Mechanism of Discordant T Wave Alternans in the In Vivo Heart

632

Mechanism of Discordant T Wave Alternansin the In Vivo Heart

MASAOMI CHINUSHI, M.D., DMITRY KOZHEVNIKOV, M.D.,EDWARD B. CAREF, PH.D., MARK RESTIVO, PH.D.,

and NABIL EL-SHERIF, M.D.

From the Cardiology Division, Department of Medicine, State University of New York Health Science Center,and New York Harbor Veterans Affairs Healthcare System, Brooklyn, New York, USA

Discordant T Wave Alternans. Introduction: Compared to concordant T wave alternans (CA), discor-dant T wave alternans (DA) may be associated with an increased dispersion of repolarization (DR) and agreater propensity to develop reentrant ventricular tachyarrhythmias. The electrophysiologic mechanismsof DA in the in vivo heart are not well understood.

Methods and Results: The mechanisms of DA were investigated in the canine anthopleurin-A surrogatemodel of long QT3 syndrome using tridimensional analysis of activation and repolarization patterns from256 to 384 unipolar electrograms. Cardiac repolarization was evaluated as the activation-recovery interval(ARI) of local electrograms. Two mechanisms for the development of DA were observed. (1) Stepwiseshortening of cycle length (CL) superimposed on preexisting DR resulted in different diastolic intervals(DI) at midmyocardial sites compared to epicardial and endocardial sites. The dispersion of DI coupledwith different restitution kinetics at those sites induced DA. (2) The dependence of conduction velocityon DI as the CL is abruptly shortened could result in differential conduction delays at mid sites. Thisenhanced the dispersion of DI between sites and, coupled with the different restitution kinetics, inducedDA. The critical step for the development of DA in both mechanisms was the occurrence of short ARIin two consecutive beats either at epicardial sites in the first mechanism or at mid sites in the secondmechanism. Sites with DA had significantly more DR compared to sites with concordant T wave alternans,and ventricular tachyarrhythmias developed mainly in the presence of DA.

Conclusion: In the in vivo heart, DA developed due to critical interaction between dispersion of DI anddifferences in restitution kinetics at different myocardial sites. The dispersion of DI could result from pre-existing DR or differential conduction delay at a critical short CL. DA is critically linked to the developmentof malignant tachyarrhythmias. (J Cardiovasc Electrophysiol, Vol. 14, pp. 632-638, June 2003)

mapping, fibrillation, reentry, long QT syndrome

Introduction

Discordant T wave alternans (TWA) occurs when the Twave alternates out of phase in adjoining spatially distinctregions of the heart. Discordant T wave alternans may be as-sociated with markedly increased dispersion of repolarization(DR) and a greater propensity to develop ventricular tach-yarrhythmias (VT).1 The electrophysiologic mechanism(s)of discordant action potential duration (APD) or TWA is notwell understood. We previously examined the mechanism ofconcordant alternans (CA) in an in vivo surrogate model oflong QT syndrome (LQT3).2 We present here a study of themechanisms of discordant alternans (DA) in the same exper-imental model and the basis for increased arrhythmogenecityof DA versus CA. Preliminary data were previously publishedin abstract form.3

Supported in part by VA MERIT and REAP grants to Dr. El-Sherif.

Address for correspondence: Nabil El-Sherif, M.D., SUNY DownstateMedical Center, Box 1199, 450 Clarkson Avenue, Brooklyn, NY 11203.Fax: 718-630-3740; E-mail: [email protected]

Manuscript received 22 January 2003; Accepted for publication 17 March2003.

Methods

Surgical Preparation

The present study was approved by the Animal StudiesSubcommittee of the local institutional review board andconformed to the Guiding Principles of the Declaration ofHelsinki. Experiments were performed on eight purpose-bredmongrel puppies (12–14 weeks old, weight 4.0–6.8 kg). De-tails of anesthesia and surgical procedure were previouslypublished.2 To slow the heart rate, complete AV block wasaccomplished by radiofrequency catheter ablation, and theventricles were paced through bipolar stainless wire elec-trodes connected to a digital stimulator (DTU-101, BloomInc., Reading, PA, USA).

Data Acquisition

Forty-eight to sixty-four plunge needle electrodes, eachconsisting of 6 to 10 unipolar electrodes, were used. Unipo-lar electrograms were simultaneously recorded from the epi-cardium (Epi), midmyocardium (Mid), and subendocardial(End) sites using three variable-gain 128-channel multiplexeddata acquisition systems (DSC 2000, INET Corp., Newark,CA, USA), allowing simultaneous recording of 256 to 384channels. Details of the recording methods, mapping tech-nique, and data acquisition system were previously reported.4

Chinushi et al. Discordant T Wave Alternans 633

Activation-recovery interval (ARI) was defined as the in-terval between the minimum first derivative (Vmin) of the QRSand the maximum first derivative (Vmax) of the T wave of theunipolar electrogram.

Drug Administration

To simulate LQT3, anthopleurin-A was used.Anthopleurin-A was dissolved in 0.9% sterile salineand administered as an intravenous bolus of 25 µg/kg,followed by a maintenance dose of 1.0 µg/kg/min.4

Stimulation Protocol

Bipolar stimulation was used to stimulate the heart. Toobtain a stable basic state, the heart was driven for 50 beatsat the basic cycle length (CL) of 1,000 msec before startingeach stimulation protocol. Two pacing protocols were used inthe present study to induce TWA: (1) abrupt shortening of theCL from a steady state of 1,000 msec in steps of 50 msec for10 to 15 seconds at each new CL; and (2) stepwise decreaseof the CL. This was accomplished by the introduction of oneor two short CLs intermediate between the long steady-stateCL and the train of constant short CLs. In the present study,the basic driven beat was defined as S1 and the rapid pacingbeats were defined as P1, P2, etc.

To study the restitution of ARI, a single premature ventric-ular stimulation was delivered after every 28th basic beat ata CL of 1,000 msec (S1S2 protocol). The S1-S2 interval wasprogressively decreased by 5 to 50 msec from 1,000 msecto determine restitution properties and effective refractoryperiods at each test site.2

Data Analysis

ARIs were measured at three different layers of the leftventricle (End, Mid and Epi), and the paced CL associatedwith the onset of local ARI alternans at each site was deter-mined. Alternans of ARI was defined as a >10 msec differ-ence between three consecutive beats. The alternans of localand surface ECG was compared. During TWA, transmuraldispersion of ARI was measured as the maximum differenceof ARI between Mid and Epi and between Mid and End. Tosimplify analysis, the most proximal and distal plunge elec-trode sites were taken to represent Epi and End layers, re-spectively, whereas the intermediate electrode site showingthe longest basic ARI site was considered to represent theMid layer. The transmural difference of ARI was measuredduring each paced beat, and the relation of surface QRS/Twave alternans to the transmural repolarization pattern wasexamined.

The S1S2 protocol was used to construct restitution curvesfor each of the three ventricular layers. Restitution curves,representing the relationship between ARI and diastolic in-terval (DI), were resolved using a single exponential decayfunction. DI was defined as the interval between the recoverytime of the last basic beat (S1) and activation of the followingS2. Data were fit using Origin 5.0 (Microcal Software, Inc.,Northampton, MA, USA) to the following equation: ARI(t) =ARImax − �ARI × exp−(t/τ ), where ARImax represents ARIduring the plateau of restitution; ARI(t) is the ARI of the DIpreceding S2; and �ARI and τ are the amplitude and timeconstant, respectively.2 ARImax, �ARI, and τ were comparedin each layer of the left ventricle, and the relationship betweenthe restitution and TWA was examined.

The role of TWA in the initiation of arrhythmias was exam-ined. To facilitate the induction of arrhythmias during TWA,the basic paced CL of 1,000 msec was sometimes changedto a longer or shorter CL.

Statistical Analysis

Statistical analysis was performed by Student’s t-test,analysis of variance (ANOVA) for multivariate and repeateddesigns, and Scheffe’s multiple range post hoc test whereappropriate. Kolmogorov-Smirnov goodness of fit test fornormal distribution was used to verify normal distribution ofdata before performing ANOVA. P < 0.05 was consideredstatistically significant.

Results

Induction of DA by Stepwise Shortening of CL

Figure 1A illustrates CA induced by abrupt shortening ofthe CL from 1,000 to 500 msec. Representative recordings ofEpi and Mid electrograms from a needle electrode in the leftventricular free wall are shown. During the basic drive, ARIwas longer at the Mid site (485 msec) compared to the Episite (413 msec). Following the first short CL of 500 msec,the ARI at both Epi and Mid sites shortened. However, thedegree of shortening of ARI at the Mid site was markedlyexaggerated compared to the Epi site (from 485 to 197 msecat Mid vs 411 to 255 msec at Epi). This could be explainedby the short DI preceding P1 at Mid (15 msec) compared tothe DI preceding P1 at Epi (89 msec). It also is explainedby the difference in restitution kinetics between Mid and Episites, as will be shown later. Following P2, the ARI intervallengthened at both the Mid and Epi sites, but relatively moreat the Mid site. The CA was associated with reversal of theARI gradient between Epi and Mid sites in alternate beats.

Figure 1B illustrates the induction of DA between the Epiand Mid sites on stepwise shortening of the CL from 1,000to 600 msec to 500 msec, followed by constant pacing at500 msec as in Figure 1A. The introduction of P1 at a CL of600 msec was associated with a longer preceding DI at bothMid and Epi sites compared to P1 at 500 msec in panel A.This resulted in less shortening of the ARI at Mid and Episites. Still, the degree of shortening of ARI at the Mid site wasgreater than at the Epi site, resulting in a shorter ARI at Midcompared to Epi. The introduction of P2 at 500 msec nowwas associated with a shorter preceding DI at Epi (137 msec)compared to Mid (175 msec). This was followed by furthershortening of the ARI at Epi but by lengthening of the ARIat Mid, thus initiating DA at a constant CL of 500 msec.

To determine the interaction between the DI and the resti-tution kinetics at the different myocardial sites, we inves-tigated the differences in the kinetics of restitution of ARIat Epi, Mid, and End sites. As we previously reported,2 thekinetics differed among sites, with the Mid sites having alonger ARImax, �ARI, and τ compared with the Epi sites.The restitution parameters of the End sites were intermediatebetween Mid and Epi sites. Figure 2 depicts the restitutioncurve of ARI at the Mid and Epi sites shown in Figure 1. Ata CL of 500 msec, the DI preceding P1 was 15 msec at Midand 89 msec at Epi. The DI of 15 msec corresponded on therestitution curve to a 290-msec shortening of ARI at Mid,whereas a DI of 89 msec at Epi corresponded to 140-msecshortening at Epi. The greater shortening of ARI at both sites

634 Journal of Cardiovascular Electrophysiology Vol. 14, No. 6, June 2003

Figure 1. A: Induction of concordant alternans by abrupt shortening of the cycle length (CL) from a control CL of 1,000 msec to 500 msec. B: Induction ofdiscordant alternans by stepwise shortening of CL from 1,000 to 600 msec, followed by constant pacing at 500 msec. Unipolar electrograms were recordedfrom the same midmyocardial (Mid) and epicardial (Epi) sites from the free left ventricular wall. Control pacing beats are labeled S1. Rapid pacing beats arelabeled P1, P2, etc. The numbers between brackets at the top of each tracing are the diastolic intervals. The numbers at the bottom of each tracing are thecalculated activation-recovery intervals. L and S represent long and short ARIs, respectively. Note that at the initiation of discordant alternans, the epicardialsite showed short ARIs in two consecutive beats. See text for details.

resulted in a longer DI preceding P2 at 500 msec and initi-ated a long-short concordant TWA. On the other hand, a CLof 600 msec, as shown in Figure 1B, was associated with alonger DI of 115 msec at Mid and 189 msec at Epi. The DI of115 msec corresponded to a shortening of ARI of 180 msec.

Figure 2. Restitution curves of the midmyocardial (Mid) and epicardial (Epi) sites shown in Figure 1. It graphically illustrates how differences in both therestitution kinetics of the activation-recovery interval (ARI) and diastolic intervals at both sites combine to explain the concordant versus discordant alternansduring different pacing protocols as shown in Figure 1. See text for details.

In contrast, DI of 189 msec at Epi was associated with lessshortening of ARI of 55 msec. The lesser degree of shorteningof ARI at Epi was associated with a relatively short DI of 137msec preceding the introduction of P2 at 500 msec comparedto the same DI of 175 msec at Mid. The differences in the DI


Figure 3. Induction of discordant alternans by abrupt shortening of the cycle length from 1,000 msec (S1) to 400 msec (P1, P2, etc.), which resulted inspatially differential conduction delay. Note that the first short cycle was associated with more conduction delay at the Mid site compared to End and Episites. This resulted in a relatively shorter diastolic interval preceding P1 at Mid compared to End and Epi, which together with the differences in restitutionkinetics explain the occurrence of two consecutive beats with short activation-recovery interval (ARI) at Mid and the onset of discordant alternans. L and Srepresent long and short ARIs, respectively. See text for details.

and the restitution kinetics at Epi and Mid sites explain thecontinued short ARI following P2 at Epi, thus initiating DAbetween the two sites.

Induction of DA Following Spatially DifferentialConduction Delay

Figure 3 illustrates a second mechanism for the initia-tion of DA. Selective recordings from Endo, Mid, and Epielectrodes on the same needle from the left ventricular freewall and the surface ECG lead aVF are shown. DiscordantTWA between the Mid site and both the Endo and Epi sitesdeveloped upon abrupt shortening of the CL from 1,000 to400 msec. At a control CL of 1,000 msec, the ARI was longestat Mid (483 msec) compared to Epi (375 msec). The ARIat End (419 msec) was intermediate between Mid and Epi.Abrupt shortening of the CL to 400 msec resulted in a slightdelay of conduction at the Epi site (stimulus spike to activa-tion of 30 msec). There was a marked delay of conductionat both End and Mid sites (85 and 95 msec, respectively).This resulted in a short DI of 12 msec preceding P1 at Midcompared to a DI of 66 msec at End and 55 msec at Epi sites.This, in turn, resulted in relatively more shortening of theARI of P1 at Mid compared to the ARIs at End and Epi. Thedifferential conduction delay at Mid versus Epi also explainsthe shorter DI preceding P2 at Mid (75 msec) compared toEpi (167 msec). The interaction between the difference in DIand the restitution kinetics of ARI at the three sites explainsthe initiation of discordant TWA between Mid and both Epi

and End sites. During discordant TWA, there was alternationof the ST/T wave morphology on the surface ECG lead andof the local depolarization complex on the End recording,but there was no perceptible alternation of the surface QRSmorphology.

CA Versus DA and DR

To compare the degree of DR during CA versus DA, we an-alyzed simultaneous recordings from the seven experimentsin which different sites showed CA and DA at the same pac-ing protocol. Figure 4 shows recordings from one of theexperiments. Stepwise shortening of paced CL from 1,000to 460 msec (PI) followed by constant pacing at 350 msec(P2-P10) induced CA at 10 plunge needle sites (top panel) andDA at five other sites (lower panel). DR was greater at sites ofDA compared to sites with CA, as shown in the graphic illus-tration of mean ± SEM of dispersion of ARIs in the bottompanel. In the seven experiments, the mean ± SEM disper-sion of ARI at sites of DA (87 ± 32 msec) was significantlygreater than at sites of CA (56 ± 21 msec; P < 0.01).

In 7 of the 8 experiments, one or more VT episodes de-veloped following TWA. In six experiments, VT developedin the presence of DA. In one experiment, VT develop fol-lowing a paced drive in which all mapped sites showed CA.Figure 5 illustrates the development of VT from an experi-ment in which sites with marked DA were induced by abruptdecrease of the pacing CL from 1,000 to 500 msec. TheVT was initiated by a premature focal discharge from a left


Figure 4. Simultaneous recording from two different Mid/Epi sites from one of the experiments. Note that the same pacing protocol of stepwise shorteningof cycle length from 1,000 to 460 to 350 msec induced concordant alternans (top recording) and discordant alternans (bottom recording). The graph at thebottom illustrates the mean ± SEM of dispersion of activation-recovery intervals (ARIs) from 10 plunge needle sites that showed concordant alternans andfive other sites that showed discordant alternans during the same pacing protocol from the same experiment. Sites with discordant alternans had greaterdispersion of ARI compared to sites with concordant alternans.

ventricular endocardial site that induced areas of functionalconduction block and reentrant excitation (Fig. 6).

Discussion

The present study describes, for the first time in an in vivoheart, two closely related electrophysiologic mechanisms forthe development of discordant repolarization alternans. In thefirst mechanism, stepwise shortening of CL superimposed onpreexisting nonuniform DR resulted in different DI at Midsites versus Epi and, occasionally, End sites. The dispersionof DI, coupled with different restitution kinetics at these sites,induced DA. In the second mechanism, the dependence ofconduction velocity on DI as the CL is shortened could re-sult in spatially differential conduction delay at Mid sitescompared to Epi and End sites. This also would enhance thedispersion of DI between these sites. Thus, in both situationsthe dispersion of the DI coupled with the difference in resti-

Figure 5. Mid/Epi recordings showing marked discordant alternans that was associated with the onset of ventricular tachyarrhythmia following a prematurebeat (marked by asterisk). The numbers at the top of the tracing are the calculated activation-recovery interval (ARI). The numbers at the bottom are thedispersion of ARI between Mid and Epi sites.

tution kinetics is the key for the initiation of DA. The criticalstep for the development of DA by either mechanism is theoccurrence of short repolarization (ARI) in two consecutivebeats either at Epi sites in the first mechanism (Fig. 1B) or atMid sites in the second mechanism (Fig. 3).

In a previous study, Pastore et al.1 investigated the mecha-nism of DA by recording optical action potentials in a guineapig Langendorff preparation in which the endocardial surfacewas cryoablated, leaving a thin rim of epicardium. During in-cremental rapid pacing, transition of CA to DA was observed.In contrast to the development of transmural DA in our in vivomodel, in this guinea pig model, DA occurred between epi-cardial apical and basal sites. DA also was associated with CLalternans, which caused QRS alternans in addition to TWA.A similar observation was reported in a computer simulationstudy in a two-dimensional sheet of cardiac tissue.5 A steep(slope > 1) APD restitution curve promoted CA and TWAwithout QRS alternans. When pacing was performed from a


Figure 6. Selected isochronal activation maps and local electrograms of the premature beat that induced the tachyarrhythmia shown in Figure 5. Thenumbers in the upper left corner of the maps refer to sections selected from a traditional representation of the activation maps of five transverse sections ofthe ventricles labeled 1 to 5 from base to apex (see reference 4). Activation is represented as closed isochrone contours of 20 ms and are labeled 1,2,3, etc.,instead of 20, 40, 60 ms, to facilitate analysis of the spread of activation. The heavy solid lines represent arcs of functional conduction block. The asterisksrepresent the site of earliest excitation, and from this site the activation wavefront advanced in a circuitous route represented by the continuous dashed arrow.The electrograms are selected along the reentrant pathway to illustrate continuous diastolic bridging, and their sites are marked on the maps.

single site, DA occurred only when the pacing rate was suf-ficient to engage conduction velocity (CV) restitution, pro-ducing both QRS alternans and TWA. Tissue heterogeneitywas not required for this effect. In the absence of steep APDrestitution and CV restitution, sustained DA did not occur.In another simulation study, DA developed in spatially ho-mogenous tissue by one of two mechanisms: (1) interactionof CV and APD restitution at high pacing frequencies; or(2) through the dispersion of DI produced by ectopic foci.6

The present study investigated an in vivo heart with sig-nificant preexisting DR, which is a more realistic model tostudy repolarization alternans. This may explain why some ofthe present observations differ from those of previously pub-lished reports. Although interaction between CV and kineticsof restitution of repolarization at different myocardial sitescould induce discordant repolarization alternans, the latterwas not associated with CL or QRS alternans. This suggeststhat the degree and/or spatial extent of conduction delays orconduction block that induce DA in the in vivo heart couldbe limited and do not necessarily manifest as alternans ofsurface QRS. On the other hand, DA that developed follow-ing stepwise shortening of CL did not involve CV restitutionbut rather was the result of interaction between dispersion ofDI and different restitution kinetics of cardiac repolarizationat different myocardial sites. In theory, however, DA due toinhomogeneity would be transient and eventually becomesCA. On the other hand, DA induced by the combination ofCV restitution and steep APD restitution would be sustainedindefinitely.

Several studies have suggested that DA is associated withsignificant DR compared to CA.1,7 However, direct compari-son between CA and DA under similar circumstances was not

investigated. In the present study, a majority of experimentsexhibited sites of CA and DA in the same heart during thesame pacing protocol, which could be explained by differ-ences in preexisting DR at different sites. This allowed quan-titative comparison, which showed that sites with DA hadsignificantly more DR compared to sites with CA. Previousstudies reported that VT occurred only in the presence of DAand not CA.1 In the present study, as well as in our previousstudies,2 VT could follow both CA and DA, depending pri-marily on the degree of underlying DR. Our observations aremore consistent with the mechanisms of initiation of VT inthe setting of repolarization alternans. Most studies1,5,6 agreethat the arrhythmia is secondary to wave break and reentrantexcitation that could develop during a regular rhythm at a crit-ical short CL, as shown in our previous report,2 or is inducedby a premature stimulus that infringes on the underlying DR,as shown in Figures 5 and 6 in the present study.

Study Limitations

Although the canine anthopleurin-A model is considereda suitable surrogate for the LQT3 syndrome, abnormalitiesof the Na+ channel inactivation are fixed in the clinical syn-drome8 and are not dependent on drug binding to the channelas in the experimental model. Because of the low-affinitybinding of anthopleurin-A to the Na+ channel at depolarizedpotentials and the decreased binding at short CL,9 the repo-larization alternans induced by fast pacing is less stable andpersists for a relatively short period of time. Nevertheless,during transient DA, increased DR still enhances suscepti-bility to reentry. Further, the model may not be suitable foranalysis of the kinetics of restitution during successive short


CLs. However, this will not significantly affect the restitutionkinetics of the first short cycle. In the clinical setting, induc-tion of TWA does not require the degree of abrupt shorteningof CL used in the present study. This distinction is importantfrom a mechanistic point of view because ionic currents thatdetermine action potential alternans during abrupt change inCL may be different from those that determine restitution.10

Finally, there are differences in restitution kinetics in differentspecies, as well as the presence and distribution of M cells.Thus, caution should be exercised in extrapolating some ofthe data to other animal species or to the clinical setting.

References

1. Pastore JM, Girouard SD, Laurita KR, Akar FG, Rosenbaum DS: Mech-anism linking T-wave alternans to the genesis of cardiac fibrillation.Circulation 1999;99:1385-1394.

2. Chinushi M, Restivo M, Caref EB, El-Sherif: Electrophysiologic ba-sis of arrhythmogenicity of QT/T alternans in the long-QT syndrome.Tridimensional analysis of the kinetics of cardiac repolarization. CircRes 1998;83:614-628.

3. Chinushi M, Kozhevnikov D, Caref EB, El-Sherif: Mechanism of thearrhythmogenicity of discordant QT/T alternans. J Am Coll Cardiol2001;37:92A.

4. El-Sherif, N, Caref EB, Yin H, Restivo M: The electrophysiologicalmechanism of ventricular tachyarrhythmias in the long QT syndrome.Tridimensional mapping of activation and recovery patterns. Circ Res1996;79:474-492.

5. Qu Z, Garfinkel A, Chen PS, Weiss JN: Mechanisms of discordant al-ternans and induction of reentry in simulated cardiac tissue. Circulation2000;102:1664-1670.

6. Watanabe MA, Fenton FM, Evans SJ, Hastings HM, Karma A: Mecha-nisms of discordant alternans. J Cardiovasc Electrophysiol 2001;12:196-206.

7. Tachibana H, Kubota I, Yamaki M, Watanabe T, Tomoike H: DiscordantS-T alternans contributes to formation of reentry: A possible mechanismof reperfusion arrhythmias. Am J Physiol 1998;275:H116-H121.

8. Bennett PB, Yazawa K, Makita N, George AL Jr: Molecular mechanismfor an inherited cardiac arrhythmia. Nature 1995;376:683-685.

9. El-Sherif N, Fozzard HA, Hanck DA: Dose-dependent modulation ofthe cardiac sodium channel by the sea anemone toxin ATX-II. Circ Res1992;70:285-301.

10. Saitoh H, Bailey JC, Surawicz B: Alternans of action potential dura-tion after abrupt shortening of cycle length: Differences between dogPurkinje and ventricular muscle fibers. Circ Res 1988;62:1027-1040.

Documents

Mechanism of Discordant T Wave Alternans in the In Vivo Heart