Author + information
- Received August 25, 2011
- Revision received October 13, 2011
- Accepted November 1, 2011
- Published online March 13, 2012.
- Stefan A. Mann, PhD⁎,
- Robyn Otway, PhD⁎,
- Guanglan Guo, PhD⁎,
- Magdalena Soka, BSc(Hons)⁎,
- Lina Karlsdotter, MBiomedSci⁎,
- Gunjan Trivedi, BSc(Hons)⁎,
- Monique Ohanian, BMedSci(Hons)⁎,
- Poonam Zodgekar, MSW, GradDipGenCouns⁎,
- Robert A. Smith, PhD†,
- Merridee A. Wouters, PhD‡,
- Rajesh Subbiah, MBBS, PhD§,
- Bruce Walker, MBBS, PhD§,
- Dennis Kuchar, MD§,
- Prashanthan Sanders, MBBS, PhD∥,
- Lyn Griffiths, PhD†,
- Jamie I. Vandenberg, MBBS, PhD⁎,¶ and
- Diane Fatkin, MD⁎,§,¶,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Diane Fatkin, Victor Chang Cardiac Research Institute, Lowy Packer Building, 405 Liverpool Street, P.O. Box 699, Darlinghurst, NSW 2010, Australia
Objectives The aim of this study was to evaluate the role of cardiac K+ channel gene variants in families with atrial fibrillation (AF).
Background The K+ channels play a major role in atrial repolarization but single mutations in cardiac K+ channel genes are infrequently present in AF families. The collective effect of background K+ channel variants of varying prevalence and effect size on the atrial substrate for AF is largely unexplored.
Methods Genes encoding the major cardiac K+ channels were resequenced in 80 AF probands. Nonsynonymous coding sequence variants identified in AF probands were evaluated in 240 control subjects. Novel variants were characterized using patch-clamp techniques and in silico modeling was performed using the Courtemanche atrial cell model.
Results Nineteen nonsynonymous variants in 9 genes were found, including 11 rare variants. Rare variants were more frequent in AF probands (18.8% vs. 4.2%, p < 0.001), and the mean number of variants was greater (0.21 vs. 0.04, p < 0.001). The majority of K+ channel variants individually had modest functional effects. Modeling simulations to evaluate combinations of K+ channel variants of varying population frequency indicated that simultaneous small perturbations of multiple current densities had nonlinear interactions and could result in substantial (>30 ms) shortening or lengthening of action potential duration as well as increased dispersion of repolarization.
Conclusions Families with AF show an excess of rare functional K+ channel gene variants of varying phenotypic effect size that may contribute to an atrial arrhythmogenic substrate. Atrial cell modeling is a useful tool to assess epistatic interactions between multiple variants.
Genetic variation is increasingly recognized to be a significant determinant of human disease. Over the past decade, genome-wide association studies have sought to identify common genetic variants that affect susceptibility to common complex disorders, but the variants identified have generally had only modest individual effect size and collectively explain relatively little of observed heritability (1). With the advent of new sequencing technologies, it is now feasible and affordable to sequence the entire human exome to look for rare gene-coding sequence variants (1–3). Data analysis has become the major challenge, with thousands of variants detected in every individual person.
Recently, exome sequencing of 237 ion channels in persons with idiopathic epilepsy demonstrated that rare deleterious variants in known Mendelian disease genes were present not only in affected cases but also in a majority of age-and race-matched healthy subjects (4). These important observations clearly demonstrate that single deleterious variants cannot be assumed to be disease-causing and that the total variant burden needs to be considered. Variants in the same gene, or in different genes, can be expected to show epistasis, namely, have additive, neutralizing, or synergistic actions that are nonintuitive and unpredictable based on knowledge of the individual variant effects (5). Consequently, development of in silico methods for modeling the biological effects of multiple variants is critically required to derive meaningful information from genomic sequence data.
Here we have utilized atrial cell modeling in an analysis of variants in multiple cardiac potassium (K+) channel genes in familial atrial fibrillation (AF). Genetic factors have an important role in the pathogenesis of AF, but the genes involved and mechanistic links with atrial arrhythmogenesis are incompletely understood. Given the fundamental importance of K+ currents in atrial repolarization, cardiac K+ channel genes have been considered strong candidates, and mutations in 8 genes have been associated with AF in families or in sporadic cases (6–9). Several common variants that modify susceptibility to AF in the general population have also been identified, including an intronic variant in KCNN3, that encodes the calcium-activated small conductance K+ channel, SK3 (10). We resequenced genes encoding all the major cardiac K+ currents in a cohort of persons with familial AF and coding sequence variants identified were evaluated in healthy control subjects. Novel variants were characterized using patch-clamp techniques, and an atrial cell model was used to assess the effects of multiple simultaneous variations of K+ channel activation on atrial action potential (AP) properties. Our data show that multiple K+ channel variants are frequently present in families with AF and can contribute to an arrhythmogenic atrial substrate.
Study subjects comprised 80 persons (56 males), 26 to 90 years of age (mean 55 years of age) with a family history of AF, defined by AF in 2 or more first-degree relatives. None of the families studied had other inherited cardiac or systemic disorders that would account for AF. Families in which isolated affected members had concurrent risk factors for AF, such as hypertension, were not excluded. All subjects provided informed written consent and were evaluated by history and physical examination, electrocardiography, and transthoracic echocardiography. Two hundred-forty healthy subjects (83 male), 16 to 91 years of age (mean 53 years of age), with no history of cardiovascular disease comprised a control group. All participants were of Caucasian ethnicity. Protocols were approved by St. Vincent's Hospital human research ethics committee.
Protein-coding sequences of the KCND3, KCNIP2, KCNA5, KCNQ1, KCNH2, KCNE1, KCNE2, KCNE3, KCNE4, KCNE5, KCNJ2, KCNJ4, and KCNJ14, genes were polymerase chain reaction amplified from genomic deoxyribonucleic acid (DNA) and sequenced using Big Dye terminator (version 3.1, Applied Biosystems, Foster City, California) and ABI PRISM 3730 DNA Analyzer (Applied Biosystems). Variants identified in AF probands were evaluated in control subjects by sequencing or high-resolution melting, using SensiMix HRM (Quantace, London, United Kingdom) or Lightcycler 480 HRM Master (Roche Diagnostics, Mannheim, Germany) mastermix and a Lightcycler 480 Instrument (Roche Diagnostics).
Chinese hamster ovary (CHO) cells were transfected with wild-type (WT) or mutant K+ channel cDNA clones and currents were recorded using conventional patch-clamp techniques (see Supplementary Methods).
In silico modeling
Code for the Courtemanche atrial cell model (11) was downloaded from the CellML repository and converted into Matlab M-code using cellular open resource (COR) (12). The model was solved using the Matlab ode15s solver with a maximal time-step of 1 ms. Models were equilibrated for 10 s of simulated time at a pacing rate of 1 Hz, and the duration of the last AP (APD) was measured from the time of the peak to the time of 90% repolarization (APD90). The sensitivity of the model to changes in repolarizing K+ currents was estimated (13). The model was solved for 1,000 consecutive runs. In each run, the 5 K+ conductances (gK1, gto, gKr, gKs, and gKur) were individually scaled by a random number drawn from a log-normal distribution centered around a mean value of 1 (SD 22%). In each of 1,000 runs, the APD90 of the 10th AP was determined and saved with the corresponding set of 5 scaling factors. After 1,000 runs, matrices containing APD90 values and scaling factors were log-transformed, centered on their mean values, and normalized to their means. These values were used as inputs for the partial least squares function PLSREGRESS in the Matlab Statistics Toolbox. The output of this function is an array of correlation coefficients that gives the model sensitivity for each K+ current. To determine the impact of altered repolarization reserve, the sensitivity analysis was repeated while increasing or reducing all K+ current densities in 10% increments in the range 140% to 60%. All possible combinations of changes in each of the 5 K+ currents were simulated at 90%, 100%, and 110% of their respective original values (35 = 243 combinations). For each combination the model was solved 1,000 times for 10 s, and the mean (and SD) of APD90 were determined. Mean (and SD) values for APD at 60% repolarization (APD60), plateau potential between 20 ms and 80 ms after AP onset, and maximum slope during repolarization were also determined.
The proportions of common, uncommon, and rare variants in AF probands and control subjects were compared using chi-square analysis. The statistical significance of differences in the number of alleles between these 2 groups was determined using Student's unpaired t test. The robustness of comparisons of the number of alleles to normality assumptions required for t tests to be valid was assessed by also comparing AF probands with control subjects using the nonparametric Mann-Whitney U test. All p values <0.05 were considered statistically significant.
K+ channel gene variants identified
Coding regions of 13 K+ channel genes that comprise the α- and β-subunits of the Ito, IKur, IKr, IKs, and IK1 currents were resequenced in probands from 80 AF families. Nineteen nonsynonymous variants in 9 genes were found (Table 1). Each of these 19 variants was evaluated in 240 healthy control subjects and was classified according to the minor allele frequency (MAF) in the control population; 4 variants were common (MAF >10%), 4 were uncommon (MAF 1% to 10%), and 11 were rare (MAF <1%). Nearly all AF probands and control subjects had at least 1 common variant allele, with similar numbers of variant alleles in each group (Table 2). The distribution of uncommon variants was also similar in AF probands and controls. Of the 11 rare variants, 6 were novel and only seen in persons with AF. Rare variants were present in a higher proportion of AF probands than control subjects (18.8% vs. 4.2%, p < 0.001), and the mean number of rare variants was greater (0.21 vs. 0.04, p < 0.001). The difference in the number of alleles between groups was also significant when a nonparametric method was used.
Functional effects of single variants
Novel variants were characterized by whole-cell voltage clamp assays (Online Figs. 1 to 4, Online Table 1). The functional effects of novel and previously identified variants (14–27) are summarized in Table 3. Most variants had demonstrable electrophysiological effects, with 9 of the 19 variants also having experimentally validated or predicted effects on protein phosphorylation or protein-protein interactions.
Multiple K+ channel gene variants present in AF families
Sixty of the 80 AF probands (75%) had 2 or more variants. In families IF and HF, the novel variants G568V KCNA5 and E444K KCNH2 showed good segregation with disease amongst family members tested (Fig. 1A) and were absent from control subjects. The G568V KCNA5 variant had a gain-of-function effect on IKur (Online Fig. 1), whereas the E444K KCNH2 variant reduced IKr (Online Fig. 2). In genotype-positive subjects in both families, additional functional K+ channel variants were present. For other AF families, there was a maximum of 5 variants present (Fig. 1B).
Effects of multiple variants on atrial repolarization
The cumulative effects of ion channel variants are difficult to assess, and direct experimental evaluation of every different combination is impractical. Hence, to get a quantitative understanding of how multiple small changes in K+ currents could affect atrial repolarization, we used the Courtemanche atrial cell model (11). We first determined the sensitivity of APD to changes in the 5 K+ currents (Ito, IKur, IKr, IKs, IK1) (Online Fig. 5). This analysis provides a distribution of APD90 (Fig. 2A), which for our simulations was 302.3 ± 39.3 ms. This spread is comparable to variations observed in recordings from isolated atrial myocytes (28) that may result from different gene expression levels and/or disruption of normally tight electrotonic interactions between cells. The 2 most important determinants of APD90 are IK1 and IKr in the Courtemanche model, with sensitivity coefficients of 0.59 and 0.46, respectively, while changes in IKur have almost negligible effect (sensitivity coefficient 0.02) (Fig. 2B).
We extended the sensitivity analysis to investigate small random variations in K+ currents combined with single or multiple variants that each have a modest effect. Results for variations in IKr and IKur (as seen in family IF) are shown in Figures 2C to 2E. As predicted by the sensitivity coefficients in Figure 2B, changes in IKr had significant effects on mean APD90 (Fig. 2C) but minimal effect on dispersion of the APD90. Changing IKur by ± 20% results in minimal changes in mean APD90 (Fig. 2D) but does significantly alter APD90 dispersion. In family IF, the K+ variants are predicted to increase IKur (Online Fig. 1) and reduce IKr (Table 3). We modeled this scenario by increasing IKur by 20% and decreasing IKr by 20%. The resulting APD90 histogram demonstrates that these 2 changes in current densities have an additive effect but not in a simple linear manner (Fig. 2E).
We also calculated sensitivity coefficients for each of the K+ currents in the context of a single variant that has a modest effect. This enables predictions about a “second hit,” e.g., from an ion channel-blocking drug. When levels of IK1 or IKur were systematically varied, there were significant changes to the sensitivity coefficients for other currents compared to baseline (Online Figs. 6A and 6B). Conversely, when IKr or IKs was systematically varied, the sensitivity coefficients for other currents were not significantly altered (Online Figs. 6D and 6E).
Finally, we examined the functional consequences of multiple variations in baseline K+ current levels. This is the putative situation in families HF, HW, and BN where a number of variants were present, including some in accessory β-subunits that can influence the properties of several channels (29). We simulated all possible combinations of changes in K+ channel currents at 90%, 100%, and 110% of baseline (i.e., 35 = 243 possible combinations). In Figure 3A, the clustering of white and black IK1 and IKr segments apparent at the “short” and “long” ends of the circular diagram, respectively, clearly shows that these current components are the main determinants for APD90 in the Courtemanche model (Fig. 2B). A 10% reduction in all repolarizing currents did not increase APD90 to the greatest extent. Rather, a 10% reduction in IK1, IKr, and IKs and 10% increase in IKur and Ito was the combination that gave the longest APD90. This superficially counterintuitive result is a consequence of highly nonlinear relationships among the effects of variations in each of the K+ currents on APD90. In Figure 3B, where these variations have been sorted according to their effects on dispersion (i.e., SD of the APD90 histograms), it is apparent from clustering of black segments in Ito and IKur at the less variable end that these 2 currents are the main determinants of APD90 variability, as described by the SD of the APD90 histograms. Notably, in the middle regions of both Figs. 3A and 3B, there is no apparent clustering, indicating that a large number of different combinations can give a “normal” value. Furthermore, >54% of the 243 combinations give mean APD values that are within ±10 ms of the normal APD90 (302 ms), whereas only 4% give APD90 values >25 ms different from the normal APD90. This perturbation of APD90 is comparable to that observed for single variants with very large effect (e.g., 50% loss of function or 100% gain of function) (Online Fig. 5).
The effects of multiple K+ channel variations on other AP parameters, including APD60, the plateau potential, and the maximum slope of repolarization, were also evaluated. It was found that Ito and IKur are the major determinants of the plateau potential (Online Fig. 7C). Interestingly, combinations of high IKur/low Ito or low IKur/high Ito give intermediate plateau levels, but Ito mainly determines the plateau level variability (Online Fig. 7D). The maximal slope of repolarization appears to be mostly set by IKur, and to a lesser extend by IKr and IKs levels (Online Fig. 7E), with the SD of the repolarization slope showing no obvious patterns (Online Fig. 7F).
Application of the partial least squares sensitivity analysis to the scenario where multiple conductances are varied enables estimation of the phenotypic variation one could expect if 1 or more of the 5 variants were absent or an additional hit, such as drug-block, was added. For example, in the case of the group of 5 variants that gave the longest APD90 (decreased IK1, IKr, IKs, increased Ito, IKur) the order of sensitivity (ranked from most sensitive to least sensitive) was IK1, IKr, Ito, IKs, and IKur.
There has been intense recent interest in rare variants as a cause of “missing” heritability in complex disorders (1–3). Although individual rare variants might occur in relatively few cases, it has been suggested that different rare variants might collectively account for a substantial proportion of phenotypic traits in populations. For example, rare variants in genes related to lipid metabolism are more frequent in persons with extremes of cholesterol levels (3).
At least 70% of rare variants in gene coding sequences are estimated to have functional effects, with ∼20% being strongly deleterious and ∼50% mildly deleterious (2). In disorders with Mendelian patterns of inheritance, single rare variants of major effect size are regarded as sufficient to cause disease. The role of less deleterious rare variants in human diseases has been very little investigated. With the exponential increase in exome sequencing for human genetics studies, investigators are being confronted increasingly by this problem.
A surprisingly low prevalence of single K+ channel gene mutations has been found in cohorts of AF families (27,30). Only 1 variant, R14C KCNQ1, described in an earlier analysis of our first 50 AF probands (27), was clearly pathogenic due to its close correlation with affection status in the family, absence from a control population, and substantial functional effects. Two novel variants reported here, G568V KCNA5 and E444K KCNH2, had family segregation data that were consistent with pathogenicity (albeit limited in family HF) and were not detected in ethnically matched controls. For both variants, the in vitro electrophysiological effects appeared relatively modest, suggesting that the rare variants might not be the sole cause of AF in these families. The limitations of cellular assays need to be taken into account, however, as demonstrated recently in a mouse model of the D1275N SCN5A mutation in which there were minimal effects in vitro, but striking abnormalities of cardiac rhythm and contraction in vivo (31). In this study, we found that AF probands and controls subjects had similar prevalence of common and low frequency variants in the 13 K+ channel genes evaluated, but that there was an excess of rare variants in the AF probands. We hypothesized that rare K+ channel variants, in combination with common variants, cumulatively affect atrial repolarization properties and propensity for AF.
Combinations of variants may have additive, opposing, or synergistic actions, and the net effects of multiple variants are difficult to predict or demonstrate experimentally. Recent advances in in silico cardiomyocyte modeling have provided a powerful tool for evaluating simultaneous perturbations of multiple ion channels. For example, the Courtemanche atrial cell model has been used previously to simulate effects of single gene mutations (27), electrolyte changes (32), and ion channel changes associated with atrial remodeling (33) on atrial AP characteristics. We used this model to systematically analyze the consequences of all possible combinations of small variations in the 5 K+ channels, Ito, IKur, IKr, IKs, and IK1. Our modeling method enabled us to predict the nonlinear interactions between different K+ channel variants at baseline and with variants of small effect. Importantly, our data indicate that uniform decreases or increases in K+ current densities do not produce the most dramatic effects on repolarization. Moreover, small variations in multiple K+ conductances can produce significant prolongation or shortening of atrial repolarization, increased dispersion of repolarization, or changes in the AP shape that could facilitate AF.
A limitation of this study is that only cardiac K+ channel variants were evaluated. Although we modeled 243 possible combinations of K+ current changes, the specific combinations of greatest clinical relevance remain to be determined.
As clearly demonstrated in exome sequencing data, functionally deleterious variants in numerous sodium and calcium ion channels as well as other cardiomyocyte components are present in every person, and it is the net effect of all these variants together with acquired “environmental” factors that contribute to the electrophysiological properties of the atrial wall. Interpretation of human genome sequence data requires global perspectives that look beyond the single variant. Accordingly, integration of comprehensive genetic screening, cellular electrophysiological data, atrial cell modeling, and systems biology approaches will be instrumental in achieving these goals. Further delineation of genetic signatures for AF is an important step toward personalized approaches to risk stratification and therapeutic decision making.
The authors thank Olivia Baddeley, Haley Crotty, Jessica Hansen, Louise Lynagh, Zara Richmond, Nan Yang, and Kathryn North for assistance with clinical data and blood sample collection; Matthew Law for statistical advice; and the following physicians for referral of study probands: Ruth Arnold, Tim Carruthers, Suchitra Chandar, Richard Cranswick, Michael Feneley, Peter French, David Gray, Dean Guy, Deborah Hayes, Peter Hayes, Christopher Hayward, William Heddle, Pramesh Kovoor, Anne Keogh, Jo-Dee Lattimore, Jim Leitch, Alex Levendel, Harry Lowe, Peter Macdonald, Matthew Pincus, David Ross, Jonathan Silberberg, Suresh Singarayar, Charles Thorburn, John Uther, Paul West, and Thomas Yeoh.
For a supplemental Methods section, tables, and figures, please see the online version of the article.
This work was supported by the National Health and Medical Research Council of Australia, Canberra, Australian Capital Territory; the National Heart Foundation, Melbourne, Victoria; the Estate of the Late R.T. Hall, Sydney, New South Wales; the Roth Foundation, Sydney, New South Wales; the Sylvia and Charles Viertel Charitable Foundation, Melbourne, Victoria; and the St. Vincent's Clinic Foundation, Sydney, New South Wales, Australia. Dr. Sanders has served on the advisory board of Bard Electrophysiology, Biosense-Webster, Medtronic, St. Jude Medical, Sanofi-Aventis, and Merck; has received lecture fees from Biosense-Webster, St. Jude Medical, and Merck; and has received research funding from Biosense-Webster, Boston-Scientific, Biotronik, Medtronic, St. Jude Medical, and Merck. Dr. Vandenberg has received consultancy fees from Lundbeck Australia Pty. Ltd. All other authors have reported they have no relationships relevant to the contents of this paper to disclose. Drs. Mann, Otway, and Guo contributed equally to this work.
- Abbreviations and Acronyms
- atrial fibrillation
- action potential
- action potential duration
- Chinese hamster ovary
- deoxyribonucleic acid
- minor allele frequency
- Received August 25, 2011.
- Revision received October 13, 2011.
- Accepted November 1, 2011.
- American College of Cardiology Foundation
- Cohen J.C.,
- Kiss R.S.,
- Pertsemlidis A.,
- Marcel Y.L.,
- McPherson R.,
- Hobbs H.H.
- Fatkin D.,
- Otway R.,
- Vandenberg J.I.
- Courtemanche M.,
- Ramirez R.J.,
- Nattel S.
- Obenauer J.C.,
- Cantley L.C.,
- Yaffe M.B.
- Ehrlich J.R.,
- Zicha S.,
- Coutu P.,
- Hebert T.E.,
- Nattel S.
- Bezzina C.R.,
- Verkerk A.O.,
- Busjahn A.,
- et al.
- Anson B.D.,
- Ackerman M.J.,
- Tester D.J.,
- et al.
- Gentile S.,
- Martin N.,
- Scappini E.,
- Williams J.,
- Erxleben C.,
- Armstrong D.L.
- Westenskow P.,
- Splawski I.,
- Timothy K.W.,
- Keating M.T.,
- Sanguinetti M.C.
- Abbott G.W.,
- Goldstein S.A.N.
- Otway R.,
- Vandenberg J.I.,
- Guo G.,
- et al.
- Feng J.,
- Yue L.,
- Wang Z.,
- Nattel S.
- Ellinor P.T.,
- Moore R.K.,
- Patton K.K.,
- Ruskin J.N.,
- Pollak M.R.,
- MacRae C.A.
- Watanabe H.,
- Yang T.,
- Stroud D.M.,
- et al.
- Severi S.,
- Pogliani D.,
- Fantini G.,
- et al.