The Interaction of Activation-Repolarisation Coupling and Restitution

CIRCULATIONAHA/2008/785352 1 The Interaction of Activation-Repolarisation Coupling and Restitution Properties in Humans Ben Hanson 1, PhD, Peter Sutton 2, PhD, Nasser Elameri 3, Marcus Gray 4, PhD, Hugo Critchley 4, MB, PhD, Jaswinder S Gill 3, MD, FRCP, Peter Taggart 2, MD, DSc, FRCP. Author Affiliations: 1: Department of Mechanical Engineering, University College London, UK 2: The Hatter Institute, Department of Cardiology, University College Hospitals, UK 3: Department of Cardiology, Guy s and St. Thomas Hospital, London, UK. 4: Brighton and Sussex Medical School, UK. Short title: Activation-repolarisation coupling and restitution. Address for correspondence: Dr Peter Taggart The Hatter Institute, Department of Cardiology, University College Hospital, Grafton Way, London WC1E 6DB Telephone: 00 44 20 380 9888 Fax: 00 44 20 380 9505 Email: peter.taggart@uclh.org

CIRCULATIONAHA/2008/785352 2 Abstract Background Dynamic modulation of repolarisation is important in arrhythmogenesis. An inverse relation exists in myocardium between activation time (AT) and action potential duration (APD). We hypothesised that resulting gradients of APD and diastolic interval (DI) interact with restitution properties and modulate the timing of repolarisation. Methods and Results: Activation-recovery intervals (ARI) were acquired from reconstructed noncontact unipolar electrograms from the left ventricular endocardium in 9 patients (7M) with normal ventricles. At a basic paced cycle length (median 450 ms) ARIs shortened along the path of activation with a mandatory reciprocal increase of DIs. In the median patient this range of DIs started at 230 ms at the site of earliest activation and increased to 279 ms at the site of latest activation at a basic cycle length of 450 ms. Four consecutive standard S1-S2 restitution curves were performed. At sites with a longer ARI (and therefore shorter DI) close to the site of stimulation, premature stimulation produced more shortening of ARIs therefore the time course of restitution was steeper than at more distal sites. At normal heart rate the decrease in ARIs along the conduction pathway compensated for later activation. Thus, dispersion in repolarization time (RT) is smaller than dispersion in ARI in a heart with a steep, negative AT-ARI relationship. This protective effect is lost in hearts without such a relationship. In the patients with a steep AT-ARI relationship at basic cycle length, this relation is lost after premature stimulation and is a function of prematurity. Thus dispersion in RT is larger after shortly-coupled extra stimuli in patients with a steep AT- ARI relationship.

CIRCULATIONAHA/2008/785352 3 Conclusions: A complex interplay exists between activation-repolarisation coupling and restitution properties, largely driven by ARI and diastolic interval gradients. This plays a significant role in the dynamics of repolarisation in humans. Key Words: Electrophysiology; restitution; human heart; conduction; repolarisation; dispersion of repolarisation.

CIRCULATIONAHA/2008/785352 4 Introduction Studies in tissues 1,2, animal hearts 3,4 and humans 4,5,6,7 have shown that as ventricular activation propagates, action potential duration (APD) or activation recovery intervals (ARI), shorten progressively. The resulting inverse coupling between APD and activation time is thought to limit dispersion of repolarisation. Both APD and activation time are strongly cycle length dependent and during premature activation several dynamic processes are engaged including APD and conduction velocity restitution 8,9. Activation / repolarisation coupling generates a gradient of APD along the activation path. We hypothesised that this in turn may create a reciprocal gradient of the subsequent diastolic intervals and thereby interact with restitution. Restitution properties have been shown to modulate the timing of activation and repolarisation in simulation studies 9,10,11,12,13,14 animal models 15,16,17,18 and may be critically important in arrhythmogenesis 12,14,18,19,20,21. In humans, it is unclear at present how APD / activation time coupling and restitution properties interact, and how this affects the dynamic aspects of activation and repolarisation at premature stimulation. Reconstructed electrograms obtained by noncontact mapping of the human ventricular endocardium have been validated as an accurate representation of the electrogram obtained using contact electrodes 6. We have recorded activation recovery intervals (ARI) from the left ventricular endocardium in patients with normal ventricles during a standard restitution pacing protocol, in order to examine the interaction between activation / repolarisation coupling and restitution properties. Methods

CIRCULATIONAHA/2008/785352 5 Patients Nine patients, 3 female, aged 24 to 68, median 59 who were undergoing radiofrequency ablation procedures for supraventricular arrhythmias were studied using noncontact mapping. 4 patients had atrial fibrillation, 4 supraventricular tachycardia and one fascicular tachycardia. Individual patient details are given in Supplemental Table 1. The study was approved by the Guy`s and St Thomas Hospitals Ethics Committee and written informed consent was obtained from all patients. Antiarrhythmic drugs were discontinued for 5 days prior to the study. All studies were performed in the left ventricle prior to the routine clinical procedure. Mapping Procedure The noncontact mapping system (Ensite 3000, Endocardial Solutions Inc) consists of an array of 64 electrodes mounted on an inflatable 7.5ml, 9F balloon catheter together with a recording/graphics workstation. Its use for electroanatomical mapping of the left ventricular endocardium has been described 22. The mapping array was introduced into the left ventricle via the retrograde route through the aortic valve or via a transeptal puncture and across the mitral valve. The array was positioned in the middle of the chamber with the pigtail portion in or near the apex and deployed to obtain a stable position. An anatomy of the left ventricular chamber was obtained using a roving deflectable ablating catheter (Boston scientific Slazer (R)). The validity of the map was checked by examining the map of sinus rhythm propagation within the chamber. No patient was known to have ventricular scar or have disordered conduction due to bundle branch abnormality. The waveform data on the Ensite system was digitised and stored at 1200Hz. Data can potentially be obtained from 3000 points around the chamber. We selected 64 points evenly distributed on the endocardial surface of the chamber from which data were analysed. These points gave a good spatial distribution without an overwhelming amount of data.

CIRCULATIONAHA/2008/785352 6 Stimulation Protocol Pacing was established from the right ventricular apex at a pulse width of 2ms and stimulus strength of 2 x diastolic threshold. After a 3 minute stabilisation period a standard restitution curve was constructed by interposing a test beat after each sequence of 9 basic beats (basic cycle length for median subject = 450ms). The test beat coupling interval was decremented by 50ms to 300ms, and then by 10ms to loss of capture. Four consecutive restitution curves were constructed for each subject under identical conditions at the same basic paced cycle length. For the analysis of restitution data, parameters were identified from each curve and mean values are presented. Illustrations present either the most typical example or a diagrammatic representation of averaged data. Analysis of Data Two methods are available for measuring ARIs, the traditional Wyatt method and an alternative method 23,24,25,26,27. Both have been the subject of theoretical scrutiny 28,29. We obtained similar overall results by both methods (though absolute values of ARI differed) and report here values obtained using the traditional method. Activation and repolarisation times were identified by a computer algorithm based on the time derivative of the recorded signal (dv/dt). These points were individually verified by eye to ensure they had been placed appropriately. Electrograms in which the T waves were inadequately defined were excluded from the analysis. The moments of activation (A) and repolarisation (R) times are shown on figure 1 for the test beat and two preceding basic beats. Activation time (AT) and repolarisation time (RT) were measured from the stimulus time, defined as t = 0. The activation to repolarisation interval (ARI) for a beat is given by the time difference between AT and RT, such that the repolarisation time RT = AT + ARI. Across the ventricle, the local times of activation and repolarisation vary; dispersion of AT and dispersion of RT are

CIRCULATIONAHA/2008/785352 7 defined as the time difference between earliest and latest times. Diastolic interval (DI) for the test beat was calculated as the difference between the measured activation interval A 1 A 2 and the basic beat ARI measured from the preceding beat A 0 R 0. The local activation and repolarisation times were combined with endocardial geometry data to produce animations comparing propagation of basic and premature beats as shown in supplementary data. Statistical Analysis Continuous data that were approximately normally distributed are presented as mean+/- SD. To determine relationships between activation times and the basic beat ARIs, normal least-squares linear regression was performed for all basic beats recorded; coefficients reported in the table are median values. Student s paired t-test was used to determine statistical significance of differences in restitution properties between earliestand latest-activating regions in the ventricle. A paired t-test was also used to determine statistical significance of differences between various parameters for basic beats versus early beats. Exponential models were applied to the recorded ARI restitution data, having the form: ARI( DI) = ARI max e C DI + C 0 1 The model parameters, ARI max, C 0, C 1, were selected and optimised to fit the data to minimise the square error between the model and experimentally recorded ARI values (a linear scale was used to determine error). Here, parameter C 1 defines the value of DI below which restitution begins to have a significant effect on ARI, and C 0 defines the curvature and maximum slope.

CIRCULATIONAHA/2008/785352 8 Statement of Responsibility The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written. Results Activation time ARI relationship at the basic cycle length A total of 36 AT-ARI relationships for the basic paced cycle length were generated in the 9 patients; a typical example is shown in Figure 2a. An inverse relation was observed between local ARIs and activation times: electrode sites activating earliest had longer ARIs and electrode sites activating later had shorter ARIs. ARI values for all patients were 231 +/- 20ms (mean +/- SD) at earliest activation sites vs 203 +/- 22ms at the latest activation sites, p<0.001. Linear regression of ARI on AT resulted in a correlation that varied between patients (median correlation: slope = 0.41, R 2 =0.73). Linear regression coefficients for all patients are presented in Supplemental Table 2. ARI / Diastolic interval at the basic cycle length At the basic cycle length, the progressive shortening of ARIs along the activation path resulted in a corresponding lengthening of the following DI as shown in Figure 2b (Example, patient 5: from 203 ms to 258 ms, range 55 ms at a basic paced cycle length of 400ms). Under a pacing protocol with constant cycle length a reciprocal relationship is mandatory since any variation in ARI must cause an equal and opposite variation in DI. Therefore plots of local ARI vs DI had a linear regression coefficient of -1. Interaction of the ARI / Diastolic Interval regression slope with ARI restitution On the ventricular endocardium there existed a gradient of increasing diastolic intervals along the activation path. In order to examine the interaction between this DI-ARI relationship and ARI restitution we employed a standard restitution pacing protocol. For

CIRCULATIONAHA/2008/785352 9 each test beat, as performed in figure 2b, local ARIs were plotted against their preceding DI and a linear regression line was drawn. Figure 3 shows the combined results at each test coupling interval throughout a restitution protocol (one example shown). The right hand side of figure 3 shows the basic cycle length, having long DIs (range 229 276 ms), and as with figure 2b the regression line has a slope of 1. As the test beat coupling interval shortened, DI decreased and the restitution slope acted to reduce ARIs. Electrode sites having shorter DIs exhibited a larger reduction in ARI compared to electrode sites having longer DIs. This resulted in the progressive flattening of the regression lines as test beat coupling interval shortened: values for the individual regression slopes in this example are given in Supplemental Table 3. From the relationships described in figures 2a and 2b, the earliest activating sites exhibited the longest basic cycle length ARIs and shortest DIs, with the opposite scenario for latest-activating sites. In order to compare the time course of restitution of early vs late-activated sites we plotted mean ARI and DI values for the electrode sites with the 8 shortest and 8 longest DIs selected at the basic cycle length and followed their response to premature test pulses through a standard restitution protocol. Figure 4 shows results from one example patient (No. 1). Exponential curves fitted to the 2 data sets obtained showed that electrode sites at which the basic cycle length ARI was longer had greater amplitude (66 +/- 26 ms vs. 52 +/- 23 ms p<0.005) and followed a steeper time course of restitution. Results for all patients are given in Supplemental Table 4. AT Restitution AT increased at short test pulse intervals. An animation included in supplemental data provides a demonstration of the difference between activation in a basic beat and a closely-coupled test beat. The increase in AT between the RV stimulus site and early activated LV sites where DI was short, was much greater than between early activated

CIRCULATIONAHA/2008/785352 10 LV sites and later activated LV sites where DI was longer (Figure 5a). As activation propagated, these local activation delays summated to generate a cumulative delay along the propagation pathway as shown in figure 5b. Thus the majority of the increase in activation time at the short coupling intervals occurred between the pacing site in the right ventricle and the earliest activated site in the left ventricle with a relatively small further increase during propagation over the left ventricular endocardium. This was a consistent finding, the effect being seen for each patient, i.e. at the basic cycle length the time from stimulus site in the RV to the earliest activating electrode site in the LV for all patients was 60 +/- 15 ms. The time for propagation from the earliest activated LV sites to latest activated LV sites was 54 +/- 13 ms. At short coupling intervals the time from RV stimulus to the earliest activating LV electrode site increased by 28 +/- 9 ms, whereas the time for propagation across the LV increased by 7 +/- 5 ms. Individual values are shown in Supplemental Table 5. Interaction of AT-ARI coupling with ARI and AT restitution An example of activation and repolarisation times for all electrodes during two consecutive basic beats is shown in figure 6a. The time to the first recorded activation represents latency from the S1 stimulus (53 ms), activation was then dispersed over the following 65 ms. Repolarisation started at the site of activation. The convergence of the lines labelled A1 and R1 highlights the progressive shortening of ARIs along the course of propagation in keeping with the negative correlation between ARI and activation time shown in figure 2a. This created reciprocal lengthening of the following diastolic intervals, indicated by the divergence of lines R1 and A2. Figure 6b is a similar representation for a basic beat followed by a closely-coupled test beat. Activation of the test beat (A2) is slightly slower than for the basic beat (71 ms vs. 65 ms); the line Y is drawn parallel to A1 as a visual aid to the slower activation of A2.

CIRCULATIONAHA/2008/785352 11 Values for all patients are provided in Supplemental Table 5. Despite the increasing gradient of DI along the propagation path, there is no increasing gradient of ARI. Dispersion of AT and Repolarisation For the basic beat, the mean value for dispersion of repolarisation was 53 +/- 19 ms (coefficient of variation = 0.18 +/- 0.06). In order to quantify numerically the relationships illustrated in figure 6, basic beat dispersion of repolarisation for each patient (Table 1) was plotted against their AT-ARI regression slope (from Supplemental Table 2). There was found to be a strong correlation, R 2 = 0.87 as shown in figure 7a. Therefore the weaker the AT-ARI correlation, the greater the dispersion of repolarisation. Since the duration of repolarisation also depends on the duration of activation, which varied between patients (33 ms to 69 ms), we normalised for this variation. We considered the decrease in dispersion of repolarisation as a fraction of the preceding activation dispersion. From the linear relationships above, the fractional reduction in dispersion should be equal to the slope of the AT-ARI regression line (delta ARI / delta AT). We scaled each patient s activation dispersion down by a factor equal to the slope of the AT-ARI regression line. This gave a prediction of the repolarisation dispersion that was closely related to the measured dispersion, as shown in figure 7b, R 2 = 0.86. For the early beat, dispersion of activation was increased by a small amount compared to the basic beat as shown in figure 6b and Supplemental Table 5. The relationship of decreasing ARI with AT was no longer present, indicated by the lines A2 and R2 being parallel, offering no reduction in the dispersion of repolarisation. Therefore the dispersion of repolarisation was longer for the early beat than for the basic beat (in the example of figure 6b: 84 ms vs 50 ms). Values for all patients are given in Table 1.

CIRCULATIONAHA/2008/785352 12 In patients exhibiting a strong AT-ARI correlation at basic cycle length, the dispersion of RT is less than the dispersion of ARI. Across all patients dispersion of RT was 53 +/- 19ms and dispersion of ARI was 53 +/-10 ms at long coupling interval. As restitution engages, the AT-ARI relationship has been shown to diminish resulting in the loss of this protective mechanism. At short coupling intervals the dispersion of RT is considerably greater than the dispersion of ARI (across all patients: 75 +/- 19 vs. 60 +/- 13 ms at shortest cycle length). Discussion These studies using non contact mapping in humans demonstrate a complex interaction between activation-repolarisation coupling and restitution properties, and provide insight into the effect of this interaction on the timing and dispersion of activation and repolarisation. Specifically we have shown that; (1) ARIs shortened progressively along the propagation path of activation creating a negative ARI gradient; (2) the ARI gradient generated a reciprocal gradient of increasing diastolic intervals for the following beat; (3) AT restitution delayed activation of test beats at short S1-S2 intervals thereby further increasing the range of diastolic intervals prior to the test beat; (4) The steepness of ARI restitution varied systematically along the path of propagation, being steeper at electrode sites with longer basic beat ARIs and shorter DI, ie. closer to the activation site, as shown in figure 4. The consequences of these effects were that at long coupling intervals when diastolic intervals fell on the restitution plateau, a negative AT-ARI correlation tended to reduce dispersion of repolarisation; a strong correlation was present between the regression coefficient of the AT-ARI relationship and dispersion of repolarisation R 2 =0.87. As S1- S2 shortened the situation changed when the diastolic intervals fell on the slope of ARI and AT restitution: then a complex interaction occurred between the AT-ARI relationship

CIRCULATIONAHA/2008/785352 13 and ARI and AT restitution resulting in an increased dispersion of repolarisation at the shortest intervals. A diagrammatic representation of these effects is provided in supplementary data: figure SD3. AT-ARI coupling A negative correlation between activation time and either ARI or APD has previously been reported in experimental models 1,2,3,4 and humans 4,5,6,7. The correlations of ARI with AT found in our study are consistent with values reported in these studies. A likely mechanism is local electrotonic current flow via gap junctions between cells which tends to equalise action potential durations. This coupled with the later repolarisation of cells downstream and electrotonic current flow from cells downstream to cells upstream would facilitate the AT-ARI gradient 30,31,32. A further illustration in supplemental data, SF1, shows this relationship still in evidence where some local short-circuiting of activation through the Purkinje system may have occurred. It can be seen in figure 6a that repolarisation of the basic beat commences at the earliest activated electrode sites where ARIs are longest. It has been a point of discussion as to whether repolarisation propagates from the opposite direction to the activation sequence, or commences at the site of activation. In all our patients repolarisation started at the site of activation, as can be seen in the animation in the supplemental data section. Similar concordant activation / repolarisation sequences have been observed in pigs 3, canines 33 and humans 32,34. As pointed out by Yue and colleagues 6 if an AT / repolarisation slope is more negative than 1, repolarisation should complete earliest at sites of latest activation, and if the slope were more positive than -1 the AT / repolarisation sequence would be concordant. In our studies the slopes were all more positive than 1. AT-ARI coupling and ARI restitution

CIRCULATIONAHA/2008/785352 14 One study in humans showed a flattening of the AT-ARI regression slope at a medium and a short steady-state pacing cycle length suggesting a cycle length dependence 6. The effect of AT-ARI coupling that we observed was to generate a gradient of decreasing ARIs along the course of activation which in turn created a reciprocal gradient of the following DI. Diastolic interval and steady state APD are major determinants of the time course of APD restitution 8,35 suggesting the possibility that AT- ARI coupling may generate a range of restitution profiles along the path of activation. In this study we examined the effect of a restitution pacing protocol on the ARIs and corresponding DIs along the propagation pathway. The interactions are illustrated in the supplemental data section, figure SF2. At long coupling intervals the DI was long such that the following ARI fell on the restitution plateau. The DI gradient then exerted little or no effect on the subsequent ARI. As the S1-S2 coupling interval shortened the situation was different as DIs shortened and the restitution slope was engaged. It has been shown experimentally that the steepness of the APD restitution curve is partly dependent on the steady state APD whereby a longer basic beat APD is associated with steeper restitution 35. Therefore the gradient of ARIs generated during activation may be expected to result in a steeper restitution time course closer to the activation site where ARIs are longer. An alternative explanation might be that at either a very short basic cycle length, or in shortcoupled premature beats, the amount of IKs current is probably so large that it permits very fast repolarization leading to short APDs. However, this also prevents the lateactivated myocytes from shortening further due to electrotonic interaction. These myocytes already have short APDs due to the large IKs conductance. This phenomenon was observed in our study: the earliest activated sites (having longer ARIs) followed a different time course of restitution compared to the latest activated sites (having shorter ARIs) (Figure 3). Consequently, as the test beat coupling interval was reduced, the longer ARIs shortened more than the shorter ARIs (figure 4).

CIRCULATIONAHA/2008/785352 15 Therefore, for a very premature beat, the ARIs ceased to systematically decrease along the path of activation propagation. From figure 6b, it is apparent that local variation in ARI still exists across the ventricle at short coupling intervals, however this variation of ARI was virtually independent of DI, thus the protective mechanism acting to reduce dispersion of repolarisation has been lost. The situation here is in many ways analogous to the guinea pig heart model in which heterogeneous restitution was generated by regional differences in intrinsic cellular ionic currents 16. In this model the restitution curve with the longer basic beat APD had a faster time course of restitution, ie. a steeper curve, compared with the site where the basic beat APD was shorter. As the test beat coupling interval shortened the longer APDs shortened more than the shorter APDs resulting in an alteration of the APD gradient between the two regions. Inspection of figure 2 in the study by Laurita and colleagues 16 shows a striking similarity to the ARI changes we observed (Figures 3 and 4). We cannot exclude the possibility that regional differences in ionic currents contributed to our results. However we would consider such an effect to be minor since to the best of our knowledge no systematic ionic gradients of the type present in guinea pig have been demonstrated in humans; and in addition different anatomical pathways of activation produced an overall similar pattern of AT-ARI coupling. We were not able to demonstrate a reversal of the ARI gradient at the shortest coupling intervals similar to the APD gradient reversal observed in the guinea pig model by Laurita and colleagues 16. This may be due to model dependent differences and in particular the inability in our study to acquire data at very short diastolic intervals. This was partly because the minimum diastolic interval attainable increases with increasing distance from the pacing site resulting in some degree of attenuation of the restitution profiles at the more distal sites, as seen in figure 4. In this regard it should be emphasised that the restitution profiles in the present study are not intended as

CIRCULATIONAHA/2008/785352 16 representations of local restitution curves but as representations of the restitution profile encountered by the propagating wavefront. AT-ARI coupling and activation time restitution Activation time restitution operated over a relatively narrow range of DIs, in keeping with the conduction velocity restitution curves reported by Yue et. al. 36. The critical coupling intervals below which ARIs/APDs start to shorten and conduction velocity starts to decrease are not necessarily the same; supplemental figure SF3 illustrates the separate effects. As seen in figure 5a the7 shortest DIs were attained near the activation site compared to late activated sites due to the time occupied by the spread of activation. Thus the impact of conduction velocity restitution on AT is greater closer to the activation site; AT restitution resulted in only a small (but significant) increase in activation time between early and late-activated sites. On the basis of these results we would speculate that factors that increase ARI and AT restitution would be expected to increase the dispersion of repolarisation at a short coupling interval. Methodological Considerations Electrograms recorded by the noncontact method we employed have been validated as providing a highly accurate representation of the local electrogram recorded using contact electrodes 6. The validation for the use of non-contact mapping for repolarisation study has been found to be less strong than the correlation reported elsewhere for activation 6,27. However this has provided sufficient resolution for studies examining the construction of restitution curves 36 and the AT-ARI relationship 6. The traditional method (Wyatt) of measuring ARIs from these waveforms, although having strong experimental and theoretical validation 23,24,25,28,29, has been suggested to underestimate the ARI for electrograms with positive T waves 3,26,27. In our study, the

CIRCULATIONAHA/2008/785352 17 ARIs observed at the point of refractoriness in some cases reached as short as 140ms, which concurred with the ARIs reported by Ramdat Misier et. al. 37 using VF intervals as a surrogate measure of refractoriness in patients with VT/VF. An alternative method has been used by some workers whereby the repolarisation moment of the ARI is measured as dv/dt min of the T wave rather than dv/dt max as in the traditional method. However, since the global distribution of T wave morphology has been shown to remain constant with only small changes limited to border zones 6 over cycle lengths from steady state to refractoriness, any difference between the two methods should remain consistent over the range of perturbations we employed. We obtained qualitatively similar results by both methods and report here results using the traditional method. Clinical Implications The timing of activation, repolarisation and dispersion of repolarisation are important in the initiation and maintenance of reentrant ventricular arrhythmias 38,39,40. The dynamic interplay that we have demonstrated between AT-ARI coupling and restitution properties suggests that their effects on both activation and repolarisation are closely linked and may play an important role in arrhythmogenesis. Our results also suggest that strong AT-ARI coupling may be protective at the basic cycle length by limiting dispersion of repolarisation, but that the situation may be much more complex at short coupling intervals due to interaction with restitution. Several studies have defined APD and conduction velocity restitution conditions that facilitate wavebreak, and in particular the protective effect of flattening APD restitution 9,12,19. Our results suggest that any variation in the slope of APD restitution, for example by autonomic effects 41 or drugs may also modulate the time course of repolarisation and dispersion of repolarisation by interaction with AT-APD coupling. Our results are more likely to apply to conduction initiated by an ectopic focus rather than normal anterograde activation of the ventricle via the Purkinje system, since ARI dispersion depends heavily on a gradient of AT and the latter is greater during unifocally-initiated propagation though ventricular muscle.

CIRCULATIONAHA/2008/785352 18 The dynamic interplay that we report here may be a fundamental property of wave propagation and therefore integral to the behaviour of restitution properties in humans. Limitations Our studies were confined to the left ventricular endocardium and therefore do not provide information on transmural gradients or other regions of the heart. In the interpretations offered we have not taken account of regional differences in intrinsic electrophysiological properties. In the protocol we employed the basic beat and the premature test beat were initiated from the same pacing site. A premature beat arising from a location different from the basic beat would be expected to modify the direction of the gradients we observed. ln addition many other factors are likely to play an interactive role with the parameters we have studied including gap junctions 42, anisotropy 43 and other time dependent mechanisms such as myocardial memory 18,44. Conclusions We have shown that AT-ARI relationship generates gradients of ARI and DI along the propagation pathway resulting in systematic variation in the restitution profile encountered by a propagating wavefront. This may be a fundamental property of human myocardium of importance in arrhythmogenesis. Funding Sources MAG and HDC were supported by the Wellcome Trust via a Programme Grant to HDC. Acknowledgments (None) Conflict of Interest Disclosures

CIRCULATIONAHA/2008/785352 19 None.

CIRCULATIONAHA/2008/785352 20 References 1. Burgess MJ, Steinhaus BM, Spitzer KW, Green LS. Effects of activation sequence on ventricular refractory periods of ischemic canine myocardium. J Electrocardiology 1985:18:323-329. 2. Osaka T, Kodama I, Tsuboi N, Toyama J, Yamada K. Effects of activation sequence and anisotropic cellular geometry on the repolarization phase of action potential of dog ventricular muscles. Circulation. 1987; 76: 226 236. 3. Gepstein L, Hayam G, Ben Haim SA. Activation-repolarization coupling in the normal swine endocardium. Circulation. 1997; 96: 4036 4043. 4. Yuan S, Kongstad O, Hertervig E, Holm M, Grins E, Olsson B. Global repolarization sequence of the ventricular endocardium: monophasic action potential mapping in swine and humans. Pacing Clin Electrophysiol. 2001; 24: 1479 1488. 5. Franz MR, Bargheer K, Rafflenbeul W, Haverich A, Lichtlen PR. Monophasic action potential mapping in human subjects with normal electrocardiograms: direct evidence for the genesis of the T wave. Circulation. 1987;75:379-386. 6. Yue AM, Betts TR, Roberts PR, Morgan JM. Global dynamic coupling of activation and repolarization in the human ventricle. Circulation. 2005;112:2592-2601. 7. Chauhan VS, Downar E, Nanthakumar K, Parker JD, Ross HJ, Chan W, Picton P. Increased ventricular repolarisation heterogeneity in patients with ventricular

CIRCULATIONAHA/2008/785352 21 arrhythmia vulnerability and cardiomyopathy: a human in vivo study. Am J Physiol Heart Circ Physiol 2006;290:H79-H86. 8. Boyett MR, Jewell BR. A study of the factors responsible for rate-dependent shortening of the action potential in mammalian ventricular muscle. J Physiol. 1978;285:359 380. 9. Qu Z, Weiss JN, Garfinkel A. Cardiac electrical restitution properties and stability of reentrant spiral waves: a simulation study. Am J Physiol. Heart Circ Physiol 1999;45:H269-H283. 10. Nolasco JB, Dahlen RW. Agraphic method for the study of alternation in cardiac action potentials. J Appl Physiol 1968;25:191-196. 11. Karma A. Electrical alternans and spiral wave breakup in cardiac tissue. Chaos 1994;4:461-472. 12. Garfinkel A, Kim Y-H, Voroshilovsky O, Qu Z, Jong RK, LeeM-H, Karagueuzian, Weiss J, Chen P-S. Preventing ventricular fibrillation by flattening cardiac restitution. Proc Roy Acad Sci 2000;97:6061-66. 13. Clayton RH, Taggart P. Regional differences in APD restitution can initiate wavebreak and reentry in cardiac tissue: A computational study. Biomed Eng Online 2005;4:54 14. Kelderman RH, ten Tusscher KHWJ, Nash MP, Hren R, Taggart P, Panfilov AV. Effect of heterogeneous restitution on VF organization in a model of the human ventricles, Am J Physiol. 2008; 294:H764-H774.

CIRCULATIONAHA/2008/785352 22 15. Rosenbaum DS, Kaplan DT, Kanai A, Jackson L, Garan H, Cohen RJ, Salama G. Repolarization inhomogeneities in ventricular myocardium change dynamically with abrupt cycle length shortening. Circulation. 1991;84:1333 1345. 16. Laurita KR, Girouard SD, Rosenbaum DS. Modulation of ventricular repolarization by a premature stimulus: role of epicardial dispersion of repolarization kinetics demonstrated by optical mapping of the intact guinea pig heart. Circ Res. 1996; 79: 493 503. 17. Laurita KR, Girouard SD, Akar FG, Rosenbaum DS. Modulated dispersion explains changes in arrhythmia vulnerability during premature stimulatio of the heart. Circulation 1998;98:2774-2780. 18. Huang J, Zhou X, Smith WM, Ideker RE. Restitution properties during ventricular fibrillation in the in situ swine heart. Circulation 2004;110:3161-3167.J 19. Weiss JN, Garfinkel, Karagueuzian HS, Qu Z, Chen PS. Chaos and the transition to ventricular fibrillation: a new approach to antiarrhythmic drug evaluation. Circulation 1999;99:2819-2826. 20. Pak H-N, Hong SJ, Hwang GS, Lee HS, Park S-W, Ahn JC, Ro YM, Kim Y-H. Spatial dispersion of action potential duration restitution restitution kinetics is associated with induction of ventricular tachycardia/fibrillation in humans. J Cardiovasc Electrophysiol 2004;15:1357-1363. 21. Pastore JM, Laurita KR, Rosenbaum DS. Importance of spatiotemporal heterogeneity of cellular restitution in mechanism of discordant alternans. Heart Rhythm 2006;6:711-719.

CIRCULATIONAHA/2008/785352 23 22. Gornick CC, Adler SW, Pederson B, Huack J, Budd J, Schweitzer J. Validation of a new noncontact catheter system for electroanatomic mapping of the left ventricular endocardium. Circulation 1999;99:829-835. 23. Wyatt RF, Burgess ML, Evans AK, Lux RL, Abildskov JA, Tsutsu T. Estimation of ventricular transmembrane action potential durations and repolarization times from unipolar electrograms. Am J Cardiol. 1981;47:488 24. Millar CK, Kralios FA, Lux RL. Correlation between refractory periods and activation-recovery intervals from electrograms: effects of rate and adrenergic interventions. Circulation. 1985; 72: 1372 1379. 25. Haws CW, Lux RL. Correlation between in vivo transmembrane action potential durations and activation recovery intervals from electrograms:effect of interventions that alter repolarization time. Circulation. 1990;81:281-288. 26. Chen PS, Moser KM, Dembitsky WP, Auger WR, Daily PO, Calisi CM, Jamieson SW, Feld GK. Epicardial activation and repolarizatin patternsin patients with right ventricular hypertrophy. Circulation 1991;83;104-118. 27. Yue AM, Paisey JR, Robinson S, Betts TR, Roberts PR, Morgan JM. Determination of human ventricular repolarization by noncontact mapping: validation with monophasic action potential recordings. Circulation. 2004; 110: 1343 1350.

CIRCULATIONAHA/2008/785352 24 28. Steinhaus BM. Estimating cardiac transmembrane activation and recovery times from unipolar and bipolar extracellular electrograms: a simulation study. Circ Res. 1989; 64: 449 462. 29. Coronel R, de Bakker JMT, Wilms-Schopman FJG, Opthof T, Linnebank AC, Belterman CN, Janse MJ. Monophasic action potentials and activation recovery intervals as measures of action potential duration: Experimental evidence to resolve some controversies. Heart Rhythm 2006;3:1043-1050. 30. Toyoshima H, Burgess M. Electrotonic interaction during canine ventricular repolarization. Circ Res 1978;43:348-356. 31. Hoffman BF. Electrotonic modulation of the Twave. Am J Cardiol. 1982;50 32. Conrath CE, Wilders R, Coronel R, de Bakker JMT, Taggart P, de Groot JR, Opthof T. Intercellular coupling through gap junctions masks M cells in the human heart. Cardiovasc Res 2004;62:407-414. 33. Janse MJ, Sosunov EA, Coronel R, Opthof T, Anyukhovsky EP, de Bakker JM, Plotnikov AN, Shlapakova IN, Danilo P Jr Tijssen JG, Rosen MR. Repolarization gradients in the human left ventricle before and after induction of short-term cardiac memory. Circulation 2005;112:1711-1718. 34. Conrath CE, Opthof T. Ventricular repolarisation: An overview of (patho)physiology, sympathetic effects and genetic aspects. Prog Biophys Mol Biol 2006;92:269-307.

CIRCULATIONAHA/2008/785352 25 35. Elharrar V, Surawicz B. Cycle length effect on restitution of action potential duration in dog cardiac fibers. Am J Physiol. 1983;244:H782 H792. 36. Yue AM, Franz MR, Roberts PR, Morgan JM. Global endocardial electrical restitution in human right and left ventricles determined by noncontact maping. J Am Coll Cardiol 2005;46(6):1067-1075. 37. Ramdat Misier AR, Opthof T, van Hemel NM, Vermeulen JT, de Bakker JMT, Defauw JJAM, van Capelle FJL, Janse MJ. Dispersion of refractoriness in noninfarcted myocardium of patients with ventricular tachycardia or ventricular fibrillation after myocardial infarction. Circulation 1995;91:2566-2572. 38. Han J, Moe GK. Nonuniform recovery of excitability in ventricular muscle. Circ Res. 1964; 14: 44 60. 39. Kuo CS, Munakata K, Reddy CP, Surawicz B. Characteristics and possible mechanism of ventricular arrhythmia dependent on the dispersion of action potential durations. Circulation. 1983; 67: 1356 1367. 40. Gough WB, Mehra R, Restivo M, Zeiler RH, el Sherif N. Reentrant ventricular arrhythmias in the late myocardial infarction period in the dog, 13: correlation of activation and refractory maps. Circ Res. 1985; 57: 432 442. 41. Taggart P, Sutton P, Chalabi Z, Boyett MR, Simon R, Elliott D, Gill JS. Effect of adrenergic stimulation on action potential duration restitution in humans. Circulation 2003;107;285-89.

CIRCULATIONAHA/2008/785352 26 42. Poelzing S, Akar FG, Baron E, Rosenbaum DS. Heterogenious connexin 43 expression produces electrophysiological heterogeneities across the ventricular wall. Am J Physiol Heart Circ Physiol 2004;286:H2001-H2009 43. Spach MS, Barr RC. Ventricular intramural and epicardial potential distributions during ventricular activation and repolarization in the intact dog. Circ Res. 1975; 37: 243 257. 44. Rosen MR. What is cardiac memory? J Cardiovasc Electrophysiol 2000;11:1289-1293.

CIRCULATIONAHA/2008/785352 27 Legends for Figures Figure 1. Representative non-contact electrograms, indicating activation and repolarisation times and the method of determining diastolic interval (see text). Figure 2a. Example of the inverse correlation between activation recovery intervals (ARI) and activation times (AT) on the left ventricular endocardium for the basic beat (Patient 5, basic cycle length = 400 ms). Figure 2b. The reciprocal shortening of activation recovery intervals (ARI) and lengthening of diastolic interval (DI) along the propagation path results in an ARI / diastolic interval relationship with a slope of -1 for the basic beat. In the example shown, slope = 1.00, R 2 = 0.97 (Patient 5, basic cycle length = 400 ms). Figure 3. Activation recovery intervals (ARIs) over the endocardium plotted against their diastolic intervals at each of the pacing intervals during a standard restitution pacing protocol. One curve is presented, representative of the four recorded for that patient (No. 1). Linear regression lines are shown for ARI vs diastolic interval at each test pulse interval; values for the slopes are shown in Supplemental Table 3. Regression lines for the basic beat and long test pulse intervals (right) have steep slopes which flatten progressively as the test beat coupling interval shortens and restitution is engaged (left). Figure 4. In order to compare the time course of restitution of early vs late-activated sites, activation recovery intervals (ARIs) and diastolic intervals (mean +/- SD for both) are presented for the 8 electrodes at which activation times were longest and for the 8 electrodes at which activation times were shortest (for the basic beat). These are shown for each coupling interval during a standard restitution pacing protocol. One curve is

CIRCULATIONAHA/2008/785352 28 presented, representative of the four recorded for that patient (No. 1). Early activating electrode sites, having longer ARIs for the basic beats follow a steeper time course of restitution than late activating electrode sites which have shorter ARIs. Exponential curve models are shown; parameters for these models are presented in Supplemental Table 4. Figure 5a. Activation time differences between the RV pacing site and the earliest LV activation site (lower panel), between this site and mid activation sequence (middle panel) and between mid activation and the site of latest LV activation (top panel). These values are plotted against the diastolic interval at the respective sites. Activation time prolongs markedly at early-activated sites where diastolic interval is short, and less so at later-activated sites where diastolic interval is longer. (Patient No. 1) Figure 5b. Activation times are shown for each coupling interval during a standard restitution protocol for one example patient (No. 1). Mean data are shown for the 8 electrode sites at which activation times were longest and for the 8 electrode sites at which activation times were shortest. These are plotted against their local diastolic intervals. Activation time prolongs at short pacing intervals at the early activated sites, but shows only a small further increase between the earliest and latest activated sites. Values are presented in Supplemental Table 5. Figure 6a. Diagrammatic illustration of real data against time on the x-axis (ms). Line A1 shows activation time at each electrode site together with the corresponding repolarisation time (R) for 2 consecutive basic beats, with stimulus S 1 at t=0. Times for each individual electrode site are distributed along the y-axis according to the sequence of their basic beat activation time (ms), thus the y-axis represents progression along the basic beat propagation path.

CIRCULATIONAHA/2008/785352 29 Figure 6b. Diagrammatic illustration using real data as in figure 6a, for a basic beat followed by a closely coupled test beat. The dashed line Y is drawn parallel to the activation time line A1 as a visual aid (see text). Dispersions of activation and repolarisation times (ms) are labelled underneath the figure. Figure 7a Measured dispersion of repolarisation for basic beats plotted against the AT-ARI regression coefficient for all patients. Figure 7b Measured dispersion of repolarisation for basic beats plotted against the dispersion predicted from activation dispersion and the strength of the AT-ARI relationship.

CIRCULATIONAHA/2008/785352 30 Table 1. Dispersion of Repolarisation for Basic Beat and Early Beat for all patients Basic Beat Earliest CL Captured P CI Dispersion Mean Coeff. CI Dispersion Mean Coeff. RT Var RT Var 1 450 50 279 0.18 220 84 250 0.34 2 450 47 275 0.17 200 69 268 0.26 3 400 38 279 0.14 200 79 262 0.30 4 450 78 319 0.24 Insufficient Early Beat Data 5 400 39 264 0.15 210 41 249 0.17 6 450 86 319 0.27 250 109 294 0.37 7 450 59 285 0.21 210 67 219 0.31 8 500 73 297 0.25 240 74 255 0.29 9 450 30 323 0.09 220 75 326 0.23 Mean 444 53 290 0.18 219 75 265 0.30 SD 32 19 21 0.06 18 19 32 0.06 P value <0.005 <0.01 0.0005 CI = Coupling Interval (ms), RT = Repolarisation Time (Activation Time + Activation Recovery Interval) (ms), Coeff. Var = Coefficient of Variance (Standard deviation as a proportion of mean value). P values presented are for comparison between Basic Beat and Early Beat results. All data from Patient 4 excluded from analysis due to insufficient early beat data.

Figure 1 DI A 0 R 0 A 1 A 2 R 2 0 200 400 600 800 1000 1200 Time, ms

Figure 2A 200 190 Slope = -0.80 R 2 = 0.94 ARI 180 170 160 150 0 10 20 30 40 50 Activation Time

Figure 2B 200 190 Early activating electrode sites 180 ARI 170 160 150 140 Late activating electrode sites 200 220 240 260 Diastolic Interval

Figure 3 Global DI - ARI relationship changing with premature coupling interval 250 Early activating electrode sites ARI 200 150 ARI Late activating electrode sites 100 50 100 150 200 250 Diastolic Interval

Figure 4 240 220 Early activating electrode sites ARI 200 180 160 140 Amplitude = 71 Late activating electrode sites Amplitude = 46 120 0 50 100 150 200 250 Diastolic Interval

Figure 5A 40 Latest-activated Site 30 0 50 100 150 200 250 300 40 Activation Time Interval 30 0 50 100 150 200 250 300 90 80 70 Earliest-activated Site 60 50 0 50 100 150 200 250 300 Diastolic Interval

Figure 5B 180 Activation Time, ms 140 100 60 Dispersion = 71 Late activating electrode sites Early activating electrode sites Dispersion = 65 20 50 100 150 200 250 Diastolic Interval

Figure 6A Activation - Repolarisation of Two Consecutive Basic Beats Time of activation of 64 electrode sites 120 100 80 60 Latest Activated Earliest Activated A1 R1 A2 R2 0 200 400 600 800 Time ARI DI ARI Figure 6B Y Time of activation of 64 electrode sites 120 100 80 60 Latest Activated Earliest Activated A1 R1 A2 R2 0 200 400 Time 65 50 84 71 Dispersion of Activation and Repolarisation

Figure 7 A Repolarisation Dispersion, ms 80 60 40 20 y = 67x + 82 R 2 = 0.87 0-1 -0.8-0.6-0.4-0.2 0 Slope of ARI/AT relationship B Repolarisation Dispersion, ms 80 60 40 20 0 y = 0.93x + 20.2 R 2 = 0.86 20 40 60 80 Predicted Repolarisation Dispersion, ms