.38	3.19E-01
CACNB2	−0.10	2.75E-01	−0.43	1.95E-02	−0.33	1.51E-02
CAMK2N2	0.30	3.13E-01	1.41	1.14E-02	1.11	1.32E-01
CACNA1H	−0.08	3.22E-01	0.44	2.93E-02	0.52	2.84E-02
CACNA2D1	0.05	3.33E-01	0.28	2.75E-02	0.23	1.57E-01
FKBP14	0.06	3.50E-01	−0.72	2.33E-03	−0.78	3.29E-02
FKBP5	0.05	3.52E-01	0.59	1.44E-02	0.55	1.26E-02

Cntl, control; FC, fold change; OE, overexpression; Scr, scramble. Bold values denote statistical significance at Q < 0.05.

Figure 5.

SYNE2α1 OE or SYNE2 KD changes calcium cycling in iCMs. A. Single representative Calcium transient by Fura-2AM staining in non-transduced (neg cntl), AAV6-GFP transduced controls, and GFP-SYNE2α1 OE iCMs paced at 0.5 Hz. B-D. Peak Calcium amplitude, time to peak 90%, and time to baseline 90% quantified in multiple cells showing SYNE2α1 OE leading to significantly faster calcium release and calcium reuptake vs. the GFP control cells (N=138 neg cntl, 140 GFP cntl and 141 GFP-SYNE2α1, violin plots with 25^th 50^th, and 75^th percentiles, p-values by Kruskal Wallace ANOVA with Dunn’s posttest). E. Single representative Ca²⁺ transient in scr and SYNE2 siRNA KD iCMs. F-H. Peak Ca²⁺ amplitude, time to peak 90%, and time to baseline 90% quantified in multiple cells showing significantly faster calcium release without any effects on calcium concentration or reuptake in SYNE2 KD iCMs (N=146 Scr siRNA and 147 SYNE2 siRNA, violin plots with 25^th 50^th, and 75^th percentiles, p values determined by Mann-Whitney t-test). I, J. Lower expression levels in several calcium related genes such as SLC8A1, PLN, ATP2A2, CASQ2, ATP2B2 in GFP-SYNE2α1 OE iCMs, (I), and higher expression levels of ATP2A2, CASQ2 and ATP2B2 in SYNE2 KD iCMs (J), assessed by qPCR normalized to PPIA expression and normalized to the control values (N=3 independent wells of iCMs per group, p-values determined by unpaired t-test ).

Next, we went back to our RNAseq data to see the calcium transient effects were consistent with the expression of calcium handling genes. SYNE2 KD or SYNE2α1 OE had concordant faster calcium release, but the effect was much larger after SYNE2α1 OE. This result is consistent with higher calcium influx form the voltage dependent calcium channel, Cav1.3, encoded by CACNA1D, as there was a 1.97-fold increase after SYNE2a1 OE (q=7.41E-03) vs. a non-significant 1.1-fold increase after in SYNE2 KD (q=0.37). There was 25% decrease in CALM2 levels after SYNE2α1 OE (q=9.87E-03) vs. 4.5% decrease after SYNE2 KD (q=2.74E-01). Decreased inhibition of the RyR2 channel by reduced expression of CALM2 encoding a calmodulin subunit is also consistent with faster calcium release. Surprisingly, RYR2 expression levels were decreased after both treatments, although this may be downstream of RyR2 protein stabilization. Meanwhile, the more rapid calcium reuptake only observed after SYNE2α1 OE is consistent with 71% decreased PLN expression, encoding phospholamban the negative regulator of SERCA2, after SYNE2α1 OE (q=2.91E-04) with no effect on PLN expression after SYNE2 KD. We validated many of the gene expression differences in calcium handling genes after SYNE2α1 OE or SYNE2 KD by qPCR (Figure 5 I, J)

Electrophysiological changes caused by SYNE2α1 OE or SYNE2 KD in iCMs

To investigate the electrophysiological changes, we used optical mapping in iCM monolayers. For GFP-SYNE2α1 or GFP only OE, we switched to AAV6 in order to achieve nearly 100% transduction efficiency. Comparing GFP-SYNE2α1 OE with GFP control transduced iCMs, there was significantly decreased conduction velocity (0.1844 ± 0.04575 m/s vs. 0.2178 ± 0.0324 m/s, p=0.0379, Figure 6A). This was consistent with SYNE2α1 OE leading to 52% (q=4.70E-04) and 45% (q=1.32E-02) decreased expression of GJA1 and GJA5, encoding connexin 43 and connexin 40, respectively (Supplemental Table 1). GFP-SYNE2α1 OE vs. GFP control OE had no significant effect on action potential duration (921.3±171.2 vs 847.7±154.1 ms, Figure 6A).

Figure 6.

SYNE2α1 OE or SYNE2 KD changes action potential and potassium related gene expression in iCMs. A. GFP-SYNE2α1 OE decreased conduction velocity (left) but not APD (action potential duration, right) in iCMs, assessed by optical mapping in monolayers, N= 21 GFP-cntl and 22 GFP-SYNE2α1. B. SYNE2 KD significantly decreased conduction velocity and APD in iCMs. For A and B, each symbol is an independent well, data pooled from 3 experiments, normalized to the mean of the negative control, N= 7 Scr siRNA and 7 SYNE2 siRNA. C. Single representative action potential shows decreased APD in SYNE2 KD iCMs vs. Scr. D. qPCR shows increased expression of potassium related genes such as KCNQ1, KCNC4, KCNJ12, KCNJ4, and decreased expression of GJA1 (connexin 43) in SYNE2 KD iCMs, N=3 wells per treatment (A, B, D, mean ± SD, p value determined by Mann-Whitney test t test).

SYNE2 KD vs. scramble control also decreased conduction velocity (0.1024±0.03538, p=0.0123 m/s vs. 0.1507±0.0256 m/s, Figure 6B), which is consistent with 63% and 42% decreased expression of GJA1 (q=2.31E-04) and GJA5 (q=4.43E-03), respectively (Table 2). In contrast to GFP-SYNE2a1, SYNE2 KD vs. scramble control had shortened action potential duration (1975±224.3 ms vs. 2282±93.56, p=0.0059) (Figure 6B, C); and, this effect was consistent with increased expression of multiple potassium related genes such as KCNQ1 and KCNC4, which were increased by 3.1-fold (q=1.04E-04) and 1.85-fold (q= 6.35E-04), respectively (Table 2 and confirmed by qPCR, Figure 6D).

Table 2.

Potassium and connexin related genes in SYNE2 KD iCMs.

Gene Symbol	Log2 FC (SYNE2 siRNA vs Scr siRNA)	Q Value
KCNJ12	1.29	9.60E-05
KCNQ1	1.64	1.04E-04
GJA1	−0.67	2.31E-04
KCNC4	0.89	6.35E-04
HCN2	1.16	8.21E-04
KCNMB4	0.79	1.57E-03
KCTD18	−0.68	2.49E-03
KCNJ4	0.81	2.73E-03
GJA5	−1.23	4.43E-03
KCNK13	1.36	1.26E-02
KCTD3	0.34	1.57E-02
KCNH7	0.86	2.30E-02
KCTD9	0.36	2.41E-02
KCNQ2	−0.94	2.69E-02
KCNJ3	1.03	3.27E-02
KCTD5	−0.24	3.47E-02
KCNIP3	0.72	4.66E-02

FC, fold change. Bold values denote statistical significance at Q < 0.05.

Discussion

Human GWAS studies are very powerful in their ability to identify common genetic variants associated with a disease or quantitative trait. The top GWAS variant (the one with the most significant p-value) at any one chromosome locus is used to identify the closest gene, which is assumed to be responsible for the genetic association; although, this is not always the case^{13, 14}. In addition, many GWAS loci have multiple variants associated with the trait with similar levels of significance due to linkage disequilibrium. Thus, GWAS is just the first step in the arduous journey to identify the real causal gene, the causal variant, and the downstream mechanisms for the genetic association.

Many GWAS variants are intergenic or intronic and thus do not change the protein sequence. Thus, we hypothesize that many GWAS loci are due to regulatory variants that alter gene expression. In order to identify the real causal gene for AF GWAS loci, we have turned to RNAseq of human LAA to find the genes whose expression are associated with the top GWAS variants, so called cis-eQTLs⁴. For example, we previously showed that the AF GWAS SNP rs17171731 on chromosome 5q31, nominally attributed to the WNT8A gene, was in fact a regulatory variant controlling the expression of a previously unstudied gene, FAM13B¹³. In our current study, we focused on two SYNE2 intronic SNPs, rs1152591 and rs1152595. Our LAA RNAseq eQTL study identified rs1152591 and rs1152595 among the strongest eQTLs for SYNE2 expression, and both are in LD with each other and the top GWAS SNPs at this locus. We relied on our own LAA eQTL study instead of the GTEX project, which utilized left ventricle and right atrial samples; and, we have shown that the left and right atria have some differences in their gene expression profiles and eQTLs ^{4, 15}. In addition, both of these SNPs were in regions of DNAse I hypersensitivity (Figure 2). Here, we found that rs1152591 was associated with a gene-dosage dependent effect on expression of a short isoform called SYNE2α1, due to use of an internal TSS, leading to the expression of an N-terminal truncated protein called Nesprin-2α1. In fact, rs11252591 is located only 10 bp upstream of the presumptive SYNE2α1 TSS. We validated the causality of rs11252591 by performing CRISPR-Cas9 gene editing in human iPS cells, which once differentiated into iCMs showed that the risk allele of rs11252591 led to lower expression of the SYNE2α1 isoform. Thus, we have accomplished two of our goals, the identification of the causal gene, a short isoform of SYNE2, and the causal variant, rs11252591. rs1152595 did not alter SYNE2α1 mRNA expression in iCMs after gene editing, which may be due to its greater distance away from its TSS; however, it may alter the expression of other SYNE2 transcript isoforms, which were not assessed. In addition, we did not rule out that other SNPs in LD with rs1152591 could play additional regulatory roles in mediating SYNE2α1 expression.

We next asked, how does higher levels of expression of the SYNE2α1 short isoform protect against AF? Nesprin-2 is an important LINC complex protein, which physically connects the nucleus with the cytoskeleton and participates in regulating nuclear morphology, nuclear position, and regulating mechanical cell signaling and downstream gene expression^16–18. Rare SYNE2 loss of function mutations are associated with autosomal dominant Emery-Dreifuss muscular dystrophy (EDMD); for example, the Nesprin-2 T89M missense mutation disrupts the LINC complex and leads to nuclear morphological changes¹⁹. Expression of Nesprin isoforms is highly tissue-dependent; the short N-terminal truncated Nesprin-1α2 and Nesprin-2α1 isoforms, which lack the calponin-homology domain⁵, are found almost exclusively in cardiac and skeletal muscle^{20, 21}.

Previous Nesprin-2 knockout mouse models showed increased nuclear area in cardiomyocytes²² and fibroblasts²³, with SYNE1 and SYNE2 double knockout mice developing early onset cardiomyopathy²². Previously, a very short form of SYNE2 encoding only the C-terminal KASH domain was transfected into mouse C2C12 cells, which localizes to the nuclear membrane and displaces the endogenous giant SYNE2 isoform from the nuclear membrane²⁴. Thus, the KASH only domain SYNE2 acts as a dominant negative by competing with the endogenous SYNE2 for binding to the nuclear membrane SUN proteins and leads to decreased cellular elasticity²⁵. We found similar dominant negative effects of over expression of SYNE2α1 on nuclear size and stiffness, as well as on downstream effects on a subset of the transcriptome.

Cardiomyocyte nuclei are under constant stress during contraction and are susceptible to physical strain from the sarcomeres and cytoskeleton. Disruption of the LINC complex by the dominant-negative effects of over expression of SUN1 or Nesprin-1α in the Lmna (encoding lamin A/C) KO cardiomyopathy mouse model increases longevity due to reduced tensional force transmission to the cardiomyocyte nuclei ²⁶ and decreased microtubule cytoskeleton activities at the nucelus²⁷. The dominant negative effect of expressing Nesprin-2α1 also reduces nuclear membrane rupture and DNA damage, leading to improved viability and contractility in Lmna KO myofibers²⁸. The AF risk allele of rs1152591, leading to low expression of Nesprin-2α1 in cardiomyocytes, may strengthen the connection of the cytoskeleton to the nuclear membrane and lamin network, which could increase mechanical stress during contraction, and lead to more nuclear damage and cell death that may promote AF susceptibility. Direct human evidence linking the nuclear membrane to AF is demonstrated by the high incidence of AF and atrial tachycardia in subjects carrying LMNA mutations²⁹.

The LINC complex plays a crucial role in mechanotransduction, which translates biophysical forces into biochemical signals that regulate gene expression. Impaired gene expression changes in response to biomechanical stimuli was reported in Nesprin-1/2 knockout mice¹⁴. Depletion of STEF/TIAM2, a binding partner of the full length Nesprin-2, significantly decreases nuclear stiffness and reduces YAP/TAZ-regulated genes³⁰. Based on prior studies, it may be inferred that changes in nuclear stiffness may influence cytoskeletal stiffness^{31, 32}, which in turn impacts cardiomyocyte interaction with surrounding ECM molecules and gene transcription. Furthermore, sinoatrial node specific knockout of Yap signaling in mice leads to sinoatrial node dysfunction, cardiac fibrosis, and altered expression of calcium related gene and cell-cell communication³³. Here, we demonstrated that both SYNE2α1 short isoform OE and all SYNE2 isoforms KD led to a significant decrease in nuclear stiffness and inhibited YAP-TEAD signaling in the cardiomyocytes, again showing the dominant negative activity of SYNE2α1. Given the complexity of atrial fibrillation and the scarcity of research about YAP/TAZ signaling in atrial fibrillation, further work on Hippo/Wnt/GPCR-Yap signaling in atrial cardiomyocytes/fibroblasts and downstream targets is needed.

Calcium handling has been shown playing a critical role in AF induction and maintenance^{34, 35}. The LINC complex is involved in excitation-contraction (EC) coupling in cardiomyocytes³⁶, with Nesprin-3 KD causing faster decay of the calcium transient, briefer contractions, and decreased cell shortening. Additionally, Nesprin-2 has been found to regulate calcium/calmodulin mediated nuclear transport in fibroblasts³⁷. Our study revealed expression differences in several calcium related genes such as CACNA1D, SERCA2, PLN, which may play a role mediating some of the calcium transient differences between the SYNE2 KD and SYNE2α1 OE iCMs. We also found colocalization of Nesprin-2α1 with SERCA2/SR, which has previously been observed in skeletal muscle²¹. Thus, there may be a direct interaction of Nesprin-2α1 with the calcium handling proteins in SR. We hypothesize that decreased Nesprin-2α1 expression, due to the AF risk alleles at this locus, may increase AF susceptibility by direct effects in the SR and accompanied changes in the expression and/or function of key ion channels.

Action potential duration and conduction velocity have been reported to influence AF pathogenesis³⁸. In our study, we observed increased expression of multiple potassium channels genes and decreased expression of gap junction proteins such as Connexin-43, which may lead to the conduction velocity or APD changes after SYNE2 KD. Previously, Nesprin-2α1 was found to interact with microtubules through kinesin³⁹, which interacts with KCNQ1 to regulate I_ks function and influence action potential kinetics particularly during β1-adrenergic stimulation in cardiomyocytes⁴⁰ and HL-1 cells⁴¹. Thus, Nesprin-2α1 may play a role in ion channel regulation, altering cellular electrophysiology and leading to electrical remodeling that may increase AF susceptibility.

There are several limitations in our study. First, there was very low expression level of Nesprin2-α1 in our lab grown less mature iPSC-derived atrial polarized cardiomyocytes, thus for our gene expression and electrophysiology studies we used the relatively more mature iCMs from Cellular Dynamics which are more ventricular like. Second, although we identified calcium cycling and action potential changes in the SYNE2 KD or SYNE2α1 OE iCMs, we were not able to successfully perform patch clamp studies to reveal different ion channels functions due to cellular fragility. Thus, we were not able to directly link SYNE2α1 expression with changes in cardiomyocyte ion currents. Third, since transfection efficiency of iCMs with GFP-SYNE2α1 only produced about 10% GFP positive cells, we needed to use AAV6 to overexpress GFP-SYNE2α1 in iCMs, which may alter iCM baseline characteristics. However, to account for this we always compared GFP-SYNE2α1 transduced cells with control cells transduced with AAV6 to express GFP.

In conclusion, we demonstrated that risk allele of the AF-associated SNP rs1152591 decreased SYNE2α1 expression, and that knockdown of all isoforms of SYNE2 or OE of the SYNE2α1 isoform in cardiomyocytes altered nuclear membrane mechanical properties and influenced many other genes such as those involved in calcium and potassium handling. Accordingly, we found that SYNE2 KD or SYNE2α1 OE in iCMs have downstream effects on calcium cycling and action potential dynamics, which are changes that may directly modify AF risk in the patients that are carriers of the risk allele.

Nonstandard Abbreviations and Acronyms

AF

Atrial fibrillation

cis-eQTL

cis-expression quantitative trait locus

iCMs

Human stem cell-derived induced cardiomyocytes

KASH

Klarsicht-ANC-Syne-homology

KD

Knockdown

LAA

Left atrial appendage

LD

Linkage disequilibrium

LINC

The linker of nucleoskeleton and cytoskeleton

Scr

Scramble

SUN

Sad1p/UNC-84

TSS

Transcription start site