Next-Generation Sequencing in Extreme Atrial Fibrillation Trait
Next-Generation Sequencing in Extreme Atrial Fibrillation Trait
This is the first NGS study for common AF. We conducted this study based on the hypothesis that the genetic determinants of extreme cases in common diseases (with extreme phenotypes) may be similar to those of rare diseases; thus, rare variants or mutations may be the underlying genetic determinants for specific patients, according to the 'rare disease rare variant' hypothesis. Interestingly, we identified several novel variants that may be disease causing for patients with extreme AF harbouring these mutations.
Applications of NGS in non-familial common diseases may not be conceptually possible because hundreds or thousands of possible variants may be identified by in-depth whole exome sequencing. Validation and determination of these variants as common variants or rare mutations by genotyping in independent populations is necessary, which requires intensive labour and high costs. We here provided an alternative approach to focus on extreme cases of common diseases and consider them to be rare diseases based on the observation that patients at the extremes of trait distributions seem to be more likely to carry loss-of-function alleles or mutations with large effects. Extremes cases are commonly defined as those having extreme values of a quantitative trait (e.g., cholesterol level). The prevalence of these extreme cases is much lower, compared with that of those having mild or moderate abnormality of the quantitative trait. Although AF is not a quantitative trait, we hypothesised that those patients with AF having a very high attack frequency are similar to those having extreme values of a quantitative trait, based on the observation that the prevalence of patients with AF with a very high attack frequency (more than once per day) is much lower than that with low to moderate attack frequency. Then the approach is similar to find mutations in rare diseases or Mendelian disorders (to find the novel responsible variant or mutation in probands).
Another more common approach is to sequence those with disease onset at younger age, who are thought to have a primarily genetic cause or harbour variants with large effects. In a large Danish follow-up cohort, the risk for lone AF was substantially increased in young-aged persons given a first-degree relative affected also at young age (<30 years). It is common to find patients who had asymptomatic AF attacks many years before the diagnosis of AF has been made. In other words, the age of AF disease onset may be many years earlier than the age of diagnosis. Therefore, it is highly possible that many of our patients with extremely symptomatic AF might have had asymptomatic AF attacks when they were young.
We further focused on the most probable candidate genes a priori, e.g., in the present study, we focused on published AF genes with GWAS signals, and did not perform genome-wide screening that would be likely to yield a very large number of possible variants. Importantly, although genetic variants identified by GWAS are supposed to be common variants, genes containing common variants with mild to modest effects on complex traits may also contain rare variants or mutations with larger effects. For example, 11 of 30 genes implicated as carrying common variants associated with lipid levels also carry known rare alleles of large effect identified in Mendelian dyslipidaemias, including ABCA1, PCSK9 and LDLR. Therefore, in a true disease-causing gene, rare variants may colocate with GWAS-defined or repeatedly replicated common variants. Recently targeted exon sequencing on candidate ionic channel genes, e.g., KCNA5, SCN5A, SCN3B and KCND3, for AF have also been reported. Interestingly, we did not find any novel variant or mutation in these genes in our Taiwanese AF cohort.
Through the candidate-gene approach, a small or modest amount of novel and specific variants may be identified. The number of possible variants may be further decreased by (1) excluding existing or published common variants, which may be confirmed by genotyping in normal populations, e.g., in the present study, we excluded variants that could be found in the published database (dbSNP), our existing exome sequence database, normal subjects and their parents and (2) excluding variants that did not have functional significance, such as synonymous exon mutations.
It may be difficult or even impossible to directly prove that these identified patient-private mutations are the true disease-causing mutations, which pertains to most of the targeted exon sequence studies for common diseases. For most of the AF-associated SNPs from GWAS, there is still ongoing investigation to determine whether the associated genes are indeed part of the mechanism for increased risk, or whether these 'hits' are identifying trans-effects or are merely false-positives. Moreover, even the identified variant do have functional changes, they may play no role in increasing the risk of AF, the so-called 'true-true-unrelated' interpretation. Undoubtedly there are also numerous other patient-private mutations within other genes that may be the responsible mutation(s). At a minimum, it is more conceivable that among all the possible mutations in the same patient, the mutation in a previous GWAS-identified gene is more likely to be the disease-causing mutation because the encoded protein of this GWAS-identified gene may be related to the disease mechanism (although not yet proven). As mentioned before, 11 of 30 genes carrying common variants associated with lipid levels also carry mutations of large effect identified in Mendelian dyslipidaemias.
Finally, it may be questioned that we used trios with unaffected parents (no family history of AF) to identify de novo mutations, and essentially the identified mutations could not explain the 'missing heritability' of AF because these variants are de novo and were not inherited from the parents. However, we could by no means rule out the very small possibility that some of these variants could also be found in the general population with a very low frequency, accounting a very small part of the missing heritability of AF.
Among the reported susceptibility loci for AF identified by GWAS, risk variants on chromosome 4q25, e.g., rs2200733, are the most strongly associated risk variants, and these results have been consistently replicated in different populations, including our Taiwanese population. Whether rs2200733 is a marker SNP or even the disease-causing variant remains unknown. The only gene close to these risk variants is the PITX2 gene, although the location of PITX2 is still relatively far from these strong risk variants. Therefore, whether these strong risk variants on chromosome 4q25 are linked to the risk of AF through PITX2 remains an open question. Furthermore, recently Gore-Panter et al have shown that the risk allele on chromosome 4q25 was not associated with the atrial expression level of PITX2. Nevertheless, it has been reported that knockout mice with heterozygous deficiency of PITX2 have an increased susceptibility to atrial arrhythmia. More studies are warranted to define whether the identified risk allele long away from the PITX2 coding region does regulate PITX2 expression and whether PITX2 protein plays a role in the mechanism of AF.
Nevertheless, our results also shed light on whether there is a functional or disease-causing variant in the PITX2 gene, if PITX2 is indeed the disease-causing gene for AF. We did not identify any variants in the coding sequence of the PITX2 gene in our AF cases (with or without extreme phenotypes) and the non-AF Taiwanese controls, indicating that the true AF risk variant may localise in a non-coding sequence. Interestingly, we found an A-319G mutation in the 5'UTR of the PITX2 gene in one patient with AF with an extreme phenotype, which was not present in Taiwanese healthy subjects and other patients with AF. It has been well established that variants or mutations in the 5'UTR or the promoter can modulate the expression of the downstream gene, especially when they localise in the regions that are enriched with transcription factor binding and epigenetic modification. It has also been shown that promoter single nucleotide mutation(s) may be disease causing for rare diseases, such as Brugada syndrome, a rare human arrhythmia. Accordingly, in the promoter assay, we found that −319G mutant allele was associated with a significant lower PITX2 gene promoter activity, either in the basal condition or during rapid depolarisation. Interestingly, decreased PITX2 function has been shown to increase the susceptibility to atrial arrhythmia in a mouse model of PITX2 haploinsufficiency. These results raise the possibility that A-319G PITX2 gene promoter mutation is the disease-causing mutation. However, as mentioned above, more studies are still warranted to define the role of PITX2 in the mechanism of AF.
For the A+688G mutation in the first exon of the SYNE2 gene, we found that the region around this mutation is DNase I hypersensitive region. Recently, it has been shown that exons which are hypersensitive to DNase I are actively involved both in the transcription (RNA polymerase II binding) and in the alternative splicing. Then it is logical to speculate that A+688G mutation may affect the expression level of the SYNE2 gene and contribute to the risk of AF.
In conclusion, rare variants or mutations may be identified by using the NGS method in patients with extreme traits of a common disease, e.g., AF, even with a very small number of cases, as has been widely used to find mutations for rare or Mendelian diseases.
Discussion
This is the first NGS study for common AF. We conducted this study based on the hypothesis that the genetic determinants of extreme cases in common diseases (with extreme phenotypes) may be similar to those of rare diseases; thus, rare variants or mutations may be the underlying genetic determinants for specific patients, according to the 'rare disease rare variant' hypothesis. Interestingly, we identified several novel variants that may be disease causing for patients with extreme AF harbouring these mutations.
Applications of NGS in non-familial common diseases may not be conceptually possible because hundreds or thousands of possible variants may be identified by in-depth whole exome sequencing. Validation and determination of these variants as common variants or rare mutations by genotyping in independent populations is necessary, which requires intensive labour and high costs. We here provided an alternative approach to focus on extreme cases of common diseases and consider them to be rare diseases based on the observation that patients at the extremes of trait distributions seem to be more likely to carry loss-of-function alleles or mutations with large effects. Extremes cases are commonly defined as those having extreme values of a quantitative trait (e.g., cholesterol level). The prevalence of these extreme cases is much lower, compared with that of those having mild or moderate abnormality of the quantitative trait. Although AF is not a quantitative trait, we hypothesised that those patients with AF having a very high attack frequency are similar to those having extreme values of a quantitative trait, based on the observation that the prevalence of patients with AF with a very high attack frequency (more than once per day) is much lower than that with low to moderate attack frequency. Then the approach is similar to find mutations in rare diseases or Mendelian disorders (to find the novel responsible variant or mutation in probands).
Another more common approach is to sequence those with disease onset at younger age, who are thought to have a primarily genetic cause or harbour variants with large effects. In a large Danish follow-up cohort, the risk for lone AF was substantially increased in young-aged persons given a first-degree relative affected also at young age (<30 years). It is common to find patients who had asymptomatic AF attacks many years before the diagnosis of AF has been made. In other words, the age of AF disease onset may be many years earlier than the age of diagnosis. Therefore, it is highly possible that many of our patients with extremely symptomatic AF might have had asymptomatic AF attacks when they were young.
We further focused on the most probable candidate genes a priori, e.g., in the present study, we focused on published AF genes with GWAS signals, and did not perform genome-wide screening that would be likely to yield a very large number of possible variants. Importantly, although genetic variants identified by GWAS are supposed to be common variants, genes containing common variants with mild to modest effects on complex traits may also contain rare variants or mutations with larger effects. For example, 11 of 30 genes implicated as carrying common variants associated with lipid levels also carry known rare alleles of large effect identified in Mendelian dyslipidaemias, including ABCA1, PCSK9 and LDLR. Therefore, in a true disease-causing gene, rare variants may colocate with GWAS-defined or repeatedly replicated common variants. Recently targeted exon sequencing on candidate ionic channel genes, e.g., KCNA5, SCN5A, SCN3B and KCND3, for AF have also been reported. Interestingly, we did not find any novel variant or mutation in these genes in our Taiwanese AF cohort.
Through the candidate-gene approach, a small or modest amount of novel and specific variants may be identified. The number of possible variants may be further decreased by (1) excluding existing or published common variants, which may be confirmed by genotyping in normal populations, e.g., in the present study, we excluded variants that could be found in the published database (dbSNP), our existing exome sequence database, normal subjects and their parents and (2) excluding variants that did not have functional significance, such as synonymous exon mutations.
It may be difficult or even impossible to directly prove that these identified patient-private mutations are the true disease-causing mutations, which pertains to most of the targeted exon sequence studies for common diseases. For most of the AF-associated SNPs from GWAS, there is still ongoing investigation to determine whether the associated genes are indeed part of the mechanism for increased risk, or whether these 'hits' are identifying trans-effects or are merely false-positives. Moreover, even the identified variant do have functional changes, they may play no role in increasing the risk of AF, the so-called 'true-true-unrelated' interpretation. Undoubtedly there are also numerous other patient-private mutations within other genes that may be the responsible mutation(s). At a minimum, it is more conceivable that among all the possible mutations in the same patient, the mutation in a previous GWAS-identified gene is more likely to be the disease-causing mutation because the encoded protein of this GWAS-identified gene may be related to the disease mechanism (although not yet proven). As mentioned before, 11 of 30 genes carrying common variants associated with lipid levels also carry mutations of large effect identified in Mendelian dyslipidaemias.
Finally, it may be questioned that we used trios with unaffected parents (no family history of AF) to identify de novo mutations, and essentially the identified mutations could not explain the 'missing heritability' of AF because these variants are de novo and were not inherited from the parents. However, we could by no means rule out the very small possibility that some of these variants could also be found in the general population with a very low frequency, accounting a very small part of the missing heritability of AF.
Among the reported susceptibility loci for AF identified by GWAS, risk variants on chromosome 4q25, e.g., rs2200733, are the most strongly associated risk variants, and these results have been consistently replicated in different populations, including our Taiwanese population. Whether rs2200733 is a marker SNP or even the disease-causing variant remains unknown. The only gene close to these risk variants is the PITX2 gene, although the location of PITX2 is still relatively far from these strong risk variants. Therefore, whether these strong risk variants on chromosome 4q25 are linked to the risk of AF through PITX2 remains an open question. Furthermore, recently Gore-Panter et al have shown that the risk allele on chromosome 4q25 was not associated with the atrial expression level of PITX2. Nevertheless, it has been reported that knockout mice with heterozygous deficiency of PITX2 have an increased susceptibility to atrial arrhythmia. More studies are warranted to define whether the identified risk allele long away from the PITX2 coding region does regulate PITX2 expression and whether PITX2 protein plays a role in the mechanism of AF.
Nevertheless, our results also shed light on whether there is a functional or disease-causing variant in the PITX2 gene, if PITX2 is indeed the disease-causing gene for AF. We did not identify any variants in the coding sequence of the PITX2 gene in our AF cases (with or without extreme phenotypes) and the non-AF Taiwanese controls, indicating that the true AF risk variant may localise in a non-coding sequence. Interestingly, we found an A-319G mutation in the 5'UTR of the PITX2 gene in one patient with AF with an extreme phenotype, which was not present in Taiwanese healthy subjects and other patients with AF. It has been well established that variants or mutations in the 5'UTR or the promoter can modulate the expression of the downstream gene, especially when they localise in the regions that are enriched with transcription factor binding and epigenetic modification. It has also been shown that promoter single nucleotide mutation(s) may be disease causing for rare diseases, such as Brugada syndrome, a rare human arrhythmia. Accordingly, in the promoter assay, we found that −319G mutant allele was associated with a significant lower PITX2 gene promoter activity, either in the basal condition or during rapid depolarisation. Interestingly, decreased PITX2 function has been shown to increase the susceptibility to atrial arrhythmia in a mouse model of PITX2 haploinsufficiency. These results raise the possibility that A-319G PITX2 gene promoter mutation is the disease-causing mutation. However, as mentioned above, more studies are still warranted to define the role of PITX2 in the mechanism of AF.
For the A+688G mutation in the first exon of the SYNE2 gene, we found that the region around this mutation is DNase I hypersensitive region. Recently, it has been shown that exons which are hypersensitive to DNase I are actively involved both in the transcription (RNA polymerase II binding) and in the alternative splicing. Then it is logical to speculate that A+688G mutation may affect the expression level of the SYNE2 gene and contribute to the risk of AF.
In conclusion, rare variants or mutations may be identified by using the NGS method in patients with extreme traits of a common disease, e.g., AF, even with a very small number of cases, as has been widely used to find mutations for rare or Mendelian diseases.
Source...