Patients with AVNRT were enrolled in the Sichuan Provincial People’s Hospital in China from 2013 to 2020. Familial AVNRT defined that two or more AVNRT patients in a family, or 1 or more clinically diagnosed PSVT patients in a family of AVNRT proband patient. In addition, 100 unrelated ethnically matched healthy participants were recruited from the Sichuan Provincial People’s Hospital. All probands were diagnosed with AVNRT using intracardiac electrophysiological examination during treatment with radiofrequency catheter ablation, and family members diagnosed with AVNRT underwent intracardiac electrophysiological or transesophageal atrial pacing examination. Healthy participants did not have a history of cardiovascular diseases, arrhythmia, systemic immune diseases, cancers, or any other diseases known to cause arrhythmias. Whole blood samples from 20 patients with AVNRT and 100 normal control participants were collected in heparinized vacutainer tubes. Patients had signed an informed consent form before enrollment. This study was approved by the ethics committee of the Sichuan Academy of Medical Sciences and the Sichuan Provincial People’s Hospital.

Intracardiac electrophysiology recordings included atrial stimulation (burst or additional stimulation pacing) and ventricular stimulation in patients. AVNRT diagnosis is established on the basis of published standards and applicable pacing operations. The physiology of dual atrioventricular node is defined as the atrial-His (AH) interval increase of ≥50 ms after a decreasing interval of 10 ms during the additional stimulation of the single atrium or the AH interval increase of ≥50 ms after the pacing cycle length is shortened by 10 ms. If continuous AVNRT is not induced (lasting more than 30 s), the same pacing procedure was repeated with isoproterenol administration as described above.

In brief, we purified DNA from the peripheral blood using the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany). Whole-exome enrichment was performed using the SureSelect Human All Exon kit V6 (Agilent Technologies, Santa Clara, CA, United States). The genomic DNA library was sequenced using the HiSeq X and NovaSeq systems (Illumina, San Diego, CA, United States).

The sequenced DNA fragments were aligned with Human Reference Genome (National Center for Biotechnology Information Build 37) on the basis of the Burrows–Wheeler transform. The removal of duplication, realignment, and recalibration were performed with Picard tools¹ and GATK².

The single-nucleotide polymorphisms and insertion-deletion polymorphisms (indels) were performed using GATK3.7 software. The high-confidence variants were annotated with snpEff (Version 4.2)³. In addition, the annotations of all variants were further performed using 1000 Genomes Project data (2014 Oct release)⁴, the Exome Aggregation Consortium⁵, EVS⁶, the ClinVar⁷ database, and Online Mendelian Inheritance in Man⁸.

In total, 95 related reference genes as arrhythmia were selected for the analysis of rare variants in patients with AVNRT and control participants (5). These genes were considered reference genes according to our previous study (5). To increase reliability and generalizability of reference genes, data integration was used to select the genes following our previous sporadic AVNRT study. Therefore, the reference genes were identified in patients with both the sporadic and familial AVNRT and were assessed for segregation within families.

Biological process (BP) of Gene Ontology (GO) was performed by database for annotation, visualization and integrated discovery (DAVID) bioinformatics resources according to previous study (19). PPI network of candidate genes were obtained from the STRING database⁹. The images of single-cell sequencing data from healthy human cardiac tissue were obtained from the Human Protein Atlas¹⁰. The mouse phenotypes associated with pathogenic reference genes were extracted from the Mouse Genome Informatics (MGI) database¹¹.

To identify the common pathogenic mechanisms in all AVNRT cases, the 37 most likely pathogenic genes from our previously sporadic AVNRT study (5) were considered the intersection of both sporadic and familial AVNRT, and then candidate pathogenic genes were identified in patients with both familial and sporadic AVNRT, and were assessed for segregation within families.

Biological process analysis was performed by DAVID bioinformatics resources (19). The PPI network of candidate genes was obtained from the STRING database (see text footnote 9). The images of single-cell sequencing data from 24 healthy human tissues are obtained from the Human Protein Atlas (see text footnote 10). There are 51 cell types of human tissues. The mouse phenotypes associated with pathogenic reference genes were extracted from the MGI database (see text footnote 11).

To analyze the aggregate association of rare variants at the gene level, we performed gene-based burden analysis to obtain gene-level significant associations of familial AVNRT patients (n = 20) and control subjects (n = 100). Rare variants were defined as “deleterious variants” according to 1000 Genomes Project data and ExAC with MAF < 0.001, MAF < 0.01, or MAF < 0.05. Fisher’s exact test was used to evaluate gene-based burden analysis. The gene level across the genome was used to identify risk genes across different allele frequency spectrums.

The significant genes were submitted to the KOBAS3.0 web server¹² to obtain the functional gene set Reactome Pathway enrichment. The PPI network of candidate genes was obtained from the STRING database (see text footnote 9). The images of single-cell sequencing data from healthy human cardiac tissue were obtained from the Human Protein Atlas (see text footnote 10). The mouse phenotypes associated with pathogenic reference genes were extracted from the MGI database (see text footnote 11).

Protein–protein interactions network of candidate genes were obtained from the STRING database (see text footnote 9). The relationships among the screened genes were predicted by STRING database and visualized with Cytoscape v2.3 software.

In this study, a total of 20 patients and 100 control participants were included to perform WES. These 20 patients were assessed in nine families, including a total of 93 members (Figure 1). Among 20 patients enrolled in this study, the male to female ratio was 1.86, the mean age at onset was approximately 47.5 years, the heart rate at onset was approximately 176.9 beats per minute, and all the patients were free from structural heart disease (Table 1). All the patients showed typical slow-fast AVNRT, and 60.0% of the patients were successfully treated by radiofrequency ablation during the operation.

To increase reliability and generalizability of related reference genes, data integration was used to confirm the genes following our previous sporadic AVNRT study (5). Therefore, patients with both sporadic and familial AVNRT were enrolled in this study. Among the 95 related reference genes, seven candidate pathogenic genes have been identified in patients with both familial and sporadic AVNRT: CASQ2, AGXT, ANK2, SYNE2, ZFHX3, GJD3, and SCN4A (Supplementary Table 1). We found CASQ2 have distinct feature between familial and sporadic AVNRT (1 rare variant, 6/20 patients, familial AVNRT; 1 rare variant, 1/82 patient, sporadic AVNRT). Furthermore, rare variants of AGXT, ANK2, SYNE2, GJD3, and SCN4A co-segregated within one pedigree and those of CASQ2 and ZFHX3 within two and three pedigrees, respectively (Supplementary Table 1).

The bubble plot of GO-BP analysis showed that the functions of these genes were mainly associated with cardiac conduction, muscle contraction, and the release of sequestered calcium ions (Supplementary Table 1 and Figure 2A). Furthermore, PPI networks of these genes indicated that CASQ2, ANK2, and SCN4A constituted the network, and ZFHX3 interacted with SYNE2 (Figure 2B). In addition, the results of single-cell sequencing showed that the relative expression of CASQ2, ANK2, and SYNE2 was higher in cardiomyocytes than others (Figure 2C and Supplementary Figure 1), whereas the expression of SCN4A, ZFHX3, AGXT, and GJD3 was relatively lower or not expressed (Figure 2C and Supplementary Figure 1).

To further verify these gene functions, the MGI database was used to confirm their phenotype. The disruption of CASQ2, essential for Ca²⁺ storage, led to ventricular tachycardia in both mice and humans (Table 2). Moreover, the abnormal function of ankyrin-2 (ANK2) may lead to sinoatrial node disease and ankyrin-B-related cardiac arrhythmia in humans (Table 2). In addition, its abnormality increased heart rate variability and caused the abnormal sinoatrial node conduction in the mouse (Table 2). ZFHX3 was identified as a crucial risk factor for atrial fibrillation (20), SYNE2 contributed to cardiac arrhythmia (Table 2), and GJD3 caused abnormal atrioventricular node conduction (Table 2).

Based on MGI database and previous study (20), we suggested that CASQ2, ANK2, SYNE2, GJD3, and ZFHX3 were the most likely pathogenic genes for AVNRT.

To identify the common pathogenic mechanisms in all AVNRT cases, the 37 most likely pathogenic genes from our previous sporadic AVNRT study (5) were considered the intersection of both sporadic and familial AVNRT. Among these genes, five pathogenic genes were identified in patients with both familial and sporadic AVNRT: LAMC1, RYR2, COL4A3, NOS1, and ATP2C2 (Supplementary Table 2). We identified only one gene NOS1 that was totally shared the same two rare variants both in six familial (6/20 patients) and 33 sporadic (33/82 patients) AVNRT patients (Supplementary Table 2). Other genes have distinct features between familial and sporadic AVNRT: LAMC1 (1 rare variant, 2/20 patients, familial AVNRT; 4 rare variants, 7/82 patients, sporadic AVNRT), RYR2 (1 rare variant, 1/20 patient, familial AVNRT; 8 rare variants, 8/82 patients, sporadic AVNRT), COL4A3 (1 rare variant, 2/20 patients, familial AVNRT; 5 rare variants, 5/82 patients, sporadic AVNRT); ATP2C2 (2 rare variants, 2/20 patients, familial AVNRT; 5 rare variants, 5/82 patients, sporadic AVNRT; Supplementary Table 2). Furthermore, the rare variants of LAMC1, COL4A3, NOS1, and ATP2C2 co-segregated within one pedigree apart from RYR2 (Supplementary Table 2).

The BP enrichment analysis suggested that the functions of these genes were mainly associated with heart contraction and the regulation of calcium ion (Figure 3A and Supplementary Table 2). In addition, the PPI networks showed that RYR2, NOS1, and ATP2C2 constituted the network, and COL4A3 interacted with LAMC1 (Figure 3B). Moreover, the results of single-cell sequencing data showed that the relative expression of RYR2 and LAMC1 was higher in cardiomyocytes than others (Figure 3C and Supplementary Figure 2), whereas the expression of COL4A3, NOS1, and ATP2C2 was relatively lower or not expressed (Figure 3C and Supplementary Figure 2).

The biological function and phenotype of these genes were further explored using the MGI database. Homozygous mutation in the NOS1 gene led to abnormal cardiac muscle relaxation and increased heart rate in the mouse (Table 2). The disruption of RYR2 was associated with ventricular dysplasia and ventricular tachycardia in humans, whereas it is mainly associated with an increased heart rate and ventricular tachycardia in the mouse (Table 2). However, cardiac diseases were independent of the functions of LAMC1, COL4A3, and ATP2C2 (Table 2).

Considering their functions and previous study (5), RYR2 and NOS1 were likely to be causal genes for AVNRT.

In search of the underlying pathogenic mechanisms within AVNRT pedigrees, we imposed more restrictive criteria: more than two mutations and one homozygous mutation in one gene segregated at least two pedigrees. A total of 299 genes with 452 rare variants were identified (Supplementary Table 3).

As shown in Supplementary Table 3, the three AVNRT-related traits among the pathways in the Reactome databases were as follows: (1) stimuli-sensing channels, (2) RYR tetramers transport Ca²⁺ from the SR lumen to the cytosol, and (3) ion channel transport. In addition, seven pathogenic genes were identified, including TRDN, ANO6, SLC9C1, CASQ2, ATP6V0A4, SGK2, and WNK1. Remarkably, CASQ2 has been involved in AVNRT as reference genes.

Mouse genome informatics database and previous studies was further used to confirm the phenotype of these genes. The disruption of TRDN contributed to ventricular tachycardia in humans (Table 2; 21). Mice lacking ANO6 developed shortened PQ intervals (Table 2). The aberration of WNK1 led to hereditary sensory and autonomic neuropathy in humans (22). Considering their functions and previous study (21, 22), TRDN, CASQ2, and WNK1 were likely to be the common pathogenic genes in familial AVNRT.

To explore the internal relationship of the candidate pathogenic genes in this study, PPI network analysis was further constructed. Among these 14 candidate pathogenic genes, three networks were established; these genes were CASQ2, AGXT, ANK2, SYNE2, ZFHX3, GJD3, SCN4A, LAMC1, RYR2, COL4A3, NOS1, ATP2C2, TRDN, and WNK1 (Figure 4A). PPI networks indicated that the genes constituted network 1 (CASQ2, ANK2, SCN4A, RYR2, NOS1, ATP2C2, and TRDN), network 2 (SYNE2 and ZFHX3), and network 3 (LAMC1 and COL4A3).

The maximum network 1 was mainly associated with the calcium-signaling pathway using Kyoto Encyclopedia of Genes and Genomes enrichment analysis (Figure 4B). Among them, RYR2 acted as a calcium channel that released calcium ions into the cytoplasm from the SR and thus regulated cardiac muscle contraction (23). The RYR forms a complex with TRDN, junction (JTC), and CASQ instead of acting independently (24). Moreover, the mutations of RYR2 or CASQ lead to Ca²⁺ leak in ventricular tachycardia and thus contribute to Ca²⁺ waves in arrhythmogenic as a result of the increasing Ca²⁺ spark frequency and rising flux (24, 25). Particularly, CASQ2 and TRDN have also been identified in this study. Furthermore, another RYR2-related gene is neuronal NOS1, which is located in the cardiac SR and enhances contraction through NO production (26). The present studies have shown that the inhibition of NOS1 decreased RYR2 activity because of reducing Ca²⁺ sparks and shortened action potential causing arrhythmia susceptibility (9, 26). In addition, ANK2 from the SR promotes the flow of calcium ions into the plasma membrane through the inositol triphosphate receptor and NCX (27). ATP2C2 encodes calcium-transporting ATPase, removing calcium from the cytosol into the Golgi body (28). Therefore, the calcium-signaling pathway may be closely associated with the occurrence of AVNRT.