Genomic Variants in NEURL, GJA1 and CUX2 Significantly Increase Genetic Susceptibility to Atrial Fibrillation

Atrial fibrillation (AF) is the most common arrhythmia. In 2014, two new meta-GWAS identified 5 AF loci, including the NEURL locus, GJA1 locus, CAND2 locus, and TBX5 locus in the European ancestry populations and the NEURL locus and CUX2 locus in a Japanese population. The TBX5 locus for AF was reported by us in 2013 in the Chinese population. Here we assessed the association between AF and SNPs in the NEURL, GJA1, CAND2 and CUX2 loci in the Chinese Han population. We carried out a large case-control association study with 1,164 AF patients and 1,460 controls. Significant allelic and genotypic associations were identified between NEURL variant rs6584555 and GJA1 variant rs13216675 and AF. Significant genotypic association was found between CUX2 SNP rs6490029 and AF. No association was found between CAND2 variant rs4642101 and AF, which may be due to an insufficient power of the sample size for rs4642101. Together with our previous findings, seven of fifteen AF loci (<50%) identified by GWAS in the European ancestry populations conferred susceptibility to AF in the Chinese population, and explained approximately 14.5% of AF heritability. On the other hand, two AF loci identified in the Japanese population were both replicated in the Chinese population.

To date, no GWAS were reported for AF in the Chinese population despite the fact that there are over 10 million AF patients in China. Our group previously analyzed the potential association between AF and the 10 AF loci identified in the early series of GWAS and meta-GWAS in the European ancestry populations. We found that three of them, including the PITX2c, ZFHX3 and CAV1 loci, showed significant association with AF in the Chinese Han population, but other loci were not replicated in the Chinese Han population [18][19][20] . For the TBX5 locus, we reported in 2013 that a genomic variant in TBX5, rs3825214, showed a significant association with AF in the Chinese population 21 . In this study, we assessed association between AF with other meta-GWAS SNPs identified in European ancestry populations and the Japanese population, including SNP rs6584555 in NEURL, rs13216675 near GJA1, rs4642101 in CAND2 and rs6490029 in CUX2, in the Chinese GeneID population. We identified significant allelic and genotypic association between NEURL rs6584555 and GJA1 SNP rs13216675 and AF, significant genotypic association between CUX2 SNP rs6490029 and AF, but no association between CAND2 SNP rs4642101 and AF.

Results
Significant allelic association between GJA1 SNP rs13216675 and NEURL SNP rs6584555 and AF.
We carried out a case control association study for AF with four SNPs, including SNP rs4642101 within the CAND2 gene on chromosome 3p25.2, rs13216675 close to the GJA1 gene on chromosome 6q22.3, rs6584555 near the NEURL gene on chromosome 10q24.33 and rs6490029 within the CUX2 gene on chromosome 12q24.11. Our study population included 1,164 AF patients and 1,460 non-AF controls from the Chinese Han GeneID population. The average age of the case group was 2.6 years younger than the control group (61.27 ± 11.33 vs. 63.8 ± 13.54, P < 0.01). The other characteristics of the case group and the control group are summarized in Table 1. In the control population, the genotypic frequencies for all four SNPs did not deviate from the Hardy-Weinberg equilibrium (P > 0.01). The minor allele frequency (MAF) of each SNP in our control population is similar to the data for the Chinese Han population from the HapMap database ( Table 2).
The GJA1 SNP rs13216675 showed significant association with AF (observed P obs = 3.9 × 10 −3 , OR = 1.2) ( Table 2). After adjusting for covariates of age, gender, hypertension (HTN), diabetes mellitus (DM) and coronary artery disease (CAD), the association remained significant (P adj = 0.01, OR = 1.19) ( Table 2). The common allele T of SNP rs13216675 is the risk allele in the Chinese Han population ( Table 2). The significant association between SNP rs13216675 and AF remained after adjusting for multiple testing with Bonferroni correction (corrected P = 0.04) ( Table 2).
The two remaining SNPs, CAND2 SNP rs4642101 and CUX2 SNP rs6490029 did not show significant allelic association with AF in the Chinese Han population before or after adjustment for covariates (P obs and P adj > 0.05) ( Table 2).
Significant genotypic association between NEURL SNP rs6584555, GJA1 rs13216675 and CUX2 SNP rs6490029 and AF. We also performed the case control association analysis for genotypic frequencies, which may pinpoint potential genetic models under which a significant association is found for a genetic variant in contrast to allelic association analysis. We analyzed the genotypic association for each SNP under three common genetic models: an additive model, a dominant model, or a recessive model. The results are summarized in Table 3.   Table 1. Clinical and demographical characteristics of study subjects. Data are shown as mean +/− standard deviation (SD) for quantitative variables and percentage (%) for qualitative variables. *The differences between cases and controls for qualitative variables such as gender, hypertension, type 2 diabetes and CAD were analyzed by a Chi-square (χ 2 ) test. The difference for quantitative variables such as means of age was analyzed with a student t test. AF, atrial fibrillation.
For CUX2 SNP rs64990029, although no significant allelic association was found for AF, significant genotypic association was identified for AF under both the additive and the dominant models, but not under the recessive model (P obs = 7.97 × 10 −3 under a dominant model; P obs = 0.02 under an additive model) ( Table 3). The significant genotypic association between SNP rs64990029 and AF remained under the dominant model and after adjusting for covariates of age, gender, HTN, DM and CAD (P adj = 8.28 × 10 −3 ) and after further adjusting for multiple testing with Bonferroni correction (corrected P = 0.04) ( Table 3).
For GJA1 SNP rs13216675, significant genotypic association was identified for AF under an additive model and a dominant model, but not under a recessive model (P obs = 6.72 × 10 −3 , P adj = 0.01 under an additive model; P obs = 2.28 × 10 −3 , P adj = 3.04 × 10 −3 under a dominant model; P obs = 0.07, P adj = 0.14 under a recessive model)  ( Table 3). The significant genotypic association between SNP rs13216675 and AF remained after adjusting for multiple testing with Bonferroni correction (P < 0.05) ( Table 3). For CAND2 SNP rs4642101, similar to the data from allelic association analysis, we did not find any significant genotypic association with AF in the Chinese Han population.

Estimation of AF heritability explained by SNPs significantly associated with AF in the Chinese population.
For the three SNPs showing significant association with AF in the Chinese Han population (GJA1 SNP rs13216675, NEURL SNP rs6584555 and CUX2 SNP rs6490029), we estimated the heritability of AF explained by each of them. As shown in Table 4, GJA1 SNP rs13216675, NEURL SNP rs6584555 and CUX2 SNP rs6490029 explained 1.8%, 3.7% and 2.6% of AF heritability, respectively. Together, these three variants explained approximately 8.1% of AF heritability. Previously, we reported three other SNPs which also showed significant association with AF in the Chinese Han population, including SNP rs2200733 on 4q25 and near PITX2, rs2106261 on ZFHX3 locus and rs3807989 on CAV1 [18][19][20] . SNP rs2200733, rs2106261 and rs3807989 explained about 6.4% of AF heritability (1.8% for rs2200733, 1.7% for rs2106261 and 2.9% for rs3807989). Together, these six SNPs explained 14.5% of AF heritability.

Discussion
In this study, we analyzed four genomic variants associated with AF in either the European ancestry populations or the Japanese population for their association with AF in the Chinese Han population. These variants include NEURL SNP rs6584555, GJA1 SNP rs13216675, CUX2 SNP rs6490029 and CAND2 SNP rs4642101. Our study population consisted of 1,164 AF patients and 1,460 non-AF controls. Three of the four loci, the NEURL locus, GJA1 locus and CUX2 locus, were successfully replicated in the Chinese population (Tables 2 and 3). The CAND2 locus was not replicated in the Chinese population (Tables 2 and 3), which may be due to an insufficient power of the sample size for this variant.
The 2014 meta-GWAS in the European ancestry populations reported four loci for AF, including NEURL (rs12415501), GJA1 (rs13216675), TBX5 (rs10507248) and CAND2 (rs4642101). The TBX5 locus was reported in 2013 by us by studying a Chinese AF population 21 before the GWAS report. For the three remaining loci, the NEURL and GJA1 loci were significantly associated with AF in the Chinese population, whereas the CAND2 locus did not show any significant association with AF (Tables 2 and 3). Previously, we showed that only three of the 10 AF GWAS loci identified in the European ancestry populations before 2014 were significantly associated with AF in the Chinese populations [18][19][20] . Interestingly, the two AF loci reported in the 2014 GWAS in a Japanese population, namely NEURL SNP rs6584555 and CUX2 SNP rs6490029, were both replicated in the Chinese population (Tables 2 and 3). This may be related to the fact that the evolution distance between the Japanese population and the Chinese population is closer that that between the European ancestry populations and the Chinese population.
Our study has a limitation. Our study population of 1,164 AF patients and 1,460 non-AF controls has a sufficient power of 97% and 91% for genomic variants rs6584555 in NEURL and rs6490029 in CUX2, respectively. However, its power for rs13216675 near GJA1 and rs4642101 in CAND2 was 0.40 and 0.39, respectively. Therefore, lack of association between rs4642101 in CAND2 and AF may be due to the small sample size. Future studies with larger AF case control populations may be needed to further clarify the association between rs4642101 in CAND2 and AF in the Chinese Han population.
In conclusion, we found significant associations between AF and NEURL SNP rs6584555, GJA1 SNP rs13216675 and CUX2 SNP rs6490029, but not CAND2 SNP rs4642101. Together with our earlier reports, we show that among the 15 GWAS loci for AF reported in the European ancestry populations and Japanese population, seven loci (PITX2c, ZFHX3, CAV1, NEURL, GJA1, TBX5 and CUX2 loci) also confer a significant risk of AF in the Chinese Han population. Our findings provide an important understanding of the detailed genomic landscape for AF susceptibility in the Chinese Han population. Our data also suggest that although the European ancestry populations share some common susceptibility loci for AF with the Chinese population, different populations may contain their own unique susceptibility loci for AF.

Materials and Methods
Study subjects. The study subjects for this study were from the large GeneID database, which has over 80,000 study subjects with cardiovascular diseases in the Chinese Han population [22][23][24][25][26][27] . To minimize stratification of population heterogeneity, only study subjects of Han ethnic origin (by self-description) were included. A total SNP genotyping. Human genomic DNA was extracted from peripheral blood samples using the Wizard Genomic DNA Purification Kit as described previously by us 30 .
Each SNP was genotyped using a Rotor-Gene 6000 High Resolution Melt system as described by us previously 31,32 . The HRM technology is based on the different molecular physical properties of DNA molecules on the fragment length, GC content and GC distribution, which makes DNA molecules with different genotypes (two different homozygotes and the heterozygote) have different shapes and positions of its dissolution curves when heated at different temperatures. The three different genotypes for a genomic variant can then be distinguished based on their different dissolution curves. The polymerase chain reactions (PCR) for genotyping was performed in a 25 μl mixture with 2.5 μl of 10 × PCR buffer, 10 mM dNTP (0.5 μl), 25 mM Mg 2+ (1.5 μl), 5 pmol of each primer, 25 ng of genomic DNA and 0.7 μl of 5 mM SYTO9. PCR was performed on an ABI9700 System (Applied Biosystems, Foster City, CA) with a thermal profile of 95 °C for 5 minutes, 40 cycles of 95 °C for 15 seconds, 60.3 °C or other appropriate annealing temperatures for 15 seconds and 72 °C for 20 seconds and 72 °C for 10 minutes. Primers for PCR are listed in Table 5. PCR products were directly genotyped using the high resolution melting (HRM) analysis on a Rotor-Gene 6000 System (Corbett Life Science, Australia). DNA samples from 100 study subjects were randomly selected for each SNP for direct Sanger sequencing analysis and the sequencing data completely matched the HRM genotyping data. Statistical analysis. We used PLINK version 1.07 to perform the Hardy-Weinberg linkage equilibrium test in the control group as described by us previously 18,27,[33][34][35] . Pearson 2 × 2 and 2 × 3 contingency table χ 2 tests were performed with SPSS (version 17.0; SPSS, Inc., Chicago, IL) to analyze allelic association and genotypic association, respectively and to compute odds ratios (ORs) and corresponding 95% confidential intervals (CIs). Multiple logistic regression analysis was used to adjust significant covariates of age, gender, hypertension (HTN), coronary artery disease (CAD) and diabetes mellitus (DM) for AF using SPSS (version 17.0; SPSS, Inc., Chicago, IL).
We estimated the heritability of AF explained by each significant SNP using the multifactorial liability threshold model based on OR estimates using the R package as described previously 36 . The computation of heritability was based on the frequency of the risk allele, relative risk of one risk allele (Aa) over that of no risk allele (aa) (OR: Aa/aa), relative risk of two risk alleles (AA) over that of no risk allele (aa) (OR:AA/aa) and the overall prevalence rate of AF in the population. We assumed a disease prevalence estimate of 0.73% for AF in the Chinese population.
Statistical power analysis of the study population was conducted using program PS (Power and Sample size Calculations, version 3.0.43) 37 . For power analysis, we utilized the reported OR values in the European ancestry populations (GJA1 SNP rs1321667 and CAND2 SNP rs4642101) or the Japanese population (NEURL SNP rs6584555 and CUX2 SNP rs6490029) 16,17 , the minor allele frequencies for the studied variants in the Chinese population from the HapMap database and a type I error of 0.05.  Table 5. Sequences for primers used for HRM genotyping and direct Sanger sequencing analysis. HRM, highresolution melting; SNP, single-nucleotide polymorphism.