A Ubiquitous Splice Variant and a Common Polymorphism Affect Heterologous Expression of Recombinant Human SCN5A Heart Sodium Channels
Amino acid sequence variations in SCN5A are known to affect function of wild-type channels and also those with coexisting mutations; therefore, it is important to know the exact sequence and function of channels most commonly present in human myocardium. SCN5A was analyzed in control panels of human alleles, demonstrating that the existing clones (hH1, hH1a, hH1b) each contained a rare variant and thus none represented the common sequence. Confirming prior work, the H558R polymorphism was present in ≈30% of subjects. Quantitative mRNA analysis from human hearts showed that a shorter 2015 amino acid splice variant lacking glutamine at position 1077 (Q1077del) made up 65% of the transcript in every heart examined. Age, sex, race, or structural heart disease did not affect this proportion of Q1077del. Estimated population frequencies for the four common variants were 25% SCN5A, 10% [H558R], 45% [Q1077del], and 20% [H558R;Q1077del], where the reference sequence SCN5A is GenBank AC137587. When expressed in HEK-293 cells, these common variants had a more positive mid-point of the voltage dependence of inactivation than the standard clone hH1. Also, channels containing Q1077 expressed smaller currents. When H558R was present with Q1077 ([H558R]), current expression was profoundly reduced despite normal trafficking to the cell surface. Thus, four variant sequences for SCN5A are commonly present in human myocardium and they exhibit functional differences among themselves and with the previous standard clone. These results have implications for the choice of background sequence for experiments with heterologous expression systems, and possibly implications for electrophysiological function in vivo.
SCN5A encodes the voltage-dependent sodium channel α-subunit protein SCN5A, also called hNav1.5,1 found predominantly in human heart muscle. This channel is responsible for large peak inward sodium current (INa) that underlies excitability and conduction in working myocardium (atrial and ventricular cells) and special conduction tissue (Purkinje cells and others), and also for late INa that influences repolarization and refractoriness. Three complete cDNA clones for this channel hH1,2 hH1a,3 and hH1b4 differ in amino acid sequence in 5 of the 2016 positions (Table 1). In addition, these three clones differ from the deduced amino acid sequence for SCN5A obtained from the two human genome databases: Celera and the International Human Genome Sequencing Collaboration (IHGSC). Before the present study, it was not clear whether or not these differences are present in human population as common variants. From previous studies, we know that dramatic differences in current expression can be found when arrhythmia mutations are expressed in different background clones.4 This study was designed to answer the questions: What is the common background sequence for SCN5A? Do common variations affect channel function? Does it matter which Na+ channel clone (ie, background sequence) is used for functional studies of wild-type and mutated channels?
Materials and Methods
Protocols used in this investigation are more fully described in the expanded Materials and Methods section in the online data supplement available at http://www.circresaha.org.
Genotyping at Positions 558, 559, 618, and 1027
Allele frequencies for the SCN5A variants were established by direct genomic DNA sequencing of 400 reference alleles derived from the 100 Caucasian human variation panel (Coriell Cell Repositories) and the 100 African-American human variation panel (National Institute of General Medical Sciences). Protein-encoding sequences harboring these variants were amplified by polymerase chain reaction (PCR). PCR products were enzymatically treated to remove unincorporated dNTP and primers and subsequently sequenced. The resulting chromatograms were analyzed for the specific variant.
Identification, Characterization, and Quantification of Alternatively Spliced SCN5A at Position 1077
Direct DNA sequencing was performed on exon 17/18–targeted RT-PCR–generated products of messenger RNA extracted from myocardial tissue. The tissue came from autopsy specimens from sudden infant death syndrome (n=5), nonaccidental infant death (n=5), or accidental adult death victim (n=5). Additional tissue came from surgical myectomy specimens in adults with hypertrophic cardiomyopathy (n=5). The total test group contained 9 males and 11 females, and the infant group contained 5 white and 5 black subjects. Total RNA from heart tissue was isolated and first-strand cDNA synthesis was performed in triplicate on total RNA. PCR was performed, and products were purified and sequenced. The relative quantity of the two alternatively spliced transcripts were quantified. Triplicate cDNA samples from each case were subject to PCR amplification using the same PCR primers and conditions. Control templates representing homozygous Q1077 and Q1077del transcripts were synthesized and PAGE purified. The resulting radionucleotide incorporated PCR products representing (1) the transcript (135 base pairs) encoding the inclusion of glutamine at position 1077 (Q1077) or (2) the transcript (132 base pairs) that does not encode for glutamine at position 1077 (Q1077del) were separated by denaturing gel electrophoresis on a polyacrylamide/urea gel. Autoradiography was used for signal quantification of the two alternatively spliced transcripts.
Gene Expression, Mutagenesis, and Nomenclature
The SCN5A clone hH1 (GenBank M77235)2 was kindly provided by Dr Al George in prcCMV (Invitrogen). The hH1c construct (GenBank AY148488) was made from the hH1b clone (GenBank AF482988)4 by mutating the arginine (R) at 558 to histidine (H), and the isoleucine (I) at positions 618 to leucine (L). The consensus reference sequence at this writing (GenBank NM_000335 August 2003) has 2015 amino acids. In this article, we used the full-length 2016 amino acid deduced sequence from the IHGSC (GenBank AC137587, deposited April 2003) as the reference sequence for SCN5A in order to preserve the standard numbering system. Single amino acid variations are referred to by the amino acid substitution without brackets. Channel variants of the full sequence are denoted by the amino acid substitutions from the reference sequence separated by semicolons and contained in brackets as recommended by den Dunnen and Antonarakis.5 The channels SCN5A, [H558R], [H558R;Q1077del], [H558R;Q1077del;M1766L], and [Q1077del;M1766L] were made by mutagenesis at appropriate positions using a PCR-based site-directed mutagenesis kit. Transformed bacteria were grown and the DNA isolated and purified with spin-column preps. All constructs were sequenced to verify incorporation of the intended amino acid change and to confirm that no unwanted changes were introduced. These constructs were placed in an expression vector and transfected into HEK-293 cells. Cells 24 hours after transfection were used to measure macroscopic INa. Experiments were done with transient transfection unless otherwise noted. In some instances, stable colonies were selected for by addition of the G418 antibiotic. After stabilization, RNA was isolated (RNAisol from LPS) and screened by RT-PCR analysis. Colonies that tested RT-PCR–positive were then analyzed by voltage clamp.
The whole-cell patch-clamp technique was utilized to measure macroscopic INa.6 The voltage-clamp protocols are described briefly with the data and have been published previously in detail.7 Peak INa and late INa were obtained after passive leak subtraction and saxitoxin subtraction as described previously.6,7⇓ Parameter fits were obtained and one-way ANOVA with Bonferroni correction was performed to determine statistical significance among 3 or more groups of mean data. Statistical significance was determined by a value P<0.05.
The FLAG epitope was introduced between S1 and S2 in domain I for channels used in the immunocytochemistry experiments with confocal microscopy as previously described.4 The rabbit anti-Calnexin IgG was used as an endoplasmic reticulum (ER) marker to test for colocalization and was applied to transfected cells immediately after incubation with fluorescein-conjugated goat anti-mouse secondary antibody.
Sequence Comparisons for SCN5A Clones and Genotyping of Control Panels
The sequences of hH1,2 hH1a,3 and hH1b4 were compared and the amino acid differences at 5 positions are shown in Table 1, along with the GenBank Accession number where available. In a previous study of the hH1b clone,4 hH1 and hH1a were resequenced and the differences between these two clones were found to be only the 3 positions included in Table 1 rather than at the 9 positions reported previously.3
We investigated allelic frequencies from genomic DNA at the 5 positions in question from a panel of 200 human controls (100 white subjects and 100 black subjects). The most common amino acids at these 5 positions are reported as hH1c in Table 1. For positions 559, 618, and 1027, 100% of the 400 reference alleles possessed T559, L618, and R1027, indicating that each of the existing clones contains a rare variant and that none represents the common sequence (Table 1). A search of the Celera human genome database showed that the deduced amino acid sequence agreed with hH1c. The deduced amino acid sequence in the NIH International Human Genome Sequencing Consortium (IHGSC) database also agreed with the hH1c sequence except for the additional presence of glutamine (Q) at position 1077. This glutamine is present in the hH1 clone but not hH1a or hH1b.
A Common Polymorphism, H558R
The residue at position 558 hosts the common polymorphism involving a substitution of histidine (H) with arginine (R) H558R8 with a reported frequency in the population of 19% to 24%.9 We confirm that H558R is a common polymorphism among both blacks (53% homozygous for H, 40% heterozygous, and 7% homozygous for R) and whites (65%/30%/5%). The apparent higher frequency of the R-encoding allele in blacks (27% versus 20%) is not statistically significant (paired student t test, Fischer’s exact). Other less common variations also exist in the human population. One of the more common variants is the S1103Y (also annotated Y1102) variant with a reported heterozygous frequency of 13.2% among blacks.10 We found approximately the same frequency among blacks in our panel. Otherwise, no variations exceeding 5% were found in our panels.
A Common Alternatively Spliced Variant, Q1077del
The hH1 clone contained a glutamine (Q) present at both amino acid positions 1076 (the final codon of exon 17) and also at 1077 (the first codon of exon 18).2 The acceptor site sequence for exon 18 was annotated as ggggtcttttcagCAGGAATCC, where the lowercase letters represent the intronic sequences and the uppercase letters represent the 5′ exonic sequences of exon 18 (see Figure 1). The bold lowercase letters indicate the predicted “ag” acceptor site rule for splice site recognition. However, splice-site analytical tools indicate that the CAG following the ag shown above could also comprise the terminal intronic sequence and may be the preferred acceptor site for splicing, resulting in a deletion of glutamine at position 1077 (Q1077del). Does SCN5A most commonly have a glutamine (Q) at both amino acid positions 1076 and 1077, as in hH1, or only at 1076 as in hH1a and hH1b?
We sequenced RT-PCR products derived from mRNA isolated from human left ventricular myocardial specimens and generated with primers targeting exons 17 and 18. The specimens came from infants with sudden infant death syndrome (SIDS), infants with a structurally normal heart, adults with hypertrophic cardiomyopathy, and adults with a structurally normal heart. All subjects contained transcript with a “cag” in-frame insertion indicating the universal presence of alternative splicing involving this acceptor site (Figure 1). The relative abundance for each alternatively spliced transcript was quantified by autoradiography and phosphoimaging (Figure 2). Examples of RT-PCR products generated from myocardial RNA (Figure 2A) show the expected size products for control experiments with the 135-bp template containing the extra cag codon (Q1077) and for the 132-bp template lacking the extra cag codon (Q1077del). Summary data (Figure 2B) show that Q1077del was the preferred alternatively spliced variant being significantly more abundant in every group tested. The proportions were not affected by age or by the presence of SIDS or ventricular hypertrophy (Figure 2) or by sex or race (data not shown). Overall, the proportion of alternatively spliced variant containing Q1077 was 35±2.0% (range 31% to 38%, n=20) and the Q1077del transcript was 65±2.0% (range 62% to 69%, n=20). In addition, the proportion of Q1077del was the same whether RNA was obtained from right atrium, left atrium, right ventricle, or left ventricle (data not shown).
Four Common SCN5A Variants in the Human Population
Based on the predicted amino acid frequency estimates from our genomic sequencing in controls and our measurement of the frequency of the splice variant Q1077del, we estimated the population frequency of the existing clones and other full-length sequences (Table 1). These estimates assume independence between the probability of the Q1077del splice variant (65%) and the genomic variant containing H558R (30%). Note that the three clones hH1, hH1a, and hH1b used in previous studies are estimated to have a 0% frequency in the population because of the presence of a mutation in each that was not seen in our panel. The most common variant, [Q1077del] at 45%, is identical to the hH1c sequence, the Celera sequence, and also to the NCBI Reference Sequence for SCN5A (Accession no. NM_000335). The full-length sequence SCN5A is actually less frequent than [Q1077del] at 25%. Only slightly less common are the variants with the H558R polymorphism designated as [H558R;Q1077del] at 20% and [H558R] at 10% (not in Table 1).
Functional Characterization of SCN5A Variants
Constructs for expressing SCN5A and the other common variants, [H558R], [Q1077del], and [H558R;Q1077del], were made and transfected individually into HEK-293 cells for voltage-clamp study. For comparison, we also expressed hH1 contemporaneously under the same conditions with the four variants because most previous studies in the literature have used this clone. Representative currents for the four variants and hH1 are shown in Figure 3A. For those variants that expressed robust current, no significant differences were identified in the parameters of activation, current decay, or the late currents (Table 2). The midpoint of inactivation for hH1, however, was significantly negative by 7 mV to the variants lacking Q1077 ([Q1077del] and [H558R;Q1077del], Table 2). Consistent with this negative shift in inactivation, the recovery for hH1 was also significantly slower (Table 2).
The most dramatic finding was that [H558R], which contains Q1077, expressed very low current density (summary data in Figure 3B). Other constructs that contained Q1077 (hH1 and SCN5A) also tended to have reduced currents, but the difference did not reach statistical significance. RT-PCR of mRNA from cells that were stably transfected with [H558R] showed abundant SCN5A transcript, but just as with the transient transfections little or no current was observed. The β1 subunit has been shown to increase wild-type current density when coexpressed with the hH1 α subunit11 and also to increase expression of a trafficking defective SCN5A mutant.12 When [H558R] was coexpressed with the β1 subunit, increased currents were observed, although not at the levels of the other α subunits expressed alone (Figure 3). To determine if the small currents seen with [H558R] could be endogenous currents, we voltage-clamped nontransfected HEK-293 cells. In 8 cells, only 1 cell had current (100 pA) for an average density of <0.5 pA/pF compared with 7.5 pA/pF in [H558R] alone and 14 pA/pF in [H558R] cotransfected with the β1 subunit. We were unable to significantly increase endogenous currents in untransfected cells by transfection with β1 (1.0±0.5 pA/pF n=9), incubation with 100 μmol/L mexiletine for 24 hours (1.1±0.4 pA/pF n=10), or incubation at 27°C (0.9±0.4 pA/pF n=8). We conclude that [H558R] did indeed generate small currents. As SCN5A is a monomeric channel, a dominant-negative suppression mechanism would not be expected, but to test of the possibility that the two variants might interact both variants: [H558R] and [H558R;Q1077del] were coexpressed (plasmid DNA 0.75 μg each) and normal current densities were observed (data not shown). Hyperpolarization of the holding potential to −200 mV for 1 second did not elicit current, suggesting the decreased density is not caused by a large shift in the voltage dependence of fast inactivation.
Cell Trafficking of the [H558R] Variant
Some SCN5A variants with decreased current density have been shown by immunocytochemistry to have defective trafficking4,13⇓ and do not make it to the cell surface. To determine if [H558R] trafficked to the cell surface, we labeled the channel variants with a FLAG epitope and localized the channels by immunofluorescence and confocal microscopy. In each panel of Figure 4, a light microscopy image is shown allowing for identification of the nucleus and the cell surface followed by the confocal immunofluorescence image(s). A nontransfected cell gave no fluorescent signal, as expected (Figure 4A). The [Q1077del] variant that gave robust current densities showed a rim of fluorescence at the cell surface, as expected for a normally trafficking channel (Figure 4B). The [H558R] variant expressed almost no current, but also showed fluorescence at the cell surface (Figure 4C), suggesting that the lack of current was not caused by a trafficking defect.
As a positive nontrafficking control for our experiment with the [H558R] variant, we overlaid the mutation M1766L into the variants and show data from two of these experiments (Figures 4D and 4E). We had previously shown M1766L to be trafficking defective in hH1a [T559A;Q1077del] but to be normally trafficking in hH1b [H558R;L618I;Q1077del].4 When the M1766L mutation was put into [Q1077del] to make [Q1077del;M1766L] (Figure 4D), the fluorescence was restricted to the area around the nucleus without any labeling in the periphery, consistent with a trafficking defect. This eliminates the possibility that the rare variant T559A in hH1a was responsible for the trafficking defect previously described.4 An image with a marker for the endoplasmic reticulum is shown in the third panel of Figure 4D, and the fourth image in Figure 4D superimposes the second and third image to show colocalization of the channel with the endoplasmic reticulum marker. When M1766L was placed in the [H558R;Q1077del] background, the channel localizes to the cell periphery (Figure 4E), consistent with previous data and also showing that the rare variant L618I in hH1b was not responsible for “rescuing” the trafficking defect.
We have identified, constructed, and characterized the function of four SCN5A sequences: SCN5A, [H558R], [Q1077del], and [H558R;Q1077del]. These the channels most common in human hearts occur because of the frequently occurring polymorphism H558R and a ubiquitous alternatively spliced variant Q1077del. The splice variant transcript encoding Q1077del was found in every individual and in remarkably similar increased relative abundance, suggesting preferential splicing or better stability. Therefore, all persons generate SCN5A channels containing either 2015 amino acids (65% of transcripts) or 2016 amino acids (35% of transcripts) continually. These splice variants account for the difference in length between hH1 (2016 amino acids) and hH1a/hH1b (2015 amino acids).
The precise frequency and density of the four common variants in humans is not known, but we made population estimates assuming independence of the H558R and Q1077del variations at 30% and 65%, respectively. For individuals, however, the frequencies are different. The approximately 65% of individuals homozygous for H558 would be expected to have two variants: SCN5A 35% and [Q1077del] 65%. The 30% of individuals heterozygous for H558R would have all four variants: SCN5A at 18%, [Q1077del] at 32%, [H558R] at 18%, and [H558R;Q1077del] at 32%. Finally, the 5% of individuals homozygous for R558 would be expected to have approximately 35% of the [H558R] transcript with profoundly reduced current density (Figure 3).
None of the existing clones (hH1, hH1a, nor hH1b) used in previous studies represents a common sequence for SCN5A because each contains a rare variant. Most of the voltage-dependent kinetic parameters we measured for the variants and hH1 were not different (Table 2), suggesting that previous studies using other sequences likely remain valid indicators of human cardiac Na+ current function, although in this study hH1 had a more negative midpoint of inactivation. We cannot exclude more subtle kinetic effects that could have physiological importance. In addition, the function of inserted arrhythmia mutations may be different depending on the channel background sequence. For example, the Brugada syndrome mutation T1620M showed a trafficking defect only with the relatively uncommon variant R1232W in the background.13 In another example, the common polymorphism H558R was shown to affect the kinetics of the conduction disease mutation T512I.14 Also, H558R affected trafficking of the M1766L arrhythmia mutation as shown previously4 and confirmed in this study to also occur in the most common background sequence [Q1077del]. We studied the LQT3 mutation KPQ1505-1507del (ΔKPQ) in [H558R] and [H558R;Q1077del] and found the late current was present in these variants (data not shown) as described previously.6,15⇓ The rare variants found in hH1, hH1a, and hH1b are not generally thought to affect function, but the rare variation R1027Q in hH1 might account for the more negative midpoint of inactivation (Table 2).
Based on our results, individuals carrying the R558 allele may have a reduced cardiac INa density in those 35% of channels with Q1077 present. What are the consequences of reduced INa? Although the safety factor for conduction is generally high (that is, a large decrease in Na+ current must occur before conduction fails), a transgenic mouse model heterozygous for defective SCN5A had arrhythmia.16 Decreased INa underlies Brugada syndrome and conduction system disorders17; therefore, H558R might be expected to increase mortality. Allelic frequencies for H558R, however, appear to be in Hardy-Weinberg equilibrium, at least among the 100 blacks and 100 whites studied, suggesting no survival disadvantage for individuals harboring R558. This cohort of reference alleles, however, is not representative of a population-based sampling; more careful study could uncover increased mortality with H558R. Also, Q1077 is the less-abundant transcript. Individuals heterozygous for H558R might have INa densities reduced by only 17.5%, and R558 homozygotes might have current densities reduced by only 35%, less than the 50% reductions in the transgenic mouse model16 and Brugada syndrome, so the effect on mortality might be less apparent. If these in vitro findings are translatable, then carefully designed genetic epidemiology studies should demonstrate a difference. Perhaps a Brugada-like ECG or early repolarization pattern might be found among the estimated 5% of whites homozygous for R558. Interestingly, the prevalence of a Brugada-like ECG has been estimated to be as high as 2.5% in males.18 Perhaps [H558R] with Q1077 accounts for this. Finally, control and feedback elements for channel expression may exist, and actual current density in vivo may not follow these simple assumptions from experiments in heterologous expression systems.
The molecular/biophysical mechanism for the profound decrease in INa density observed in the setting of a sodium channel that harbors both R558 and Q1077 appears to be that of a nonfunctioning or poorly functioning channel. Transcript levels are abundant in both transient and stable transfections, and the channel appears at the cell membrane. The mechanism for the dysfunction, however, remains unknown. The amplitudes of the elicited INa do not allow for detailed kinetic analysis, but the small currents appear to be normal.
Whether or not the heterologous expression system for voltage-clamp studies faithfully recapitulates function of SCN5A (both expression and kinetic behavior) in the human heart is a constraint of this study and limits the confidence with which they can be applied to the clinical situation. Nonetheless, these systems are widely used to study SCN5A variants and have provided nearly all of the data to date that has been used to provide insight into how channel structure might cause arrhythmia. Even if these findings do not have direct relevance to human disease, the effect of background sequence on current expression levels has important implications for interpretation of functional studies using recombinant DNA and heterologous methods to study channel variants.
Finally, a few comments on nomenclature and numbering of the sodium channel sequence may be helpful. Although the most common variant [Q1077del] has 2015 amino acids, the 2016 amino acid sequence (AC137587) should probably be preferred as the reference sequence rather than the 2015 amino acid sequence (NM_000335). This retains the original numbering system of hH1 with 2016 amino acids and also maintains consistency with the numbering used in the majority of the literature. Alternatively the more common 2015 amino acid sequence could be used, but we suggest then that the less common variant be called [Q1077ins] and the next position termed E1078 rather than E1077, and so on (Figure 1). Thus, the recently described polymorphism in blacks that appears to predispose to ventricular ectopy would be referred to as S1103Y19 rather than “Y1102.”10
Our studies show that the background clone used in expression studies may affect function in important ways. All such studies should clearly state the background sequence of the construct used, preferably with the GenBank accession number of the reference sequence. It also may be prudent when studying arrhythmia mutations to characterize the electrophysiological phenotype using the common background channel sequences. This will always include at least two channels, with and without Q1077. Ideally the patient’s own background should be used with particular consideration for the presence of the most common polymorphism H558R.
This study was supported by an American Heart Association Northland Affiliate Predoctoral Fellowship to B.Y., a Doris Duke Clinical Scientist Development Award, Mayo Foundation Clinical Research Award, NIH HD42569 to M.J.A., and NHLBI HL66378 and HL71092 to J.C.M.
↵*Both authors contributed equally to this study.
Original received January 16, 2003; resubmission received June 17, 2003; revised resubmission received September 9, 2003; accepted September 10, 2003.
- ↵Goldin AL. Evolution of voltage-gated Na+ channels. J Exp Biol. 2002; 205: 575–584.
- ↵Gellens ME, George AL Jr, Chen LQ, Chahine M, Horn R, Barchi RL, Kallen RG. Primary structure and functional expression of the human cardiac tetrodotoxin-insensitive voltage-dependent sodium channel. Proc Natl Acad Sci U S A. 1992; 89: 554–558.
- ↵Hartmann HA, Tiedemann AA, Chen S-F, Brown AM, Kirsch GE. Effects of III-IV linker mutations on human heart Na+ channel inactivation gating. Circ Res. 1994; 75: 114–122.
- ↵Ye B, Valdivia CR, Ackerman MJ, Makielski JC. A common human SCN5A polymorphism modifies expression of an arrhythmia causing mutation. Physiol Genomics. 2003; 12: 187–193.
- ↵Yang P, Kanki H, Drolet B, Yang T, Wei J, Viswanathan PC, Hohnloser SH, Shimizu W, Schwartz PJ, Stanton M, Murray KT, Norris K, George AL Jr, Roden DM. Allelic variants in long-QT disease genes in patients with drug-associated torsades de pointes. Circulation. 2002; 105: 1943–1948.
- ↵Splawski I, Timothy KW, Tateyama M, Clancy CE, Malhotra A, Beggs AH, Cappuccio FP, Sagnella GA, Kass RS, Keating MT. Variant of SCN5A sodium channel implicated in risk of cardiac arrhythmia. Science. 2002; 297: 1333–1336.
- ↵Nuss HB, Chiamvimonvat N, Perez-Garcia MT, Tomaselli GF, Marban E. Functional association of the β1 subunit with human cardiac (hH1) and rat skeletal muscle (μ1) sodium channel a subunits expressed in Xenopus oocytes. J Gen Physiol. 1995; 106: 1171–1191.
- ↵Valdivia CR, Ackerman MJ, Tester DA, Wada T, McCormack J, Ye B, Makielski JC. A novel SCN5A arrhythmia mutation, M1766L, with expression defect rescued by mexiletine. Cardiovasc Res. 2002; 54: 624–629.
- ↵Wang DW, Yazawa K, George AL Jr, Bennett PB. Characterization of human cardiac Na+ channel mutations in the congenital long QT syndrome. Proc Natl Acad Sci U S A. 1996; 93: 13200–13205.
- ↵Papadatos GA, Wallerstein PM, Head CE, Ratcliff R, Brady PA, Benndorf K, Saumarez RC, Trezise AE, Huang CL, Vandenberg JI, Colledge WH, Grace AA. Slowed conduction and ventricular tachycardia after targeted disruption of the cardiac sodium channel gene Scn5a. Proc Natl Acad Sci U S A. 2002; 99: 6210–6215.
- ↵Chen S, Chung MK, Martin D, Rozich R, Tchou PJ, Wang Q. SNP S1103Y in the cardiac sodium channel gene SCN5A is associated with cardiac arrhythmias and sudden death in a white family. J Med Genet. 2002; 39: 913–915.