Modulation of KCNQ1 alternative splicing regulates cardiacIKs and action potential repolarizationHsiang-Chun Lee, MD, Msc,*†‡§ Yoram Rudy, PhD, FHRS,* Po-Yuan, PhD,║ Sheng-Hsiung Sheu, MD,†‡§

Jan-Gowth Chang, MD,¶ Jianmin Cui, PhD*

From the *Department of Biomedical Engineering, Cardiac Bioelectricity and Arrhythmia Center, Washington University inSt. Louis, St Louis, Missouri, †Division of Cardiology, Department of Internal Medicine, Kaohsiung Medical UniversityHospital, Kaohsiung, Taiwan, ‡Graduate Institute of Medicine, School of Medicine, Kaohsiung Medical University, Kaohsiung,Taiwan, §Department of Internal Medicine, Faculty of Medicine, School of Medicine, Kaohsiung Medical University,Kaohsiung, Taiwan, ║College of Electrical Engineering and Computer Science, National Cheng-Kung University, Tainan,Taiwan, and ¶Epigenome Research Center, Department of Laboratory Medicine, China Medical University Hospital, Taichung,Taiwan, School of Medicine, China Medical University, Taichung, Taiwan.

BACKGROUND Slow delayed-rectifier potassium current (IKs)channels, made of the pore-forming KCNQ1 and auxiliary KCNE1subunits, play a key role in determining action potential duration(APD) in cardiac myocytes. The consequences of drug-inducedKCNQ1 splice alteration remain unknown.

OBJECTIVE To study the modulation of KCNQ1 alternative splicingby amiloride and the consequent changes in IKs and actionpotentials (APs) in ventricular myocytes.

METHODS Canine endocardial, midmyocardial, and epicardial ven-tricular myocytes were isolated. Levels of KCNQ1a and KCNQ1b aswell as a series of splicing factors were quantified by using thereverse transcriptase-polymerase chain reaction and Western blot.The effect of amiloride-induced changes in the KCNQ1b/totalKCNQ1 ratio on AP was measured by using whole-cell patch clampwith and without isoproterenol.

RESULTS With 50 μmol/L of amiloride for 6 hours, KCNQ1a attranscriptional and translational levels increased in midmyocar-dial myocytes but decreased in endo- and epicardial myocytes.Likewise, changes in splicing factors in midmyocardial wereopposite to that in endo- and epicardial myocytes. In midmyo-cardial myocytes amiloride shortened APD and decreasedisoproterenol-induced early afterdepolarizations significantly.

This work was supported by the Taiwan National Science Council grantNSC 99-2221-E-037-001 (to Dr Lee) and American Heart AssociationEstablished Investigator Award 0440066N and NIH grant R01 NS060706(to Dr Cui), NIH-NHLBI grants R01-HL-049054-20 and R01-HL-033343-28, and a Leducq Foundation Award (to Y. Rudy). Address reprintrequests and correspondence: Dr Jianmin Cui, Department of BiomedicalEngineering, Cardiac Bioelectricity and Arrhythmia Center, WashingtonUniversity in St Louis, 290C Whitaker Hall, One Brookings Dr, St Louis,MO 63130-4899. E-mail address: jcui@wustl.edu.

1547-5271/$-see front matter B 2013 Heart Rhythm Society. All rights reserved.

The same amiloride-induced effects were demonstrated byusing human ventricular myocyte model for AP simulationsunder beta-adrenergic stimulation. Moreover, amiloride re-duced the transmural dispersion of repolarization in pseudo-electrocardiogram.

CONCLUSIONS Amiloride regulates IKs and APs with transmuraldifferences and reduces arrhythmogenicity through the modulationof KCNQ1 splicing. We suggested that the modulation of KCNQ1splicing may help prevent arrhythmia.

KEYWORDS Action potential; Alternative splicing; Amiloride; IKs;KCNQ1

ABBREVIATIONS AP ¼ action potential; APD ¼ action potentialduration; CL ¼ cycle length; EAD ¼ early afterdepolarization;ECG ¼ electrocardiogram; Endo ¼ endocardial; Epi ¼ epicardial;ICaL ¼ L-type calcium current; IKs ¼ slow delayed-rectifier potassiumcurrent; ISO ¼ isoproterenol; Mid ¼ midmyocardial; ORd ¼ O’Hara-Rudy; PKA ¼ protein kinase A; RT-PCR ¼ reverse transcriptase-polymerase chain reaction; TDR ¼ transmural dispersion ofrepolarization

(Heart Rhythm 2013;10:1220–1228) I 2013 Heart Rhythm Society. Allrights reserved.

IntroductionVentricular repolarization is significantly dependent on slowdelayed-rectifier potassium current (IKs), and the

contribution of IKs to action potential (AP) repolarization isenhanced during beta-adrenergic stimulation.1–4 Thisenhancement is also reflected in the phenotype of most longQT syndrome type 1, which is caused by KCNQ1 (α-subunitof IKs) mutations, with ventricular arrhythmias mostlyoccurring under physical or emotional stress. Phosphoryla-tion by protein kinase A (PKA) following beta-adrenergicreceptor stimulation greatly enhances IKs by increasingcurrent amplitude, by leftward shift in the current-voltagecurve, and by changing activation and deactivationkinetics.2,4 In large mammals, IKs also constitutes a “repola-rization reserve” that compensates for compromised other

http://dx.doi.org/10.1016/j.hrthm.2013.04.014

1221Lee et al Alternative Splicing and Cardiac IKs Currents

repolarization currents, in particular rapid delayed-rectifierpotassium current.3,4

A weaker IKs contributes partially to the longer AP inmidmyocardial (Mid) myocytes, and a prominent IKs enablesendocardial (Endo) and epicardial (Epi) myocytes to be moreresistant to early afterdepolarization (EAD) activity than Midmyocytes.5,6 In the human and the canine heart, there are 2splicing variants of KCNQ1. One is full-length, KCNQ1a, andthe other is N-terminal truncated, KCNQ1b.7,8 KCNQ1b exertsa dominant negative effect owing to a trafficking defect, andKCNQ1b transgenic mice manifest with pronounced QTprolongation and ventricular arrhythmias.7–9 The differentialexpression of KCNQ1 splicing variants has been suggested tobe responsible for differential IKs amplitude and also for APheterogeneity across the ventricular wall.9 However, whethersplicing of KCNQ1 can be altered and how such alterationsaffect cardiac arrhythmogenicity remain unexplored.

Amiloride, a potassium-sparing diuretic for the treatment ofhypertension and congestive heart failure since 1967, hasmultiple pharmacological actions. Our previous studiesshowed that splicing site selection in several genes can bealtered by amiloride.10,11 Prolonged treatment with amiloridealso exerts an antiarrhythmic effect both in postinfarction dogsand in patients with ventricular tachycardia.12,13 However,cellular mechanisms have not been determined.

Here, we study the electrophysiological effects of ami-loride on ventricular myocytes, in terms of its effect onKCNQ1 splicing variants expression and IKs densities, aswell as investigate the effect of KCNQ1 splicing on theaction potential duration (APD), isoproterenol (ISO)-induced EAD, and cardiac transmural dispersion of repola-rization (TDR). The mechanism of how amiloride affectsarrhythmogenicity is also addressed.

MethodsAn expanded, detailed methods section is provided in theOnline Supplemental Material.

ElectrophysiologyThe Xenopus oocytes expressed with human KCNQ1 andKCNE1 were used to examine the acute effect of amilorideand the dominant negative effect of KCNQ1b on currents(see Figure 2). Myocyte whole-cell patch clamp wasperformed at 36.51C with Tyrode’s solution (see Figures 3and 4). We recorded AP after 6 hours of incubation time withor without 50 μmol/L of amiloride. Myocytes with APDsafter stabilization outside the 250–450 ms range were notincluded for analysis.

BiochemistryTo study the effect of amiloride on KCNQ1 splicing, freshlyisolated canine myocytes were incubated in either M199culture medium (Sigma, St Louis, MO) or M199 with50 μmol/L of amiloride (Sigma) at 371C in a humidified5% CO2 incubator. The amiloride concentration of 50 μmol/Lwas according to the plasma level in clinical use.14 After

6 hours, cells were resuspended, washed, and homogenized inTrizol (Invitrogen, Grand Island, NY) for RNA extraction orin protein lysis buffer for protein extraction. The reversetranscriptase-polymerase chain reaction (RT-PCR) of thecontrolled RNA amount (50 ng) was performed by usingprimers for total KCNQ1, KCNQ1a, and KCNE1. Thequantification of mRNA expression levels of total KCNQ1and KCNQ1a were studied by performing real-time PCRusing a LightCycler instrument (Roche, Indianapolis, IN)with the universal probe system. Western blotting wasperformed by using specific antibodies shown in Figure 1.Differences were considered statistically significant ifP o .05 (by the Student t test).

Model simulationsThe simulations of cardiac APs are based on a modifiedO’Hara-Rudy (ORd) model of human ventricular cardio-myocytes, in which the signaling cascade from ISOapplication to PKA phosphorylation of target proteinswas incorporated.15–17 The parameters affected by PKAphosphorylation were computed by using the Heijman et almodel of the beta-adrenergic signaling pathway.16 The IKsdensities used in the simulations during baseline and underISO for different KCNQ1b/total KCNQ1 ratios were frompublished data and our experimental results.9 To determinehow the KCNQ1b/total KCNQ1 ratio affects repolariza-tion in the context of heterogeneous heart tissue,simulations on a 1-dimensional transmural wedge wereperformed.18

ResultsKCNQ1 splicing variants expression was changed byamiloride treatment in canine cardiomyocytesThe effect of amiloride on KCNQ1 splicing was tested infreshly isolated canine cardiomyocytes. The analysis ofRT-PCR products (Figure 1A) indicated that the expres-sion of KCNQ1a was differentially changed by amiloridein Endo, Mid, and Epi myocytes (Figures 1B–1H).Compared to the corresponding control myocytes thetranscript of KCNQ1a was higher in the amiloride-treatedgroup in Mid myocytes (1.39 � 0.08 fold to control) butlower in Endo (0.84 � 0.06 fold to control) and Epi(0.82� 0.02 fold to control) myocytes. The transcriptionalexpression of total KCNQ1 did not differ in Endo, Mid, orEpi myocytes between the amiloride-treated and thecorresponding control group (0.99 � 0.05, 0.98 � 0.04,and 0.93 � 0.05 folds to control, respectively). Thetranscriptional expression of total KCNQ1 and KCNQ1awere also quantified by the real-time PCR (Figure 1L). Theexpression of total KCNQ1 did not differ in Endo, Mid, orEpi myocytes between the amiloride-treated and corre-sponding control groups (1.03 � 0.04, 1.08 � 0.16, and1.03 � 0.07 folds to control, respectively). The transcriptof KCNQ1a was higher in the amiloride-treated group inMid myocytes (1.43 � 0.25 fold to control) but lower inEndo myocytes (0.75 � 0.08 fold to control). Protein

Endo Mid EpiC A C A C A

KCNQ1aTotal KCNQ1

GAPDH

KCNE1

GAPDH

C = control; A = amiloride- treated

100 kDa

60 kDaKCNQ1a

20 kDa30 kDa

KCNE1

hnRNP R100 kDa

60 kDa45 kDa

30 kDahnRNP C1/C2

20 kDaSRP 20

hnRNP B130 kDa

Endo Mid EpiL C A C A C A LEndo Mid Epi

C A C A C A

hnRNP A2

Endo Mid EpiC A C A C A

16 kDa

KCNQ1b

Figure 1 Amiloride treatment alters splicing variants of KCNQ1 and splicing factors in canine cardiomyocytes. A: Primers for the reverse transcriptase-polymerase chain reaction (RT-PCR) of KCNQ1 and the PCR products reflect the transcriptional amounts of KCNQ1a (362 bp) and total KCNQ1 (160 bp).B: The representative RT-PCR product electrophoresis of KCNQ1a, total KCNQ1, KCNE1, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) inendocardial (Endo), midmyocardial (Mid), and epicardial (Epi) myocytes. C and D: The relative transcriptional expression levels of total KCNQ1 and ofKCNQ1a vs GAPDH (n ¼ 4 for each group in the RT-PCR). E. Expression under amiloride compared with corresponding controls. F: Transcriptionalexpression of KCNE1mRNA of control and amiloride-treated groups.G:KCNQ1 and KCNE1 protein expression shown byWestern blot with a molecular massof 75 kDa for KCNQ1a and 60 kDa for KCNQ1b. Three bands of KCNE1 represent double-glycosylated (30 kDa), single-glycosylated (20 kDa), andunglycosylated (16 kDa) KCNE1 protein, respectively (n¼ 6 for each group).H:Densitometry analysis of Western blots for the effect of amiloride on KCNQ1a,KCNQ1b, and KCNE1. I: Western blots show splicing factors hnRNP-R, hnRNP-C1/C2, hnRNP-A2/B1, and SRP20 of control (part C) and amiloride-treated(part A) groups. J: Densitometry analysis of Western blots for the effect of amiloride on splicing factors.K: Coomassie blue of the same sodium dodecyl sulfatepolyacrylamide gel electrophoresis gel as in part I. L: Real-time PCR for transcriptional expression of total KCNQ1 and KCNQ1a under amiloride treatment(n ¼ 3 for each group).

Heart Rhythm, Vol 10, No 8, August 20131222

1223Lee et al Alternative Splicing and Cardiac IKs Currents

expression changes were consistent with transcriptionalchanges in KCNQ1. The KCNQ1 protein migrates insodium dodecyl sulfate polyacrylamide gel electrophoresiswith a molecular mass of 75 kDa for KCNQ1a and 60 kDafor KCNQ1b (Figure 1G). The densitometry analysisshowed that amiloride decreased KCNQ1a and increasedKCNQ1b protein expression in Endo (KCNQ1a: 0.82 �0.04 fold to control, P o .05; KCNQ1b: 1.26 � 0.10fold to control, P o .05) and Epi (KCNQ1a: 0.84 �0.04 fold to control, P o .05; KCNQ1b: 1.29 � 0.06 foldto control, P o .05) myocytes. Conversely, amilorideincreased KCNQ1a and decreased KCNQ1b in Mid myocytes(KCNQ1a: 1.24 � 0.06 fold to control, P o .01; KCNQ1b:0.85 � 0.04 fold to control, P o .05). We also examined theeffect of amiloride on KCNE1 and found that amiloride didnot change the transcriptional or translational expression inany of the myocyte types (Figures 1B, 1G, and 1H).

Multiple splicing factors may cause oppositechanges in Mid myocytes compared to Endo and EpimyocytesSplicing regulation involves a large number of RNA-bindingproteins, with SR proteins and hnRNPs being the 2 largestfamilies that function by a wide variety of mechanisms.19 Weused Western blot to determine the effect of amiloride on theexpression of several splicing-associated proteins to examinewhether the differential regulation of KCNQ1 splicingvariants in various types of myocytes by amiloride is a resultof the changes in splicing factors. The splicing factors testedinclude hnRNP-R, hnRNP-I, hnRNP-C1/C2, hnRNP-A1,SF2, and SRP20 (Figures 1I–1K), and the amiloride-induced changes in some splicing factors in Mid myocyteswere opposite to those in Endo and Epi myocytes. Forinstance, with amiloride treatment, hnRNP-R expressionsignificantly increased in Endo and Epi myocytes butdecreased in Mid myocytes while hnRNP-C1/C2 andSRP20 expression significantly decreased in Endo and Epimyocytes but increased in Mid myocytes.

Acute effect of amiloride on KCNQ1 and IKs channelsThe amiloride-induced changes in KCNQ1 expression mayaffect the electrophysiological properties of the heart (seebelow). To examine whether amiloride can also alter heartfunction by acutely altering IKs channel function, we treatedamiloride on human KCNQ1 and KCNQ1þKCNE1 (IKs)channels expressed in Xenopus oocytes. Amiloride did notchange the peak amplitudes of KCNQ1 (n¼ 4) current or IKs(n ¼ 7) (Figures 2A and 2B). The conductance-voltage(G-V) relationships of KCNQ1 current and IKs showed thatamiloride did not change voltage-dependent activation of thechannels (V1/2 for KCNQ1: amiloride-treated group −22.3�5.2 mV; control group −19.7 � 6.9 mV, P 4 .05; for IKs:amiloride-treated group 29.7 � 5.8 mV; control group 30.4� 4.9 mV, P 4 .05). There is a slope change in the G-Vrelationship of KCNQ1 after amiloride treatment (amiloride-treated group 14.7� 2.4; control group 9.5� 1.5, P ¼ .01).

However, the slope for IKs did not change with amiloride(amiloride-treated group 14.6 � 2.5; control group 13.2 �2.6, P4 .05) (Figure 2C). Amiloride also did not change theactivation time (Figure 2D). These experiments demon-strated that amiloride does not directly affect the IKs channelfunction.

KCNQ1b exerts dominant negative effect on KCNQ1current and IKsTo assess the dominant negative effect of KCNQ1b quanti-tatively, we measured the dependence of KCNQ1 and IKscurrent amplitude on the KCNQ1b/total KCNQ1 mRNAratio with a fixed amount of KCNQ1a and KCNE1 mRNAco-injection in Xenopus oocytes (Figures 2E and 2F). TheKCNQ1 and IKs current amplitudes decreased with increas-ing KCNQ1b/total KCNQ1 mRNA ratio. KCNQ1 and IKschannels contain 4 KCNQ1 subunits. It is possible that achannel containing any (1–4) KCNQ1b subunit is notfunctional. To test this possibility, we assumed that thepopulation of KCNQ1 and IKs channels containing 0–4KCNQ1b subunits follows a binomial distribution on thebasis of the KCNQ1b/total KCNQ1 mRNA ratio andonly the channels containing no KCNQ1b are functional.This model prediction of the dominant negative effectof KCNQ1b fits well with the experimental results(Figure 2F).

Introduction of the dominant negative effect ofKCNQ1b on IKs into the simulations of humanventricular myocyte APThe amiloride-induced changes in KCNQ1 splicing variantsmay affect the APs of cardiac myocytes. To examine whetheramiloride can alter the human cardiac AP, we first performeda simulation by using the ORd model.15 To define specificIKs conductance in different types of myocytes for APsimulations, we adopted a previous study result showingdifferential KCNQ1b expression across the human ventricle(see the Online Supplemental Table).9 In their quantitativeRT-PCR, KCNQ1b represented 25.2% � 2.3%, 31.7%� 1.2%, and 24.9% � 1.7% of total KCNQ1 expression inleft ventricular Endo, Mid, and Epi tissues, respectively. Weused the model prediction of the dominant negative effect ofKCNQ1b (Figure 2F) to determine human IKs conductanceon the basis of the KCNQ1b/total KCNQ1 ratio (Figure 3A).Because there is no significant differential expression of totalKCNQ1 mRNA (Figure 1C),9 the KCNQ1b/total KCNQ1ratios of 0.252 for Endo, 0.317 for Mid, and 0.249 for Epimyocytes were used. The normalized IKs conductance ofcontrol human myocytes was thus 0.31 for Endo, 0.22 forMid, and 0.32 for Epi myocytes, respectively (Figure 3A).We then used our biochemistry results (Figure 1) to identifythe normalized IKs conductance after amiloride treatment.For example, with Endo myocytes, RT-PCR showed that theKCNQ1a/total KCNQ1 ratio for amiloride-treated myocytesis 0.82 fold to control and Western blot showed that it is 0.84fold to control:

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

0:82� 0:84p ¼ 0.83. Thus, the KCNQ1a/

5 μA

KCNQ1 KCNQ1 + amiloride I + amilorideI

1S

KCNQ1a 4 ngKCNE1 3.5 ng

KCNQ1b 0 ng 1 ng 2 ng 4 ng 8 ng

2 S

2 μA

Figure 2 Acute effect of amiloride and effect of KCNQ1b on KCNQ1 current and slow delayed-rectifier potassium current (IKs). A: Human KCNQ1 and IKs(KCNQ1aþKCNE1) currents expressed in Xenopus oocytes before and after acute amiloride treatment. Currents were elicited by voltage pulses from –80 toþ60mV with 10 mV increment; the holding and repolarization potentials were –80 and –40 mV, respectively. B: KCNQ1a and IKs peak tail currents followingdepolarization to þ40 mV at various times (n ¼ 3 for KCNQ1a and n ¼ 6 for IKs). Open symbol is for currents without amiloride application during the entirerecordings (control group); filled symbol is for currents recorded before and during amiloride application, with the gray box indicating the period ofadministration (amiloride group).C: Conductance-voltage (G-V) relationships of normalized KCNQ1a and IKs tail currents with and without amiloride treatment(n¼ 5 for each). D: Voltage dependence of activation times of KCNQ1a current and IKs. τAs and τAf are obtained from fitting the activating current traces with adouble exponential function. E: IKs from oocytes with an injection of 4 ng of KCNQ1a, 3.5 ng of KCNE1, and 0–8 ng of KCNQ1b mRNA. F: IKs and KCNQ1current amplitudes depend on the KCNQ1b/total KCNQ1 mRNA ratio. Peak currents were normalized to that without KCNQ1b co-injection. The curve ofpredicted dominant negative effect (see text) is shown by the dotted line.

Heart Rhythm, Vol 10, No 8, August 20131224

total KCNQ1 ratio after amiloride treatment was estimated tobe 0.83, 1.31, 0.83 fold to control for Endo, Mid, and Epimyocytes, respectively. Because the total KCNQ1 expres-sion was not significantly different among Endo, Mid, andEpi myocytes (Figure 1C), KCNQ1b/total KCNQ1 ratios foramiloride-treated myocytes was 0.380, 0.104, and 0.378 forEndo, Mid, and Epi myocytes, respectively. Thus, thenormalized IKs conductance of human Endo, Mid, and Epimyocytes after amiloride treatment was 0.15, 0.64, and 0.15(Figure 3A), with a 0.47, 2.96, and 0.47 fold changecompared to control myocytes, respectively. These factorswere used for subsequent AP simulations. For Mid myo-cytes, AP simulations at a cycle length (CL) of 1000 msshowed that APDs were shorter in amiloride treatment(Figure 3B) and the shortening was the same in the absenceand presence of ISO (shortening was 34.4 and 29.3 ms,

representing 9.2 and 9.9% decrease, respectively).The human AP simulations demonstrated that amiloride-induced KCNQ1b/total KCNQ1 ratio changes in Midmyocytes cause APD shortening in the absence and presenceof ISO.

APD shortening in amiloride-treated Mid myocytescan be simulated by introducing KCNQ1 splicing-related IKs conductance changesTo examine the above simulation results, we compared AP inMid myocytes of control and amiloride-treated caninemyocytes. The myocytes were paced at a CL of 1000 ms(Figures 3C and 3D). APD (mean � SD) was significantlyshortened by amiloride treatment in the absence of ISO, from368 � 38 ms in control myocytes to 330 � 33 ms in

1225Lee et al Alternative Splicing and Cardiac IKs Currents

amiloride-treated Mid myocytes under baseline conditions(P ¼ .01), and the shortening was from 289 � 39 ms incontrol myocytes to 249 � 31 ms in amiloride-treated Midmyocytes in the presence of ISO (P¼ .16) (Figures 3C and 3D).

In addition, the difference in APD between control andamiloride-treated Mid myocytes increased while the pacingCL shortened (faster rate) in both experiments and modelsimulations. Experimentally, APD in amiloride-treated myo-cytes was 11.7%, 10.4%, and 7.5% shorter than in controlmyocytes at CLs of 500, 1000, and 2000 ms, respectively. Inhuman Mid myocyte AP simulations, the APD was

Vm(m

V)

μA/ μ

F

Time (ms) Time (ms)

Endo

AP IKs

Control BaselineAmiloride BaselineControl + ISOAmiloride + ISO

0 200 400-100

-50

0

50

0 200 400-100

-50

0

50

0 200 40

0.2

0.4

0 200 40

0.2

0.4

0 200 400

0.2

0.4

0 200 400-100

-50

0

50

Mid

Vm(m

V)

μA/ μ

F

Epi

Vm(m

V)

μA/ μ

F

Figure 3 Action potential duration (APD) shortening in amiloride-treated midmrelated slow delayed-rectifier potassium current (IKs) conductance changes.A: Simuratio (see text). Arrows indicate IKs densities for the cell types. B: Simulations of acttreated endocardial (Endo), midmyocardial (Mid), and epicardial (Epi) myocytes byamiloride-treated canine Mid myocytes in baseline and after isoproterenol (ISO)pacing CL of 1000 ms in canine Mid myocytes in the absence and presence of ISmyocytes and model simulation of human Mid myocytes. Increased IKs conductan

shortened by increasing IKs conductance 2.96 times(Figure 3A), and APD shortening was also enhanced at fastrates (9.7%, 9.2%, and 8.8% shorter than in control myocytesat CLs of 500, 1000, and 2000 ms, respectively) (Figure 3E).This phenomenon is due to an important IKs role in the rateadaptation of APD, by which IKs facilitates AP repolarizationacceleration and APD shortening at fast rates.3 The fre-quency dependence of amiloride induced APD shortening inboth experiments, and simulation further supports that APDshortening was derived from enhanced IKs current byamiloride treatment.

Mid Control Mid Amiloride

100 ms

50 m

V

CL = 1000 ms

BaselineISO

BaselineControl

Amiloride 100 ms

50 m

V

00

00

0

yocardial (Mid) myocytes can be simulated by introducing KCNQ1 splicing-lation showing normalized IKs densities based on the KCNQ1b/total KCNQ1ion potentials (APs) and corresponding IKs densities in control and amiloride-using the ORd human ventricular model.15 C: AP recordings of control and

(10 nmol/L) administration. CL ¼ cycle length. D: APD measurement at aO. E: Frequency dependence of APD shortening recorded from canine Midce simulates the effect of amiloride.

Heart Rhythm, Vol 10, No 8, August 20131226

Amiloride treatment reduced the occurrence of EADsunder ISO challenge in Mid myocytesFor canine left ventricular myocytes, ISO at concentrationsas low as 20 nmol/L can induce EADs.20 To examinewhether amiloride can affect the development of ISO-induced EADs at both slow and fast rates, we performedAP recordings in canine Mid myocytes with ISO challengeby using gradual titration. The threshold of ISO concen-tration for inducing EADs in our canine Mid myocytes was20 or 50 nmol/L and was not different between control andamiloride-treated myocytes. However, EADs were generatedmore frequently in control than in amiloride-treated myo-cytes (Figures 4A–4C). In both control and amiloride-treatedMid myocytes, EADs were more frequently generated atslower rates (Figure 4C).

Mid Control

Mid Amiloride

40 m

V

0.5 S

Mid Control

Mid Amiloride

CL = 2000 ms CL = 1000 ms CL = 500 ms0.1S

40 m

V

Time (ms)

-10-8-6-4-20

-100

-50

0

50

-100

-50

0

50

-10-8-6-4-20

0 200 400 6000

0.5

1.0

-10

-5

5

Time (ms)

V(m

V)I C

aL(μ

A/μF

)I K

s(μ

A/μF

)

CL = 2000 ms CL = 1000 ms

Control + 1 μM ISO Amiloride + 1 μM ISO

0 200 400 600 8000

0.5

1.0

-

0

1

To evaluate the effect of amiloride on EAD generation, weused the ORd model, in which a series of modifications in theequations were incorporated to account for beta-adrenergicstimulation (see the Online Supplemental Material). Previousstudies have shown that under steady-state pacing at CL of1000 ms, the application of 1 μmol/L of ISO led to EADgeneration even in normal Mid myocyte AP simulations,partially because ISO-induced phosphorylation is faster in L-type calcium current (ICaL) than in IKs.

17,21 Our simulationsrevealed that at slower pacing EADs appeared in both controland amiloride-treated Mid myocytes after ISO application(Figure 4D). Less EADs were generated at fast pacing rate andthey were inhibited by amiloride. The EADs were associatedwith ICaL reactivation, while the increased IKs conductancecaused shortening of APD and reduced ICaL reactivation. EAD

BaselineISO

0

0

0

0

Time (ms)

CL = 500 ms

10-8-6-4-20

0 200 4000

.5

.0

Figure 4 Amiloride treatment reduces isoproter-enol (ISO)-induced early afterdepolarization (EAD)in isolated midmyocardial (Mid) myocytes. A:Representative traces of continuous action potential(AP) recordings from canine Mid myocytes with andwithout amiloride treatment. Baseline APs are shownin gray and with ISO (20 or 50 nmol/L) in black. B:APs at pacing cycle lengths (CLs) of 2000, 1000, and500 ms with and without amiloride treatment. Base-line APs are in gray and with ISO in black. C: Themean number of EAD (�SEM) per 10 consecutivebeats during steady-state pacing. *P o .05. Thenumber of cells recorded at CLs of 2000, 1000, and500 ms are as follows: n ¼ 3, 4, 2 for the controlgroup and n¼ 3, 3, 2 for the amiloride-treated group,respectively. D: Simulations of ISO-induced EADsin human Mid myocytes. Descending rows show AP,L-type calcium current (ICaL), and slow delayed-rectifier potassium current (IKs). Columns, from leftto right, show simulations at CLs of 2000, 1000, and500 ms under ISO. Gray lines represent simulationsof control myocytes, and black lines those ofamiloride myocytes.

1227Lee et al Alternative Splicing and Cardiac IKs Currents

was not noted in either Endo or Epi AP simulations under ISOchallenge (data not shown). The simulations suggest thatamiloride-induced 2.96-fold IKs conductance increase caninhibit EAD generation in human ventricular Mid myocytesunder the challenge of high concentrations of ISO.

DiscussionThe results of the present study suggest that (1) amiloridechanges the alternative splicing of KCNQ1 and that effect maybe caused by splicing factor changes, (2) a prominent IKsdensity modulation can be attributed to a small change inKCNQ1b splicing variant expression, and (3) in Mid myo-cytes, increased IKs density shortens APD and inhibits EADsdevelopment under beta-adrenergic stimulation.

Amiloride has multiple pharmacological actions onionic channels and transporters. It inhibits the Na/Hexchanger in cultured cardiac cells and in sheep Purkinjefibers; nonetheless, Na/H exchange inhibition by 0.01–1.0mmol/L of amiloride does not cause intracellular acido-sis.22 Amiloride also suppresses the T-type calcium currentin guinea pig ventricular myocytes with a dissociationconstant Kd of 233 μmol/L and inhibits sarcolemmal Naþ/Ca2þ exchanger with half maximal effective concentrationEC50 at 0.7 mmol/L in isolated beating guinea pigatria.23,24 The amiloride concentration of 50 μmol/L usedin this study is too low to affect the intracellular pH or toinhibit the T-type calcium current and the Naþ/Ca2þ

exchanger. Amiloride (10 μmol/L) was shown to prolongAPD with prolonged treatment in canine cardiac Purkinjefibers.25 Whether amiloride could change KCNQ1 splicing(with increased KCNQ1b expression) in Purkinje fibersremains to be determined.

ECG

Control

Epi

Mid

Endo

0

0

0

0

0

TDR 55.6 ms

0 200 400Time (ms)

0

QT 379.6 ms

Figure 5 Amiloride transmural disper-sion of repolarization (TDR) owing to theslow-delayed rectifier potassium current(IKs) conductance changes. Simulatedaction potentials of (top to bottom) humanendocardial (Endo), midmyocardial(Mid), and epicardial (Epi) myocytes,together with a simulated pseudo-electro-cardiogram (ECG). Columns, from left toright, show cases of controls, amilorideeffect with IKs conductance changes, andamiloride effect with both IKs conduc-tance change and potassium (Kþ) eleva-tion. The QT interval and TDR areindicated.

Other genetic factors have been shown to modify KCNQ1transcript expression and QT interval, for example, variantsin the 3′ untranslated region of KCNQ1, variants in putativeion channel regulatory subunits such as NOS1AP, andstimulating protein 1 which is involved in transcriptionalactivation of the KCNQ1 promoter.26–28 To our knowledge,this is the first report describing drug-induced alternativesplicing of KCNQ1 in cardiac myocytes. The detailedmechanism involved in KCNQ1 splicing remains to beinvestigated.

Clinical implicationsQT dispersion, which is the difference in minimal QT intervaland maximal QT interval in the surface 12-lead electro-cardiogram (ECG), has been suggested to reflect repolariza-tion inhomogeneity in the heart and to have association withsudden cardiac death in healthy subjects and in patients withchronic heart failure.29,30 In a double-blinded clinical study,amiloride significantly reduced QT dispersion and alsoreduced ventricular extrasystoles.31 Simulated pseudo-ECGestablishes a direct relationship between ionic currents, theAP, and the morphology of electrocardiographic waveforms.It allows us to evaluate the effect of amiloride on the ECGQTinterval without performing an in vivo study (Figure 5). Weintroduced the estimated IKs conductance of Endo, Mid, andEpi in the amiloride simulations as 0.47, 2.96, and 0.47 timescontrol (Figures 3A and 5). Moreover, owing to the nature ofpotassium sparing diuretics, amiloride increases plasmapotassium by an average value of 0.5 mmol/L.12,13,31 There-fore, we further simulated APs and pseudo-ECGs withincreased extracellular Kþ level by 0.5 mmol/L (Figure 5).In AP simulations, the reduced IKs conductance caused APDprolongation with 8 ms for Endo and 4.7 ms for Epi myocytes

AmilorideIKs change

50mV

0.2 μA/μF

0 200 400Time (ms)

TDR 48.5 ms

TDR 48.7 ms

0 200 400Time (ms)

AmilorideIKs + K+ changes

QT 377.5 ms QT 369.3 ms

Heart Rhythm, Vol 10, No 8, August 20131228

(Figure 5); however, these prolongations were offset by theelevation of Kþ level (APD changes were only 0.8 ms forEndo and −0.9 ms for Epi). The above clinical study alsoobserved that the potassium elevation shortened QT inter-val.31 The heterogeneity of the IKs conductance among 3transmural layers was enlarged by amiloride (Figure 3A);nonetheless, APD differences and TDR were reducedbecause of major increase of Mid IKs and shortening ofits APD (Figure 5). The reduction of TDR by amiloridewas frequency dependent and was larger at a faster pacingrate (14%, 16.7%, and 25.8% reduction with amiloridecompared to the control at CLs of 2000, 1000, and 500 ms,respectively) (Figure S1). These results suggest thatreduced dispersion may be an antiarrhythmic mechanismof amiloride in the clinical setting.30 This study alsosuggests that the alternative splicing of KCNQ1 may bean antiarrhythmic therapeutic target.

In conclusion, amiloride induces KCNQ1 splicing changesby affecting the expression of splicing factors and the effectsshow transmural differences. This drug-induced modulation ofIKs due to the alternative splicing of KCNQ1 changes the IKstransmural heterogeneity and results in reduced transmuraldispersion, reduced EAD formation and therefore reducedarrhythmogenicity under beta-adrenergic stimulation.

AcknowledgmentsWe are grateful to Dr Richard B. Schuessler and Dr Urvi Leefor their courtesy of providing isolated canine cardiacmyocytes. We thank Dr Kathryn Yamada, Dr Jeanne M.Nerbonne, and Dr Kai-Chien Yang for technical support;Dr Jingyi Shi and Kelli Delaloye for DNA subcloning; andDr Leonid Livshitz, Dr Ali Nekouzadeh, Dr Namit Gaur, andDr Ashwin Mohan, Smiruthi Ramasubramanian, Jiajing Xu,and Mark A. Zaydman for helpful discussions. We partic-ularly thank Dr. Tom O’Hara for advice to the computationalsimulations in the ORd model.

AppendixSupplementary dataSupplementary data associated with this article can be foundin the online version at http://dx.doi.org/10.1016/j.hrthm.2013.04.014.

References1. Rocchetti M, Besana A, Gurrola GB, Possani LD, Zaza A. Rate dependency of

delayed rectifier currents during the guinea-pig ventricular action potential.J Physiol 2001;534:721–732.

2. Stengl M, Volders PGA, Thomsen MB, RLHMG Spätjens, Sipido KR, Vos MA.Accumulation of slowly activating delayed rectifier potassium current (IKs) incanine ventricular myocytes. J Physiol 2003;551:777–786.

3. Silva J, Rudy Y. Subunit interaction determines IKs participation in cardiacrepolarization and repolarization reserve. Circulation 2005;112:1384–1391.

4. Volders PGA, Stengl M, van Opstal JM, et al. Probing the contribution of IKs tocanine ventricular repolarization: key role for β-adrenergic receptor stimulation.Circulation 2003;107:2753–2760.

5. Burashnikov A, Antzelevitch C. Prominent IKs in epicardium and endocardiumcontributes to development of transmural dispersion of repolarization but protectsagainst development of early afterdepolarizations. J Cardiovasc Electrophysiol2002;13:172–177.

6. Liu D-W, Antzelevitch C. Characteristics of the delayed rectifier current (IKr andIKs) in canine ventricular epicardial, midmyocardial, and endocardial myocytes: aweaker IKs contributes to the longer action potential of the M cell. Cir Res1995;76:351–365.

7. Demolombe S, Baro I, Pereon Y, et al. A dominant negative isoform of the longQT syndrome 1 gene product. J Biol Chem 1998;273:6837–6843.

8. Jiang M, Tseng-Crank J, Tseng GN. Suppression of slow delayed rectifier currentby a truncated isoform of KvLQT1 cloned from normal human heart. J Biol Chem1997;272:24109–24112.

9. Pereon Y, Demolombe S, Baro I, Drouin E, Charpentier F, Escande D.Differential expression of KvLQT1 isoforms across the human ventricular wall.Am J Physiol Heart Circ Physiol 2000;278:H1908–H1915.

10. Chang WH, Liu TC, Yang WK, et al. Amiloride modulates alternative splicing inleukemic cells and resensitizes Bcr-Ab/T315I mutant cells to imatinib. CancerRes 2011;71:383–392.

11. Chang JG, Yang DM, Chang WH, et al. Small molecule amiloride modulatesoncogenic RNA alternative splicing to devitalize human cancer cells. PLoS One2011;6:e18643.

12. Duff HJ, Lester WM, Rahmberg M. Amiloride. Antiarrhythmic and electro-physiological activity in the dog. Circulation 1988;78:1469–1477.

13. Duff H, Mitchell L, Kavanagh K, Manyari D, Gillis A, Wyse D. Amiloride.Antiarrhythmic and electrophysiologic actions in patients with inducible sus-tained ventricular tachycardia. Circulation 1989;79:1257–1263.

14. Smith AJ, Smith RN. Kinetics and bioavailability of two formulations ofamiloride in man. Br J Pharmacol 1973;48:646–649.

15. O'Hara T, Virag L, Varro A, Rudy Y. Simulation of the undiseased humancardiac ventricular action potential: model formulation and experimental vali-dation. PLoS Comput Biol 2011;7:e1002061.

16. Heijman J, Volders PG, Westra RL, Rudy Y. Local control of β-adrenergicstimulation: effects on ventricular myocyte electrophysiology and Ca2þ-transient.J Mol Cell Cardiol 2011;50:863–871.

17. O'Hara T, Rudy Y. Arrhythmia formation in subclinical ("silent")long QT syndrome requires multiple insults: quantitative mechanistic studyusing the kcnq1 mutation Q357R as example. Heart Rhythm 2012;9:275–282.

18. Gima K, Rudy Y. Ionic current basis of electrocardiographic waveforms: a modelstudy. Circ Res 2002;90:889–896.

19. Wang Z, Burge CB. Splicing regulation: from a parts list of regulatory elementsto an integrated splicing code. RNA 2008;14:802–813.

20. Volders PG, Kulcsar A, Vos MA, et al. Similarities between early and delayedafterdepolarizations induced by isoproterenol in canine ventricular myocytes.Cardiovasc Res 1997;34:348–359.

21. Liu G-X, Choi B-R, Ziv O, et al. Differential conditions for early after-depolarizations and triggered activity in cardiomyocytes derived from transgenicLQT1 and LQT2 rabbits. J Physiol 2012;590:1171–1180.

22. Brown KK, Yee R, Grant AO, Strauss HC. Effects of amiloride on pHregulation in canine cardiac Purkinje fibers. J Pharmacol Exp Ther 1990;254:83–90.

23. Tytgat J, Vereecke J, Carmeliet E. Mechanism of cardiac T-type Ca channelblockade by amiloride. J Pharmacol Exp Ther 1990;254:546–551.

24. Floreani M, Luciani S. Amiloride: relationship between cardiac effects andinhibition of Naþ/Ca2þ exchange. Eur J Pharmacol 1984;105:317–322.

25. Marchese AC, Hill JA Jr, Xie PD, Strauss HC. Electrophysiologic effects ofamiloride in canine Purkinje fibers: evidence for a delayed effect on repolariza-tion. J Pharmacol Exp Ther 1985;232:485–491.

26. Amin AS, Giudicessi JR, Tijsen AJ, et al. Variants in the 3′ untranslated region ofthe KCNQ1-encoded Kv7.1 potassium channel modify disease severity inpatients with type 1 long QT syndrome in an allele-specific manner. Eur HeartJ 2012;33:714–723.

27. Crotti L, Monti MC, Insolia R, et al. NOS1AP is a genetic modifier of the long-QTsyndrome. Circulation 2009;120:1657–1663.

28. Luo XB, Xiao JN, Lin HX, Lu YJ, Yang BF, Wang ZG. Genomic structure,transcriptional control, and tissue distribution of HERG1 and KCNQ1 genes. AmJ Physiol Heart Circ Physiol 2008;294:H1371–H1380.

29. Molnar J, Rosenthal JE, Weiss JS, Somberg JC. QT interval dispersion in healthysubjects and survivors of sudden cardiac death: circadian variation in a twenty-four-hour assessment. Am J Cardiol 1997;79:1190–1193.

30. Cuddy TE, Halli PS, Tate RB. QT dispersion and heart rate predict the risk ofsudden unexpected cardiac death in men: the Manitoba follow-up study. PrevCardiol 2009;12:27–33.

31. Farquharson CA, Struthers AD. Increasing plasma potassium with amilorideshortens the QT interval and reduces ventricular extrasystoles but does notchange endothelial function or heart rate variability in chronic heart failure. Heart2002;88:475–480.