The cohort for this study comprised 50 patients diagnosed with arrhythmia at Xiangya Hospital, Central South University. These patients were enrolled for WES based on the attending clinician’s judgment, typically involving early-onset arrhythmia, atypical clinical presentation, or suspected genetic etiology. All participants were of Han Chinese descent and were recruited in a consecutive manner. Ethical approval for this research was obtained from the Review Board of Xiangya Hospital of Central South University (approval number: 202103427 for human specimens). Informed consent was obtained from all subjects prior to their inclusion in the study. Comprehensive clinical data, encompassing medical and family histories, were collected and documented for each participant, detailed in Supplementary Table S1.

Peripheral blood samples were collected from all study participants. Genomic DNA was extracted from collected samples using the DNeasy Blood and Tissue Kit (Qiagen, Valencia, CA, USA), following the manufacturer’s protocol. WES was conducted at the Beijing Genomics Institute (Beijing, China). The exonic regions were enriched using Agilent SureSelect Human All Exon V6 kits. Subsequently, high-throughput sequencing was performed on an Illumina HiSeq2000 platform, adhering to standard protocols.

Raw sequencing data were evaluated for quality using FastQC (Babraham Bioinformatics, 2019) (v0.11.9). Adapter sequences and low-quality bases were trimmed using TrimGalore (Babraham Bioinformatics group, 2024) (v0.6.6). The processed reads were aligned to the human reference genome (GRCh38/hg38) using the Burrows-Wheeler Aligner (Li and Durbin, 2010) (BWA-MEM, v0.7.17). After alignment, SAMtools (Li et al., 2009) (v1.11) was used to sort the reads and remove duplicates. Base quality score recalibration (BQSR) and local realignment around Indels were conducted using the Genome Analysis Toolkit (GATK, v4.1.9.0), following the best-practice guidelines. Variant calling was performed with GATK HaplotypeCaller, generating single nucleotide variants (SNVs) and small insertions/deletions (Indels) in the variant call format (VCF). Joint genotyping was applied across all samples, and a variant quality score recalibration (VQSR) approach was employed to filter potential false-positive calls. Only high-confidence variants passing the default threshold (FILTER = PASS) were retained. The resulting VCF files were normalized and filtered with following criteria: allele balance of 0.2–0.8 for heterozygous genotypes, genotype quality ≥20, and read depth (DP) ≥10 (≥5 for sex chromosomes). Filtered variants from all individuals were merged for generating the initial arrhythmia genetic dataset for subsequent analyses. We achieved a median sequencing depth of 110x for targeted exonic regions and a minimum coverage of 20x for 99% of the targeted sequences. Our data provided a high level of confidence for identifying variants with potential functional significance, setting the stage for a comprehensive analysis of the potential genetic architecture underlying cardiac arrhythmia in this cohort.

Our study focused on rare, deleterious exonic variants. A multi-step selection strategy was applied (Guo et al., 2023; Lin et al., 2025) (Supplementary Figure S1). To distinguish between potentially pathogenic variants and common polymorphisms, we utilized population frequency data from multiple databases: the 1000 Genomes Project (2,504 individuals from 26 populations), gnomAD (v2.1.1, >140,000 individuals), the ExAC database (>60,000 individuals), and our in-house cohort of 100 healthy controls. Variant calling was refined by excluding variants with population variant allele frequencies (VAF) ≥0.01 in any of these databases, as these frequently observed variants are likely to be benign polymorphisms. Variants with population frequencies not available (NA) were assigned a frequency of zero. Non-exonic variants, including intergenic, intronic, and synonymous SNPs, were also filtered out. Focusing on exonic single nucleotide variations (SNVs) and insertions/deletions (Indels) with the potential to alter amino acid sequences and protein structures, we included 42,179 SNVs and 1,945 frameshift Indels in the study. To assess pathogenicity, variants were subjected to a rigorous computational prediction process using four widely recognized algorithms: SIFT (Ng and Henikoff, 2003), MutationTaster (Steinhaus et al., 2021), Polyphen2 (Adzhubei et al., 2013) (in both HumDiv and HumVar modes), and CADD (Rentzsch et al., 2019). The cutoff values for variant pathogenicity were established at 0.05 for SIFT, 0.435 for Polyphen2-HumDiv, 0.445 for Polyphen2-HumVar, and 5 for CADD. Variants predicted benign by at least three of the five tools were categorized as consensus benign, leading to the exclusion of 18,718 variants from downstream analysis. Our procedure yielded 25,406 rare deleterious SNVs and Indels for downstream analysis. The majority were missense variants (non-synonymous variants, n = 21,910; 86.24%), followed by a smaller proportion of nonsense variants (stop gain variants, n = 1,512; 5.95%), stop loss (n = 33; 0.13%), splicing region variants (n = 6; 0.02%), and frameshift indels (n = 1,945; 7.66%). Variants were also annotated with OMIM (Hamosh et al., 2005) and ClinVar (Landrum et al., 2018) databases following their standard tutorial to assess known disease associations and clinical significance.

Protein-Protein Interaction analysis was conducted using the STRING (Szklarczyk et al., 2023) database following their tutorial (https://string-db.org/). We used the 132 r.m. genes plus genes of interest (SPP1, CCR2, and CD44) as input to assess the direct and indirect associations between proteins. The analysis parameters were set to include all active interaction sources such as text mining, experiments, databases, co-expression, neighborhood, gene fusion, and co-occurrence. The confidence score cutoff was established at a high threshold (0.7) to ensure the specificity of interactions. For visualization purposes, the disconnected nodes were hidden from the final network. We also used k-means clustering method imbedded in STRING to cluster interconnected protein nodes. The PPI network visualization highlighted the potential functional connectivity between the identified proteins and their possible collective role in the pathogenesis of cardiac arrhythmias.

GO analysis was carried out using the ClusterProfiler (Xu et al., 2024) package in R to categorize the identified r.m. genes into biological processes and cellular components. We performed the GO analysis following the recommended tutorial (https://learn.gencore.bio.nyu.edu/rna-seq-analysis/gene-set-enrichment-analysis/). P-values were adjusted by Benjamini–Hochberg procedure (Benjamini and Hochberg, 1995) with a cutoff of <0.05 considered significant. The results were visualized in bar plots displaying the -log10 adjusted P-values, which highlighted the most significantly enriched GO terms in the context of biological processes and cellular components.

Single-cell RNA sequencing data were obtained from the GEO database under accession number GSE224959 (Hulsmans et al., 2023). Analysis of the scRNA-seq data was conducted using the R package Seurat (Hao et al., 2021) (version 5.0.0), following the QC steps outlined in our previous work. Standard procedures, including data normalization, variable gene identification, scaling, and principal component analysis (PCA), were performed. Sample-wise batch effects were removed using the Seurat integration method (CCA mod). The annotation of cell populations within the normal and arrhythmia groups was performed with the canonical cell type makers and can be found in Supplementary Figure S4. Differential expression (DE) analysis between normal and arrhythmia groups was carried out for each identified cell type using the FindAllMarkers function within Seurat. This is to identify the cell-type-specific conditional DE genes. Subsequently, GO enrichment analysis was conducted for the upregulated and downregulated genes in each cell type, utilizing ClusterProfiler as described in the previous section. Cell-cell communication analysis was performed using the CellChat (Jin et al., 2025) R package (v1.6.1) following the standard tutorial.

All statistical analyses were performed using R (v4.0.0). Continuous variables were compared using Student’s t-test. Clinical association analysis was conducted to identify correlations between genetic variants and clinical characteristics (e.g., disease onset, gender, and cardiomyopathy presence). The associations were assessed using Fisher’s exact test in R with function fisher.test. Variants were plotted against their respective chromosomal positions in a Manhattan plot, generated using the qqman package in R. The threshold for significance was set at a P-value of 0.05. A forest plot was created to visualize the odds ratios (ORs) and their confidence intervals (CIs) for variants associated with specific clinical features using the forestplot package in R. We note that in some cases where mutations were absent in one subgroup, OR calculations resulted in extreme values (either 0 or +), which posed challenges for log₂ transformation in forest plot visualization. To mitigate this, OR values were capped at 0.01 for lower-bound cases and 100 for upper-bound infinite values. This capping strategy is only for data visualization purposes. Mutational hits analysis was conducted to map the distribution of mutational hotspots along the protein sequences of genes identified to harbor variants of interest. Protein length information was retrieved using the biomaRt package in R. Mutations were then positioned along the protein sequences according to their amino acid change coordinates. Each variant was plotted to reflect its position within the protein sequence and its recurrence within our patient cohort. The results of this analysis were summarized in a series of lollipop plots, with each lollipop representing a mutation and its frequency. P-values were adjusted using the Benjamini–Hochberg (Benjamini and Hochberg, 1995) false discovery rate (FDR) correction method.

In our study, 50 unrelated Chinese individuals diagnosed with cardiac arrhythmia were analyzed (Table 1). The cohort included 21 females and 29 males, with ages ranging from 9 to 49y (mean: 22y) (Figures 1A,B). Notably, 94% of patients presented with early-onset arrhythmias (age ≤35y). This predominance of young patients minimizes environmental confounding factors and emphasizes genetic determinants. Half of our cohort (n = 25) exhibited very early-onset disease (age ≤20years). As arrhythmias are frequently associated with cardiomyopathy, we analyzed these clinical relationships in our cohort (Figure 1C). The majority (72%) presented with isolated atrial arrhythmia, while ventricular arrhythmia showed a striking association with structural heart disease, with cardiomyopathy present in 50% of ventricular cases compared to merely 5.3% of atrial cases. This pattern aligns with the established clinical observation that ventricular arrhythmias typically present with more severe clinical manifestations and stronger connections to underlying structural abnormalities. Detailed clinical information is in Supplementary Table S1.

We performed WES experiments on peripheral blood samples per patient. WES data was preprocessed and filtered following our established procedures (Guo et al., 2023; Lin et al., 2025) (Supplementary Figure S1), with a focus on exonic rare deleterious variants. After quality controls, the median variant count per sample was 452, with most samples showing consistent mutational burden (Figure 1C). One outlier, CS021, harbored 1,649 variants. We observed no significant correlation between total mutation burden and patient demographics (age, gender) or clinical disease categories, suggesting that qualitative rather than quantitative genetic differences may underlie the pathophysiological mechanisms in this cohort.

We prioritized genes that were recurrently mutated (r.m.) across our arrhythmia cohort. We established a recurrence threshold of 15 (representing 30% of the cohort), and captured genes consistently implicated in cardiac arrhythmia (Figure 2A). This approach revealed 132 r.m. candidate risk genes harboring a total of 8,075 variants (Figure 2B; Supplementary Table S2), providing a foundation for understanding the genetic architecture underlying cardiac arrhythmia in our cohort. We noticed several genes exhibited very high mutation frequencies, such as PDE4DIP, MST1L, PRAMEF1, and members of the HLA family. The mutation patterns varied considerably across genes, with some displaying primarily missense variants (e.g., ADAMTS7, CACNA1B) while others showed enrichment for potentially more disruptive frameshift or nonsense mutations (e.g., IFNA10, OTOA).

Additionally, chromosomal mapping demonstrated non-uniform variant distribution, with significant enrichment on chromosomes 1, 6, and 13 (Figure 2C). The pronounced enrichment on chromosome 6 (variants: n = 822; protein coding genes: n = 1,014) primarily stemmed from extensive mutations in the HLA gene family, consistent with previous reports of high polymorphism rates in these immunologically relevant loci (Shiina et al., 2009). The notable peaks on chromosome 1 (variants: n = 1,962; protein coding genes: n = 2,013) corresponded to clusters of genes involved in cell cycle regulation and signal transduction pathways—processes critical for cardiac electrophysiology (Grune et al., 2021) and have been previously implicated in arrhythmogenic processes (Schwartz et al., 2020). Chromosome 13 exhibits the highest normalized density of recurrent variants when adjusted for the number of protein-coding genes (n = 315). Further investigation revealed 442 variants (missense, 284; nonsense, 83; and frameshift indels, 75) distributed across five genes among our cohort (Supplementary Table S2): inflammatory control (ALOX5AP), RNA metabolism (PABPC3), DNA repair and chromatin remodeling (PARP4), and cellular structural integrity (SKA3, SLAIN1). Moreover, chromosome 18 emerges with a moderate variant count but the highest proportion of deleterious predictions. All those variants were within SLC35G4, which encodes a nucleotide sugar transporter involved in glycosylation pathways. Its disruption may affect cardiac glycoprotein processing and cellular signaling networks essential for proper cardiac function.

To assess the clinical and functional relevance of these r.m. genes, we first annotated them using established disease databases. For ClinVar (Landrum et al., 2018), only 143 variants (1.77%) returned valid annotations; for OMIM database (Hamosh et al., 2005), 4,700 variants (58.2%) had at least one associated OMIM entry. Two genes harbored variants with established cardiac pathogenicity (Supplementary Table S3): (1) variants in TTN were associated with multiple cardiac phenotypes including dilated cardiomyopathy (OMIM: 604145), hypertrophic cardiomyopathy (OMIM: 613765), limb-girdle muscular dystrophy with cardiac involvement (OMIM: 608807), and myopathy with early-onset fatal cardiomyopathy (OMIM: 611705); and (2) variants in MST1 were associated with immunodeficiency syndromes (OMIM: 614868) that can present with cardiac manifestations. These findings indicate the limited direct overlap between our cohort’s mutational profile and currently catalogued pathogenic variants in cardiac disease databases, suggesting that the genetic landscape of arrhythmia may involve novel or incompletely characterized pathogenic networks. Therefore, to further explore the functional implications of these 132 r.m. genes, we performed protein-protein interaction (PPI) (Szklarczyk et al., 2021) and gene ontology (GO) analysis (Wu et al., 2021), revealing critical insights into arrhythmia pathogenesis and its potential progression to cardiomyopathy and cardiac fibrosis.

PPI network analysis identified two major gene clusters alongside three significant protein pairs (Figure 3A). The first prominent cluster consisted of mucin family genes, which encode glycoproteins critical for maintaining epithelial barrier integrity. Recent evidence suggests that mucin dysregulation affects not only gastrointestinal (Fan et al., 2023) and respiratory (Guo et al., 2023) tract protection but also influences cardiac pathophysiology through gut-heart axis disruption (Fan et al., 2023). Mucin alterations may compromise intestinal barrier function, leading to systemic inflammation and subsequent cardiac electrical remodeling. The second major cluster comprised HLA gene families, which orchestrate immune responses and antigen presentation. Variations within these genes are known to modulate the cardiac immune environment, prompting inflammatory reactions that have been recognized as significant precursors to cardiac arrhythmias (Darsee et al., 2010; Li et al., 2021), and ultimately, myocardial fibrosis. Among the identified protein pairs, the NEB-TTN interaction is noteworthy. These structural proteins maintain sarcomere integrity and myofibril elasticity; their disruption could compromise mechanical coupling between cardiomyocytes, potentially establishing a substrate for both arrhythmias and progressive fibrotic remodeling (Laitila et al., 2020; Splawski et al., 2005). The ESRRA-NCOR2 pair regulates metabolic and transcriptional processes essential for cardiac homeostasis, with dysregulation potentially exacerbating oxidative stress and fibroblast activation (Cividini et al., 2021; Van Ouwerkerk et al., 2020). The HYDIN-GPRIN2 interaction affects cellular signaling cascades that could influence both electrical conduction and fibrotic responses (Brünger et al., 2023; Pierpont et al., 2018).

GO enrichment analysis substantiated these findings (Supplementary Table S4), revealing significant enrichment of immune-related biological processes (Figure 3B) and the involvement of extracellular matrix (ECM) elements (Figure 3C). First, multiple pathways related to antigen processing and presentation were prominently represented, including MHC complex assembly and both endogenous and exogenous peptide antigen processing (Fan et al., 2023; Li et al., 2021). This immune signature suggests that genetic predisposition to aberrant immune responses may trigger chronic inflammation, a recognized driver of electrical instability and subsequent cardiac fibrosis (London, 2017). Cellular component analysis further highlighted significant enrichment in collagen-containing ECM, transport vesicles, and cytoskeletal components. These findings directly implicate ECM remodeling and fibrosis-related processes in arrhythmia pathogenesis. Additionally, enrichment in Golgi apparatus components suggests alterations in protein processing and secretion pathways crucial for ECM maintenance and immune function.

Our functional analyses reveal a genetic framework linking cardiac arrhythmia to immune dysregulation and ECM remodeling. The identified mutations likely create a substrate for both electrical disturbances and structural remodeling, establishing a molecular basis for the observed clinical progression from arrhythmias to cardiomyopathy and cardiac fibrosis in susceptible individuals (Schwartz et al., 2020; Shoureshi et al., 2024; Sohns and Marrouche, 2020; Piek et al., 2019).

To validate our genetic findings at the transcriptomic level and establish their relevance in specific cardiac cell populations, we analyzed scRNA-seq data from cardiac tissues of arrhythmia patients (n = 7, atrial fibrillation) and healthy controls (n = 5) from a public dataset (Supplementary Table S5) (Hulsmans et al., 2023). Given the consistent enrichment of immune pathways and ECM components among our 132 r.m. genes, we hypothesized that their transcriptional consequences would be particularly evident in immune and stromal cell populations. We proceeded with our established pipeline (Huang et al., 2023), and cell type annotation was confirmed by expression of canonical markers (Supplementary Figure S2). This procedure yielded 37,675 cells (arrhythmia, n = 22,532; control, n = 15,143) and identified nine distinct cell types (Figure 4A), including myeloid lineages, neutrophils, mast cells, plasma dendritic cells (pDCs), T cells, B cells, endothelial cells, fibroblasts, and mural cells. Cell-type composition exhibited significant alterations between arrhythmic and control hearts (Figure 4B). Notably, macrophages showed the most pronounced enrichment in arrhythmic samples (+13.0% difference), followed by monocytes (+4.0% difference). Conversely, structural cell type endothelial cells (−8.7% difference) were proportionally reduced in arrhythmic hearts, potentially reflecting tissue remodeling processes.

Macrophage-focused DE analysis revealed that arrhythmia-associated macrophages exhibited upregulated cytokine response and immune activities with concurrent downregulation of metabolic processes (Figure 4C; Supplementary Table S6), consistent with a pro-inflammatory phenotype. When mapping our 132 r.m. genes onto macrophage DE results (Figure 4D), we observed significant upregulation of NOTCH2NL and HLA family genes, which are reported to drives inflammation and fibrosis (Fiddes et al., 2018). Conversely, TNXB, an ECM glycoprotein that regulates collagen fibrillogenesis, showed decreased expression (Schalkwijk et al., 2001). This observed pattern suggests a shift toward pro-inflammatory states with altered ECM regulation in arrhythmia-associated macrophages.

Our comparative analysis of differentially enriched pathways across other major cell types revealed a conserved signature of elevated cytokine responses (Figure 4E; Supplementary Figure S3; Supplementary Table S6), underscoring a potential shared inflammatory mechanism driven by r.m. genes. Additionally, we identified cell-type-specific dysfunctions: T cells showed reduced fatty acid metabolism; endothelial cells exhibited compromised vasculogenesis; fibroblasts displayed increased collagen fibril organization; and mural cells demonstrated alterations in mitochondrial organization and smooth muscle contraction. The fibroblast-specific enhancement of collagen organization aligns with the ECM remodeling pathways identified in our genetic analysis, further supporting the role of fibrotic processes in arrhythmogenesis.

Previous research has established that SPP1+ macrophages play a critical role in cardiac arrhythmogenesis (Hulsmans et al., 2023). Importantly, the SPP1 protein functions as a multifaceted signaling molecule with dual roles in promoting both inflammation and fibrosis, perfectly aligning with the immune dysregulation and ECM remodeling pathways identified in our genetic analysis. To further characterize this critical cell subpopulation, we first identified SPP1+ macrophages (Figure 5A; Supplementary Figure S4). SPP1+ macrophages were markedly enriched in arrhythmic hearts (Figure 5B), indicating that the observed SPP1 signal reflects specific cellular infiltration rather than broad transcriptional upregulation across the macrophage population (Hulsmans et al., 2023).

DE analysis comparing SPP1+ macrophages to other macrophages revealed 164 upregulated genes (Figure 5C; Supplementary Table S7). Key markers include SPP1, FABP5 (fatty acid binding protein), LIPA (lysosomal acid lipase), and FN1 (fibronectin 1, ECM glycoprotein). Those makers indicate a metabolically active population involved in lipid processing, lysosomal function, fibrotic processes, and ECM remodeling. Pathway enrichment analysis of SPP1+ macrophage-specific genes revealed a complex functional profile (Figure 5D): while these cells exhibited classical M2-like macrophage signatures (Yunna et al., 2020) (e.g., lipid localization and transport, phagocytosis, lysosomal function, and wound healing), they simultaneously displayed pro-inflammatory characteristics through IL-12 signaling and neutrophil degranulation pathways. Additionally, prominent enrichment in extracellular matrix organization and cell-substrate adhesion pathways underscores their role in tissue remodeling processes. This mixed polarization profile suggests that SPP1+ macrophages represent a distinct pathological subset that combines tissue remodeling capacity with sustained inflammatory activity (Reggio et al., 2025; Palma, 2025), consistent with their proposed role in driving both immune dysregulation and fibrotic remodeling in cardiac arrhythmia.

To systematically investigate the communication networks involving SPP1+ macrophages, we inferred cell-cell interaction using CellChat (Jin et al., 2025) (Figures 5E–G). The SPP1+ macrophages undergo remarkable activation from normal to disease conditions, showing enhanced both outgoing and incoming signaling strength compared to other cell types (Figure 5E). This positions SPP1+ macrophages as central signaling hubs in the arrhythmic cardiac microenvironment. Comprehensive signaling pathway details across all cell types are provided in Supplementary Figure S5 (arrhythmia hearts) and Supplementary Figure S6 (normal controls). Next, we performed differential interaction strength analysis between conditions. SPP1+ macrophage-fibroblast communication is significantly enhanced in arrhythmic hearts (Figure 5F). Motivated by this observation, we examined the specific signaling pathways mediating SPP1+ macrophage-fibroblast crosstalk in detail (Figure 5G). SPP1+ macrophages showed enhanced outgoing signals including COLLAGEN, APP (amyloid precursor protein), MIF (macrophage migration inhibitory factor), and SPP1 pathways, all exclusively upregulated in arrhythmic conditions. Correspondingly, fibroblasts exhibited increased incoming COLLAGEN and SPP1 signaling, along with enhanced FN1 (fibronectin) and LAMININ pathways. This bidirectional enhancement creates a comprehensive network of inflammatory and ECM remodeling signals, establishing a potential positive feedback loop where activated fibroblasts may further recruit SPP1+ macrophages, perpetuating both inflammatory and fibrotic processes in arrhythmic hearts.

Another key question in arrhythmia pathogenesis is what genetic mechanisms drive the recruitment and activation of SPP1+ macrophages. Previous studies identified elevated CCR2 as critical for SPP1+ macrophage recruitment (Hulsmans et al., 2023), but the genetic underpinnings remained unclear. We hypothesized that dysregulated r.m. genes could activate the CCR2/SPP1 axis, thereby facilitating inflammatory macrophage infiltration while simultaneously promoting fibrotic remodeling. Our PPI network analysis revealed direct interactions between the CCR2/SPP1 axis and multiple r.m. genes (Figure 5H, top), particularly with the HLA gene family, which showed elevated expression in arrhythmic conditions (Supplementary Figure S3). Notably, SPP1 also displayed interactions with COL18A1 and COL6A3, two r.m. genes upregulated in fibroblasts (Supplementary Figure S3), establishing a molecular bridge between macrophage-driven inflammation and cardiac fibrosis (Zhang et al., 2019). We further explored CD44, another key regulator of SPP1+ macrophages (Palma, 2025), and found extensive PPI connections within r.m. genes including additional ECM components (Supplementary Figure S7). This suggests a potential mechanism through which SPP1+ macrophages may directly mediate inflammatory-fibrogenic network, two critical hallmarks in arrhythmia progression and persistence (Shoureshi et al., 2024; Sohns and Marrouche, 2020). Correlation analyses within macrophages confirmed positive associations between CCR2 and the r.m. genes HLA-DR5 and HSPA6 (Figure 5H, bottom), key players in immune modulation and stress response (Grune et al., 2021; London, 2017). The negative correlation with PABPC1, involved in mRNA stability, suggested translational stress in arrhythmia-associated macrophages, further supported by concurrent HSPA6 upregulation (Maxwell et al., 2021).

In summary, our single-cell analysis offers transcriptomic validation of our genetic findings and provides mechanistic insights into how r.m. genes might drive cardiac arrhythmia through dual inflammatory and fibrotic pathways. The SPP1+ macrophage emerges as a central cellular mediator, linking immune dysregulation and ECM remodeling—the two key pathways identified in our r.m. gene functional analysis. These findings suggest that genetic variants in immune and ECM-related genes may predispose certain arrhythmia patients to develop cardiomyopathy through SPP1+ macrophage-mediated inflammation and fibrosis.

Building on our pathway and transcriptomic findings, we conducted exploratory analyses to investigate associations between r.m. genes and clinical features (Figures 6A,B), and to identify conserved mutational hotspots that might contribute to arrhythmogenesis (Figure 6C). We aimed to complement our WES and single-cell insights, providing additional molecular insights into how specific mutations within r.m. genes might contribute to cardiac arrhythmia through disruption of immune regulation and ECM homeostasis.

We identified several significant genotype-phenotype associations. Notably, ADAMTS7 mutations were enriched in very early-onset arrhythmia patients (OR = 9.71 [2.38–47.74], P-value <0.001). ADAMTS7 encodes a secreted metalloproteinase recognized as a risk locus for coronary atherosclerosis (Bengtsson et al., 2017; Mizoguchi et al., 2021). These mutations likely alter ECM composition and affect vasculature behavior (Mizoguchi et al., 2021), potentially accelerating arrhythmia development. Gender-specific associations revealed that SLC9B1 mutations, which encode a sodium-hydrogen exchanger regulating cellular pH, were exclusively found in female patients (P-value = 0.017). These variants may disrupt intracellular pH balance and consequently affect cardiac electrical activity (Patel et al., 2023). The presence of mutated SLC9B1 only in females may be influenced by sex-specific genetic expression and hormonal interactions, potentially associated with estrogen signaling (ESRRA is recurrently mutated in our cohort), which can modify cardiac ion channel functionality (Van Ouwerkerk et al., 2020). Conversely, PCDH11Y, located on chromosome Y, was exclusively identified in male patients (P-value <0.001). Interestingly, OTOA mutations showed associations with both later onset (OR = 0.17 [0.04–0.68], P-value = 0.009) and female predominance (OR = 3.41 [0.92–13.58], P-value = 0.045). The age-association parallels observations in OTOA-related hearing loss (Sugiyama et al., 2019), possibly reflecting age-dependent expression patterns, while gender-specific regulation may underlie increased female susceptibility to OTOA variants. We did not identify genes enriched in patients with concurrent cardiomyopathy (n = 9), either due to limited statistical power from the small subgroup or suggesting arrhythmia patients may share a conserved genetic predisposition toward structural heart disease.

Our analysis further revealed domain-specific mutational hotspots in several key r.m. genes (Figure 6C). In mucin family genes (MUC4, MUC17), mutations predominantly affected extracellular serine/threonine-rich tandem repeats, potentially disrupting glycosylation patterns critical for mucosal barrier function. PABPC1 harbored mutations within RNA recognition motifs that could compromise mRNA stability and protein synthesis (Maxwell et al., 2021)—consistent with the translational stress observed in our single-cell analysis. In HLA family genes (HLA-DRB1, HLA-DRB5), mutations clustered in peptide-binding regions, potentially altering antigen presentation capacity and immune responses. HSPA6 mutations affected domains essential for ATPase activity, potentially impairing stress response functions (Maxwell et al., 2021)—particularly relevant given its correlation with CCR2 in arrhythmia-associated macrophages. NOTCH2NL displayed mutations in EGF-like domains critical for Notch signaling (Fiddes et al., 2018). SLC9B1 mutations concentrated in transmembrane domains, potentially disrupting ion transport (Patel et al., 2023), while GPRIN2 mutations might interfere with G protein-coupled receptor interactions affecting downstream signaling (Brünger et al., 2023; Pierpont et al., 2018).