• CHAOTIC CARDIAC DYNAMICS VOLUME I A Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy @ Michael Raymond Guevara Department of Physiology McGill University Montreal February, 1984 ABSTRACT Experiments are described in which the electrical activity of a spontaneously beating aggregate of embryonic chick heart cells is altered by intracellular injection of a periodic train of current pulses. The coupling patterns set up between the electronic stimul a tor and the aggregate are de se ri bed as a function of the stimulus amplitude and the stimulation frequency. Various periodic and nonperiodic patterns are found, most of which resemble dysrhythmic patterns seen clinically in the human heart. The response of the aggregate to an isolated current pulse delivered at various phases of its spontaneous cycle is also investigated. 1t is shown that knowledge of this single pulse response suffices to explain and even predict the response of the aggregate to periodic stimulation. In certain circumstances, the dynamics seen during periodic stimulation can then be identified as "chaotic ... 0 , , RESUME Nous d~crivons des exp~riences ob 1 'activit& ~lectrique d'agregats, spontanement actifs, de cel1ules cardiaques ct•embryons de poulets est modifi&e par 1 'application repet&e, au nfveau intracellulaire, d'fmpulsions electriques. Les reponses de couplage etab1ies entre le stimulateur electronique et 1 'agregat sont decrites a l'aide d'une fonction dependant de l'ampl itude et de la fr&quence de stimulation. Diverses reponses, periodiques ou non, sont obtenues, la plupart s'approchant des reponses arrhythmi ques observees, en cl i ni que, dans 1e coeur humai n. a une etudions aussi la reponse d'un agregat impulsion appliquee a diverses phases de son cycle spontane. Cette reponse stimulation isolee peut alors permettre d'expliquer~ predire, la reponse de 1 'agr~gat a la Nous stimulation a la et meme de p~riodique. certains cas, l'activite electrique sous stir.lUlation periodique peut etre qualifiee de "chaotique11 ( traduit par J. Bel air} • Dans To my Mother, Dorothy Guevara (nee Telfer) c 0 . ACKNOWLEDGEMENTS 0 I wish to thank Dr. Leon Glass for encouraging me to pursue graduate studies and for advising on the work presented in this thesis. The experimental results were obtained in the laboratory of Dr. Alvin Shrier, who introduced a novice into the rather arcane world of cardiac electrophysiology. Dr. John Clay guided me through the intricacies of ionic modelling of cardiac cells. The work recounted in this thesis grew directly out of two influences in my 1ife. I must thus first acknowledge the medical and nursing staff of the emergency room of the Queen Elizabeth Hospital of Montreal (1975-1978), who encouraged and assisted me in my early attempts to comprehend cardiac dysrhythmias; I must then thank Leon Glass· and Michael C. t~ackey for showing me that there existed analytical tools with which one could attack the probl er:1. The expert laboratory assistance of Diane Colizza, Ken Rozansky, Glen Ward, Richard Brochu, Robert Lowsky, and Howard Dubarsky has always been very much appreciated and is gratefully acknowledged. I thank John Knowles, Guy Isabel, and Albert Hagemann for their practical help in matters electronic and mechanical. Nelson Publicover encouraged me and gave valuable advice in the early stages of setting up computer systems. Dr. J.S. Outerbridge was always available for consultation concerning numerical techniques. A special note of acknowledgement must go to Peter Krnjevic, whose advice on computers and computing was 0 i V invaluable, and at 0 ~1 times freely and cheerfully given. I thank Sandra James and Christine Pamplin for typing the thesis at very short notice in their usual careful style. photographed the figures. Guy L Heureux expertly 1 I wish to thank Professor Michael C. Mackey, Dr. Jacques Belair, Diane Colizza, and my fellow graduate students Carl Graves, Ralf Siegel, and Ehud Isacoff for being always willing to lend an attentive (if somewhat skeptical!) ear. Finally, I am again grateful to my wife Diane, this time for her patience and understanding during the trying circumstances surrounding the final stages of completion of this manuscript. I owe a special debt of gratitude to my mother, Dorothy Guevara, who, by dint of her perseverence and sacrifice, has been the person most responsible for putting me into a position that allowed carrying out the work described in this thesis. During the second half of my tenure as a graduate student, I held a research traineeship awarded by the Canadian Heart Foundation. Earlier financial assistance was obtained from research grants awarded jointly to M.C. Mackey and L. Glass by the Natural Sciences and Engineering Research Council of Canada, and to A. Shrier and L. Glass by the Canadian Heart Foundation. I once again gratefully acknowledge the support provided me by the aforesaid individuals and institutions, without which the work described in this thesis would not have been possible. V FOREWORD 0 t4y main goal in writing this thesis is to suggest to the reader that, even though the dynamics of the heart is very complex, there now exists a mathematical framework which begins to approach and perhaps even encompass that complexity, and in which that complexity naturally arises. mathematics of Chaos 11 11 • Part of this framework includes the The term chaos is used here in a technical sense which does not deviate too much from its everyday meaning of disorder or confusion. However, these synonyms are somewhat misleading, since chaotic dynamics, though highly complex, is deterministic and has an intricate highl y-order.ed internal structure. It may thus be better termed "ordered chaosu. The layout of the thesis is as follows. In CHAPTER 1, I recount some of the early work done in describing the various patterns of mechanical and electrical activity that can be seen in the heart. I next show that these behaviours can be described mathematically and briefly introduce the topic of chaotic dynamics. In CHAPTER 2, the experimental response of a cardiac oscillator to perturbation with single pulses of current is described; in CHAPTER 3, I show that an ionic model assembled from voltage cl amp data reproduces many of these experimental results. The experimental response of the cardiac oscillator to stimulation with a periodic train of current pulses is next detailed in CHAPTER 4. 0 vi In CHAPTER 5, I show that the response to single pulses dec ri bed ·; n CHAPTER 2 0 can be used to predict the experimental response to periodic stimulation that was described in CHAPTER 4. presents some overall conclusions. vii Finally, CHAPTER 6 TABLE OF CONTENTS 0 PAGE CHAPTER 1. INTRODUCTION: REGULAR AND IRREGULAR CARDIAC RHYTHMS, THEIR 1-1 MATHEMATICAL DESCRIPTION, AND CHAOTIC DYNAMICS 2. PHASE RESETTING OF THE RHYTHM OF SPmJTANEOUSLY BEATING 2-1 AGGREGATES OF EMBRYONIC CHICK VENTRICULAR CELLS BY A CURRENT PULSE OF BRIEF DURATION 3. THE IONIC BASIS OF SPONTANEOUS ACTIVITY AND PHASE RESETTING IN THE AGGREGATE: 3-1 NUMERICAL INVESTIGATION OF A PARTIAL ~IODEL 0 4. PERIODIC STIMULATION OF SPONTANEOUSLY BEATING AGGREGATES OF EMBRYONIC CHICK VENTRICULAR CELLS ~HTH 4-1 CURREUT PULSES OF BRIEF DURATION 5. PREDICTION OF THE RESPONSE OF THE AGGREGATE TO PERIODIC 5-1 STif.1ULATION FROt4 ITS RESPONSE TO STI!'tlULATION WITH SINGLE PULSES 6. CONCLUSIONS 6-1 7. BIBLIOGRAPHY 7-1 viii 0 11 one can perhaps say that the nonlinear field also presents a kind of a paradise of potential possibilities, once the 11 11 theoretical knowledge of these numerous phenomena is supplemented by adequate means for their experimental real i zati on. 11 Nicholas Minorsky, 1960 0 ix 0 0 0 CHAPTER 1 INTRODUCTION: REGULAR AND IRREGULAR CARDIAC RHYTHMS, THEIR MATHEMATICAL DESCRIPTION, AND CHAOTIC DYNAMICS "Hypothesis has its right place, it forms a working basis; but it is an acknowledged makeshift, and, as a final expression of opinion, an open confession of failure, or, at the best, of purpose unaccompl i shed." Thomas Lewis, 1920 1-1 The heart is a nonl inear 0 that is, any real ist~c dev~ce; mathematical description of it must employ nonlinear mathematics. Perhaps the best direct evidence for this statement comes from voltage clamp experiments (of the sort outlined in CHAPTER 3 below) which show that the processes controlling the flow of ionic currents through the membrane of space-clamped cardiac cells are highly nonlinear. These ionic currents are responsible for the generation of spontaneous ~lectrJcal activity in the heart, for the initiation of mechanical contraction, and (together with the flow of current from cell to cell) for the spread of electrical and mechanical activity throughout the heart. It has become increasingly evident over the last lOO years that nonlinear systems (both experimental and mathematical) can display exceedingly complex behaviour. More recently~ it has become cl ear that this dynamics can become so complex that be described as Chaotic 11 11 • a can In fact, nonperiodic time series occurring in some deterministic mathematical systems have lead to the latter being labelled "chaotic". It has also been known for about 100 years that the dynamics displayed by the heart can be quite complex, even nonperiodic. I show in this thesis that behaviour approaching the degree of complexity seen in the ~ntact heart can occur in a very siinple preparation of cardiac origin. Furthermore, I identify this complex behaviour with chaotic dynamics. This introductory chapter is divided into three parts. 1-2 In the first part, I describe patterns of electrical activity that are c commonly seen in the diseased heart. There is still no consensus as to the mechanisms by which these various dysrhythmias occur, even though most of them were originally described before the year 1900. In the second part of this chapter, I indicate how one may begin to think of the electrical activity of the heart in a mathematical way. The applicability of mathematical concepts such as equilibrium points, limit cycles, quasiperiodic orbits, and strange attractors to the description of cardiac events is outlined. The first three of these constructs are treated in a cursory way; the last one is examined in greater detail. Since strange attractors occur in chaotic systems, in the third part of this chapter I next describe what is meant by chaotic dynamics, and recount the evidence- accumulated largely within the past five years - for the existence of chaotic dynamics in several different systems. The mathematical and physical literature is stressed, since there is at present little hard evidence for the existence of chaotic dynamics in biological systems. 1-3 PART A. CARDIAC DYSRHYTHMIA$ 0 (i) Some Patterns of Activity Seen in Cardiac Tissue a. 1 :1 Pattern During normal sinus rhythm, all myocytes in the heart are subjected to a periodically arriving electrical stimulus. For cells situated outside the dominant centre of the sinoatrial node, which is the pacemaker of the heart (Bleeker et al., 1980}, this stimulation is due to the conducted cardiac impulse that originates in and spreads from the dominant centre. Even the cells forming the dominant centre are subjected to periodic stimulation, since the activity of the sinoatrial node can be affected by the Q activation and contraction of the ventricles (see CHAPTERS 2, 4). During nonnal sinus rhythm, there is a one-to-one (1 :1} synchronization between the electrical activity in any two particular areas of the heart, with a more or less fixed time delay or latency between the activation of cells in the two spatially separated locations. While there are indeed a few cardiac dysrhythmias that do not necessarily result in the loss of 1:1 synchronization (e.g. sinus bradycardia or tachycardia, atrial or junctional or ventricular escape rhythm or tachycardia, first degree atrioventricular (AV) heart block, pre-excitation syndrome}, a loss of 1:1 synchronization is seen in most dysrhythmias. 1-4 This desynchronization is generally due to one of two factors: i) 0 the emergence of a subsidiary, ectopic, or triggered pacemaker or the establishment of a micro-reentrant circuit whose activity may compete and interfere with the output of the sinoatrial node (e.g. parasystole, premature contractions}; ii) block of conduction of the propagated cardiac impulse (e.g. sinoatrial exit block, AV block, bundle branch block). When loss of 1:1 synchronization occurs, the temporal pattern of activation seen at any given location in the heart may rema~n periodic (e.g. 2:1 AV block), or it may become nonperiodic (e.g. ventricular fibrillation). I now detail some of the patterns that can be seen in the heart when overall 1:1 synchronization lost. ~s Most of the patterns mentioned below were originally described in the late nineteenth century. b. n:l Patterns Loss of 1 :1 synchronization is often seen at high heart rates. Bowditch in 1871, Kronecker and Stirling in 1874, and von Basch in 1879 all reported that sufficiently rapid stimulation of the frog ventricle with a train of single induction shocks produced a 2:1 pattern of block, where only every second stimulus would provoke a contraction {Gaskell, 1900}. Higher grades of n:l block (n 0 1-5 ~ 3, where n is an integer) were also observed. 0 Schiff had earlier reported in 1850 that mechanical stimulation of the surface of the heart with a needle at a high rate would produce a contraction only after a definite number of stimuli (Hoff, 1941-42}. These n:l (n) 2} patterns can also be seen at more normal heart rates. Gaskell (1883) slit the atrium of the tortoise so as to leave a thin bridge of connecting tissue between the sinus venosus and the ventricle. There was a greater than normal delay in the propagation of contraction across the damaged atrium (analogous to first degree AV block). to 2:1 block; Extension of the section led still further continuation of the slit produced 3:1, 4:1, 5:1, 6:1 and even higher grades of n:l block (Gaskel1, 1883, 1900}. Finally, if the slit was made long enough, complete block resulted; 0 with the emergence of a ventricular pacemaker, atrioventricular dissociation was established. Application of a screw-cl amp to the atrio-ventricular groove in the frog or to the atrio-ventricular bundle in the dog ( Gaskell , 1882; Kent, 1893; Gaskel1, 1900; Erlanger, 1906; Lewis, 1920}, cooling of the atrioventricular region (Zahn, 1912}, stimulation of the vagus (Gaskell, 1882; McWilliam, 1888a; Lewis, 1920), infusion of toxic materials such as digitalis or aconitine (Cushny, 1897; Cushny, 1909-10; Lewis, 1920), asphyxia (Lewis and Mathison, 19101911; Lewis, 1920), or ligation of the coronary arteries (Cohnheim and van Schulthess-Rechberg, 1881; See, Bochefontaine, and Roussy, 1881) can produce atrioventricular block. 1-6 The final common pathway of all these interventions is presumably a decrease in the ability 0 of tissue in or below the atrioventricular junction to conduct the cardiac impulse. Clinically and experimentally, 2:1 and 4:1 block are the most common forms of n:l AV heart block; 3:1 and 5:1 patterns are rarely seen (Lewis, 1920; Besoain-Santander, 'Pick, and Langendorf, 1950; S1ama et al., 1978). In certain cases, there appears to be a direct transition from a 2:1 pattern to a 4:1 pattern of block, without going through an intermediate pattern of 3:1 block (e.g. Urthaler et al., 1974; James, Isobe, and Urthaler, 1979). c. n+l :n and Associated Patterns Occasional drcpping or skipping of beats can also be observed during the transi~ion from a 1:1 to a 2:1 pattern of block (Gaskell, 1882, 1900; Engelmann, 1896, 1896-97; Wenckebach, 1899; von Kries, 1902; Erlanger, 1906; Hay, 1906; Mobitz, 1924). In the case of AV block, each skipped ventricular beat is preceded by n conducted beats, which display a gradually increasing PR interval on the electrocardiogram. These n+l :n Wenckebach (or Mobitz type I) cycles can recur in a periodic pattern (i.e. ~ienckebach repeated n+l :n cycles with n fixed) or in a nonperiodic fashion. While n:l patterns had been demonstrated in response to fast driving of cardiac tissue in the nineteenth century, it appears that it was not until the early years of this century that n+l :n AV b1 ock was 1-7 found in response to fast atrial pacing (van Kries, 1902; Erlanger, 0 1906). The n+l :n Wenckebach pattern can also be observed in cardiac tissues other than the atrioventricular node, such as strips of ventricular muscle (Trendelenburg, 1903). Other patterns that have been associated with the n+l :n Wenckebach pattern have been described. These include n+2:n and n+3:n patterns (Cushny, 1899-1900), the 2n-1 :n patterns of reverse Wenckebach (Roberge and Nadeau, 1969), and the 2n+2:n and 2n+l :n patterns of alternating Wenckebach types A and B respectively ( Sl ama et a1 • , 1979). d. 2n:2m Patterns A pattern seen only rarely during the progression of first 0 degree AV block to 2:1 AV block is a 2:2 pattern, in which there is one ventri~ular beat for each atrial beat, but with an alternation of the PR interval back and forth between two fixed values (Lewis and Mathison, 1910-1911). Electrical alternans, an alternation from beat to beat in the morphology of the electrocardiographic complexes, was also described around 1910 (Hering, 1908, l910a; Kahn and Starkenstein, 1910; Lewis, 1910-1911). A 2:2 or alternans pattern had been described much earlier in the behaviour called pulsus al ternans, which is an alternation in the strength of the peripheral arterial pulse (Traube, 1872). 1-8 I call patterns displaying the alternans phenomenon 2:2 patterns, since the basic 0 unit that repeates in time consists of 2 stimuli and 2 responses. Other patterns of the form 2n:2m have also been seen in cardiac tissue. For example, a 4:2 pattern can be seen in the periodically stimulated sinoatri al node ( Kerr and Strauss, 1981 ) and in periodically stimulated Purkinje fibre (Jalife and Moe, 1979a}; a 6:2 pattern is not infrequently seen in cases in atrioventricular block during atrial flutter (Besoain-Santander, Pick, and Langendorf, 1950; Slama et al., 1978; Slama et al., 1979). e. N:M Patterns with N < M All of the N:M patterns (N,M positive integers) listed above 0 had N~f-1 tissue. and in most instances occured in presumably quiescent However, they can also be found in situations in which non-quiescent tissue is involved. For example, periodic stimulation of the atrium can lead to 1:1, 2:2, and n+l :n patterns in the sinoatrial nod~ (Kerr and Strauss, 1981; Bonke et al., 1982). In some bigeminal rhythms, an atrial, junctional, or ventricular extrasystole occurs for each impulse of sinoatrial origin, with a fixed coupling interval between each impulse of sinoatrial origin and the coupled extrasystole (1 :1 pattern). While the origin of extrasystoles has not been firmly established (enhanced automaticity vs. micro-reentrant circuit vs. triggered automaticity), recent work has shown that many extrasystolic '1-9 patterns similar to those observed clinically can be generated by 0 periodic subthreshold input to a slow parasysto1 ic" focus U'loe 11 et al., 1977; Jalife and t·loe, 1979a). In particular, N:t~ patterns with N < M occur when there is escape of the driven site if the frequency or the strength of the driving stimulus falls to below a critical value. I shall not go into the historical aspect of these· patterns, since this has been adequately covered in the encyclopaedic tome of Scherf and Schott (1973). It suffices to say at this point that patterns of the form 1:1, 2n:2m, 1:n, and n:n+l have been described • . f. Nonperi odic Patterns Many of the above-mentioned patterns (e.g. n:l, n+l :n, 1 :n) can occur in a periodic fashion. For example, maintained 2:1 AV block is quite common in cases of atrial flutter (Lewis, 1920). More complex periodic patterns that may be described as "mixtures" of the more basic patterns have also been documented. For instance, alternation of a 2:1 and a 3:1 pattern leads to a periodic 5:2 pattern, and alternation of a 5:4 and a 4:3 pattern leads to a periodic 9:7 pattern. However, nonperiodic patterns are often seen in the heart. One of the earliest irregular cardiac phenomena to be described was ventricular fibrillation. This pattern of activity was first described by Erichsen in 1842 and was provoked by faradic 0 1-10 stimulation of the heart by Hoffa and Ludwig in 1850. 0 Another early description of irregular behaviour was by Bowditch in 1871. who found irregularly dropped beats during periodic electrical stimulation of the frog ventricle (Hoff, 1941-42). However, Kronecker and Stirling repeated Bowditch's experiments in 1873, and could find no irregular responses. They ascribed the irregular responses seen by Bowditch to oxide buildup on the contacts of a relay (Hoff, 1941-42). Nevertheless, irregular response of the ventricle to direct electrical pacing or to input arising in the sinoatrial node does occur. For example, in textbooks on cardiac dysrhythmias, one invariably comes across electrocardiograms in which there are irregular mixtures of various n+l:n or n:l cycles in a single clinical tracing {e.g. 1920; Lewis, Katz, 1946; Bell et, 1971; Phillips and Feeney, 1973; Mandel, 1 980; Se hamroth, 1980; Chung, 1983) • (ii) Spatiotemporal Considerations Since the heart is a spatially distributed structure, the dysrhythmic patterns described above involve both space and time. For example, during 2:1 AV block, cells within either the atrial muscle or within the ventricular muscle may be said to be responding in a 1:1 fashion to the input presented to them. However, while this is occurring there are celis within the atrioventricular node in which there is a 2:2 pattern (Watanabe and 0 1-11 Dreifus, 1980). 0 Thus, different temporal patterns of activity may coexist in different regions of the heart. One can perhaps then speak of spatial bistability or mu1tistability in instances where there are two or more coexisting stable periodic patterns in different areas of the heart. Another example of this kind of behaviour is bundle branch alternans, in which there is 2:1 conduction in each bundle branch, but with conduction to the ventricle occurring alternately through one bundle branch and then the other. of activation in the ventricle. This produces a 2:2 pattern A simple biological analogue of this situation was constructed by t·1ines ( 1914), who mechanically produced longitudinal dissociation of the conducting pathway. A pattern of 2:1 conduction in one bundle branch and a 1:1 pattern of conduction in the other would also produce a 2:2 pattern in the ventricle (Bandura and Brody, 1974; Cohen et al., 1977). Gaskell {1882} proposed that alternans of the ventricle would result if there were two populations of cells in the ventricle, one responding in a 1:1 fashion, the other in a 2:1 fashion. To further complicate matters, the population of cells with the 2:1 rhythm may be composed of two subpopulations, responding on alternate stimuli (Mines, 1914). "Localized fibrillationu can occur in a circumscribed area, while contraction proceeds relatively normally elsewhere (Garrey, 1924; Moe, Harris, and Wiggers, 1941; Harris and Guevara Rajas, 1943; Downar, Janse,and Durrer, 1977). 1-12 The existence of different dynamics in different parts of the heart can usually be ascribed to 0 the existence of spatial inhomogeneities in the heart. Spatial asyrrmetry, a special form of spatial inhomogeneity, may also be playing a role. For example, periodic stimulation of cardiac tissue can lead to 1:1 conduction for antegrade propagation, but block of retrograde conduction (Engelmann, 1894; Cranefield, Klein, and Hoffman, 1971 }. ( i'f i) Bi stability It is thus possible to see more than one temporal pattern of activity in different areas of the heart at the same time. The heart is also capable of displaying more than one pattern of overall activation without altering its basic physiol agical state. The most dramatic example of this is ventricular fibri11ation, which can be made to appear or disappear more or less instantaneously (so that the change in dynamics cannot be ascribed to a change in the physiological condition of the heart) by delivery of a single pulse of mechanical (McWilliam, 1887; Pennington, Taylor, and Lown, 1970; Yakaitis and Redding, 1973; Befeler and Aranda, 1977; Forester, 1978; Lawn, Verrier, and Blatt, 1978) or electrical (Prevost and Bate1li, 1900; Wiggers and Wegria, 1940; energy. activity: f,1ines, 1914; Kouwenhoven and Mi1nor, 1954-1955) In this case the heart displays one of two modes of the stable periodic activity of normal sinus rhythm or the maintained chaotic activity of ventricular fibrillation. 1-13 The heart can also be shown to be capable of supporting two 0 different modes of periodic activity ( bistability"). 11 For example, ventricular tachycardia is routinely converted to normal s'inus rhythm by a precordial thump or by electrical cardioversion. Another example is given in the early work of Hines (1913a), who demonstrated that induction of a single properly timed extrasystole could convert a 2:1 pattern into a 1:1 pattern, and that intermission of two or three stimuli in a train of stimuli could convert a l :1 pattern into a 2:1 pattern. 11 ••• In the words of t4ines: over a quite considerable range of frequencies of excitation, there exists two possible equilibria, stable so long as the heart continues beating regularly and without interruption." Bistability of two different periodic behaviours has also been 0 seen in diseased human ventricular myocardium. Injection of a stimulus during periodic activity demonstrating 11 bimodal" action potentials can convert the action potentials to ones showing only a unimodal component (Gilmour et al., 1983). Similar behaviour can be seen in ionic models of cardiac tissue (Guevara, unpublished). Finally, there can be bistability between the stable quiescent condition of asystole and the stable periodic condition of normal sinus rhythm or ventricular tachycardia. mechanical (e.g. A single pulse of YakaHi s and Reddi ng, 1973) or el e<;trical {e.g. Cranefield, 1977) energy can cause spontaneous activity to commence in a quiescent cardiac system. l-14 (iv) 0 Unified Theories of Cardiac Oysrhythmias In summary, a great plethora of patterns can be seen heart. ~n the If anyone needs to be convinced of this fact, I suggest that they simply open any textbook on clinical electrocardiography; these books are detailed compendia of the modes of electrical activity available to the heart. The patterns range from the regularity of normal sinus rhythm to the irregularity of atrial or ventricular fibrillation. Often, a specific mechanism is identified with a particular dysrhythmia. Attempts to provide a unifying hypothesis for patterns with apparently different origins have been previously made. Most notable have been those of Oecherd and Ruskin (1946), Roberge and Nadeau (1969), El-Sherif, Scherlag, and Lazzara (1975), and Moe et a1. {1977). The work described in this thesis is an effort to extend these earlier endeavours. The approach of three of the above four groups of investigators (Oecherd and Ruskin, 1946; Roberge, Bhereur and Nadeau, 1971; Moe et al., 1977) included using the response of cardiac tissue to a single stimulus to predict the response to periodic delivery of that same stimulus. Indeed, this approach seems to have been first enunciated by Cushny and Matthews (189}) in their report concerning electrical stimulation of the heart. Cushny and Matthews used electrical stimulation to mimic irregularities in the cardiac rhythm that Cushny had previously 0 1-15 seen in response to administration of digitalis or aconitine 0 {Cushny, 1897; Cushny, 1899-1900). 11 Cushny and Matthews stated that in order to gain any real insight into them [i.e. the irregularities, M.G.], it was absolutely necessary to study first the comparatively simple dev~ ati ons caused by single stimuli • 11 Thus, I now turn to consideration of the effect of a single st~mulus on the heart. (v) The Response of Cardiac Tissue to Premature Stimulation The history of direct electrical stimulation of the heart goes back to at least the early nineteenth century, when Aldini and a handful of his cont~Jporaries stimulated the hearts of animals {including decapitated criminals) using Galvanic current provided by the colur.m of Volta (Aldini, 1803, 1804). Electrical stimulation could provoke a cardiac contraction and so Aldini (1803) cl ai rvoyantly suggested: 11 Gal vani sm affords very powerful c. means of resus)tation in cases of suspended animation under common circumstances. & c. The remedies already adopted in asphyxia, drowning, when combined with the i nf1 uence of Gal vi ni sm, wi 11 produce much greater effect than either of them separately.~~ Subsequently, more systematic investigations by Bowditch, Kronecker and Stirling, and Marey established that cardiac tissue has a refractory period following a contraction, during which stimulation cannot produce a second contraction {Hoff, 1941-42). 1-16 Furthermore, the response to a stimulus is seemingly all-or-none. 0 (According to Hoff (1941-42), the Abbe Fontana had already realized these facts by 1785.) If the ventricle is prematurely stimulated outside of its refractory period, a disturbance in the rhythm of contraction of the ventricle results: either an interpolated beat (Wenckebach, 1903; Laslett, 1909-1910) or a compensatory pause {Engelmann, 1895; Cushny and Matthews, 1897) occurs. Similar behaviours are seen if there is a spontaneous premature ventricular contraction of endogenous origin {Lewis, 1920). A more complex response is seen if the atrium rather than the ventricle is prematurely stimulated. The spontaneous cyclic activity of the sinoatrial node is affected in such a way that the returning atrial cycle can be fully compensatory, partially compensatory, or fully reset (Engelmann, 1896-97; Cushny and Matthews, 1897; Henckebach, 1903; Lewis, 1920). More recent v-10rk has shown that the perturbation to the rhythm of the sinoatrial node caused by a prenature atrial contraction depends on the phase in the sinus cycle at which the premature contraction is induced {Bonke, Bouman, and van Rijn, 1969; Bonke, Bouman, and Schopman, 1971; Klein, Singer, and Hoffman, 1973; Strauss et al., 1973; Miller and Strauss, 1974; Steinbeck et al., 1978; Kerr et al., 1980). Similar phasic effects on the sinoatrial node are found if its spontaneous activity is perturbed by a subthreshold pulse of current {Sano, Sawanobori, and Adaniya, 1978; Jalife et al., 1980}. 0 1-17 This "phase resetting" of the spontaneous activity of the 0 sinoatrial node also occurs in response to delivery of a single vagal volley (Brown and Eccles, 1934a,1934b; Dong and Reitz, 1970; Levy et al., 1969; Greco and Clark, 1976; Jalife and Hoe, 1979b; Spear et al., 1979; Jalife et al., 1983}. Phase resetting has also been described in several other spontaneously active cardiac tissues (Weidmann, 1951; Klein, Cranefield, and Hoffman, 1972; DeHaan and Fozzard, 1975; Jalife and Moe, 1976; Scott, 1979; Ferrier and Rosenthal, 1980; Guevara, Glass, and Shrier, 1981; Ypey, van ~1eerwijk, ~~eerwij k . and DeHaan, 1982; Clay, Guevara, and Shrier, 1984; van et al • , 1984). If the atrioventricular node is presented with an earlierthan-expected input arising from a premature activation of the atrium, it will conduct the cardiac impulse to the ventricle with a velocity that is slower than normal. The response (i.e. the conduction time through the AV node} can be systematically investigated as a function of the degree of prematurity of activation of the AV node. In this way, the AV nodal recovery curve can be constructed (Mobitz, 1924; Ashman, 1925; Decherd and Ruskin, 1946). Thus, the response of several different parts of the heart to premature stimulation with a single stimulus (a propagated wave of contraction or an electrical pulse) has been known for some time. I call the response to delivery of a single current pulse the single-pulse response. I show in CHAPTER 5 that knowledge of the single-pulse response is indeed indispensAble in understanding the 0 response to periodic stimulation of the particular cardiac system 1-18 that I have been investigating. c Not only are many of the patterns described earlier in this section seen in this cardiac preparation, but their existence is predicted from the single-pulse response. Thus, it appears that the suggestion of Cushny and Matthews made in 1897 was indeed a valuable one. The experimental work described below in CHAPTERS 2 and 4 is carried out on spontaneously oscillating cardiac tissue. Within this century, a considerable body of literature has been built up on the mathematical description of oscillating (or just excitable) systems, and on the effects of periodic stimulation on such systems. In the next section I first show how one may begin to look· at the h1:!art from a mathematical perspective. I then give a sampling of very recent results (both theoretical and experimental) that demonstrate that Chaotic dynamics can result from the 11 11 periodic stimulation of systems which are excitable or which oscillate spontaneously. 1-19 0 PART B. (i) MATHEMATICAL DESCRIPTION OF SOME CARDIAC PHENOMENA Equilibrium Points. Limit Cycles, Quasiperiodic Dynamics, and Strange Attractors in Cardiac Electrophysiology The experimental work I report on in this thesis is carried out on a spontaneously oscillating cardiac preparation. Since the preparation is effectively isopotential (i.e. space-clamped), ~t can be described by a system of ordinary differential equations. Equations describing the electrical properties of space-clamped cardiac tissue are high-dimensional, nonlinear systems of ordinary differential equations, which are formulated using results from voltage-clamp experiments (e.g. Noble, 1962; f'.t::Allister, Noble, and Tsien, 1975; Beeler and Reuter, 1977; Yanagihara, Noma, and Irisawa, 1980; Bristow and Clark, 1982; Irisawa and Noma, 1982; Clay, Guevara, and Shrier, 1984). A distributed system such as the heart, where variables such as the transmembrane potential and activation and inactivation variables of the various ionic currents are functions of both time and space, must be modelled by a syster.1 of partial differential equations. In an N-dimensional system of ordinary differential equations there are N variables x1 , x2 , ••• , xN. At any one point tin time, the state of the system is completely specified by the values of these N variables at that time. The path traced out by the N- dimensional state-point (x 1 (t), x2 (t}, ••• ,xN(t)) as time 1-20 0 progresses can be thought of as a trajectory or an orbit in the Ndimensional phase space of the system. Essentially four types of asymptotic (i.e. t ~ oo) behaviours can evolve in a bounded, dissipative system of ordinary differential equations. The state-point of the system may generically tend to ( i) an attractor of dimension zero (a stable equilibrium point, fixed point, singular point, or steady state}; {ii) a periodic orbit of dimension unity (a stable limit cycle); (iii) an orbit lying on a toroidal hypersurface (a quasiperiodic orbit); or {iv) an attractor whose fractal or Hausdorff dimension 01ori, 1980) is greater than its topological dimension (a strange attractor). The state point may wander around in an apparently random fashion in a system in which there are no stable periodic orbits, but rather an infinity of unstable orbits (Pikovskii and Rabi novich, 1978; Ueda, 1979, 1980a, 1980b). Asymptotic approach to a stable equilibrium point is well known in cardiac electrophysiology. For example, it generally occurs after a single action potential is ·induced in quiescent tissue. An ionic model of quiescent tissue must therefore possess at least one stable equilibrium point (e.g. Beeler and Reuter, 1977). A limit cycle was defined by Poincare in 1881 to be a closed curve in the phase space of a system of ordinary differential equations (Minor sky, 1962). t·1ovemen t of the state point of the system along the limit cycle trajectory results in a periodic 1-21 .o time series for the system variables. Trajectories with initial conditions sufficiently close to an asymptotically stable 1imit cycle approach the cycle as t • ""· Therefore, no trajectory sufficiently close to the 1imit cycle is also a closed trajectory. The physicist and engineer van der Pol (1926, 1940) was probably the first person to think of the cardiac cycle as a relaxation oscillator (a special case of a limit-cycle oscillator). Numerical simulation suggests that asymptotically stable limit cycles exist in ionic models of spontaneously active cardiac cells (e.g. Noble, 1962; Scott and Kang, 1974; McAllister, Noble, and Tsien, 1975; Yanagihara, Noma, and.Irisawa, 1980; Bristow and Clark, 1982; I ri sawa and Noma, 1982) • Quasiperiodic motion is nonperiodic motion, whose Fourier spectrum has a finite number of characteri stk frequencies which are rationally independent of each other C' i ncommensurate11 ) • The trajectory followed by the state-point of the system can be thought of as lying in a toroidal hypersurface embedded in a higherdimensional space. Although the time series of a variable undergoing quasiperiodic dynamics is nonperiodic, any two trajectories that start out with initial conditions close to one another do not diverge from each other very rapidly. the two trajectories start out within a distance In fact, if o of each other, a point in time is eventually reached when the two trajectories return to within a distance e of each other, with e q. However, a system with quasiperiodic dynamics can be 1-22 ~ c structurally unstablet with the quasiperiodic motion being destroyed by an infinitesimally small change of the system parameters {e.g. Moser, 1969). Other systems showing quasiperiodic dynamics seem to be structurally stable (e.g. Franceschini, 1983; Thoulouze-Pratt, 1983). There do not seem to be any published reports labelling any phenomenon seen in the heart as a manifestation of quasiperiodic dynamics. I present evidence in CHAPTER 4 below that complete heart block with atrioventricular dissociation is indistinguishable from quasiper1od1c dynamics. f11ovement of the state-point of the system along d strange at tractor results in a nonperi odic time series for any of the system variables (e.g. the transmembrane potential). Although the state-point of the system never returns to a location in phase space previously visited, its motion takes place in an invariant volume of the phase space. Unlike the case of quasiperiodic dynamics, there is "sensitive dependence on initial conditions.. , with trajectories that are initially close diverging away from each other exponentially with time (Guckenheimer, 1979b; Ruelle, 1979). This exponential divergence means that at least one of the Liapunov exponents is positive (Benettin, Galgani, and Strelcyn, 1976; Nagashima and Shimada, 1977; Shimada and Nagashima, 1978, 1979; Geisel, Nierwetberg, and Keller, 1981 ). Furthermore, the fractal or liausdorff dimension of a strange attractor is generally nonintegral, and is less than its topological dimension {Mori, 1980; 0 1-23 0 Packard et al., 1980; Russel, Hanson, and Ott, 1980; Froehling et al., 1981; Farmer, 1982; Greenside et al., 1982; Grassberger and Procaccia, 1983; Termonia and Alexandrowicz, 1983). Systems that admit strange attractors may or may not be structurally stable (Guckenheimer and Holmes, 1983). Indeed, some strange attractors are stable, others not {Kaplan and Yorke, 1979). Note that exactly . what is meant by a strange attrac tor (or even just an attrac tor) is a matter of definition which is presently the subject of some debate (e.g. Ruelle, 1980, 1981; Guckenheimer and Holmes, 1983}. The concept of Strange attractot··" was initially introduced by 11 Ruelle and Takens (1971 ). As is the case for quasiperiodic dynamics, there do not appear c to be any published reports linking the presence of irregular dynamics in a cardiac system to the existence of a strange attractor in the phase space of a model of that system. In CHAPTER 4 below, I show experimental tracings from a periodically stimulated cardiac oscillator that are nonperiodic in time. In CHAPTER 5, I present theoretical and numerical ar·guments which suggest that these tracings reflect the presence of a strange attractor. (ii) Electrical and Electronic Models of the Heartbeat In their pioneering study, van der Pal and van der r~ark (1928, 1929) modelled the heart as a system of three coupled oscillators - 1-24 0 one each for the sinoatrial node, the atrium, and the ventricle. They wired together three neon bulb relaxation oscillators with unidirectional coupling, so that the "sinoatrial node" could affect the u atrium' (but not vice versa), and so that the "atrium" would affect the "ventricle" (but not vice versa}. This simple circuit produced patterns remarkably reminiscent of normal sinus rhythm, first degree heart block, 3:2, 2:1, 5:2, 3:1 (and higher grade n:l) heart block, and·complete heart block. This approach of van der Pol and van der t1ark to modelling the heartbeat stemmed from earlier experimental and theoretical work on driving an electronic or electrical oscillator with a sine wave generator (Appleton, 1923; van der Pal, 1927; van der Pal and van der r~ark, 1927). It was found that the driven oscillator could be made to modify its spontaneous activity so as to synchronize with or lock onto the driving sinusoid. In this circumstance, for each cycle of the sine wave generator, there is one cycle of the driven neon bulb or triode valve oscillator, with a fixed phase difference ("phase angle") between the wavefonns of the driving and driven oscillators (_1 :1 synchronization, entrainment, or phase-locking). As the ratio of the frequency of the sine-wave generator to the intrinsic frequency of the driven oscillator would be changed, the phase·angle would also change. Eventually, as the driving frequency became large with respect to the instrinsic frequency of the driven oscillator, the one-to-one pattern of synchronization or entrainment would be lost, and would be replaced by periodic n:l 0 1-25 0 coupling patterns {"frequency demul tip 1icati on" ) or nonperi odic ("quasiperiodic der ~1ark, 11 ) dynamics (van der Pol, 1927; van der Poland van 1927, 1929; t1inorsky, 1962; Hayashi, 1964). Thus, the normal sinus rhythm and first degree block observed in the electrical model of the heart of van der Pol and van der tvlark corresponds to 1 :1 synchronization, the 2:1, 3:1, and higher grades of n:l block correspond to frequency demultiplication, and complete heart block with atrioventricular dissociation corresponds to quasiperiodic dynamics or periodic dynamics with a very long period {see CHAPTERS 4 and 5). The approach of using electrical and electronic analogues of the heart to model normal cardiac activity and cardiac dysrhythmias 0 has continued through the years down to the present day (Bethe, l940-41a, 1940-4lb; Grant, 1956; Chebotarev, 1968; Roberge, Nadeau, and James, 1968; Roberge and Nadeau, 1971; Li~ko Nadeau, 1971; Sideris, 1976; ~~oul opoul os, 1977; ~~adeau, 1969; Bhereur, Roberge, and and Landahl, 1971; Roberge, Bhereur, and Padmanabhan, 1977; Keener, 1983a). Sideris and The work of Roberge, ~Jadeau, and Bhereur is expecially noteworthy, since they brought to bear on the problem a two-pronged attack that combined electronic modelling with physiological experimentation. The richness of the electrical modelling approach is underscored by the fact that a model consisting of eight neon bulb oscillators produces behaviours similar to a score of phenomena- both normal and pathological experimentally observed in the heart (Sideris, 1976; 0 1-26 Sideris and 0 ~1oul opoul os, 1977). The various periodic patterns described in these analogues (e.g. normal sinus rhythm with or without first de~ree heart block, 3:2 block, n:l block) presumably correspond to the existence of an asymptotically stable limit cycle in the phase space of the equations describing the particular analogue. The way in which limit cycles are born and die. as the various patterns come and go as a parameter (e.g. atrial frequency or degree of atrioventricular coupling) is changed was not investigated. A single limit cycle can arise or disappear via a Hopf bifurcation or a reverse Hopf bifurcation respectively (a translation of Hopf's original 1942 paper is found in Howard and Kopell, 1976); a pair of limit cycles 0 can arise de novo or coalesce via a saddle-node bifurcation (Minorsky, 1962; Sotomayor, 1973); a pre-existing stable limit cycle can become unstable producing a new stable 1imit cycle of approximately twice the original period in its immediate vicinity via a period-doubling bifurcation (Brunovsky, 1971; Ruelle, 1973). More complicated behaviours of a more global nature can also occur ( Ruell e and Takens, 1971; Guckenheimer, 1979a; Guckenheimer and Holmes, 1983}. With the possible exception of the study of Bethe (1940-41a, 1940-41b}, period-doubling bifurcations were apparently not observed in the studies on electrical and electronic analogues listed above. This thesis shows that period-doubling bifurcations can exist in periodically stimulated cardiac tissue. 0 1-27 0 (iii) Bistability and Hysteresis Bistability occurs in a system when there are two coexisting stable attractors. For example, two stable equilibrium points, one stable equilibrium point and one stable limit cycle, or two stable limit cycles can simultaneously exist; distinct fonns of bistabil ity. these are thus three Experimental and theoretical work by Appleton and van der Pol (1922) and by van der Pol (1922) showed that the latter two types of bistability could exist in a triode valve oscillator. In the case of the co-existence of a stable equilibrium point and a stable 1imit cycle, 11 oscillation hysteresis11 is seen. In this phenomenon, as a parameter is changed in one direction and then in the reverse direction, oscillation appears and disappears at two different values of the parameter. Experimental evidence for oscillation hysteresis was. found by Appl eton and van der Pol (1922) in a triode valve oscillator, and has been recently found in a neural membrane (Guttman, Lewis, and Rinzel, 1980). Two other related phenomenon can also be seen when a stable equilibrium point and a stable 1imit cycle coexist. If the system is not oscillating, it can be made to do so by injection of a stimulus of sufficiently high amplitude. This procedure was successfully carried out by Appleton and van der Pol {1922) in a triode valve circuit by using an electromagnetically-induced electromotive force. In this case one speaks of an oscillator with 0 1-28 0 11 hard self-excitation" {e.g. Minorsky, 1962). Hard self-excitation has also been seen in cardiac tissue where it has been termed "triggered activiti• (e.g. Cranefield, 1977; Jalife and Antzelevitch, 1980). Conversely, if the system is oscillating, the osciilation can be annihiliated by a single well-timed stimulus (Winfree, 1980). This has been recently seen in three different cardiac tissues- the sinoatrial node (Jalife and Antzelevitch, 1979), Purkinje fibre (Jalife and Antzelevitch, 1979,1980), and diseased human ventricular myocardium (Gilmour et al., 1983). As mentioned earlier, van der Pol (1922) found experimental evidence for the coexistence of two different periodic solutions in an unforced triode valve oscillator. 0 In addition, investigation of the sinusoidal fore i ng of the equations that van der Pal ( 1926) devel aped to model his electronic oscillator {Cartwright and Littlewood, 1945; Hayashi, Shibayama, and Nishikawa, 1960; Littlewood, 1960; Grasman, Veling, and Willems, 1976; Flaherty and Hoppensteadt, 1978; Guckenheimer, 1980b; Levi, 1981) and of a piecewise linear approximation to them (Levinson, 1949) have revealed that two stable limit cycles can coexist at one set of parameter values. When two stable periodic orbits exist simultaneously, one of two different periodic patterns will appear, depending on the particular initial conditions chosen. In an experimental situation, where one parameter such as the forcing frequency or amplitude is gradually increased or decreased, this bistability could manifest itself as hysteresis. 0 1-29 (An experimental 0 system that demonstrates hysteresis or memory is one in which the behaviour seen at a particular set of parameters depends on how that parameter set was approached.) In a system of ordinary differential equations, it seems to me that hysteresis must be due to some form of bistability. Bistability of two periodic orbits can also be seen in numerical simul ati ons of a forced Ouffi ng-van der Pol .osc ill.ator (Kawakami, 1982) and in the forced Brusselator.(Kai and Tomita, 1979). Evidence for the coexistence of two attracting periodic orbits has been found in several physical systems, including experiments involving Rayleigh-Benard convection {Gollub and Benson, 1978) and experiments using lasers {Atecchi et al., 1982). 0 t,1odel s of 1aser systems also show coexistence of two stable 1 imi t cycles (Antoranz et a1 ., 1982; Arecchi et al., 1982), as do simplified models of convective fluid flows (e.g. Fowler and McGuinness, 1982). Hysteresis where one or the other of two periodic patterns can appear depending upon the prior history of the system has been documented in cardiac muscle (Mines, 1913a; Moulopoulos, Kardaras, and Sideris, 1965; El-Sherif et al., l977a; Guevara et al., unpublished: see CHAPTER 4). Bando et al., 1979; It also occurs in an electronic model of a cardiac pacemaker cell {Roberge, Bhereur, and Nadeau, 1971) and in many other analogue systems {Hayashi, 1964). It is unclear to me at this time whether or not the hysteresis seen in all five of the above cardiac studies can be ascribed to 0 1-30 0 bistability in a set of ordinary differential equations. Bistability of two periodic orbits 1n cardiac systems has however definitely been seen in two separate cases; in each instance, a periodic pattern was converted into a different periodic pattern by application of a brief stimulus (lv1ines, 1913a; Gi'lmour et al., 1983). There is also evidence for tristability in the human heart: two successive precordial thumps converted ventricular tachycardia of one morphology first into ventricular tachycardia of another morphology and then into normal sinus rhythm (Pennington, Taylor, and Lown, 1970: Fig. 4}. Tristability has also been seen in a model of a laser (Arecchi et al., 1982). Finally, if the basins of attraction of the two periodic orbits are very much intertwined, behaviour that appears aperiodic may result, since small perturbations ("noise11 ) in the system will make the state-point of the system skip back and forth from the basin of attraction of one limit cycle to that of the other (Flaherty and Hoppensteadt, 1978). A similar hoppingu mechanism 11 operating between two attracting domains can apparently produce a broadband power spectrum that decays algebraically (Arecchi and Lisi, 1982; Ben-Jacob et al., 1982). Coexistence of two stable equilibrium points leads to a third form of bistability. In the heart, this produces the phenomenon of .. two stable states of resting potenti al 11 ( Wiggi ns and Cranefi el d, 1974, 1976; Shrier and Clay, unpublished), in which a brief perturbation can cause the state-point to move from one equilibrium 1-31 0 point to the other. This bistability of two equilibrium points has been seen in physical systems (e.g. Gibbs, r•1cCall, and Venkatesan, 1976). Tristability of three equilibrium points has been seen in optical systems (e.g. Cecchi et al., 1982), but is unl ike1y to exist in cardiac tissue, since curr.ent-vol tage characteristics have a simple N-shape (see CHAPTER 2). Quasiperiodic and periodic motions can also coexist in Rayleigh-Benard convection (Gollub and Benson, 1978} and in models of Rayleigh-Benard convection (Curry, 1979). I know of no obvious cardiac anal ague of this behaviour. Coexistence of a periodic orbit and a strange attractor is also possible (e.g. Grebogi, Ott, and Yorke, l983b; Guckenheimer and Holmes, 1983), as is the coexistence of one or more stable equilibrium points with a strange attractor (e.g. Nagashima and Shimada, 1977). As outlined earlier in this chapter, one can generally repeatedly fibri11 ate and defibrillate a healthy heart at· this fact may imply such a coexistence. will: Coexistence of two strange attractors can also occur (e.g. Leven and Koch, 1981; Grebogi, Ott, and Yorke, 1982; Arecchi and Li si, 1982; Arecchi et al., 1982) and can lead to the production of a time series whose power spectrum falls off in a 1/fa fashion, with a a positive real number (Arecchi and Lisi, 1982; however, see also Beasley, D'Humieres, and Huberman, 1983; Voss, 1983). It is interesting to note in this context that strange attractors often have a similar falloff in spectral content, and that l/f 0 1-32 0 fluctuations are seen in many biological membranes and in the beatto-beat interval of the heart (Kobayashi and Musha, 1982). Recently, formulation of the problem of periodic stimulation of the van der Pal and other oscillators in terms of one- and twodimensional maps has yielded many new insights. Bistability turns out to be due to the existence of two stable periodic orbits on such return maps (Guckenheimer, 198Gb; Levi, 1981; Glass and Perez, 1982; Perez and Glass, 1982; Guckenheimer and Hol mes, 1983; Guevara et al., 1983: see CHAPTER 6; Glass 0 1-33 et~., 1984). 0 PART C. CHAOTIC DYNAMICS In this section, I give a brief survey of the field of chaotic dynamics. I stress two areas: ( i) the experimental observation of dynamics that has been described as chaotic; and (ii} the numerical investigation of mathematical models of experimental systems. The works surveyed below come largely from the physics literature. Note that almost no mention is made of chaotic dynamics in conservative systems, since I am interested in this thesis in a biological system that is dissipative. The review articles by Chirikov {1979) and by He11eman (1980b), and the book by Lichtenberg and Lieberman (1983) can be consulted with regard to chaotic dynamics in Hamiltonian systems. the mathematical literature. I also 1argely neglect The books (and the references contained therein) by Collet and Eckmann (1980}, Gumowski and 11ira (1980), and Guckenheimer and Holmes (1983) can be used as entry points into that literature. The review articles of Feigenbaum (1980b, 1983}, Eckmann (1981 ), Hofstadter (1981), Ott (1981 ), and Swinney (1983} and the conference proceedings edited by Gurel and Ross1er (1979}, Hel1eman (1980a), and Campbell and Rose (1983) may also be of some general interest to the uninitiated. In the last two paragraphs, I have used the adjective "chaotic" and the term 11 Chaotic dynamics11 11 • The tenn 11 Chaos11 or Chaotic dynamics 11 is 1oosel y used in the mathematical and scientific literature and has different meanings to different 0 1-34 0 people. Perhaps its most common usage is to indicate the presence of a nonperiodic time series of some variable measured in an experiment or generated in a numerical simulation. In this case, the investigator often assumes that the observed nonperiodic dynamics is the reflection of a determi ni stically nonperi odic orb1t in the phase space of the system being studied. For example, in a numerical study of a system of ordinary differential equations, this would mean the appearance of a strange attractor in the phase space of the system. (Although a quasi periodic orbit will also generate a nonperi odic time series, the term chaotic dynamics is not usually applied in this case.) Note that the concepts of chaos and randomness still give rise to some confusion in the literature 0 {e.g. Kozak, Musho, and Hatlee, 1982 versus Karney, 1983). Simply based on examination of its appearance, a computergenerated time series can never be said to be nonperiodic, since this would involve carrying out the simulation for an infinitely 1ong time. However, it can be proven mathematically that some dynamical systems, for certain well-chosen values of the parameters and initial conditions, have periodic solutions of arbitrarily long V period and even solutions which are nonperiodic (e.g. Sarkovskii, 1964; Li and Yorke, 1975; "Stefan, 1977; Pikovskii and Rabinovich, 1978). Numerical simulation of these systems on an ideal digital computer would theoretically result in a periodic time series for the system variables, since a digital computer is a finite-state machine (e.g. Stein and Ulam, 1964; Mayer-Kress and Haken, 198la; 1-35 0 Conrad and Rossler, 1982; Levy, 1982; Lichtenberg and Lieberman, 1983}. However, the effects of "noise", either (e.g. ~nternal quantum fluctuations in semiconductor devices, bombardment of semiconductor junctions by disintegration products of radioactive elements present in trace quantities) or external {e.g. cosmic ray bombardment) to the machine, may very well convert this finitestate-induced periodicity into nonperiodicity. However, the characteristic time-seal e of the computer-induced artifact is generally much different from the dominant frequency of the nonperiodic or chaotic oscillation. Numerical calculation of the Liapunov exponents and of the fractal dimension is often carried out to demonstrate the presence of chaotic dynamics. 0 Numerical artifact also occurs duri ~g the simulation of an orbit that can be mathematically proven to be periodic. For example, since a digital computer has finite resolution, the computed state-point of a set of differential equations will seldom be exactly at any point on the (mathematically) correct limit_cycle orbit. Instead, the simulation will produce a periodic orbit that will be an approximation to the limit cycle, with finitely many points on the orbit. In addition, starting with initial conditions off of the limit cycle, the state-point will end up on this finitedifference approximation to the real limit cycle in finite time. In the case of the real (i.e. mathematical) 1imit cycle, the approach to the 1imit cycle is only asymptotic for initial conditions not on the limit cycle itself. 0 1-36 0 Some experimental systems are also said to be chaotic or to display chaotic dynamics because an experimental time series appears to be nonperiodic. Again, there are several fundamental problems in stating that a physical or biological system generates a deterministically nonper'iodic output. I shall only state what I see as the most fundamental objection. All physical and biological systems are made up of molecules, atoms, and more fundamental constituent particles that are governed by the laws of quantum mechanics. If one accepts the statistical interpretation of quantum mechanics, then deterministic nonperi odic dynamics cannot occur in experiments. I illustrate this statement with an example taken from cardiac e1ec trophys i o1ogy. Close inspection of the potential dHference measured across a supposedly quiescent cardiac membrane reveals that the membrane is not truly quiescent: there is low-amplitude voltage noise present (e.g. DeFelice and DeHaan, 1977; DeHaan and DeFelice, 1978a, 1978b). Similarly, observation of the cyclic output from any cardiac oscillator shows that the activity is not.strictly periodic: there are slight variations in the interbeat interval from cycle to cycle (e.g. Bouman et al., 1982). If one accepts the fact that the transmembrane potential is generated by the opening and closing of single ionic channels, there are at least four levels at which one can think of a stationary cardiac membrane in the absence of environmental fluctuations: 1-37 0 (i) One can construct a deterministic set of Hodgkin-Huxley like differential equations and then add a stochastic term to represent membrane noise. ( ii) One can build an inherently stochastic model us~ng the statistical properties of single channels (experimentally measured or deduced from the macroscopic rate constants) and then reconstruct the macroscopic action potential with its now 1nherent microscopic voltage fluctuations (e.g. Cl ay and DeFel ice, 1983; DeFel ice and Cl ay, 1983). (iii} In principle, one could build a quantum mechanical model of each of the several ionic channels, obtain the statistical properties of the currents from the wave functions of the ions flowing through the channels, and 0 then reconstruct the macroscopic behaviour. (iv) One could replace the (Copenhagen) probabilistic interpretation of quantum mechanics with a deterministic interpretation that provides a statistical description at the molecular level for the single channel behaviour, and then reconstruct the macroscopic action potential (incidentally winning a Nobel prize!). Only in case (iv) above can one truly say that the irregular, nonperi odic dynamics displayed by the cardiac system is deterministic. Nevertheless, it may well be that, in the same way that one speaks of cardiac oscillators, understanding that some 0 1-38 0 microscopic voltage f1 uctuations are present, one can also speak of cardiac strange attractors. In mathematical systems, deterministically nonperiodic dynamics generally occurs at one particular combination of the various parameters in the system. An infinitesimal change in any of the system parameters can change the qualitative form of the dynamics. Thus, systems of differential equations are often structurally unstable at values of the parameters where a quasiperiodic orbit or a strange attractor exists (e.g. see t·1oser, 1969; Guckenheimer and Holmes, 1983). This fact therefore also adds another twist to the question: "Should deterministically nonperiodic dynamics be observable in physical and biological systems?n 0 Problems with measurement also cloud the issue. Since measurements can be made only to a finite number of decimal places, can nonperiodic sequences· ever really be demonstrated in experimental work? The old conundrum of whether the length of a particular stick is rational or irrational again raises its head. Again, this measurement problem is a fundamental one involving quantum mechanics (in particular, the uncertainty principle). While nonperiodic dynamics with sensitive dependence on initial conditions may be taken as the litmus test for chaotic dynamics, chaotic dynamics has several other features that can be illustrated with the simple example that I present in section (i)b below; and which are more amenable to experimental 1-39 observation. 0 I have found it convenient to divide up the systems that display chaotic dynamics into six main classes. I now describe experimental and numerical evidence for the existence of chaotic dynamics in each of these six classes of systems. (1) Chaotic Dynamics in Periodically Forced Oscillating Systems a. Chaotic Dynamics in Mathematical Models of Periodically Forced Limit-Cycle Oscillators Periodic forcing of limit-cycle oscillators can lead to complicated dynamics which can even be nonperiodic. Among the earliest studies in this area were those of Cartwright, L~ttlewood, and Levinson previously mentioned in which the van der Pol oscillator was forced with a high-frequency sinusoid. In fact, the behaviour of the periodically forced van der Pol oscillator is chaotic ( Hol mes and Rand, 1978; Levi, 1981; Guckenheimer and Holmes, 1983). ~Jumerical studies using digital computers have revealed the existence of chaotic dynamics in several other periodically-forced limit-cycle oscillators. (i) These include: the Brusselator- a two dimensional model of an oscillating chemical reaction (Tomita and Kai, 1978a, 1978b; Kai and Tomita, 1979; Tomita and Kai, 1979; 1-40 Broomhead, 0 (ii) ~tCread~e, and Rowlands, 1981; Ka~, 1981) a Duffing-van der Pal oscillator (Coullet, Tresser, and Arneodo, 1980; Kawakami, 1982); {iii) the Bonhoeffer-van der Pal (BVP} or Fitzhugh-Nagumo (FN) equations - a simplified two-dimensional model for excitable biological membranes modified to produce spontaneous activity (Guevara et al., 1983); (iv) simple impulsively-forced, two-dimensional limit-cycle oscillators (Zaslavsky, 1978; Guevara and Glass, 1982; Hoppensteadt and Keener, 1982); (v) the Hodgkin-Huxley equations- a four-dimensional model of quiescent squid axon modified to produce spontaneous activity (Guevara et al., 1983); (vi) the 14cAll ister-Nobl e-Tsien equations - a nine-dimensional model of spontaneously active cardiac Purkinje fibre (Guevara, unpublished). One of the earliest studies to find very complicated behaviour in a periodically forced oscillator was that of Ueda, Hayashi, and Akarmatsu {1973), who simulated a sinusoidally forced resonant circuit cantai ni ng a negative-resistance element on an anal oglie computer. 0 1-41 b. ~ Chaotic Dynamics in a Prototypical Peri odical1 y Forced Lim1tCycle Oscillator I will now present and discuss the behaviour of a simple mathematical system that illustrates several features of chaotic dynamics. A more complete description of this work can be found in Guevara and Glass (1982). Investigation of this system has also been carried out by Hoppensteadt and Keener (1982). The system considered is a mathematical model of a simple two-dimensional limit-cycle oscillator that is periodically forced by a train of impulses. There are striking similarities between the behaviour of this periodically forced oscillator and that of the periodically forced cardiac system described in CHAPTER 4. 0 consideration of the dynamics displayed by Indeed, th~s model preceded and thus has proven invaluable in guiding' the experimental work detailed in CHAPTER 4. The oscillator is described in polar coordinates by the equations: d~ at = 2'lf ( 1-1) dr (ff = ar(l-r), where 9 is the angular coordinate {-= < ~ < ~}, coordinate and a is a positive real number. 1-42 r is the radial The unit circle forms a limit cycle that is globally attracting for all initial 0 conditions except for the equilibrium point at the origin. Note that the unperturbed oscillator has unit period and its state can be parameterized by an angular coordinate $ q, = \P Z1r (modulo 1 ). ( 1 -2) The variable$ is called the phase of the oscillation (0 ( q, 1 ). < Perturbation away from the limit cycle results in relaxation back to the stable limit cycle at a rate that depends on the parameter a. I consider the limiting case of a In this case, + ~. following perturbation away from the limit cycle, there is an instantaneous relaxation back to the 1imi t cycle along a radial direction. In what follows I consider perturbations consisting of impulses of magnitude b which are directed parallel to the x-axis (Fig. 1-1 ). Thus, the effect of a single impulse is to instantaneously reset the phase of the oscillator. Calling q, the old phase of the oscillator immediately preceeding the perturbation, and e the new phase of the oscillator immediately following the perturbation one has 0 = g( q, 'b) ' 1-43 ( 1 -3) 0 Figure 1-1. The limit-cycle oscillator and the effect on the oscillator of stimulation with an impulse of magnitude b. The unit circle forms a limit cycle which is globally attracting for all initial conditions except for the origin. The stimulus of amplitude b instantaneously resets the phase of the oscillator from a phase$ prior to stimulation to a phase following the perturbation. Identifying the x-axis variable with membrane potential , stimuli with b 0 > 0 may be regarded as being analogous to depolarizations, stimuli with b are analogous to hyperpolarizations. < 0 1-44 0 Figures (1-1} to (1-3) are slightly modified from figures originally drafted by Leon Glass and Burt Gavin. 0 0 11 -I~ If 11 -IN 11 where the function g is called the phase transition curve or PTC 0 (Kawato and Suzuki, 1978; Kawato, 1981). trigonometry, an analytic expression for (Guevara and Glass, 1982; eqn. (5)). 5 different values of Using some simple g(~,b) can be obtained Figure l-2A shows the PTC for o. Consider now the effect on the oscillator of delivering a periodic train of impulses, with a timeT between stimuli. Let ~4 I be the phase of the oscillator immediately preceding delivery of the ;tll stimulus. The phase of the oscillator immediately preceding delivery of the next stimulus will be given by <jli+l = g( cf>pb) + T ( 1 -4) • This equation is a one-dimensional finite-difference equation, depending on the two parameters b and r. Thus, the two assumptions of impulsive perturbation and of infinitely fast relaxation back to . the limit cycle following a perturbation allow the investigation of the dynamics of the periodically forced two-dimensional system to be reduced from a two-dimensional to a one-dimensional problem. Equation (1-4} describes a map T:~;+~i+l which is called the Poincare map (also called the first return map or the phase advance 1-46 0 Figure 1-2. A. The phase transition curve (PTC) for five different values of b• B. . The Poincare map for three different values of held fixed at b = -1 .30. 0 1-47 < with b 0 1·0 . . . . - - - - - - - - - - A e o·5 b = -1·1 b= -0·9 0·5 0·0 1·0 cp 0 B T=0·35 4't+t 0·5 0·0 0·5 cpl 1·0 map). 0 The dynamics in response to periodic stimulation (i.e. the evolution of the <P; starting with some found by iterating eqn. (1-4). init~al value $ ) 0 can be For fixed b, the Poincare map is obtained by vertically translating (modulo 1} the PTC by an amount r, as shown in Fig. 1-28. Numerical iteration of the Poincare map at many different combinations of band r yields Fig. l-3A. In each "phase-locking zone" 'in the (b,r) parameter space that is labelled ~J:f~, periodic dynamics occurs in which for every N impulses, the state-point of the oscillator makes Mcomplete revolutions about the origin. Note that most of the area of the (b,r) parameter plane is filled with periodic dynamics. Figure l-3B shows a small region of Fig. 1-3A at an expanded scale. In the unlabelled region is found periodic and nonperiodic dynamics. Indeed, in this region, there are an infinity of phase-1 ocki ng zones of very suall extent, in . ~1hich dynamics of arbitrarily high period can be found. ~~=i~ periodic There are also, theoretically, some values (b,-r) in the unlabelled region at which nonperiodic behaviour with sensitive dependence on initial conditions occurs. Contiguous with and lying below the 4:M zones of Fig. 1-3B are zones of the form 8:~1. In fact, there are infinitely ma.ny 2":t•1 zones with arbitrarily high n. More recent computations have shown that there is also a 3:3 zone in the unlabelled region (Glass and Belair, unpublished). Between the accumulation boundary of the 2°:M zones and the 3:3 zone is a 1-49 0 Figure 1-3. Phase-locking zones resulting from periodic delivery of impulses of magnitude b at a frequency .- 1 • The areas not labelled contain dynamics that is phase-locked and dynamics that is not. A. Phase-locking zones for N ~ M, N ~ 3. The borders between the 1 : 0 and 1 : l , 1 : 0 and 2: 0 , 2: 0 and 2: 1 , 2: 1 and 2:2, and 2:2 and 1:1 zones are determined analytically, as are the borders of the 1:0 and 1:1 zones down to b 0 = 0. The other borders are found numerically. B. Phase-1 ocking zones for N ~ t4, N " 4 over a more limited region of (T,bl parameter space than shown in panel A. The upper borders of the 3:2 and 4:3 zones are uncertain, as indicated by the interrupted lines (same for panel A). In panel A, the borders of the 3:1, 3:2, and 2:1 (for b "1.05) zones were determined numerically by Lean Glass, as were the borders of the 3:2 and lower 4:3· zones in panel B. 0 l-50 2 5 ....----.----..,.. ------r----, A 0 1·5 b 1·0 0·5 00 0·25 0·50 T' 0 B 1·3 b 1·2 1·1 10 region which is sometimes called the chaotic zone or chaotic region 0 in which phase-locking zones of the form 2n:M and 2n+l :M are found, as well as regions of "banded chaos"," ser:1ipedodici ty", or "noisy periodicity'' (Lorenz, 1980). Bistability of periodic orbits also occurs in the unlabelled region, with phase-locking zones overlapping so that two different N:M patterns can be found at the same point in the (b,T) parameter plane (Hoppensteadt and Keener, 1982; Glass and Belair, unpublished). In contrast, for b<l, only periodic and quasiperiodic dynamics exist. In addition, bistabil ity and sensitive dependence on initial conditions do not occur for b<l. that the dynamics is chaotic only for b>l. Thus, one might say Note however, that for b sufficiently large, only periodic dynamics is found (Fig. 1-3). Thus, in this model, chaotic dynamics exists only at intermediate 0 stirnul us amp 1 i tu des. The phenomena 1 ogy seen in different systems undergoing chaotic dynamics shows certain common features. I now identify six of these features, illustrating them with reference to the particular case of a periodically forced 1 imi t-cycl e asci 11 a tor. ( i) The dynamics is deterministic. The mathematical equations describing the system contain no stochastic terms: once the initial conditions are set, the evolution of the dynamics is predestined. (ii) The dynamics is densely packed. c 1-52 An exceedingly small change in any of the properties of the oscillator itself 0 or in the frequency or amplitude of the periodic stimulation can lead to qualitative changes (" bifurcations (iii) 11 } in the dynamics. The dynamics can display "sensitive dependence on initial conditions". Exceedingly small changes in initial conditions (such as the exact phase in the spontaneous cycle at which the driving oscillator is turned on) can lead to one of two or more different behaviours, without any alterations to either the forcing or the forced oscillator. However, these different behaviours are often unstable. (iv) 0 Bistability can be present. One of two stable periodic behaviours occurs, depending on i nit~ al conditions. (v) The dynamics which displays sensitive dependence on initial conditions is nonperiodic. For certain combinations of stimulation frequency and amplitude, the driven oscillator displays a nonperiodic time series; that is, one that does not repeat in finite time. Also, periodic sequences of arbitrarily long length can be found, which are impossible to distinguish from aperiodic sequences in numerical and experimental work. Quasiperiodic dynamics- which also has a nonperiodic time series- can also be found. (vi) The dynamics becomes chaotic by fo11 owing one of a few 1-53 well~characterized 0 "routes to chaos" (Eckmann, 1981 ). That is, there is some stereotyped sequence of bifurcations from one type of periodic or quasiper1odic behaviour to another that precedes the onset of · nonperiodic or turbulent behaviour. For example, as one moves vertically down through the (T,b) plane of Figs. 1 -3A and 1-38 for T = 0.65 starting from b = 2.0, the period-doubling route to chaos is found; zones of the form 2°:M with an increasingly higher value of n are found, until finally the chaotic region is entered. The three major routes to chaos that have been described to date are the quasiperiodic route (Ruelle and Takens, 1971; ~lewhouse, Ruell e, and Takens, 1978), the period- doubling route 01ay, 1976; 0 Feigenbaum, 1978), and the intermittent route (Mannevi11e and Pomeau, 1979). routes to chaos have also been described (e.g. instability" : Langford et al • , 1980; 1980; "crises": 11 Other Soft-mode Iooss and Langford, Grebogi, Ott, and Yorke, 1982; Ikezi, deGrassie, and Jensen, 1983}. Thus, when I speak of chaotic dynamics, I mean complex, tightly-compressed dynamics that displays several or all of the above-named features, the most striking of which are perhaps the nonperiodicity and sensitive dependence on initial conditions. 1-54 c. Experimental Work on Periodically Forced Oscillators There have not been very many experiments that have shown chaotic dynamics in a periodically forced osc ill ati ng system (unless one regards the simulation of a forced oscillator on an analogue or digital computer as an experiment on a very complex electronic osci11 ator). One of the earliest experiments to show what is now generally regarded as chaotic dynamics was one involving Rayleigh-Benard convection. This form of hydrodynamic flow occurs when a fluid layer that is held between two horizontal plates is heated from below in the presence of a gravitational field. induced at a critical Rayleigh number. Periodic motion is Periodic modulation of the temperature difference between the two plates (a parametric modulation) leads to nonperiodic dynaQics (Gollub and Benson, 1978). Curry {1978) developed a model for Rayleigh-Benard convection in which the system equations (a system of partial differential equations derived from the Navier-Stokes equations for incompressible fluid flow and the heat conduction equation) were reduced by using separation of variables and truncation of higherorder Fourier modes to a 14-dimensional system of ordinary differential equations. Small amplitude modulation of the dimensionless Rayleigh number in this model yields results qualitatively similar to those seen in the experiments of Gollub 0 1-55 and Benson (Curry, 1979}. 0 Other experiments ~n which nonperiodic time series have been found have been carried out on periodically forced neural (Hayashi, Nakao, and Hirakawa, 1982; Hayashi et al., 1982) and cardiac (Guevara, Glass, and Shrier, 1981; ·oscillators. Glass et ai., 1984) The study of Gollub and Benson (1978) showed the quasiperiodic route to chaos; Guevara, Glass, and Shrier (1981) showed the period-doubling route to chaos; Hayashi, Nakao, and Hirakawa (1982) based their identification of chaos on the theorem of Li and Yorke (1975); Glass et al. (1984) showed the perioddoubling and intermittent routes to chaos. In certain circumstances, experimental work (Glass et al., 1984: see CHAPTER 5) or numerical simulation (e.g. Tomita and Kai, 0 1978a, 1978b, 1979; Guevara et ·al., 1983) shows that the periodically-forced oscillator yields a 11 Single-humped11 map of the type that is known to be capable of generating chaotic behaviour " (Sarkovskii, 1964; Li and Yorke, 1975; Collet and Eckmann, 1980). An alternative approach to direct numerical simulation of the trajectory in phase space of the forced oscillator is to numerically study the behaviour of one-or two-dimensional difference equations that can be obtained from the system equations by making certain approximations (Zaslavsky, 1978; Coullet, Tresser, and Arneodo, 1980; Guckenheimer, 1980b; Broomhead, r~1cCreadie, and Rowlands, 1981; Levi, 1981; Alseda, Llibre, and Serra, 1982; 0 Glass and Perez, 1982; Perez and Glass, 1982; Guevara 1-56 and Glass, 1982; 0 Belair and Glass, 1983). obtain a one-dimensi anal Shrier, 1981; m~p It is also possible to in experimental { Guevara, Glass et al., 1984: Glass~ and see CHAPTER 5) and modelling (Guevara et al., 1983) work which allows prediction of the response of an oscillator to periodic delivery of a pul satil e stimulus given its response to that same stimulus delivered at various phases in its cycle. 0 0 1-57 {ii) Chaotic Dynamics in Periodically Forced Quiescent Systems As indicated in the last section, chaotic dynamics can result from periodic forcing of a system that intrinsically oscillates. Chaotic dynamics can also be produced by periodically forcing a quiescent system. In this case, there is generally a sequence of bifurcations that takes place as the stimulus intensity or frequency is changed, first leading to the appearance of various periodic or quasiperiodic behaviours and then to the appearance of a strange attractor. (A caveat: in the physics 1iterature, the term n oscillator" is often applied to systems which are not spontaneously oscillating but which nevertheless have a characteristic frequency. These are often systems that have a single stable equilibrium point with complex eigenvalues; a damped oscillatory return to the equilibrium point occurs if the system is perturbed from rest. . The classic example of this kind of system is the damped simple harmonic oscillator.) Physical systems which do not oscillate spontaneously and which apparently show chaotic dynamics experimentally in response to periodic fore i ng include ( i) (ii) {iii} {iv) the damped pendul urn (Koch et al • , 1983); the buckled beam {Moon and Holmes, 1979; Moon, 1980); the cantilivered rod (Moon, 1980); acoustical systems ( Lauterborn and Cramer, 1981; Smith, Tejwani, and Farris, 1982; Smith and Tejwani, 1983}; 1-58 0 (v) the varactor-LRC resonator (Linsay, 1981; Hunt, 1982; Jeffries and Perez, 1982; Perez and Jeffries, l982a, 1982b; Testa, Perez, and Jeffries, 1982; Brorson, Oewey, and Li nsay, 1983; Cascai s, Di 1ao, and Noronha da - Costa, 1983; Ikezi; deGrassie, and Jensen, 1983); (vi) the Josephson junction 0·1iracky, Clarke, and Koch, 1983); {vi 1) electronic analogues of the Josephson junction ( Yeh and Kao, 1982) and of the damped pendulum (D'Humieres et al., 1982}. I know of no experimental work on the periodic forcing of a 0 quiescent biological system which has dynamics. demonstl~ated chaotic Chaotic behaviour has also been found in studies carried out on numerical models of periodically forced quiescent systems. These include models of (i) the damped pendulum (Leven and Koch, 1981; Mclaughlin, 1981; Arneodo et al., 1983); ( ii) a particle lying in an anharmonic potential well (Huberman and Crutchfield, 1979; Crutchfield and Huberman, 1980; Arecchi and Lisi, 1982; Steeb, Erig, and Kunick, 1983); (iii) a particle in a spatially periodic potential (Huberman, Crutchfield, and Packard, 1980); 1-59 0 ( iv) the Josephson junction (Huberman, Crutchfiel d,. and Packard, 1980; Ben-Jacob et a.l., 1982; Yeh and Kao, 1982; Miracky, Clark, and Koch, 1983 (and references therein)); (v} (v'i) (vii) the buckled beam (Holmes, 1979}; the cantilevered rod (Moon, 1980); the varactor LRC resonator (Brorson, Dewey, and Linsay, 1983; Rollins and Hunt, 1983}; (vi i i) (ix) neural membranes (Jensen et al., 1983}; excitable membranes (Rossler, Rossler, and Landahl, 1978; Keener, 1981 a, 1981 b). 0 Work on the forced Duffing equation was initially carried out by Ueda, Hayashi, and Akamatsu (1973) and later extended by Ueda (1979, 1980a, 1980, 1981 ) , Ueda and Akamatsu (1981), Hol mes and associates (Ho1mes, 1979; Guckenheimer and Ho1mes, 1983; Holmes and Whitley, 1983 {and references contained therein)), and Sate, Sano, and Sawada {1983}. In particular, chaotic dynamics arises due to the presence of transversal homoclinic orbits {Guckenheimer and Ho1mes, 1983). As in the case of periodically forced oscillators, one-dimensional maps are a great aid in understanding the very intricate behaviours observed when a quiescent system is periodically forced (e.g. Perez and Jeffries, 1982; Rollins and Hunt, 1982; Turschner, 1982; Holmes and Whitley, 1983). Keener (l98la, 198lb) has shown that a simple two-factor 1-60 (excitation-refractoriness) model of the quiescent but excitable atrioventricular node can produce chaotic dynamics under periodic stimulation. Unlike most of the reports cited above, this study was an analytical one that relied on the iterative properties of a class of one-dimensional maps (Keener, 1980). Iteration of human atrioventricular nodal recovery curves yields the same class of maps and therefore a similar kind of chaotic behaviour (Guevara, unpub 1 i shed.). (iii) Chaotic Dynamics in Systems That Are Not Periodically Forced A third class of systems in which chaotic dynamics has been exp~rimentally forced. demonstrated are systems which are not periodically In these systems there is a sequence of bifurcat~ons to various periodic or quasiperiodic behaviours culminating in nonperiodic dynamics as some parameter in the system is changed. This parameter might be the Rayleigh number in experiments involving the dynamics of fluids or the flow rate of reactants into the reaction chamber in experiments involving chemical reactions. Hydrodynamic systems belonging to this third class of systems were among the first systems in which chaotic dynamics was demonstrated. One hydrodynamic system which has been studied a great deal is circular Couette flow, where a f1 uid rotates in the space between two concentric cylinders, the outer one being hollow and usually fixed in space, the inner one solid and rotating. the bifurcation parameter- the Rayleigh number- is increased, 1-61 As 0 there is a sequence of bifurca~ions . involving periodic and quasiperiodic motions that eventually culminates in a nonperiodic motion, as evidenced by an elevation in the base-line of the velocity power spectrum C'broadband noise11 ). Since the mid-1970's there have been many studies on this system looking at nonperiodic dynamics (Gollub and Swinney, 1975; Sw.inney and Gollub, 1978; Fenstermacher, Swinney, and Go11ub, 1979; Fenstermacher et a1 ., 1979; Gorman, Reith, and Swinney, 1980; L'vov and Predtechensky, 1981; Swinneys 1983 (and references contained therein)). Modelling of circular Couette flow produces power spectra similar to those seen experimentally (Shennan and Mclaughl in, 1978) and can produce complicated periodic and nonperiodic dynamics (L'vov and Predtechensky, 1981 ) • Another hydrodynamic system which has been investigated is Ray1 eigh--:Benard convection taking p1 ace in water ( Fenstemacher et al., 1979; Gollub.and Benson, 1979; Gollub and Benson, 1980; Gollub, Benson, and Steinman, 1980; Giglio, Musazzi, and Perini, 1981 ) , in oi·l ( Berge et al • , 1980; Duboi s and Berge, 1981 ; Duboi s, Berge, and Croquette, 1982}, in liquid helium (Ahlers and Behringer, 1978a, 1978b; Libchaber and Maurer, 1978; Maurer and Libchaber, 1979, 1980; Arneodo et al., 1983; Haucke and Maeno, 1983), and in mercury (Fauve, Laroche, and Libchaber, 1981; Libchaber, Laroche, and Fauve, 1982; Libchaber, Fauve, and Laroche, 1983). In many of the above studies, the transition to turbulent behaviour was identified with the onset of chaotic dynamics. 0 1-62 0 Deterministic numerical models of Rayleigh-Benard convection have been successfully investigated. As in experiment, there are successive bifurcations to various periodic or before nonperiodic dynamics sets in. quasiperiod~c orbits One of the earliest models ever to demonstrate deterministic nonperiodic dynamics was that of Lorenz (1963). This reduced three-dimensional model for Rayleigh- Benard convection was generated in a manner similar to that described earlier for another model of Rayleigh-Benard convection (Curry, 1978). There has been intensive theoretical and numerical investigation carried out on the Lorenz model; much has been discovered but a full description of the dynamics of the model is yet to be attained {see for example: Shimizu, 1978; Shimizu and 0 ~lorioka, Rossler, 1977; Morioka and 1978; t·~anneville and Pomeau, 1979; Shimada, 1979; Yorke and Yorke, 1979; Franceschini, 1980; Sparrow, 1982 (and references contained therein)). Extensions to the original simplistic model of Lorenz were subsequently made by Nclaughlin and ~1artin (1974, 1975), by Curry (1978, 1979), and by Franceschini ( 1983); these resulted in more complex models which also demonstrated chaotic dynamics. On the other hand, Rossler (1976a, 1976d} found that simplification of the Lorenz equations so that there was only one nonlinear cross term instead of two produced a system which was still capable of generating chaotic behaviour. It is perhaps surprising that the Lorenz equations arise in models of the segmental disc dynamo {Knobloch, 1981 ), the laser (Haken, 1975; Graham, 1976), and the water wheel (Sparrow, 1-63 1982). One final cautionary note: a recent paper proposes that some of the features of Rayleigh-Benard convection can be accounted for by a stochastic (i.e. not deterministic) model {Greenside et a1 • , 1982}. Other early observations of chaotic dynamics were made in chemical systems, most notably the Belousov-Zhabotinsky reaction {Olsen and Oegn, 1977; Schmitz, Graziani, .and Hudson, 1977; Rossler and Wegmann, 1978; Hudson, Hart, and Marinko, 1979; Schmitz, Renola, and Garrigan, 1979; Yamazaki, Oono, and Hirakawa, 1978, 1979; Vid~ et al., 1980; Pomeau et al., 1981; Roux et a1 ., 1981; Turner et al., 1981; Simoyi, Wolf, and Swinney, 1982; Epstein, 1983; Roux, 1983; Roux, Simoyi, and Swinney, 1983; Swinney, 1983; other references can be found in the last four references cited}. Modelling of the reaction steps and numerical simulation of the resulting equations leads to nonperiodic dynamics (Tomita and Tsuda, 1979; Turner et al., 1981 ). Even very simplified reaction kinetics can produce· dynamics similar to that seen experimentally {Rossler, 1976b; Kuramoto, 1978; 1983). Pikovsky, 1981; Rinzel and Tray, Dynamical behaviours resembling those seen in these chemical systems have been seen in a molluscan neuron treated with the potassium-channel blocker 4-aminopyridine (Holden, Win1ow, and Haydon, 1982); similar behaviours can also be seen in ionic models of cardiac tissue (Guevara, unpublished). Once again, in many of the studies mentioned in this section, the dynamics can be {at least partially) understood by 0 1-64 0 consideration of a one-dimensional map (e.g. hydrodynamic turbulence: Lorenz, 1963; Yorke and Yorke, 1979; Berge et al., 1980; Sparrow, 1982; chemical reactions: Pikovsky, 1981; Pomeau et al., 1981; Simoyi, Wolf, and Swinney, 1982; Rinzel and Troy, 1983; Roux, 1983; Roux, Simoyi, and Swinney, 1983; cardiac systems: Guevara, unpublished). (iv) Chaotic Dynamics 1n Systems of Two or More Coupled Oscillators A fourth situation in which chaotic dynamics can result is when two oscillators are coupled together bi di recti onally. Unlike the case of unidirectional coupling considered in section {i) above, with bidirectional or mutual coupling there is no driven oscillator and no driving oscillator; instead, the activity of each oscillator influences the other. Chaotic activity has been found experimentally in a system of two mutually coupled electronic oscillators (Gollub, Brunner, and Danly, 1978; Ito et al., 1983) and in a system of two coupled chemical reactors (Schreiber, Kubicek, and Marek, 1980). It can a1 so be seen in rnodel s of two coupled e1 ectronic oscillators (Gollub, Romer, and Socolar, 1980; Ita et·al., 1983; Thoulouze-Pratt, 1983), two or more coupled chemical cells (Rossler, l976c; Fujisaka and Yamada, 1978; Yamada and Fujisaka, 1978; Schreiber, Kubicek, and i~arek, 0 1-65 1980; Schreiber and 1•1arek, 0 1982a, 1982b), and two coupled cardiac cells (Guevara, unpublished). Chaotic dynamics also exists in an analytic treatment of two coupled van der Pol oscillators {Belair, 1982), and in numerical modelling of two coupled BVP oscillators {Guevara, unpublished). (v) Chaotic Dynamics in Systems With Time Delays The four sets of systems considered so far have been modelled by systems of ordinary differential equations. The evolution of the behaviour of the state-point of the system simply depends on its present position in phase space. 0 However, if there are time delays in the system, this is no longer the case and time-delay differential equations must be used to model the system. Chaotic dynamics has been experimentally demonstrated in an acoustical system (Kitano, Yabuzaki, and Ogawa, 1983), in optical systems (Gibbs et al., 1981; Hopf et al., 1982; Nakatsuka et al., 1983} and in electronic systems (Ikezi, deGrassie, and Jensen, 1983), all of which possess time delays. In fact, in each instance, the time delay is said to be crucial to the generation of the chaotic dynamics. Modelling work suggests that time delays can induce chaotic behaviour in physiological systems (Mackey and Glass, 1977; Glass and ~1ackey, 1979b; an der Heiden, t1ackey, and Wal ther, 1981; r·1ackey and an der Heiden, 1982), -~n optical ring cavities ( Ikeda, 1979; 0 1-66 0 Ikeda, Da1do, and Akimoto, 1980; Hopf et al., 1982; Ikeda and Akimoto, 1982; Ikeda, Kondo, and Akimoto, 1982; Nakatsuka et al., 1983}, and in acoustical feedback systems (Kitano, Yabuzaki, and Ogawa, 1983). Simple models, in which one-dimensional maps again play a role, provide insight into the complex dynamics (an der Heiden, f'.1ackey, and Walther, 1981; an der Heiden and t·1ackey, 1982; Ikeda, Kondo, and Akimoto, 1982). It is interesting to note that the Lorenz equations for Rayleigh-Benard convection can be reinterpreted as the equations of a particle subjected to a force which, at any given time, depends both on the present position of the particle and its position at all previous times {Shimi zu and f.torioka, 1978). Note that even though a system with a time delay is of infinite dimension, there can be finite dimensional structures such as limit cycles and strange attractors. The fractal dimension of the strange attractor can be as low as 2.13 {Farmer, 1982). (vi) Chaotic Dynamics in Spatially-Distributed Systems Chaotic dynamics can also occur in spatially-distributed systems where the variables are tunctions of both time and space. In fact, several of the systems mentioned so far (e.g. hydrodynamical systems and the heart) are distributed systems which should properly be modelled by partial differential equations. Chaotic dynamics has been shown to exist in models of systems 1-67 0 formulated as partial differential equations (e.g. Moon and Holmes, 1979; Keener, 1981a; Bishop et al., 1983a, 1983b; Moon, Huerre, and Redekopp, 1983). Interesting phenomena such as spatial period- halving bifurcations and the suppression of temporal chaos by spatial structure can be seen (Bishop et al., 1983a). The converse phenomenon of spatial chaos and temporal periodicity has also been seen in modelling work (Kopell, 1980). {vii) Concluding Remarks Concerning Chaotic Dynamics Within the last five or so years, chaotic dynamics has been seen experimentally in many systems: 0 physical (hydrodynamic, el ectrohydrodynamic, acoustical, electronic, and optical), chemical, and biological (neural and cardiac). i~athematically, chaotic dynamics can be produced by one- or two-dimensional finite difference equations, systems of three or more ordinary differential equations, time-delay differential equations, integrodifferential equations, and partial differential equations. Chaotic dynamics can be seen in mathematical models of many systems in which it is experimentally observed. It is also seen in models of systems where the experimental determination is yet to be made. For example, the Einstein field equations admit chaotic solutions in the mixmaster cosmological model (Chernoff and Barrow, 1983}. This thesis grew out of the conviction that much of the irregular dynamics appearing in the clinical electrocardiogram might be the result of chaotic dynamics. In what follows, I set out to con vi nee the reader that this might indeed be so. 1-68 0 CHAPTER 2 PHASE RESETTING OF THE RHYTHM OF SPONTANEOUSLY BEATING AGGREGATES OF EMBRYONIC CHICK VENTRICULAR CELLS BY A CURRENT PULSE OF BRIEF DURATION 0 u in order to gain any real insight ••• , it was absolutely necessary to study first the comparatively simple deviations caused by single stimuli • 11 A. Cushny and S. Matthews, 1897 1. c INTRODUCTION A single premature beat can dramatically affect the rhythm of the heartbeat. An appropriately timed stimulus delivered to a normal healthy heart can abolish rhythmicity by inducing ventricular fibrillation {Mines, 1914), as can a premature beat of endogenous origin {Smirk, 1949). A single premature stimulus can annihiliate spontaneous activity in the isolated sinoatrial node, in depolarized Purkinje fibre, or in diseased human ventricular myocardium (Jalife and Antzelevitch, 1979; Jalife and Antzelevitch, 1980; Gilmour et al., 1983); it can also terminate triggered activity in canine Purkinje fibres or in fibres of the simian mitral valve (Cranefield and Aronson, 1974; Cranefield, 1976). Wit and However, the effect of a premature stimulus is usually not quite so drastic; the result more generally seen is a transient alteration in the cardiac rhythm, followed by a reestablishment of the rate and rhythm initially present. As mentioned in CHAPTER 1, it has been known for a long time that a premature beat resets the phase.of the oscillatory activity of the sinoatrial node or a subsidiary pacemaker to an extent that depends upon the prematurity of the beat. For instance, a premature atrial contraction that arrives sufficiently late in the cycle of the sinoatrial node will invade and capture it, thus shortening the sinus cycle, whereas one arriving early enough encounters refractoriness and entrance block and will instead 0 2-1 produce an electrotonic depolarization in the node that will delay 0 the time of appearance of the next spontaneous action potential (Klein, Singer, and Hoffman, 1973; Kerr et al., 1980}. A similar situation occurs in pacemakers other than the sinoatrial node {Klein, Cranefield, and Hoffman, 1972; Ferrier and Rosenthal, 1980; Gilmour et al., 1983). Within the last decade, there has been renewed interest in the phase-resetting of spontaneously active cardiac tissue produced by premature stimulation with current pulses. Sano, Sawanobori, and Adaniya (1978) demonstrated that the cycle length of the rabbit sinoatrial node could be either prolonged or shortened by extracellular injection of a depolarizing current pulse. Jalife et al. (1980) studied the response of strips of kitten sinoatrial node subjected to pulses of current delivered across a sucrose gap. The sign and magnitude of the change produced in the sinus cycle length depends on the phase of the cycle at which the pulse is injected, as well as on the polarity, amplitude, and duration of the current pulse. For example, a depolarizing pulse delivered early in the cycle prolongs the cycle length, while the same pulse delivered later on in the cycle shortens the cycle length. Weidmann (1951} injected current pulses into Purkinje fibres during diastolic depolarization, and was the first to demonstrate clearly that a depolarizing input could prolong the cycle length of spontaneously active tissue, while a hyperpolarizing input could shorten the cycle length. 0 Jalife and Moe (1976) later carried out a more 2-2 systematic study using canine false tendons in a sucrose gap 0 apparatus. The above experiments have two factors in common which complicate the analysis, interpretation, and modelling of the results. Firstly, the pacemaker is a distributed structure consisting of a heterogeneous population of cells in which propagation effects contribute to produce an asynchronous activation of the individual cells of the pacemaker. For example, in the sinoatrial node, there are transitional cells between the pacemaker proper and the surrounding quiescent atrial myocardium (Pevet-Masson, 1979). Secondly, the stimulus is a broad invading wavefront of depolarization which does not affect all cells equally. For example, electrotonic shortening of action potential duration can occur in some cells of the sinoatrial node in response 0 to a relatively late premature atrial contraction that is blocked in the peri nodal fibres ( Mil 1er and Strauss, 1974; Dorticos et al., 1978; Steinbeck et al., 1978}. A premature beat that is early enough to encounter entrance block has a fractionated wavefront which invades and captures only a minority of the cells in the pacemaker. Other cells display a subthreshold depolarization, the magnitude of which varies from cell to cell. Furthermore, shifts in the site of the dominant pacemaker can occur in response to premature stimulation (Bonke, Bouman, and van Rijn, 1969; Bonke, Bouman, and Schopman, 1971; Klein, Singer, and Hoffman, 1973). These complications remain in studies of the sinoatrial node {Sano, 2-3 Sawanobori, and Adaniya, 1978; Jalife 0 et~., 1980} and of spontaneously active Purkinje fibre (Jalife and Moe, 1976) in which extracellular injection of current was employed. Even in cases where intracellular injection of current was carried out ( sinoatrial node: IJshiyama and Brooks, 1974; Purkinje fibre: Weidmann, 1951), population and conduction effects necessarily complicate the response. Finally, ionic modelling of propagated action potentials requires the numerical integration of a partial differential equation. Computer programs to carry this out are not readily available, as well as being very time consuming (and therefore also expensive) to run. To obviate these complicating factors, I have decided to investigate the-phase resetting of a cardiac oscillator in which the presence of a heterogeneous cellular population and propagation effects do not appear to play a significant role. The preparation employed is the embryonic chick ventricular heart cell aggregate, whose phase-resetting behaviour has been hitherto studied (DeHaan and Fozzard, 1975; Scott, 1979; Guevara, Shrier, and Glass, 1980; Ypey, van Meerwijk, and DeHaan, 1982; van Meerwijk et al., 1984). Since the cells that make up the aggregate are virtually isopotential {DeHaan and Fozzard, 1975; OeFelice and DeHaan, 1977; DeHaan and DeFelice, 1978a, 1978b; Ebihara et al., 1980; Mathias et al., 1981), the aggregate can be modelled by a system of ordinary differential equations {Guevara et al., 1982; Shrier and Clay, 1982; Shrier et al., 1983; Clay, Guevara, and Shrier, 1984: 0 2-4 see CHAPTER 3 of this thesis). 0 The aggregate may be viewed as an analogue of the dominant centre of the sinoatrial node, which is the pacemaker of the heart and consists of about 5000 cells displaying virtually simultaneous electrical activity (Sleeker et al • , 1980). In what follows, I first outline the methods used to prepare the aggregate and to study its response to premature electrical stimulation. Next, I summarize its basic el ectrophysiol ogical properties and describe its response to intracellular injection of a single pulse of current. Finally, I discuss the experimental observations, offer some interpretations, and indicate the implications of this work for normal and abnormal electrical activity in the intact heart. 0 2. METHODS (1) Tissue Culture: a. Technique Aggregates were prepared following the techniques described in McDonald, Sachs, and DeHaan (1972) and in Sachs and DeHaan (1973) with minor modifications. Dissociation of the parent tissue into single cells is obtained· by a multiple cycle procedure 2-5 (DeHaan, 1967, 1970), and reaggregation of single cells into heart 0 cell aggregates is obtained by a gyration process ( ~1oscona, 1961, 1965; Fischman and Moscona, 1971}. White Leghorn chick embryos were incubated for 7 days at a temperature of 37°C and a relative humidity of 85%. the embryos were between stage 29 and stage 31. At this point, Each embryo was removed from its shell and decapitated under sterile conditions. The thorax was opened and the heart {together with the attached great vessels) gently removed. The heart was transected by a transverse cut made just below the level of the atria. The atria were discarded and the apical portions of the ventricles of four to twelve enbryos were then snipped into small fragments with a fine iridectomy scissors. 0 The fragments were then transferred into a 25 ml Erlenmeyer f1 ask containing 10 ml of di ssoc i ati on medi urn at 37° C and a magnetic stirring bar; the flask was then tightly sealed with a silicone stopper. The f1 ask containing the medium, ventricular fragments, and magnetic stirring bar was then placed on a magnetic stirring table (Thermolyne: Type 7200). After ten minutes of combined mechanical agitation and chemical digestion at ambient temperature, the supernatant, which contains few viable myocytes (Josephson and Sperelakis, 1982}, was discarded. A second seven-minute dissociation cycle was then carried out in the same way, using 3.3 ml of fresh dissociation medium, but the supernatant (now containing many isolated myocytes) was poured into a 50 ml 2-6 centrifuge tube containing 20 ml of enzyme-inactivating medium 0 maintained at 37°C. Two more dissociation cycles were carried out, each time with 3.3 ml of fresh dissociation medium. Before transferring the supernatant to the centrifuge tube at the end of each cycle the contents of the flask were gently agitated by sucking the contents up and down in a pipette in order to help break up the surviving ventricular fragments. At the end of the last dissociation cycle, the entire contents (including undigested fragments) of the flask were transferred to the centrifuge tube. The contents of the centrifuge tube were then poured into the barrel of a 20 ml syringe, the end .of which had been previously fitted with a polycarbonate filter with 12.0 llm-diameter pores OJuclepore: No. 110616). Gentle pressure applied to the plunger of the syringe yielded a suspension of single cells, including myocytes, red blood cells, and fibroblasts. The suspension of single cells was then compacted into a pellet by centrifugation at about 170 g for 15 minutes. The pellet was resuspended in 1 ml of pre-gassed (10% 02 , 5% C0 2 , 85% N2 ) maintenance medium. After the cells were evenly suspended, a Pasteur pipette was used to place an aliquot of the single cell suspension on a haemacytometer slide. After the cell density in the suspension was determined, appropriate volumes of the suspension were aliquoted out into two to six flasks, each containing 3 ml of pre-gassed maintenance medium. The resultant number of eel 1 s in each flask was from 5 x 10 5 to 7 x 10s. The 2-7 f1 asks were then placed on a gyratory table (New Brunswick Scientific Co: Model G2) which rotated at 70 r.p.m. with a stroke of about 1~ cm and which was enclosed in an incubator maintained at 37°C. b. Media The dissociation medium consisted of 5.25 x l0- 5 g/ml crystalline lyophilized trypsin (Worthington Biochemical, 245 U/mg) and 5 x l0- 6 g/ml deoxyribonuclease I (Worthington, 9.1 x 10 4 U/mg) in a ca++-Mg++-free, phosphate-buffered, balanced salt solution: (millimolar) NaCl 116.0, KCl 5.4; NaH 2 P0 4 0.44, Na 2 HP0 4 0.95, dextrose 5.6. In some cultures the trypsin was replaced with ·collagenase 5 x 10-s g/ml (Sigma, Type 1, 146 U/mg). Aggregates formed from trypsin-dissociated cells will be referred to as trypsin-dissociated aggregates; those formed from collagenasedissociated cells will be called collagenase-dissociated aggregates. rhe pH of the dissociation medium was adjusted to 7.3 with either 1 N HC1 or 1 N NaOH. The maintenance medium, a modification of medium 818A {OeHaan and Fozzard, 1975) consisted of 2% horse serum (Kansas City Biological), which was heat-inactivated by placing in an oven for 30 minutes at 57°C, 4% fetal bovine serum {Grand Island Biological (GIBCO}), and 20% medium 199 {GIBCO) in a bicarbonate-buffered balanced salt sol uti on. 0 The final concentrations (millimolar) were 2-8 approximately: NaCl 116.0, KCl 1.3, CaC1 0 0.9, NaHC0 3 20.0, dextrose 5.5. {Schering: 2 1.8, MgSO'+ 0.8, NaH 2 P0 4 The antibiotic gentamicin sulphate Garamycin, 10 mg/ml) was also added to the medium to yield a final concentration of 5 x 10-s g/ml. The enzyme-inactivating medium was the same as the maintenance medium, but with the following exceptions: 0% fetal bovine serum, 10% horse serum, and approximately 4 mM KCl. All solutions were filtered with a sterile filter having a 0.45 ~m­ diameter pore size (Nalgene: No. 245-0045). Preparation of aggregates and formulation of solutions was carried out by the technical staff mentioned earlier in the acknowledgements. (11) Electrophysiology After two to four days of gyration culture, the 3 ml of maintenance medium, now containing several score reaggregates of cardiac cells, was poured into a 35 mm x 10 mm plastic tissue culture dish (Becton, Oickinson: Falcon 3001}. The aggregates firmly adhered to the bottom of this dish within 20 minutes. The dish was placed in the well of a heater plate lying on the stage of a dissecting microscope. The temperature was continuously monitored by a thermistor probe (YSI: Model 43TD) and was maintained to within one degree of 36°C by a switching regulator built around a temperature regulator integrated circuit (National: 0 2-9 LM3911N}. 0 Slight adjustment of the controller was needed at the beginning of and occasionally during each experiment. Non-toxic mineral oil (Witco Chemical: Klearol) was layered out on top of the medium to prevent evaporation. The medium was gassed from above by a toroidal gassing ring at a flow rate of 200 ml/min with a gas mixture of 5% C0 2 , 10% 02 , and 85% N2 • The bicarbonate buffer in the medium maintained the pH at about 7.2 or 7.3. Phenol red in the maintenance medium (0.04 mg/ml} provided a rough continuous estimate of the pH. Under these conditions, more than 95% of the aggregates in a dish will beat spontaneously (r~Donald, Sachs, and OeHaan, 1972; Sachs and OeHaan, 1973). Mean aggregate diameter was taken to be the mean of the minor and major axes in the horizontal plane. 0 These were measured with an optical graticul e placed in the microscope eyepiece. Oi ameters could be estimated accurately and repeatedly to within one half of a minor division of the graticule (± 19 ~m). The volume of an aggregate was calculated assuming it to be a sphere with a diameter equal to the computed mean diameter. Using a micromanipulator, one cell of an aggregate was impaled with a machine-pulled (Sutter Instrument: Model P-77 Brown-Flaming) glass microelectrode filled with 3t4 KCl (20-100 Pl\1 resistance). Transmembrane potential was registered using a negative capacitance compensated amplifier which had a gain of x50. In these experiments, the capacity compensation was adjusted by bringing the amplifier just to the edge of oscillation. 2-10 The medium in the dish was maintained at virtual ground by being coupled to a current-to-voltage converter (10-100 mV/nA) through an agar salt bridge and a chlorided silver wire. Pulses of current were injected into the aggregate through the same microelectrode used for recording the membrane voltage and their amplitudes measured by the current-to-voltage converter. nearest~ nA. Currents were measured to the A programmable stimulator (Frederich Haer: Pulsar 4i) was used in conjunction with a 22 Mn resistor to deliver current pulses. In some experiments, current pulses were injected in a chopped fashion, at a frequency of 1kHz. Within several minutes of the initial impalement, the 11 Sealing in process (Draper and Weidmann, 1951; Shrier and Clay, 11 1982} was complete, and spontaneous activity with stationary action potential parameters could be ·obtained. Voltage and injected current waveforms were monitored on a digital oscilloscope {Nicolet: Model 206) and recorded on an FM instrumentation recorder at 3% ips {Hewl ett Packard: Hodel 3964A; 3dB frequency response at 3% i ps: OC-1250 Hz) for 1ater offl i ne analysis. Impal ements could · sometimes be maintained for several hours. The protocol for investigating the response of spontaneous beating to perturbation with a pulse of current was as follows. A pulse of current was delivered at a certain coupling interval after a spontaneous beat. This current pulse altered the interbeat interval of the perturbed cycle. After allowing ten beats to take place to allow recovery back to control activity, another pulse was 0 2-11 injected at a coupling interval that was automatically incremented (by a multiple of 1 msec). cycle could be scanned. In this way, the entire spontaneous In some experiments, the coupling interval was not systematically incremented, but was instead randomly varied. The above protocol could then be repeated for a different current amplitude, duration, or polarity. (iii) Data Analysis Onl ine and some offl ine analysis was carried out using the digital oscilloscope to measure i nterbeat intervals and coupling intervals. The sampling frequency was usually set to 1kHz, and since the scope memory contained 4096 samples, approximately 4 seconds of data could be displayed on the oscilloscope screen. Interbeat intervals could therefore be reproducibly measured with a precision of ± 1 msec. The bulk of the offline analysis was performed by an automated system. (i.e. 1:4 Magnetic tapes were played back at 15/16 ips real time), 1ow-pass filtered at 500 Hz, and the voltage waveform then sampled at 250 Hz by a Z80-based microprocessor system (Cromenco: System III; California Data Corporation: 12 bit analogue-to-digital converter). AD-100 The digitized waveform was transferred over an RS-232 serial line at 9600 baud to a minicomputer (Hewlett-Packard: Model HP1000 Series F) and stored on digital magnetic tape. Interbeat intervals were extracted out of 2-12 the digitized waveform by a pattern recognition program. Raw data and processed curves were output on a video terminal (Lear-Siegler: r1odel ADM-3A fitted with a graphics board (Digital Engineering: Model RG-512}} or on a digital plotter (Hewlett-Packard: Model 9872$). With the exception of Fig. 2-lA, all of the experimental voltage traces in this thesis were obtained by playing back the tape-recorded signal to the digital osc ·ill oscope through a 1ow-pass filter (Krohn-Hite: Model 3323: 4-pole RC filter, 24dB/octave attenuation) with a cutoff frequency set at twice the oscilloscope sampling frequency. The contents of the oscilloscope memory were then reconverted to an analogue signal using a digital-to-analogue converter internal to the oscilloscope and sent to an analogue X-Y plotter (Hewlett-Packard: 3. RESULTS (f) Spontaneous Actfvity l~odel 7015B}. In this study. stable spontaneous electrical activity was recorded from 60 trypsin-dissociated and 12 collagenase-dissociated aggregates, with mean diameters ranging from 95 ~m to 228 ~m. Aggregates in this size range are composed of approximately 6008,000 cells (Clay, DeFelice, and DeHaan, 1979}, 75-85% of which are 2-13 myocytes (Sachs and DeHaan, 1973}. Sachs and DeHaan (1973) found that extracellular space amounted to 20.4 ml/100 gm of aggregate tissue, while Clapham (1979) found that it amounted to 3.6 ml/100 ml of tissue volume. than 95 ~m Aggregates with diameters of less were not used, since it is known that beat-to-beat fluctuation in interbeat interval increases dramatically with decreas.ing size for aggregates containing fewer than 125 cells (Cl ay and DeHaan, 1979). {An aggregate containing 125 cells is about 60 Aggregates with diameters larger than 228 ~m in diameter.} were not studied, because of the possible existence of ~m voltage inhomogeneity and interstitial potassium accumulation in such large preparations. A 200 ~m diameter aggregate has about 2/3 of its cells in the outermost 3 cell layers (Clay and Shrier, 0 198la). Necrotic cells are not seen in aggregates with diameters of less than 250 ~ {Williams and DeHaan, 1981). Fig. 2-1A shows a typical recording of spontaneous electrical activity. In a separate study carried out on other aggregates obtained from the same cultures as used in these experiments, the visually monitored beat rate (mean ± S.D.) of 104 trypsin- dissociated aggregates was found to be 53.3 Guevara, and Shrier, 1983). ± 32.3 min-1 (Colizza, While larger aggregates tend to beat more slowly than smaller aggregates, there is a large scatter in the beat rates of different aggregates of the same size. The action potentials of 14 trypsin-dissociated aggregates were found c 2-14 Figure 2-1. A. Tracing of transmembrane voltage recorded during spontaneous unperturbed activity in a trypsindissociated aggregate, illustrating the regularity of beating. All experimental traces in this thesis are from trypsin-dissociated aggregates unless otherwise labelled. Since this impalement lasted for several hours, and since there are drifts in the recording system, the zero of transmembrane potential difference can only be estimated. This consideration also holds true for most of the other voltage tracings in this thesis. 0 B. A single action potential from the same period of spontaneous ac ti vi ty as shown in pane 1 A, with OS • = 31 mV, MOP = -88 mV, APO = 145 msec, V max = 72 V sec- 1 • Panel A was obtained by playing back the tape recorder (at !.i real time) directly onto an analogue X-Y plotter. Panel B was obtained by transferring the contents of the digital oscilloscope memory (sampling frequency 10kHz) through a digital-to-analogue converter (internal to the scope) to the analogue X-Y plotter. (Aggregate #1: diameter = 114 urn. ) 2-15 0 lsec B ] O.lsec 0 V -som • to have (mean± S.D.) maximum upstroke velocity ( Vmax) of 120.9 ± 0 35.4 V sec-1, maximum overshoot potential (OS) of 28.0 action potential duration (APO: diastolic potential) of 171.5 ± ± 4.1 mV, time from upstroke to maximum 27.6 msec, and maximum diastolic potential {MOP} of -90.8 ± 6.3 mV {Colizza, Guevara, and Shrier, 1983}. Fig. 2-18 shows an action potential demonstrating these typical parameters. The fast-channel blocker tetrodotoxin (TTX} at a concentration of 10- 7 g/ml or the slow channel blocker D600 at a ·concentration of 10- 5 g/ml abolishes spontaneous activity in these aggregates, as does i ne rea sing the external potassi urn concentration to 4.5 mM (Col izza, Guevara, and Shrier, 1983}. A general review of other properties of the trypsin-dissociated aggregate preparation can be found in OeHaan and OeFelice {1978b). 0 Aggregates formed from collagenase-dissociated cells have been shown to have action potential parameters which are not too different from those listed above for aggregates formed from • trypsin-dissociated cells: the one major exception is Vmax (mean± s.o., 25 aggregates), which is 24.3 ± 13.4 V sec- 1 (Colizza, • Guevara, and Shrier, 1983). Reduced values of Vmax have also been found in collagenase-dissociated aggregates in two other studies (Mackenzie and Standen, 1982; van Meerwijk et al., 1984). Furthermore, TTX {10- 7 g/ml} did not block spontaneous activity in six out of fourteen collagenase-dissociated preparations studied (Colizza, Guevara, and Shrier, 1983). Thus, the fast sodium 0 2-16 channel does not appear to be functionally present in these 0 collagenase-dissociated aggregates to the extent tha~ it is in trypsin-dissociated aggregates, nor does it appear to be necessary for spontaneous action potential generation. Furthermore, collagenase-dissociated aggregates tend to beat faster than trypsin-dissociated aggregates of the same size (Colizza, Guevara, and Shri er, 1983). Figure 2-2A shows the i~ interbeat interval IBI; plotted as a function of the interval number i for 826 consecutive beats recorded during a period of unperturbed spontaneous activity. The interbeat interval (or cycle length) is the time between crossings of 0 mV on the upstrokes of two successive action potentials. For these 826 beats, the mean interbeat interval was 686 msec and the standard deviation (S.D.) was 10.2 msec, giving a percentage coefficient of variation C (= 100 S.D./mean} of 1.48%. shows the interval histogram. Fig 2-28 Irregularly beating aggregates {as judged by observing the f1 uc tuati ons in i nterbeat interval on the oscilloscope) were occasionally encountered, but were not used to obtain phase-resetting data. One such aggregate had C = 4.92%. Thus, the beat-to-beat regularity of the 7-day embryonic chick ventricular aggregate (C ~ 1-5%) appears to be better than that of a single isolated embryonic chick ventricular cell, for which C"' 20% {Clay and DeHaan, 1979), but is not quite as good as that of the isolated rabbit sinoatrial node, for which C < 1% (Jongsma and 0 2-17 Figure 2-2. A. Interbeat interval IBI 4 plotted vs. interval number i from a period of unperturbed activity. Number of beats = 826. an interbeat interval is B.. The precision in measuring ± 1 msec. Unsmoothed interval histogram obtained from the data shown in panel A. Intervals are sorted into bins and the number of intervals n in a given bin is plotted against the interbeat interval IBI in msec. Number of interbeat intervals = 825, mean interbeat interval = 686 msec, standard deviation = 10.2 msec, coefficient of variation = 1.48%, bin size = 2 msec. This aggregate appeared to have a typical beat-to-beat variability, as judged by triggering on the upstroke of the action potential and observing successive interbeat intervals on the oscilloscope. This ·was done in all impale- ments to obtain a rough estimate of the variability of the interbeat interval. 2-18 0 c. Scattergram for the same data used in panel A showing each of 824 interbeat intervals plotted as a function of the immediately preceding interbeat interval. The straight line plotted through the data is a least-squares !it to the data; it has a slope of 0.92 and a coefficient of determination (r2) o. of 0.922. Plot of the first 51 serial correlation coefficients Rj for the data shown in panel A. E. Plot of the Rj (0(j,50) for the same data, but with the IBI.; randomly shuffled. Note the· 1ack of ' any carrel ati on for j ;Jrl. Adjacent points ( j, Rj) and. {j +1, Rj +l) are connected together by straight lines. 0 (Aggregate #2: 2-19 diameter= 105 ~m.) A 0 750 IBii 675 (m sec) B D 100 2 r I; l I t n R·J 50 Q 0 650 c IBI (msec) E 750 / IBii+l 700 (m sec) 1 R· 0 J ,, IBii (msec) 0 0 750 750 0 50 Tsjernina, 1982). 0 Figure 2-2C is a scattergram showing each of 824 interbeat intervals plotted against the previous interbeat interval. Thus, an interbeat interval that is longer (shorter) than average tends to be followed by one that is also longer (shorter) than average, but less so. A straight line fit through the data resulted in a line of slope less than 1 in each of four trypsin-dissociated aggr.egates. whose spontaneous activity was analyzed. The serial correlation coefficients (Perkel, Gerstein, and Moore, 1967; Jongsma et al., 1975) R.J =---------------------------- were calculated, where Rj is the jth serial correlation coefficient, N is the total number of interbeat intervals, IBii is the ith interbeat interval, and 11IT is the mean interbeat interval. The coefficients Rj (1 < j < 50) were all positive and fell more or less monotonically with increasing j (Fig. 2-20). Random shuffling of the IBI; destroys the dependence of IBI 1 on IBii-l and reduces the Rj ( l<<j <50) to values close to zero (Fig. 2-2E). Most aggregates displayed occasional "bursts" (sudden 2-20 increases in the spontaneous beat rate). c of bursts of varying le~gths Figure 2-3 shows examples from four different aggregates. The increase in intrinsic frequency is accompanied by a marked increase in the rate of diastolic depolarization; there is typically a small change in the maximum diastolic potential and little or no change in the threshold potential. Some bursts developed and died away gradually (Fig. 2-3A,B); qthers were more paroxysmal (Fig. 2-JC,D). Bursting could encompass anywhere from 3-5 action potentials (Fig. 2-3A} to approximately 100 action potentials (Fig. 2-30). (Bursting was not present during collection of the data used to make up Fig. 2-2.) (ii) 0 The Response to Perturbation with Single Depolarizing Pulses Phase-resetting data was obtained at one and only one d~polarizing current amplitude in 9 trypsin-dissociated aggregates, and at two. or more amplitudes in another 17 trypsin-dissociated aggregates; these aggregates were taken from 20 cultures. Since the phase-resetting responses of only 8 collagenase-dissociated aggregates were studied, I will not here report on their response. Figure 2-4 shows the effect of injecting a single depolarizing current pulse of 20 msec duration and 6.5 nA amplitude into a spontaneously be·ati ng aggregate. Depending on the phase of the cycle at which it falls, this brief pulse of current can either prolong (Fig. 2-4, upper trace} or shorten (Fig. 2-4, lower trace) 0 2-21 0 Figure 2-3. 11 Bursti ng behaviour recorded in four different 11 aggregates. Only the first half of the burst that occurred in panel D is shown. These voltage tracings were obtained by transferring the contents of the digital oscilloscope memory to the analogue X-Y plotter. Since the interval between samples was 2 msec (panels A,B) or 5 msec (panels C,O), the upstrokes of the action potentials do not appear to be as rapidly rising as they really are (contrast with Fig. 2-lA, which was obtained without any digitization process). This artifact is present in many of the experimental voltage tracings following in this thesis. (A: aggregate #1, diameter = 114 11m; B: #3, diameter = 190 11m; C: aggregate #4, aggregate diameter = 152 11m; 0: aggregate #5, diameter= 95Jlm.) 2-22 0 A 0 ]mv -50 1sec B D 0 0 Figure 2-4. Effect on the spontaneous rhythm of an aggregate produced by injecting a 6.5 nA amplitude current pulse of 20 msec duration. The stimulus artifact is the off-seal e rapid vertical deflection in the tracing and can be used as a marker for the time during which current is injected. The number to the 1eft of these and other voltage traces in this chapter is the coupling interval in milliseconds. · 0 Upper trace: T1 = 527 msec. T0 = 469 msec, tc = 170 msec, The perturbed cycle length T1 is greater than the control cycle length T0 • Lower Trace: msec. T0 = 467 msec, tc = 180 msec, T1 = 305 The perturbed cycle length T1 is less than the control cycle length T0 • Note that in both cases the stimulus has 1i ttl e effect on the post-stimulus cycle. (Aggregate #1: diameter= 114 2-23 ~m.) 0 PULSE AMPLITU DE 6.5nA COUPLIN G INTERVA L (m sec) 0 170 ] mV 180 ] -50 1 sec the interbeat interval of the perturbed cycle. 0 I call this interbeat interval the perturbed interbeat interval, and denote it by T1 • The coupling interval ( tc) is defined to be the time from 0 mV on the upstroke of the action potential immediately preceding the stimulus to the beginning of the stimulus; the interbeat interval of the spontaneous cycle immediately preceding the perturbed cycle is denoted by T0 • I refer to the cycle immediately preceding (following) the perturbed cycle as the control (poststimulus) cycle. Note that there is almost no residual effect of the stimulus on the interbeat intervals of the cycles following the perturbed cycle (however see section (vii) below). Figure 2-5 (left panel) shows the effect of delivering depolarizing pulses of 5 nA amplitude and 20 msec duration at 0 coupling intervals that are systematically incremented in 10 msec steps. The current waveform remained constant as the coupling interval was changed. Ten spontaneous beats are allowed between trials in order to give the preparation enough time to recover back to its control activity. Stimuli delivered at coupling intervals of less than about 100 msec had little effect on the interbeat interval due to the low membrane resistance during phase 2, and the voltage tracings are therefore not shown. A stimulus delivered later in the cycle {110 msec < tc < 170 msec) causes a prolongation in the perturbed interbeat interval beyond the control value, ~hile an identical stimulus delivered still later in the cycle (t) 180 c msec) causes an abbreviation of the interbeat interval. 0 2-24 The 0 Figure 2-5. Phase resetting at 3 different current amplitudes (left panel: 5 nA; middle panel: 6.5 nA; right panel: 8 nA) produced by a 20 msec duration current pulse. The coupling interval tc is steps. (e.g. tc incre~ented. in 10 msec For stimuli sufficiently late in the cycle ~ 200 msec in the right panel) the upstroke of the stimulated action potential is obscured by the stimulus artifact. Note that the transition from prolongation of cycle length to abbreviation of cycle 0 length occurs abruptly in the right panel (between tc = 160 msec and tc = 170 msec). One can also see from the rightmost panel that the stimulus produces little effect on the interbeat interval of the poststimulus cycle (tc = 170, 180, 190 msec). aggregate as in Figs. 2-4, 2-6, 2-7, 2-8. (Aggregate #1: diameter = 114 2-25 0 ~m.} Same 0 PVLS £ A.MPLJTI-'~!: ~ ,. ... 6.5 nA I J_"'o" CONTROL " ~~ \ ' 100m tu COUP LING ltHE RVA L fm•.c l 110 j~~L-~~ 120 130 r~J .J i _j~~t 140 150 160 110 \J\ ___]'\ J\Jl ___~__J\ _ ___0JL_r'L _j \_ L_ _ 180 I !H) 200 210 I'\ _)1\\JI!L__o \._..-...- _ i\ J lJ '\ _ _f \ IU\ . ~ /: 220 230 ___ j\ _ _J\j~ 1\Jl 1\~ j \\ _J I _ j\ J L _ f\__ S\JU\\__-_I\ju\___ _i\ju"\__11'\ --' /f\__ \_ _f\J~L~--j _}\jf\____ .J\Jf\ ii \~-__,f\uU _j\j~L-- _j\J~\-~~- _j\J~-~- J \j - J\J~L -_ j 2<0 250 _ j\ /~\ ~ \._. .--.- ---- i" ----' IL _J~ i ·.__.. .! I prolongations are associated with a decrease in the slope of 0 diastolic depolarization following the end of the current pulse. This decrease becomes progressively greater as tc increases for approximately tc (T 1/T 0 = 1.12) ~ 180 msec; however, the maximum prolongation occurs for tc = 150 msec. If one thinks of the spontaneous activity as being generated by diastolic depolarization to a fixed threshold voltage, there are at 1east two other factors that influence the time to threshold following a stimulus: {i) the take-off potential becomes more positive with increasing \; increases with increasing \· (ii) the membrane slope resistance These two factors conspire to make the voltage attained at the end of a pulse more positive with increasing tc. Thus, for tc increasing in the range 150 msec " tc ( 180 msec, progressively decreasing interbeat intervals can be obtained despite a progressively decreasing slope of diastolic depolarization, since the membrane voltage is being brought increasingly closer to threshold at the end of a current pulse. Stimuli falling before the occurrence of MOP can prolong the action potential duration (time from upstroke to t4DP); this is a separate cause for prolongation of the interbeat interval. Figure 2-5 {middle panel) demonstrates that increase in the amplitude of the current pulse to 6.5 nA causes an increase in the maximal prolongation and abbreviation observed, as well as a decrease in the coupling interval at which prolongation of cycle length changes into abbreviation of cycle length. c 2-26 The transition from prolongation to abbreviation also occurs more abruptly (i.e. 0 over a shorter range of coupling intervals) at this higher stimulus intensity. These effects are even more pronounced as the amplitude of the current pulse is further increased to 8 nA, as shown in Fig. 2-5 (right panel). At this current level, the transition from prolongation of cycle length to abbreviation of cycle length takes place w;th a change in tc of 1ess than 10 msec. Figure 2-6A shows 10 randomly .selected superimposed cycles fran a control period of unperturbed activity, while Fig. 2-68 shows superimposed tracings taken from ten repeated trials with a 16 nA amplitude pulse at each of two fixed coupling intervals. At this current level, the transition from prolongation to abbreviation of cycle length is very abrupt, and can occur when c tc is changed by as little as 1 msec (see section (vi) below). The responses shown in Fig. 2-68 are highly repeatable, in that there is only a slight degree of variability apparent in the response from trial to trial. The same holds true for current amplitudes of less than about 5 nA in this aggregate. However, at current amplitudes greater than about 5 nA but below about 10 nA, there can be considerable scatter in the response in repeated trials carried out at a fixed coupling interval, if tc is in the range where the response is rapidly changing fr~~ maximal prolongation to maximal abbreviation of cycle length. This effect is considered further in section (vi) below. For a much larger current amplitude, the above effects 0 2-27 0 Figure 2-6. A. 10 superimposed voltage tracings randomly selected from a period of unperturbed control activity immediately preceding the collection of the data shown in panel B. The upstrokes of the first of the two action potentials occurring in each trace. are synchronized. B. 10 superimposed voltage tracings of perturbed activity in response to a single 16 nA amplitude, 20 msec duration current pulse at each of two different coupling intervals (tc = 130 msec and tc = 140 msec). (Aggregate #1: diameter= 114 2-28 0 ~m.) 0 200 msec A B CONTROL 0 ]mv -50 COUPLING INTERVAL (m sec) 130 140 continue to become more pronounced, except that the size of the c maximum prolongation obtainable decreases and graded action potentials (Kao and Hoffman, 1958) can be elicited (Fig. 2-7). Prolongation in cycle length by an early stimulus (e.g. Fig. 2-7: \ = 80 msec} is now attained by an increase in action potential duration, which is partially nullified by a decrease in the diastolic time (time from MOP to next upstroke) of the perturbed cycle. This decrease in di as tolie tii1Je is associated with an ~~DP of the perturbed action potential that is more negative than control, and with an increase in the slope of diastolic depolarization. The diastolic time of the cycle following the perturbed cycle is increased for stimuli falling sufficiently late in the cycle. Figure 2-8 demonstrates that the current pulse waveforrfl does not change as the coupling interval is increased. Note that this current amplitude ( 24 nA) is small in comparison to the peak sodium current, which is about 3000 nA (see CHAPTER 3}. Effects qualitatively similar to those I have just described are seen if the pulse duration is increased instead of the current ampl i tude. (iii) The Response to Perturbation with Single Hyperpolarizing Pulses Figure 2-9 shows the response to hyperpolarizing input at two different pulse amplitudes. These responses were qualitatively 2-29 0 Figure 2-7. Phase resetting at a higher current amplitude (24 nA) than that shown in Figs. 2-5 and 2-6. Stimuli delivered earlier than about 100 msec now have an appreciable effect on the action potential duration. Note also the appearance of graded action potentials at, for example, tc = 120 or 130 msec. Thus, the coupling interval which marks the border between prolongation of. cycle length and abbreviation of cycle 1ength is now at about tc = 120 msec. {Aggregate# 1: diameter= 114 0 2-30 ~m.} 0 /~',, ' ' CO NT RO L --=-=-- 1-s0o'"\( I 20 0m ," " COUPLING IN TU V A l (m sec} 40 /i ;'"'-' 50 -- -' 60 70 J 80 )f\_ r11\ ' ... ... .-' N ______; 90 Q lOO \'- --- "- no 12 0 ~I\ I __ _. ) '~ J\_ 130 14 0 IS O loo 17 0 18 0 '~ J \ \ c Figure 2-8. Superimposed current pulse waveforms (unfiltered) from the first five trials of the experimental series shown in Fig. 2-7. A negative deflection corresponds to injection of a depolarizing current. current waveform{±~ The noise in the nA peak-to-peak) originates largely in the virtual ground current measurement circuitry and in the tape recorder. The current waveform is typical of that seen in most experiments; a time of about 5 msec is required for the current to 0 rise to or fall from its steady-state injected value. (Aggregate# 1: diameter= 114 JJm.) 2-31 0 0 20 0 m sec . similar to those seen in 8 other aggregates. 0 The response is the opposite to that produced by a depolarizing input, in that an early stimulus now abbreviates the cycle while a later one lengthens the cycle. For the smaller stimulus strength (Fig. 2-9, left panel), the abbreviation obtained is negligible(- 1%). With increasing stimulus intensity, the maximal prolongation and shortening observed increase, and the transition from prolongation to shortening of cycle length {which takes place in the region of the action potential upstroke) moves to a larger value of tc and becomes more abrupt {Fig. 2-9, right panel: tc tc = 560 msec and = 0 msec). At the smaller stimulus intensity (Fig. panel), the maximal prolongation occurs at about tc = 2-9, left 520 msec {T/To = 1.09); lesser prolongations are obtained for\ ) 520 msec. At the higher stimulus intensity, the degree of prolongation obtainable continues to increase for stimuli delivered right up to the end of the cycle. . A stimulus occurring before MOP is attained abbreviates the cycle length largely by producing a decrease in action potential duration (e.g. Fig. 2-9, right panel: tc = 80 msec), while one falling soon after MOP shortens the cycle by increasing the slope of diastolic depolarization beyond the control value (e.g. Fig. 29, right panel: tc = 200 msec). This abbreviation of cycle length occurs even though the potential at the end of the current pulse is more negative than it would have been otherwise (i.e. in the absence of the pulse). A pulse delivered later in the cycle (e.g. 2-32 0 Figure 2-9. The effect of delivering a 20 msec duration hyperpolarizing pulse on spontaneous rhythmic activity in an aggregate. Determinations with coupling intervals 20 msec apart were made at two current levels (left panel: 4.5 nA; right panel: 23 nA); every second trial is shown. The latest pulse in each trial (tc = 560 msec) is delivered very close to the end of the spontaneous cycle. For the higher current amplitude, there is an abrupt transition from prolongation of cycle length at tc = 560 msec to shortening of cycle length at tc 0 =0 msec. The sharp vertical deflection on the 11 of'P' of the stimulus artifact in some traces in the right panel is due to oscillation or "ringing" in an overcompensated amplifier. Since the chopped mode of current injection was used during this experiment, the effective values of the currents are quoted. (Aggregate# 6: diameter= 228 2-33 ~m.) 4. S oA 0 C O N TR l~mv ot C O U P liN (\ \ ) _ _ \ _ ; \._ r L ___; G t:Rv.} -. l INTtts ee H1 0 40 80 12 0 160 200 0 -fL_}\L ~fLJ\_ ~~l _j}-J\_ ~1\~ _j),_J\_ J~ i \- r'\L nl _;~L.........-' (\ ___.)''jI f L _1\]r--I'l L 240 280 J2 o 360 400 440 1\ f\ ----- ~ J !I 480 L -'~rA _}\~((\_ _j\~~l _ ! \ \. .. ._ -, r-J\-' L 1/ 0 I /\ ' ~ y \ {\ ' \ '-- I _!'~~'\_ Fig. 2-9, right panel: \ 0 = 520 msec) produces a prolongation even though the slope of diastolic depolarization is increased beyond control following the pulse and the take-off potential is more positive than at smaller coupling intervals, since the change in voltage produced by the pulse is greater (due to the i ne rea se in slope resistance). (iv) The Perturbed Cycle Length Function The phase resetting data of Figs. 2-5 and 2-7 can be plotted in a normalized fashion as shown in Fig. 2-lOA. perturbed interbeat interval T1/T normalized coupling interval t/T 0 is plotted as a function of the 0 0 The normalized • {The normalized coupling interval is also called the (old) phase and is denoted by cp.) I call a functional fit through this data the perturbed cycle length function and denote it by T1/T 0 respect to T 0 , = g(cp). Normalization is with since the intrinsic interbeat interval of the perturbed cycle would have been approximately equal to T in the 0 absence of stimulation (see Fig. 2-2C), and since R1 is the largest serial correlation coefficient Rj for j)l (Fig, 2-20}. In CHAPTER 5, I show that the response of the aggregate to periodic stimulation can be predicted given the function g{!J>). Figure 2-108 shows that increasing the stimulus duration has qualitatively the same effect as increasing the stimulus amplitude (different experiment from Fig. 2-lOA). 0 2-34 Since the stimulus Figure 2-10. Plots of the normalizea phase-resetting data: c the normalized perturbed interbeat interval T1 /T 0 plotted vs. the normalized coupling interval tc/T 0 • A. The effect of i ne re as i ng the amp1i tu de of a fixed duration {20 msec) depolarizing pulse from 5 nA (e) to 8 nA ( +) and then to 24 nA ( • ) • Pot nts obtained at 8 nA for tc plotted. ~ 130 msec are not For\ sufficiently large, the point (tc/T 0 , T1/T 0 ) lies somewhere in the stippled region between the two diagonal lines. This data is from the same experiment as that shown in Figs. 2-4 to 2-8. 0 B. The effect of increasing the duration of a fixed amplitude (10 nA) depolarizing pulse from 10 msec (e) to 40 msec (•). The points 1n this panel and in panels C and D were obtained by stimulating at random phases in the cycle. Not all data points obtained have been plotted. Same aggregate as in panels B and D. The intrinsic interbeat interval of this aggregate was about 470 msec. c. Symmetry in the response to a depolarizing hyperpolarizing (0) (e) and pulse of duration 10 msec and ampl i tu de 10 nA. 0 2-35 c D. Asymmetry in the response to a hyperpolarizing (D.) and depolarizing ( _.) pulse of duration 20 msec and amplitude 10 nA. Same aggregate as in panels Band C. (A: Aggregate # 1, diameter = 114 wm; Aggregate# 7, diameter= 228 wm.) 2-36 B-D: 1.5 1.5 A .. ..•..• c • 1.0 4-:::: 1.0 ·-· ...·. '· .. ...... I ") . . . • ••• • 0.5 0.5 • 20 M$eC 5 1.0 0.5 0.0 10nA 10 msec 8 nA • 24 + • D H 0.5 0.0 1.0 Jj_ To 1.5 1.5 B 1.0 ····~--= D • •• ..... '-:. . .. • 1.0 ... • 0.5 0.5 lOnA o.o 0 0.5 10 • 40 m sec 1.0 lOnA 0.0 0.5 20m$eC • D H 1.0 artifact obscures the ac.tion potential upstroke when the stimulus 0 is delivered late enough in the cycle to be a threshold stimulus (e.g. Fig. 2-68, lower panel), the interbeat interval of the perturbed cycle can only be estimated to within one half of the pulse duration for such cycles. Thus, for tc/T 0 sufficiently large, the point (t /T , T1/T 0 ) falls somewhere in the stippled c 0 regions of Fig. 2-10A,B. Figures 2-10C and 2-100 contrast the response to depolarizing and hyperpolarizing inputs. For stimuli of small enough strength (i.e. amplitude or duration), the response to a hyperpolarizing stimulus is close to being the mirror image of the response to a depolarizing stimulus (Fig. 2-lOC). However, for stimuli of higher amplitude or longer duration the mirror symmetry is lost (Fig. 2- 0 100). (v) Long Delays; Triggered Activity The phase-resetting data presented in Figs. 2-4 to 2-7 were all taken from one aggregate. These results were typical of the response to depolarizing current pulses seen in 14 other aggregates studied at two or more pulse amplitudes. However, much longer prolongations of interbeat interval could be obtained in 2 other aggregates, both of which had spontaneous eye 1 e 1eng ths of greater than 1 second. (Only 3 preparations studied had interbeat intervals of greater than 1 second.) 0 2-37 Figure 2-11 shows part of a phase-resetting experiment in one of these 2 slower-beating 0 aggregates. There is a range of coupling intervals for which prolongations of greater than 50% above the control cycle length {i.e. T1 > 1.5 T0 ) can be produced; in contrast, prolongations of greater than 41% were never seen in any of the other 15 aggregates which had intrinsic periods of less than 1 second. prolongation at tc The very long = 570 msec is accompanied by oscillatory behaviour of the transmembrane voltage in the subthreshold range of potentials. Long delays with subthreshold oscillatory activity can only be produced in a narrow range of coupling intervals and stimulus amplitudes. An increase or decrease in the coupling interval of as little as 10 msec can destroy the effect (Fig. 2-11). 0 The response is variable, in that the long prolongation is of a different length and might be seen at a slightly different coupling interval if the phase-resetting run is repeated keeping the stimulus amplitude and duration fixed. Indeed, there can be significant fluctuation in the response when repeated trials at a fixed coupling interval are carried out: threshold is attained at one or other of the maxima of the subthreshold oscillation (Figs. (tc (T 1 = 620 msec) = 5T 0 ), and 2-12,2-13). Figure 2-14 shows one of the longest prolongations observed shows that more than one cycle of subthreshold oscillation can occur before threshold is attained. Even though I have been able to obtain prolongations amounting to several cycle lengths (e.g. Fig. 2-14: tc 0 2-38 = 620 msec), 0 Figure 2-11. Phase resetting in a preparation with a long intrinsic interbeat interval (i.e. > 1 sec) produced by a 9 nA amplitude, 20 msec duration depolarizing pulse. Note the 1ong prolongation produced at tc = 570 msec that is accompanied by oscillatory activity in the pacemaker range of potentials. (Aggregate# 8: diameter= 170 2-39 0 ~m.) PULSE AMPLITUDE 0 9nA ,......_....., lsec COUPLI NG • ~;,:;t· /~UllliJ Ls}· 7 520 wu m 560 570 ill~ 580 J 590 600 610 0 620 Jll W lllu vL UJJuJJL JL} ~ '-' 0 Figure 2-12. Ten repeated trials at a fixed coupling interval of 570 msec. The pulse amplitude was 9 nA and the pulse duration 20 msec. Threshold is attained either at the first crest of the subthreshold oscillatory activity (the group of 7 action potentials labelled "a") or at the second crest (the group of 3 action potentials labelled 11 b11 ) . Same aggregate as in Figs. 2-11, 2-13, 2-14. (Aggregate# 8: diameter= 170 um.) 2-40 0 PULSE AMPLITUDE 9nA COUPLING INTERVAL E a ) ( b > (msec:) l 0 mV 570 -so lsec: 0 0 Figure 2-13. Thirty-one cons~cutive trials at a fixed coupling interval of 570 msec; consecutive trials are vertically displaced. The current pulse amplitude was 9 nA and the pulse duration 20 msec. indicate the stimulus artifact. The arrows Again, the action potentials following the stimulus cluster into two groups, labelled nau and 11 b11 • The tracings in Fig. 2-12 were taken from this group. (Aggregate# 8: 0 diameter= 170 2-41 ~m.) 0 PULSE AMPLITUDE. 9nA + COUPLING I N TERVAL (ms ec) 570 t 0 lsec < a ) ( b ) Figure 2-14. Three trials from a phase-resetting experiment in which the coupling interval was changed in 10 msec steps. The pulse amplitude was 6.5 nA, the pulse duration 20 msec. The action potential following delivery of the stimulus fires on the first (tc msec), second ( tc = 570 msec), or third ( tc = msec) crest of the subthreshold oscillation in c membrane voltage. (Aggregate# 8: diameter= 170 2-42 0 ~m.) = 600 620 0 PULSE AMPLITUDE 6.5nA COUPLING INTERVAL (ms ec) 600 570 620 ) / i..-..J lsec 0 I have not been able to completely abolish spontaneous activity in 0 the aggregate with a single 20 msec duration depolarizing current pulse at an external potassium concentration of 1.3 mM. This is in spite of systematic searches in many aggregates changing the coupling interval by as little as 1 msec and changing the stimulus amplitude by as little as a fraction of a nanoampere. However, there were a few atypical cultures in which most aggregates were found to be not beating or in which impalement of a beating aggregate caused it to become quiescent. Upon irnpal ement, most of these aggregates were found to be resting at a potential in the plateau range of potentials. Injection of a large enough pulse of current into such an aggregate could provoke sustained triggered ac t i vi ty ( Fi g • 2-15 ) • Triggered activity could also be obtained (but rarely) if spontaneous activity in a normal preparation was stopped by · injecting a constant depolarizing bias current. Triggered activity can however often be seen in aggregates which are quiescent during the transient seal i ng-i n process which immediately fall ows 11 impalement. 11 During this process in such aggregates, the membrane potential hyperpolarizes with time, until a point is reached where spontaneous cyclic action potential generation commences. The sealing-in process can be modeled as a gradual diminuition of a depolarizing bias current (Shrier and Clay, 1982). Theory predicts that a single well-timed current pulse should be capable of extinguishing triggered activity (Best, 1979; Jalife 0 2-43 Figure 2-15. Sustained activity triggered by a 13 nA amplitude, 20 msec duration depolarizing pulse in a spontaneously beating aggregate that bee ame quiescent upon impalement. activity. A 4 nA pulse was too small to trigger Triggered activity was also obtained in five other aggregates successfully ~npaled in this experiment (all aggregates from the same culture). (Aggregate# 9: diameter= 200 2-44 0 ~m.) 0 L-....1 lsec c and Antzelevitch, 1979, 1980; Winfree, 1980). This was not attempted in the experiments in which triggered activity was found in the aggregate. (vi} Apparent Discontinuities in the Phase-Resetting Response As stimulus intensity is increased, the transition between prolongation of cycle length and abbreviation of cycle length becomes more and more abrupt for either a depolarizing (Fig. 2-5) or a hyperpolarizing (Fig. 2-9) pulse. For example, Fig. 2-68 showed an abrupt transition from prolongation to shortening as tc was changed from 130 to 140 msec. To investigate this transition further, a stimulus of arnpl i tu de 27 nA and duration 20 msec was delivered 11 times at a fixed coupling interval of 142 msec into another aggregate. Figure 2-16. shows that one of two responses occurred ( all-or-none depolarization 11 11 ): one at a prolonged interbeat interval, and the other at a shortened interbeat interval. Responses with intermediate values of interbeat interval were not observed. In other experiments in which as many as 50 trials at a fixed coupling interval were carried out. a similar dichotomyin the response was observed. Since there is spontaneous fiuctuation in the unperturbed interbeat interval of several milliseconds from beat to beat (as well as in the shape of the action potential), the stimulus is actually being delivered in a narrow range of phases even though the coupling interval tc is 2-45 Figure 2-16. Superimposed tracings from 11 repeated trials with a 27 nA amplitude, 20 msec duration current pulse at a fixed coupling interval of 142 msec. Eight of these trials produced an advance in the time of occurrence of the next beat, while the other 3 produced a delay. 5/5 trials attempted at tc = 141 msec produced only delays, while 7/7 at tc = 143 msec produced only advances. The unperturbed interbeat interval of this aggregate was about 615 msec. Thus the discontinuity in the response T1 is about 2/3 of the spontaneous cycle length. (Aggregate# 10: diameter= 149 2-46 ~m.) 0 PULSE AMPLITUDE 27nA COUPLING INTERVAL (m sec) 0 142 ] mV -50 200msec c fixed from trial to trial. There is also a small degree of fluctuation in the trigger point on the action potential upstroke, which will serve to enhance this effect. For the experiment shown in Fig. 2-16, only prolongation of interbeat interval was seen at tc = 141 msec; at tc = 143 interval was observed. msec, only abbreviation of interbeat Figure 2-17 shows that in some circumstances a change in tc of as 1i ttl e as 1 msec can suffice to transform prolongation into abbreviation. The coupling interval where this rapid transition occurs is just a bit less than the action potential duration. I thus say that there is an experimentally observed discontinuity in the response T1 as a function of the coupling interval tc, and thus also in the perturbed cycle length function g(~). However, if the stimulus amplitude used in Fig. 2-16 or Fig. 2-17 is reduced, intermediate values of T1 can be obtained, with a large scatter from trial to trial at a fixed coupling interval. Figure 2-18 shows an example: over a range of coupling intervals of 20 msec, almost all intermediate values of T1 could be obtained. Thus, the perturbed interbeat interval T1 appears to be a continuous function of the coupling interval tc at this level of stimulation. Increase in stimulus amplitude leads to a response similar to the apparently discontinuous response shown in Fig. 2-6 or Fig. 2-17. The amplitude at which the transition from cycle delay to cycle advance (in response to a 20 msec duration depolarizing 0 2-47 c Figure 2-17. Superimposed tracings from 10 repeated trials at a coupling interval of 183 msec (upper trace) and 184 msec ( 1ower trace). The stimulus amplitude was 11 nA, and the stimulus duration 20 msec. Note the absence of intermediate values of T1 , even though T0 is fluctuating substantially from trial to trial at a fixed coupling interval. Here, the discontinuity in T1 is about 0.9 of the spontaneous cycle length. The upstrokes of the stimulus artifacts have been retouched in this figure, since misadjustment of the balance control on the amplifier produced a biphasic stimulus artifact. (Aggregate# 11: diameter= 132 urn.) 2-48 0 . PULSE AMPLITUDE 11 nA COUPLING INTERVAL (msec) 183 ] m~ -50 184 500 msec Figure 2-18. Ten superimposed voltage tracings at each of three coupling intervals for a current pulse of amplitude 5 nA and duration 20 msec. Notice the large scatter in the response at a given coupling interval. The current waveform was constant from trial to trial at a fixed coupling interval. A burst occurred during collection of the data at tc = 230 msec, as evidenced 0 by the decrease in T0 for three trials. (Aggregate# 2: diameter= 105 2-49 0 ~m.) PULSE AMP LITU DE 5 nA COU PLIN G INTE RVA L (m sec) 220 0 230 240 .5sec c pulse) no longer appeared to be continuous was generally found to lie between about 5 nA and about 15 nA. This variability is not unexpected, due to the different sizes and presumably differing electrophysiological properties of different aggregates. For faster-beating aggregates, the coupling interval at which the transition occurs is smaller than for the 2 more slowly beating aggregates in which long delays could be obtained. For the faster beating aggregates, the transition occurs at about the maximum diastolic potential (e.g. Figs. 2-6, 2-16, and 2-17); for the slower beating aggregates, it can occur much further into phase 4 {Fig. 2-11). Figure 2-19 shows that with high-amplitude hyperpolarizing input, the· apparent discontinuity appears during the plateau phase, just after the action potential upstroke - all11 or-none repolarizatiorl' (Weidmann, 1951, 1956; Cranefield and Hoffmann, 1958; Noble and Hall, 1963; Vassale, 1966; McAllister, Noble,and Tsien, 1975; Beeler and Reuter, 1977). The slower- beating aggregates that display a subthreshold oscillation can have more than one apparently discontinuous jump in the plot of the perturbed cycle length function, since they only fire on one or another of the maxima of the subthreshold oscillatory activity (Fig. 2-14); action potential s are not seen between groups 11 a" and 11 0 b in Figs. 2-12 and 2.13. 11 2-50 ·- Figure 2-19. Phase-resetting in response to a high-amplitude (64 nA), 20 msec duration hyperpolarizing current pulse. There can thus be an abrupt abolition of the action potential by a strong hyperpol ari zi ng stimulus delivered during the plateau of the action potential. A slight change in coupling interval is sufficient to induce the effect ("all-or-none repolarization" }. 0 (Aggregate# 12: diameter= 132 2-51 ~m.) - PULSE AMPLITUDE 64nA ~· COUPLING I N"rERVAL (msec) 6 8 38 46 500 m sec (vii) Effects on the Post-Stimulus Cycles A cycle that is shortened by a depolarizing stimulus falling 1ate enough in its cycle tends to be followed by a postextrasystolic cycle that is longer than control, while a cycle that is prolonged by an early stimulus tends to be followed by a poststimulus cycle that is briefer than control. These effects can last for several cycles following stimulation, but with a gradually decreasing effect on cycle length. At lower stimulus strengths (producing responses such as those shown in Fig. 2-5), for a maximal prolongation of approximately 20% (i.e. T1/T 0 = 1.2), the post-stimulus cycle is typically shortened by about 2-5%, while for a shortening of about 50% (i.e. T1/T 0 = 0.5), the post-stimulus cycle is typically lengthened by about the same amount. In the former (latter) case, the decrease (increase) in the duration of the post-stimulus cycle is associated with a decrease (increase) in its diastolic period. At a stimulus amplitude high enough to produce a graded action potential (e.g. Fig. 2-7: tc = 120 msec), a shortened cycle also has a post-stimulus cycle with an increased diastol ic time; however, the decrease in the duration of the graded action potential can cause the actual post-stimulus cycle length APO + diastolic time) to decrease. (= The cycle following the post- stimulus cycle has a normal APD, but an increased diastolic time, and thus is lengthened beyond control. Appreciable decreases in post-stimulus cycle length are also apparent when a very long delay is produced (e.g. Fig. 2-14). 2-52 (vfi} Pulse-induced Rapid Repetitive Activity Rapid repetitive activity in response to a depolarizing current pulse was a rare phenomenon (incidentally seen in only 2 aggregates from two different cultures subjected to high-amplitude current pulses}. I present results from the aggregate in which the effect was more dramatic. Figure 2-20A shows the induction of repetitive activity at a rate faster than the spontaneous beat rate by a single 20 msec duration pulse. The increase in rate is mediated by an increase in the slope of diastolic depolarization. The beat rate gradually declines back to normal. The exact form of the response varied from trial to -trial carried out at fixed tc. The effect was seemingly not dependent on the coupling interval of the stimulus, in that it could be elicited at phases throughout the spontaneous cycle. It is unlikely that this effect is due to dislodging of the electrode and a subsequent speeding up of the . rate due to the resultant "1 eakage", si nee the maximum di astolic potential remains relatively unaltered. In some trials, there was a s11ght decrease in the absolute value of the MOP. The two aggregates in which this phenomenon was seen came from cultures in which aggregates appeared to be contracting normally under the microscope, and which had the usual high percentage of spontaneously beating aggregates. Furthermore, the action potential parameters measured from the aggregates in which the rapid activity was seen were all within the normal range. 2-53 Figure 2-20. Rapid repetitive activity induced by a single 20 msec duration current pulse. A. Depolarizing pulse of amplitude greater than 100 nA, with tc = 130 msec. The exact current is not known due to saturation of the current measurement circuitry. The rapid activity was not seen at a current amplitude of two thirds of this amplitude; only a single extrasystole occurred. (Aggregate# 13: 0 B. diameter= 190 urn.) Hyperpolarizing pulse of amplitude 37 nA, with tc = 8 msec. Upstroke of stimulus artifact retouched. (Aggregate# 11: diameter= 132 urn.) 2-54 A B 1 sec Rapid repetitive activity was also .seen in two aggregates in response to a high-amplitude hyperpolarizing stimulus (Fig. 2-208). However, in these cases, the phenomenon could only be provoked by stimuli ,falling during the action potential. The phenomenon was accompanied by a decrease in the absolute values of the overshoot potential and the maximum diastolic potential, both of which gradually recovered back to normal {Fig. 2-208}. 4. DISCUSSION (i) B1phasic Nature of the Response to Premature Stimulation It is not surprising that a brief depolarizing input can temporarily accelerate a cardiac oscillator. However, it is less commonly appreciated that a subthreshold depolarizing pulse, if . delivered early in the spontaneous cycle, can transiently prolong the interbeat interval or that an early hyperpolarizing pulse can shorten the interbeat interval. This is so in spite of the fact that the biphasic nature of the response to a current pulse has been demonstrated in the sinoatrial node (Ushiyama and Brooks, 1974; Sano, Sawanobori, and Adaniya, 1978; Jalife et al., 1980), in Purkinje fibre (Weidmann, 1951; Klein, Cranefield, and Hoffman, 1972; Jalife and Moe, 1976), in diseased human ventricle {Gilmour et al., 1983}, and in heart cell aggregates (DeHaan and Fozzard, 0 2-55 1975; Scott, 1979; Guevara, Shrier, and Glass, 1980; Glass, and Shri er, 1981; Ypey, van Meerwij k et al., 1984). r~eerwij Guevara, k, and DeHaan, 1982; van Electrotonic depol ari zati on of a pacemaker produced by a blocked premature beat can produce a prolongation in the cycle length. This has been found in the sinoatrial node {Kerr et al., 1980) and in a focus produced by depolarization of the right bun~le branch {Ferrier and Rosenthal, 1980). A biphasic response of the sinoatrial node to ventricular activity during atrioventricular dissociation has also been documented {Roth and Kisch, 1948; Rosenbaum and Lepeschkin, 1955). In addition, the sinoatrial node has been shown to have a biphasic response to vagal stimulation {Jalife and Moe, 1979b), and Moe et al. {1977) have· inferred a biphasic response of a junctional parasystol ic focus in a human being from analysis of the electrocardiogram. Several invertebrate neural pacemakers also have a biphasic response to premature stimulation {Winfree, 1977; Ayers and Selverston, 1977,1979; Guttmann, Lewis, and Rinzel, 1980). Winfree {1980) has pointed out that the response to premature stimulation of nearly all biological oscillators that have been studied is biphasic. Ionic models for cardiac oscillators (McAllister, Noble, and Tsien, 1975; Bristow and Clark, 1982; Difrancesco and Noble, 1982a; Guevara et al., 1982; Clay, Guevara, and Shrier, 1984; see also CHAPTERS 3 and 6 of this thesis) and for neural oscillators {Best, 1979; 1983: Guttmann, Lewis, and Rinzel, 1980; Guevara et al., see CHAPTER 6 of this thesis) display a biphasic response to 2-56 premature stimulation, as do simpler 1imi t cycle models ( Greco and c Clark, 1976; Scott, 1979; Winfree, 1980; Guevara and Glass, 1982; Hoppensteadt and Keener, 1982; Barbi and Holden, 1983; Guevara et al., 1983: see also CHAPTER 6 of this thesis)and electronic models of cardiac pacemakers (Roberge and Nadeau, 1965; Roberge, 1968; Roberge, Bhereur, and Nadeau, 1971). Furthermore, the integrate- and-fire (slow pacemaker depolarization to a fixed threshold voltage) model that is often used to explain spontaneous phase 4 depolarization of cardiac pacemakers does not yield a biphasic response to either a depolarizing or a hyperpolarizing stimulus ( Wi nfree, 1980) • Many of the electrophysiological studies listed above show that the transition point from prolongation to delay occurs at earlier coupling intervals as the intensity of a depolarizing stimulus is increased. Scott {1979) offers a model for this using the Bonhoeffer-van der Pol {BVP) equations (FitzHugh, 1961), while I describe the ionic basis below in CHAPTER 3. I also indicate in the following section how the biphasic nature of the response to premature stimulation may be responsible for a number of interesting synchronization phenomena. (ii) The Synchronism of Spontaneous Activity in the Aggregate: Mutual Synchronization Electrical activity occurring in widely separated cells within an aggregate is virtually synchronous. 2-57 During spontaneous .·-· activity, the action potential upstrokes in two cells separated by 165 ~m occur within 45 ~sec of each other (Clapham, 1979). Based upon the response of the aggregate to injection of sma11 amplitude current pulses, DeHaan and Fozzard (1975) could find no evidence to contradict the assumption that the aggregate could be considered to be a spherical synctium. Mathi as et al. ( 1981) came to a similar conclusion using impedance measurements. For events with frequency content below about 10Hz, Clapham (1979) and Clay, OeFelice, and DeHaan ·(1979) could find no experimental evidence of deviation from isopotentiality. Furthermore, membrane voltage noise or subthreshold oscillations recorded from two widely separated cells in aggregates made quiescent by the addition of TTX is virtually identical (DeFelice and DeHaan, 1977; DeHaan and DeFelice, 1978a, 0 1978b). Visual observation of the aggregate carried out on numerous occasions shows no evidence of gross disorganization in the spread of contractile activity following perturbation with a single current pulse. In light of the above facts, I feel justified in treating the aggregate, for ~ purposes, as a single card i ac osc i1 1ator. Isolated ~ocytes are found to have a large range of mean interbeat intervals, with many cells beating irregularly (DeHaan and Hirakow, 1972; Jongsma et al., 1975; Jongsma and Tsjernina, 1982). The variability in the interbeat interval of an isolated myocyte is much greater than that of a single cell in a monolayer or of a single cell in the isolated sinoatrial node (Jongsma and 0 2-58 Tsjernina, 1982). When assembled together to form an aggregate, the single cells synchronize their individual rates to achieve a single common rate. As the number of cells in the aggregate increases from 1 to 125, the beat-to-beat variability in interbeat interval decreases, with the coefficient of variation C dropping from about 20% to about 3% {Clay and DeHaan, 1979). connection, it is noteworthy that the dom~nant In this centre of the sinoatrial node is made up of a population of approximately 5000 cells {Sleeker et al., 1980). In the·aggregate, as in the dominant centre of the sinoatrial node, each cell is controlled by the overall activity of the other cells, but each cell can be presumed to play as important a role in determining the overall activity as any other cell • 0 The flo·w of current from cell to cell is presumably the mechanism by which this mutual synchronization occurs within the aggregate. This has been shown to be the case in the sinoatrial node {Sano, Sawanobo~i, and Adaniya, 1978; Tuganowski, Kopee, and Tarnowski, 1981). Two isolated myocytes (DeHaan and Hirakow, 1972) or aggregates {Scott, 1979; Ypey, Clapham, and DeHaan, 1979; Clapham, Shrier, and DeHaan, 1980; Clapham and DeHaan, 1982; Ypey, van Meerwijk, and DeHaan, 1982) will synchronize their rates if coupled together. In the 1atter case, there is a rapid fall in the coupling resistance between the two aggregates within a few minutes of coupling, presumably as a result of insertion of gap junctions into the membrane area common to the two aggregates {Clapham, 2-59 . Shrier, and DeHaan, 1980). .,.-..., '-"' Indeed, calculation suggests that a single gap junction should be more than sufficient to allow mutual synchronization of two single cells (Clapham, Shrier, and DeHaan, 1980; Noble, in discussion appended to DeHaan, 1982). Normally, each cell within an aggregate has of the order of 10 6 nonj uncti anal membrane particles and approximately 10 4 particles clustered into identifiable nexuses (Williams and DeHaan, 1981). Upon treatment with cycloheximide, an inhibitor of protein synthesis, organized gap junctions cannot be demonstrated ultrastructurally, but yet mutual synchronization is maintained (Will iams and DeHaan, 1981). These authors conclude that the enduring synchronization is due to the existence of 11 isolated channels scattered throughout the area of closely apposed plasma membranes r~echanical, 11 • as well as electrical , effects may contribute to the process of mutual synchronization. It has been known for over two hundred years that mechanical stimulation of the heart can provoke a premature contraction (Hoff, 1941-42). For example, gentle prodding of an aggregate with a blunt glass probe can induce a contraction. Stretch of the sinoatrial node in mammals (James and Nadeau, 1963; James, 1967; James et al., 1970; Brooks and Lu, 1972; Ushiyama and Brooks, 1977) and of other cardiac pacemakers in a wide range of phyla (Jensen, 1971) influences the rate of spontaneous activity. Diederich (1969) has concluded that acceleration of the sinoatrial node can be produced by stretch of the node as a result of contraction of the right atrium against a 2-60 closed tricuspid valve. c Ushiyama and Brooks (1977) have demonstrated that sinusoidal stretching of strips of right atrium containing the sinoatrial node can lead to entrainment of the node. Hashimoto et al. (1967) and James (1967, 1973) have proposed that the arterial pulse in the sinoatrial nodal artery may act as the mechanical feedback arm of a servomechanism in which sinoatrial activity is modulated by the ventricular output. The extent to which the mechanical stimulation of any one cell in the aggregate by the contraction of the rest of the aggregate is responsible for synchronization has not been investigated. Segers (1946b) showed that mechanical coupling with a non-conducting paraffin bridge of two separate pieces of contracting ventricle could result in synchronization. 0 Pollack (1977) has suggested than stretch of sinoatrial cells contributes to their mutual synchronization. However, in one study in which aggregates were treated with cytochalasin B (which blocks mechanical but not electrical activity} no effect on the regularity of beating was reported (Sachs, McDonald, and Springer, 1974}. Finally, the onset of mechanical contraction follows the action potential upstroke with a delay that is not inconsiderable. The spontaneous rate of the upper half of the sinoatrial node is generally faster than that of the lower half, with the rate of the intact node being intermediate to the rates of the two halves (Lu, Lange, and Brooks, 1965; Sano, Sawanobori, and Adaniya, 1978; 0 2-61 Bouman et al., 1982). Similar results are obtained if a functional b1ock between the two ha1 ves is created by p1 ac i ng the node in a sucrose-gap apparatus (Tuganowski, Kopec, and Tarnowski, 1981). However, coupling together of two isolated myocytes can produce a synchronized state with a final common rate that is higher than, intermediate to, or lower than the two initial rates (OeHaan and Hi rakow, 1972). The experiments described in this paper suggest one mechanism by which mutual entrainment might occur. simpler case of two coupled cells. mutually synchronized. Consider first the Assume that they have become That is, not only do they share the same common rate, but there is also a fixed latency between the upstroke of one cell and that of the other, and the action potential durations are approximately equal. 0 Assume also that cell A is intrinsically faster than cell B (Fig. 2-21A,8) and that, upon synchronization, the intrinsically faster cell (cell A) 1eads the intrinsically slower cell (cell 8} and that the final common rate is intermediate to the two uncoupled rates. Si nee the up stroke of the action potential of A occurs late in the-cycle of B, cell A sources a depolarizing current to cell B that advances the time of occurrence of the next action potential of B. Conversely, the occurrence of the upstroke of the action potential of B early in the cycle of A produces an early depolarizing stirnul us that results in a slowing down of A. Cell A also sources a hyperpolarizing current to cell 8 during the repol ari zati on phase of its action potential, which will tend to speed up the activity in B. 2-62 Thus, A 0 Figure 2-21. Schematic diagram illustrating synchronization of two cardiac oscillators. One oscillator (panel A} has an intrinsic frequency higher than that of the other {panel B). The action potential duration of the slower cell is longer. Upon coupling (panel C), the two oscillators synchronize to an intermediate rate, .with the intrinsically faster oscillator (labelled A) leading the intrinsically slower oscillator (labell~d B). The time between upstrokes of A and B is exaggerated for clarity of presentation. 2-63 0 A 0 c may be said to be overdriving B, while at the same time, B is underdriving A. The reader is cautioned that the experimental coupling of two cells does not always result in a final common rate that is intermediate to the two initial rates (OeHaan and Hiral<ow, 1972). Ionic modelling studies also suggest that an intermediate rate does not necessarily have to occur upon synchronization (Berkinblit et al., 1975). However, it is possible to use the phase-resetting behaviours in response to single pulses to spell out how these more complicated cases may occur (Ypey and van Meerwijk, 1980; Ikeda, 1982). During synchronized activity, some cells in an aggregate presumably have intrinsic rates slower than the aggregate rate, others faster. However, there is an essential difference between overdrive or underdrive spontaneously occurring in an aggregate, and that which results in response to a periodic train of current pulses (see CHAPTER 4). In the former case, the perturbation occurs throughout the entire cycle, and not just for a brief period of time as· in the latter case. Due to the great degree of synchronism present in an aggregate (Clapham, 1979), large potential differences between neighbouring cells can only occur during the upstroke and rapid repolarization phases of the action potential. During the slower plateau and diastolic depolarization phases, only small potential differences can exist. Thus, large currents will fl a.-~ from cell to cell for brief periods of time 2-64 during phase 0 and phase 3, while smaller currents will flow for 0 longer periods of time during phase 2 and phase 4. To further complicate matters, gap junctions formed between coupled amphibian embryonic blastomers show increased resistance with increased transjunctional potential (Spray, Harris, and Bennett, 1979). Thus, several different effects may all act to improve the beat-tobeat regularity of an aggregate; at the present time, it remains unclear to me whether or not current flow during one particular phase of the cycle may be said to be mainly responsible for mutual synchronization. Figure 2-20 shows that there is a roughly linear dependence of any i nterbeat interval on the immediately preceding one. Random shuffling of the intervals destroys this dependence (Fig. 2-2E). Furthermore, si nee the Rj fall monotonically, this suggests that the effect is brought about by a action within a few beats. mec~anism that can alter its An interbeat interval that is longer (shorter) than average is followed by one that is also longer (shorter) than average, but less so. It is interesting to note that an interbeat interval that is artificially prolonged (shortened} by application of a current stimulus is followed by a post-stimulus cycle that is shorter (longer) than control. Perhaps the same mechanisms are operating in the two cases. The bursting phenomenon shown in Fig. 2-3 occurred even in aggregates which had been successfully impaled for several hours. The most likely explanation is that there is a temporary partial 2-65 loss of impalement, ~ith subsequent resealing of the membrane around the microelectrode tip. Partial loss of impalement leads to a small depolarizing leakage current which can presumably be sufficient to significantly increase the beat rate {Shrier and Clay, 1982). Athias et al. (1979) found 2 types of periodically occurring irregularities in the beat rate of neonatal rat heart cells. In their type II arrhythmian, the beat-to-beat variation 11 was due to alteration in the level of the threshold voltage, and not to alteration in the rate of diastolic depolarization, as found in the aggregate. Phenomena similar to that shown in Fig. 2-3 also occur in the sinoatrial node, and have been attributed to shifts in pacemaker site within the node or to the effect of a premature stimulus of unknown origin (Sano, 1966). One alternative hypothesis that might explain the bursting seen in these experiments is that it is a normal phenomenon that one might expect in any population of many coupled oscillators. Indeed, in his firefly machine.. that consisted of 71 coupled neon 11 bulb oscillators, Wi nfree ( 1980) found that .. it is not unusual to see small groups of neon oscillators temporarily escape entrainment ...... Thus, occasional bursting in the cells that form the aggregate may be a periodic revolution against conformism that is somehow necessary to allow that conformism to become reestablished. 0 2-66 (iii) Apparent Discontinuity of the Phase-Resetting Response For a sufficiently strong stiQulus, be it depolarizing or hyperpolarizing, there is an abrupt transition between prolongation of the perturbed cycle length and shortening of the cycle length (Figs. 2-16, 2-17, 2-19). This effect has been previously seen in the sinoatrial node in response to vagal (Oong and Reitz, 1970; Iano, Levy, and Zieske, 1972; Greco and Clark, 1976; Jalife and Moe, 1979b) and electrical (Sano, Sawanobori, and Adaniya, 1978; Jalife and Moe., 1979b; Jalife et al., 1980) stimulation, as well as in Purkinje fibres (Jalife and i•1oe, 1976; Jalife and Moe, 1979a}, a focus in the right bundle branch (Ferrier and Rosenthal, 1980), and heart cell aggregates (Scott, 1979; Guevara, Glass, and Shrier, 1981) in response to electrical stimulation. I have found, in agreement with the studies of these other investigators, that the transition from prolongation to abbreviation of cycle length is continuous at a low stimulus level and occurs at increasingly smaller coupling intervals as the depolarizing pulse intensity is increased. At higher stimulus levels, the response becomes apparently discontinuous, in that intermediate values of T1 are not seen. There are fundamental problems in ascertaining whether or not the perturbed cycle length T1 is truly a discontinuous function of the coupling interval tc, as is suggested by Figs. 2-16 and 2-17. If the response is continuous but very steeply changing in the range of coupling intervals lying between those at which maximal 2-67 prolongation and maximal shortening of the interbeat interval are produced~ trials. intermediate values may not be seen, even after very many If they were seen only very rarely, the possibility of artifactual causes of the response such as a spontaneous fluctuation in the properties of the aggregate or a partial loss of impalement could not be ruled out. This difficulty is compounded by the fact that the i nterbeat intervals would be expected to fluctuate over a wide range of values from trial to trial in the critical range of coupling intervals if the response were indeed continuous (as in Fig. 2-18, but over a range in tc of 1 msec and not 20 msec). However~ the fact that intermediate responses were not seen in Fig. 2-16, even in the presence of considerable fluctuation in T0 , can be taken as circumstancial evidence for the 0 discontinuity of the response. The waveforms shown in Figs. 2-16 and 2-17 are very similar_ to those seen in quiescent cardiac tissue {e.g. Myerburg, Gelband, and Hoffman, 1971; Sasyniuk and Mendez, 1971), when a change in the coupling interval of 1 msec will produce either a propagated response or a subthreshold event in response to premature stimulation. This similarity is not surprising, since the ionic currents responsible for the phenomenon in the aggregate are also present in quiescent cells; the currents involved in generating spontaneous activity in the aggregate are not involved in this response (see CHAPTER 3). Investigation of the response of an ionic model of the 2-68 - aggregate to perturbation with current pulses suggests that the response is fundamentally continuous, but steeply changing, at lower stimulus amplitudes (Clay, Guevara, and Shrier, 1984: CHAPTER 3 of this thesis for further details). see Although the response is also continuous in the (noise-free) model at intermediate amplitudes, the transition from prolongation to abbreviation can be very steep. In fact, at a sufficiently high amplitude, increments in the coupling interval of as little as 1 ~sec are not sufficient to demonstrate continuity. This is largely due to the threshold-like behaviour of the fast sodium current INa' which produces a quasi-threshold phenomenon (FitzHugh, 1955; FitzHugh, 1960). Investigation of the quiescent giant axon of the squid at 35°C (Cole, Guttman, and Bezanilla, 1970) and of the Hodgkin-Huxley equations appropriate for squid axon at 35°C {Cole, 1958; Cole, Guttman, and Bezanilla, 1970; Clay, 1977) show similar continuous responses, demonstrating that depolarization is not an all-or-nothing affair for that particular non-propagated action potential. The addition of realistic membrane noise to ionic models of either the squid axon (at 6°C) or the aggregate (at 37°C) would presumably reproduce the apparent all-or-none threshold phenomenon that occurs experimentally in both these preparations. (iv) Depression of Cardiac Pacemakers The disturbance in the arterial pulse that we now attribute 2-69 to the presence of an extrasystole has been known to medical 0 practitioners for a long time. The compensatory pause following a premature ventricular contraction had been described and explained by the late nineteenth century. It was also known by then that the returning cycle following a premature atrial contraction could be equal in duration to the spontaneous sinus cycle length, fully compensatory~ or of i ntennedi ate duration ( Lewi s, 1920). The three zones of reset, incomplete interpolation, and complete interpolation are routinely encountered in the course of clinically determining the sinoatrial conduction time (see for example Dhingra et al • , 197 5). The reset zone results from invasion and capture of the sinoatrial node by a late premature contraction. The zone of incomplete interpolation results from an electrotonic subthreshold 0 depolarization which delays the time of the next sinoatrial activation (Kerr et al., 1980). The perturbation is subthreshold, since it results from an early premature beat that encounters entrance block. This del ay or depressi on 11 11 of spontaneous impulse formation in the sinoatrial node had been earlier clinically inferred (Pick et al., 1951; Dressler, 1966). Similar depression of subsidiary pacemakers can also be seen, both clinically and experimentally (Pick et al., 1951; 1972; Klein, Cranefield, and Hoffman, Klein et al., 1973; Goel, Han, and Rogers, 1974; and Lane, 1978; Loeb et al., 1979). Kennelly Rarely, oscillations in membrane potential in the pacemaker range of potentials similar to c 2-70 those of Figs. 2-11 to 2-14 have been observed experimentally (Klein, Cranefield, and Hoffman, 1972; Scott, 1979). I ascribe these oscillations to the presence of a spiral point in the pacemaker range of potentials {see section (viii) below). al. Loeb et {1979) have recorded dramatic depression of spontaneous activity in ectopic atrial foci in response to a single premature atrial contraction. However, it is not clear from the above reports whether or not the effect is phase-dependent, as it is in the case of the aggregate. My studies extend these other observations that spontaneous activity in tissue of ventricular origin is subject to temporary depression or suppression by premature subthreshold stimuli. Prolongations in cycle length of as much as approximately thirty percent can be seen in all aggregates for an appropriately chosen pulse strength, while much longer-lasting suppression was seen in 2 out of 3 more slowly-beating aggregates studied (Figs. 2-11 to 214). While alterations in conduction can sometimes be invoked to explain the cases of depression seen in the intact heart (Parkinson, Papp, and Evans, 1941) , this cannot be the explanation in our case given the experimental circumstances. It is not generally appreciated, but has been explained by Winfree (1980).that the theory of nonlinear oscillators guarantees that temporary or permanent suppression can be seen for any limit cycle oscillator given a stimulus of the appropriate size and timing. This theorem of nonlinear dynamics hinges upon the fact that every limit-cycle oscillation has associated with it one or 2-71 more equilibrium points (also called singular points, fixed points, or steady states). Perturbation of the state-point of the system into a neighbourhood of the singular point by a subthreshold stimulus can lead to permanent cessation of firing (if the equilibrium point is stable) or to a temporary suppression of generation of action potentials (if it is unstable). However, there is no guarantee that the appropriate combination of stimulation strength and timing will necessarily occur if the stimulus is a premature contraction of endogenous origin. Suppression of automatic impulse generation is usually seen in patients with evidence of cardiac pathology or in the hypodynamic state in the experimental setting. This is in agreement with what theory would predict, since a subthreshold 0 input to the pacemaker (as would occur if there is generalized depression of conduction) is theoretically needed to elicit the phenomenon. Furthermore, the sinoatrial node and parasystolic foci are known to display entrance block for sufficiently premature stimulation, leading to a subthreshold electrotonic depolarization of the central region of the pacemaker. However, my experiments show that the phenomenon of suppression is only seen in a limited range of combinations of coupling interval and stimulus strength, perhaps accounting for the clinical rarity of the phenomenon. In the sick sinus syndrome, abnormally long values of sinoatrial conduction time {SACT) are often reported. The point has been previously made that abnormally large values of SACT may 2-72 be not in fact a reflection of prolonged conduction times into and out of the node, but may rather be evidence of a depression of nodal automaticity directly attributable to the premature stimulus (Breithardt and Seipel, 1976). reinforce this point. My observations Since sinus arrest is a hallmark of the sick sinus syndrome, one could speculate that some instances of temporary sinus arrest could be due to suppression caused by subthreshold depolarization of the node following a premature atrial contraction blocked in the perinodal fibres. The activity shown in the two lower panels of Fig. 2-14, if it occurred in nodal cells, would be clinically interpreted from the electrocardiogram as evidence of sinoatrial arrest or sinoatrial exit block. Furthermore, oscillations in the subthreshold range of potential 0 . seem to occur quite readily in tissue of sinus origin (Bozler, 1943; West, 1961; Lu, Lange, and Brooks, 1965). However, it must be noted that one attempt to produce . suppression of activity in the in vivo sinoatrial node of healthy dogs using a single stimulus was unsuccessful (Loeb et al., 1979), although another study (Jalife and Antzelevitch, 1979) demonstrated that subthreshold stimuli could temporarily suppress {and even permanently stop) activity in an in vitro sinoatrial preparation. 2-73 ( v) Anni·hn ation of Spontaneous Activity; Triggered Automaticity It has been reported that a single subthreshold stimulus of the right size delivered within a narrow range of coupling intervals can abolish spontaneous activity in the kitten sinoatrial node (Jalife and Antzelevitch, 1979). The same effect could be produced in canine and bovine Purkinje fibres made to fire spontaneously by placing in a low-potassium solution containing epinephrine (Jalife and Antzelevitch, 1979; Jalife and Antzelevitch, 1980). However, it appears that a constant bias depolarizing bias current had to be passed in at least one case to permit observation of the effect (Jalife and Antzelevitch, 1980). A more recent report shows that spontaneous activity in diseased c human ventricular myocardium can be annihilated by a critically timed stimulus (Gilmour et al., 1983}. Earlier reports indicated that triggered activity in fa1 se tendons exposed to acetylstrophanthidin (Ferrier, Saunders, and Mendez, 1973), in Purkinje fibres exposed to a low-sodium, high-calcium, TEAcontaining solution (Cranefield and Aronson, 1974), and in fibres of the simian mitral value exposed to solutions containing catecholamines and EDTA (Wit and Cranefield, 1976) could be stopped by appropriately timed premature stimuli. Theory predicts that in the case of other preparations displaying triggered activity (e.g. Segers, 1939; Reid and Hecht, 1967; Mary-Rabine et al., 1980; Gilmour et al., 1983), a single critically-timed stimulus should c also be capable of annihilating the triggered activity. Guttman, Lewis, and Rinzel (1980) showed that repetitive 2-74 c activity in a non-cardiac membrane, the giant axon of the squid (which can be made to oscillate by placing in a calcium-poor medium and passing a constant depolarizing current), could be abolished by a single critically timed hyperpolarizing or depolarizing pulse of current. This phenomenon had been earlier predicted by a numerical analysis of the Hodgkin-Huxley equations (Best, 1979). Another numerical study, this time of Noble's earlier (1962) equations describing Purkinje fibre, demonstrated triggered activity (Krinskii and Kokoz, 1973). An examination of an ionic model of two coupled cells showed that triggered activity occurring at a high frequency and at depolarized voltage levels could be induced (Kokoz, Krinskii, and Mornev, 1974). A later modelling study coupled together two simplified cells, one of which showed c spontaneous activity, the other not. Upon coupling, the pair was quiescent, but triggerable (van Capelle and Durrer, 1980}. The extent to which the sinoatrial node or any other intrinsic focus responsible for pacing the heart can be actually stopped in vivo by premature stimuli arising elsewhere in the heart is unknown. However, the question may be somewhat academic, since if the sinoatrial node were stopped due to the arrival of such an impulse, it would presumably resume spontaneous activity upon receipt of a retrograde threshold wave of depolarization arising in a subsidiary pacemaker. The same would also hold true for a primary subsidiary pacemaker, assuming that a secondary subsidiary pacemaker would take over and (again) assuming no retrograde block 2-75 of conduction. Adams-Stokes' attacks in cases of complete heart block can be sometimes attributed to the temporary cessation of ventricular activity due to the extinction of spontaneous impulse formation in the subsidiar-Y junctional or idioventricular focus responsible for driving the ventricle. There are instances on record of this extinction being immediately preceded by a single premature contraction (e.g. Parkinson~ Papp, and Evans, 1941: Fig. 7). Despite strenuous attempts to do so, I have not been able to abolish spontaneous activity in the aggregate with a single 20 msec duration pulse. A similar result has recently been found by van Meerwijk et al. {1984) using hea~t cell aggregates. This probably indicates that the steady state{s) of the system are unstable. 0 This conclusion is backed up by the ionic modelling studies of the aggregate described in the next chapter. Investigation of ionic models of sinoatrial node and Purkinje fibre demonstrates that spontaneous activity in these systems cannot be annihilated, unless modifications are made to the equations (Guevara, unpublished). It remains to be seen, however,·whether measures such as injection of a constant current, alteration of external ionic concentrations, or addition of pharmacological agents may reproducibly convert an unstable steady state into a stable steady state, thus allowing annihilation of the spontaneous activity of the aggregate with a single pulse. c 2-76 c {vi) Ionic Mechanisms The response of an ionic model of the aggregate to premature stirnul ati on has recently undergone some i nvesti gati on ( Guevara et al., 1982; Shrier et al., 1983; Clay, Guevara, and Shrier, 1984}. The model is a partial one, since it simulates only phases 3, 4, and 0 of the spontaneous activity. The parameters of the model are derived from voltage- clamp measurements of the ionic currents underlying spontaneous activity in the aggregate preparation. The currents included are the fast sodium current INa(Ebihara and Johnson, 1980), the time-dependent pacemaker current IK (Clay and 2 Shrier, 198la,198lb), the background sodium and potassium currents INa,b and IK (Clay and Shrier, ·1981a, 198lb), and the plateau 1 potassium current Ix (Shrier and Clay, 1982}. I have also investigated the phase resetting behaviour of the MNT equations (McAllister, Noble, and Tsien, 1975), which are a more complete model for spontaneously active cardiac Purkinje fibre. The results of these investigations are described in CHAPTERS 3 and 6. For the moment, it suffices to say that ionic mechanisms can explain many of the phenomena described in this chapter. 2-77 {vii) The Vulnerable Period, Repetitive Extrasystoles, and 0 Ventricular Fibrillation Perhaps the most dramatic effect of the delivery of a single current pulse to the activity. ~ocardium is the induction of fibrillatory If the ventricle is not in its refractory period, a single threshold current pulse leads to a propagated wave of depolarization. Within a narrow 11 VU1 nerable period 11 , increasing the amplitude of the stimulus produces, in addition, a second, non-driven extrasystole. Further increase in intensity results in multiple or repetitive extrasystoles (RES), which decrease in rate and die away, eventually returning control of the ventricle to the supraventricular input. Yet another increase in stimulus amplitude (to something of the order of ten times diastolic threshold) leads to sustained repetitive activity (ventricular tachycardia) which accelerates and degenerates into ventricular fibrillation (Matta, Verrier, and Lown, 1976). The mechanism underlying the above sequence of behaviours remains controversial, seventy years after the demonstration by r4ines that a single shock from an induction coil, "if properly timed', would induce fibrillation in the ventricles of an isolated cooled rabbit heart (Mines, 1914). Mines offered two mechanisms, which in today's terminology, would be called {i} reentry, and {ii} echo (e.g. Reshetilov, Pertsov, and Krinskii, 1979) or reflection (Antzelevitch, Jalife, and Moe, 1980; Jalife and Moe, 1981). In fact, reentry has been implicated in the generation of repetitive activity in response to electrical stimulation in many 2-78 c preparations of ventricular origin. These include an isolated papillary muscle- false tendon preparation {Sasyniuk and Mendez, 1971), focally cooled ventricle (Wallace and 11iynone, 1966), and in situ ventricle (El-Sherif et al., 1977a,l977b; de Bakker, Henning, and Merx, 1979; Janse et al., 1980; Wit et~., 1982). However, in many studies, single pulse stimulation apparently cannot induce fibri11ation, and a two or three pulse protocol must be employed. It may well be that had higher intensities been used, fibrillation would have been elicited by single pulses in these studies (van Tyn and ~1acLean, 1961; Spielman et al., 1978). As has been previously stated, there is generally thought to be a narrow window of vulnerability for the provocation of repetitive ventricular activity (Ferris et al., 1936; Wiggers and 0 Wegria, 1940). Clinically, this window is identified with the R-on-T phenomenon (Smirk, 1949). However, transthoracic shock, when applied to dogs, has been shown to be capable of producing repetitive extrasystoles and ventricular fibrillation at times scattered throughout the cardiac cycle (Milnor, Knickerbocker, and Kouwenhoven, 1958). Evidence has been accumulating over the last fifteen years that episodes of ventricular tachycardia are more often initiated by late premature ventricular contractions than by early extrasystoles in the in-hospital and post-hospitalization ambulatory phases of myocardial infarction (Ahuja, Gutierrez, and Manning, 1968; Bleifer et al., 1973; De Soyza et al., 1974; Kleiger et al., 1974; Winkle, Derrington, and Shroeder, 1977; Chou and 0 2-79 Wenzke, 1978; Roberts et al., 1978). Lie et al. (1975} found that many episodes of fibrillation were associated with late premature beats. However, induction of ventricular tachycardia and fibrillation does occur via the R-on-T pathway, and has been documented in at least four ambulatory patients showing el ectrocardi agraphic signs of acute myocardial ischaemia ( Gradman, Bell, and DeBusk, 1977; Hinkle et al., 1977; Reichenbach et al., 1977; Wei et al., 1979). Three of these four cases occurred in an out-of-hospital setting, and resulted in .. sudden cardiac death 11 • I have described pulse-induced repetitive activity in preparations that were less than 200 ~m in diameter (Fig. 2-20). Similar observations were previously made in quiescent·and in spontaneously active aggregates (Parshintsev, 1973). c I corroborate the observation made in this earlier study that the phase of the ·cycle at which the depolarizing stimulus is injected is not critical in inducing the phenomenon and that a stimulus amplitude of many times diastolic threshold is needed. The mechanism underlying the behaviour cannot be classical reentry, since the size of the preparation is too small to· support a long-loop reentrant pathway. Pulse-induced repetitive discharge has also been described in small strips of atrial muscle in which it is 11 most improbable that a circus movement can exist' {Dawes and Vane, 1 1951). Alternative mechanisms to explain the rapid activity include "refl ection11 , which has been recently experimentally demonstrated c 2-80 (Antzelevitch, Jalife, and Moe, 1980; Jalife and 0 11 echo" (a 1imi ti ng case of 11 r~oe, 1981), and reverberation") which has been shown to be possible in an ionic model of two coupled cells (Reshetilov, 1974; Reshetilov, Pertsov, and Krinskii, 1979). It is not likely that the activity I see is triggered, since it occurs in the normal range of membrane potential. One alternative which cannot immediately be discarded is that the depolarizing current pulse decoupl ed the cell in which the mic roel ectrode was embedded from its neighbours by increasing the resistance of the gap junctions that are presumably responsible for the synchronous activity normally observed (Spray, Harris, and Bennett, 1979). However, it is unreasonable to expect that, in all cases seen both by us and by Parshintsev (1973), the decoupled cell had an intrinsic rate faster than the overall rate of the aggregate. Furthermore, as already stated, one cell in an aggregate apparently has of the order of 10 6 more gap junctions than needed for electrical synchronization. Transjunctional potentials of the order of 40 mV are the maximum possible in the response of Fig. 2-20 and probably lead to at most a tenfold increase in gap-junctional resistance (Spray, Harris, and Bennett, 1979). Thus, the possibility of cellular decoupling appears slight. Also, repeated visual observation of the aggregate through the microscope showed no sign of disorganized mechanical activity. Finally, the tracings in Fig. 2-20A show no evidence of major decrease in the magnitude of the maximum diastolic potential, as 2-81 would be expected if the electrode were slipping out of the cell 11 0 11 (the converse of the sealing i n phenomenon de se ri bed and modelled 11 in Shrier and Clay (1982)): 11 the increased rate is due to an increase in the rate of diastolic depolarization. However, the voltage traces shown in Parshintsev {1973) show significant decrease in the amplitude of the maximum diastolic potential which might indicate a "slipping out.. of the electrode. This would cause a small leakage cur.rent which would be then responsible for a speeding up of the beat rate. Subsequent"sealing in" would gradually restore the originally prevailing situation. The fact that I only saw the phenomenon four times may be due to the fact that the large currents needed to elicit the phenomenon could not be obtained with most of the microelectrodes used in this study. 0 Rapid repetitive activity also can result if a monolayer of chick embryonic heart cells is subjected to a countershock-type electric field stimulation, and has been ascribed to a prolonged depolarization of the membrane potential probably due to electromechanical deformation of the cell membrane (Jones et al., 1978). This was not seen in the aggregate. The existence of any connection between the repetitive activity seen twice in response to a depolarizing stimulus in a ventricular tissue culture preparation and that seen in the intact ventricle and associated with paroxysmal ventricular tachycardia and fibrillation is purely speculative. Another finding of this study which might have more to do c 2-82 with paroxysmal tachycardia and fibrillation than the phenomenon 0 described above is the abrupt trans1tion shown in Figs. 2-6, 2-16, and 2-17. If a single stimulus is applied to the ventricle with effectively the same timing and amplitude as shown in these figures, there will be a great deal of asynchrony produced, since some cells will immediately fire, while others will only suffer a subthreshold depolarization. The respo~se in neighbouring cells will differ because the properties of adjacent cells, as well as the timing and strength of the current stimulus delivered to these cells, will be slightly different. This induced spatiotemporal inhomogeneity will predispose the ventricle to reentrant circuit formation. Stimulation of isolated strips of cardiac muscle can indeed 0 produce repetitive firing via a mechanism resembling the one outlined above, with the activity in two neighbouring cells being approximat~ly 1970: 180° out of phase with each other (Lu and Brooks, Fig. 3}. In this case, the immediate responses of the two cells to the current pulse, which must not be too high nor too low in amplitude and which must be correctly timed, are very similar to the two responses shown in Fig. 2-16. This experimental fact fits in well with the proposed mechanism, since the discontinuous response of Fig. 2-16 is not seen at smaller amplitudes {e.g. Fig. 2-5) or at higher amplitudes (Fig. 2-7). Thus, the discontinuous response found in the aggregate only at intermediate amplitudes may well explain why multiple extrasystoles, tachycardia, and 0 2-83 fibrillation are often only seen at intermediate stimulus 0 intensities with critically-timed pulses delivered towards the end of the effective refractory period (e.g. Matta, Verrier, and Lawn, 1976). However, this mechanism offers no obvious explanation for the experimental fact that, as the pulse amplitude is increased, first only single non-driven extrasystoles, then multiple extrasystoles, and finally fibrillation make their appearance. The explanation may well have to do with the fact that, with increasing stimulus intensity, the point of discontinuity moves to a shorter coupling interval, and so the duration of the evoked action potential decreases. It is interesting to note in this context that a numerical model consisting of 1000 interconnected quiescent cells shows rapid 0 repetitive activity in which there are two suboooulations of cells activated approximately 180° out of phase with respect to each other {Herschl eb, var1 der Twee1, and :·!eijl er, 19~2). These authors note that the characteristic frequency of ventricular fibrillation is about twice that of ventricular tachycardia; they also report that spectral analysis of electrocardiograms recorded during ventricular fibrillation reveals a subharmonic at one half of the dominant fibrillation frequency in~ of the 874 cases analyzed. 0 2-84 (viii} Topological Considerations 0 a. Type 1 and Type 0 Phase Resetting In phase-resetting experiments such as those described above for the aggregate, a stimulus of fixed intensity and waveform is delivered at different points in the spontaneous cycle. perturbation caused by the stimulus (gener~lly) The produces a transient change in the instantaneous frequency of the oscillator. The magnitude of the change depends both on the strength of the stimulus and on the phase in the spontaneous cycle at which it is injected. The instantaneous frequency is defined to be the reciprocal of the time between two consecutive reference or marker 0 events of the process. In these experiments, I have taken the marker event to be the zero-crossing on the action potential upstroke. When a stimulus is injected, the next event is taken to be the first positive-going zero-crossing that occurs during the stimulus or following the termination of the stimulus. As t+®, the instantaneous frequency recovers back to its unperturbed value, but the oscillator is left phase-shifted with respect to an unperturbed control (Fig. 2-22). The number of cycles needed for the effect of the stimulus to effectively wear off varies considerably from one biological oscillator to another. The ventricular heart cell aggregate generally recovers from the effects of a perturbation within one cycle following stimulation (e.g. Fig. 2-22). 2-85 0 Figure 2-22. A typical phase-resetting experiment in the aggregate. The heavy vertical bars represent the occurrence in time of the marker event - .:~ zero-crossing of potential on the action potential upstroke. trace is an unperturbed control. The upper The 1 ower trace shows the effect of a single depolarizing stimulus delivered at a coupling interval tc. (i~} In this case, the depolarizing stimulus is delivered relatively early in the cycle, producing a phase delay; thus the ~Ti are negative. 2-86 0 ~Ti is positive if the phase of the oscillator is advanced, negative if it is delayed. c Note that 0 0 Figure 2-22 shows an example of phase-resetting in the 0 A plot of the phase shift aggregate. ~T, • M.;' ( modul o • ( 2-1) 1) as a function of the old phase ( 2-2) i~ transient phase response curve, which 1s denoted is called the by PRC.; (Kawato and Suzuki, 1978; Kawato, 1981). In the limit ' 0 the curve is called the steady-state phase response curve ;~, (PRC ) (lC) . are 6~~ and~~~ (i~) • is called the steady-state phase shift. functions not only parameters. of~. The but also of the stimulus Note that several other ways of plotting phase- resetting data are also termed phase response curves {Winfree, 1980). .th- trans1ent • h A p1ot of th e 1 p ase ~ 1. ' =~ + I <P; 6c/> ,• ' ( mod 1) ( 2-3) as a function of q, is called the i~ phase transition curve, ~1hich is denoted by PTC.; (Kawato and Suzuki, 1978; ' c the limit 1~, Kawato, 1981). In q,i• is called the new phase, the eventual phase, or the latent phase, and the corresponding plot of 2-87 ~i '(i+oo) vs~ is c called the steady-state phase transition curve (PTC ) or the new 00 phase-old phase curve (Winfree, 1980). Note that the eventual phase is not necessarily defined for all values of~ (see below). Hypersurfaces of dimension N-1 connecting points in the Ndimensional phase space of a limit cycle model which all have the same value of eventual phase are called isochrons. The one- dimensional locus of points to which the state-point of the system is taken immediately following a perturbation (for all can be called the perturbed cycle. ~. 0<$,1) It is the intersection of the perturbed cycle (for a particular set of stimulus parameters) with the isochronal surfaces that gives a particular PTC • There has 00 been some mathematical work concerning the topology of isochronal surfaces (Guckenheimer, 1975; Kawato and Suzuki, 1978; Kawato, 0 1981; Winfree, 1980); these surfaces can have a very tortuous geometrical structure (Winfree, 1980; Glass and Winfree, 1984). In a system of ordinary differential equations with only one equilibrium point and one asymptotically stable limit cycle, PTCoo is defined and continuous for all strengths· (Kawato, 1981}: itself. ~at almost all stimulus PTCoo it is a mapping of the circle into The winding number or topological degree of PTC can be 00 defined to be its average slope (Winfree, 1980}. Examination of the available biological phase-resetting data 16d Winfree (1977, 1980) to conclude that only two types of phase resetting are to be found experimentally: PTCoo is either of topological degree one or of topological degree zero. c 2-88 0 Figure 2-23 shows the change in the form of PTC 2 for the aggregate as the stimulus amplitude is increased. Due to the very fast recovery following perturbation, the curve PTC 2 is an excellent approximation to PTC 00 • At the two lower stimulus intensities (5 nA and 6.5 nA), type 1 phase resetting is found in the aggregate, whilst type 0 phase resetting is found at the highest stimulus intensity (24 nA). At a stimulus current of 8 nA, the response to a single pulse is almost at the point at which it becomes discontinuous (Fig. 2-5, right panel ; Fig. 2-6). Thus, for a pulse ampl i'tude somewhere between 8 nA and 16 nA in this aggregate, PTC 2 (as calculated from eqn. (1-3)) becomes discontinuous. This is probably due to a breakdown in the assumption that spontaneous activity in the aggregate is a reflection of the existence of a limit cycle in the phase space of a deterministic set of differential equations that model the behaviour of the aggregate (see CHAPTER 3). A stochastic term must be added to the model; better still, the model should be completely reformulated along the lines indicated in CHAPTER 1 as an inherently stochastic one based on the behaviour of single channels in the membrane. Thus, use of the term PTC is called into question at this level of stimulation, since the assumptions under which it was derived (a limit cycle existing in a continuous system of ordinary differential equation) probably no longer hold in the range of stimulus amplitudes at which this discontinuous transition c 2-89 c Figure 2-23. Second transient phase transition curve (PTC 2 ) at four different pulse amplitudes: C. 8 nA D. 24 nA. A. 5 nA B. 6.5 nA Data taken from the experimental traces shown in Figs. 2-5 and 2-7. Type 1 phase resetting occurs in panels A and B, while type 0 phase resetting occurs in panel D. region ~~0.4 Repeated trials in the reveal that the curve is continuous in panel B as indicated by the dashed lines. Experimental determinations for $larger than about 0 0.5 were not carried out in this experiment. dashed lines for • ~ 0.5 are approxfmations based on the results of other experiments. (Aggregate# 1: diameter= 114 urn.) 2-90 0 The 0 B A ........ .... ···· ••• • • • • • • • • ••••• • 5nA 0 6.5 nA 0 0 0 0 c ...................................... ••••• D 8 nA 0 0 0 24 nA 0 0 from prolongation to shortening of cycle length occurs. The difference in T1 between the two responses seen can be significantly less than one spontaneous cycle length (e.g. Fig. 2In addition, the lengths of the post-stimulus cycles are very 16}. similar in the two cases. Thus, in PTC; (i~2), discontinuity of condiderably less than 1: there is also a type 0 phase resetting is not occurring. In the aggregate, as stimulus strength is increased, PTCoo is first type 1 and then type 0. In the limit of zero stimulus strength, there is no effect of the stimulus on the rhythm, I 8~~• + 0, and so PTC <» approaches the diagonal line 1 $ 00 = ~. At finite but suffic·iently small stimulus strength (e.g. Fig. 2-lOC), PTCoo is type 1 and monotone increasing. For a higher stimulus strength, it remains type 1 but becomes non-monotonic (e.g. F1g. 2238}. As the stimulus strength increases, ·the dip in the curve responsible for the non-monotonicity deepens (e.g. Fig. 2-25A,B,C). Eventually, type 0 phase resetting occurs (Fig. 2-230). This sequence {type 1 (monotonic) +type 1 (non-monotonic) +type 0} occurs in several different periodically forced limit-cycle oscillators (see CHAPTER 6). There is reason to believe that the preparation from which the curves shown in Fig. 2-23 were obtained has only one equilibrium point which is unstable and which lies in the plateau range of potentials (Clay, Guevara, and Shrier, 1984; Fig. 2-24A). ·As the stimulus amplitude is increased from low levels which give type 1 phase resetting, the perturbed cycle eventually intersects 0 2-91 the stable manifold ("null space.. ) of this point. c At this one particular amplitude, there is one value of tc for which the trajectory of the system would (in a noise-free situation) asymptotically approach the equilibrium point and spontaneous oscillation would cease. Further increase in amplitude results in type 0 phase resetting. Preliminary investigation of ionic models of space-clamped cardiac tissue that possess only one equilibrium point in the plateau potential range (Guevara, unpublished) suggests that the voltage at the end of a current pulse must be very close to the voltage of this equilibrium point in order for the state point of the system to be close to the stable manifold of the equilibrium This fact is borne out in the experiments in faster-beating point. aggregates, since the membrane potential must be pushed into the plateau range of potentials (Fig. 2-7) before type 0 phase resetting (Fig. 2-230) is seen. When behaviour such as that shown in Figs. 2-11 to 2-14 occurs, the situation is more complex. Ionic modelling indicates that slower-beating preparations may have an equilibrium point of a spiral nature (i.e. with complex eigenvalues) in the pacemaker range of potentials (Clay, Guevara, and Shrier, 1984; Fig. 224B,C). If this equilibrium point is the only equilibrium point in the system (Fig. 2-24C), then type 0 phase resetting occurs for a much smaller amplitude depolarizing pulse than required if the only equilibrium point in the system were to lie in the plateau 2-92 c::J Figure 2-24. Schematic steady-state current-voltage (IV) relations for three different aggregates. The curves are shifted in a hyperpolarizing direction as one moves from panel A to panel C, corresponding to a decrease in the beat rate. The intersection of the IV characteristic with the horizontal axis I=O gives the equilibrium point(s) of the system. The middle equilibrium point in panel B is a saddle point. Perturbation of the state point of the system into a neighbourhood of the equilibrium point in panel C would lead to subthreshold oscillatory activity in the pacemaker range of potentials if that point had 0 complex eigenvalues. The same holds true for the most negative equilibrium point in panel B. In panels A and C, the isochrons of the single unstable equilibrium point foliate the entire phase space. In panel B, the separatrix associated with the middle equilibrium point, which is a saddle, presumably winds into the other equilibrium points. Thus, the topology of the isochronal hypersurfaces in this case is considerably more complex than in the simpler cases shown in panels A and C. 2-93 I A 0 / B / c I I I / ----·------. ___.... ', I range of potentials (Fig. 2-24A}. However, this equilibrium point in the pacemaker range of potentials can coexist with two other equilibrium points: one equilibrium point is the previously mentioned point in the plateau range of potentials, the other is a saddle point (Fig. 2-24B}. In this case, for i ntennedi ate stimulus amplitudes, a topological degree cannot be assigned to the phase-resetting, since T1 becomes unbounded as tc approaches one critical value which results in perturbation of the state-point onto the separatrix hypersurface which forms the stable manifold of the saddle point (Clay, Guevara, and Shrier, 1984; Glass and Winfree, 1984}. Thus, one cannot say with certainty that the type of behaviour shown in Fig. 2-11 lies close to the border between type 1 and type 0 phase-resetting. 0 For example, in the other slowly beating preparation in which long delays were seen (not the one whose phase-resetting response is shown in Fig. 2-11}, long delays were produced with a monotonic approach to threshold, showing no evidence of subthreshold oscillatory activity in the pacemaker range of potentials. This suggests the existence of three equilibrium points, one being a saddle. The considerations outlined in this section probably also apply to the phase-resetting ·study of van Meerwij k et al • ( 1984}. Phase resetting experiments carried out on sinoatrial node ( Dong and Reitz, 1970; Levy, Iano, and Zieske, 1972; Greco and Clark, 1976; Sano, Sawanobori, and Adaniya, 1978; Jalife and Moe, 1979b; Jal ife et al., 1980 ), a focus in the right bundle branch 2-94 (Ferrier and Rosenthal, 1980), Purkinje fibre (Jal ~fe and Moe, 1976; Jal ife and t1oe, 1979a), and heart cell aggregates ( Scott, 1979; Guevara, Glass, and Shrier, 1981; van r1eerwijk et al., 1984) all show apparent discontinuities when T1/T 0 is plotted as a function of tc/T 0 • However, in the majority of these reports, increments in tc of substantially greater than 1 msec were employed. Thus, some of these curves might actually have been continuous had finer increments in tc been used. In most of these experiments, the discontinuity in T1/T 0 is less than 1, and voltage tracings showing graded action potentials in response to a depolarizing pulse are not shown. These graded action potentials appear to be necessary for type 0 phase resetting 1n ionic models with only one equilibrium point in the plateau range of potentials 0 (Guevara, unpublished). In light of the above considerations, I caution against the i nferral of the ex1 stence of Type 0 phase resetting simply based on an examination of curves of T1/T 0 plotted against t/To (Scott, 1979; Winfree, 1981). It is unclear from Fig. 1D of Jalife and Antzelevitch (1979) whether type 0 phaseresetting is seen in the sinoatrial node; however, one curve in Winfree (1983b) plotted from data obtained by Jalife and Salata shows type 0 phase resetting. It is unclear to me whether type 0 phase resetting has been seen in Purkinje fibre (Jalife and Antzelevitch, 1979, 1980}. 2-95 b. Discontinuities in Phase Resetting Using the definition of the phase transition curve given . {i.;;:oo) earlier, real discontinuties can appear in plots of PTC.; obtained from systems of ordinary differential equations possessing only one equilibrium point, even when PTC 00 is continuous. SiDple two-dimensional limit-cycle oscillators can be used to demonstrate this point (Kawato and Suzuki, 1978; Kawato, 1981; Holden, 1983). Barbi and These discontinuities arise when the perturbed cycle intersects the event surface. Using My definition of an • event, the event surface is the demi-hyperplane V~= 0, V 0 ~0. Fig. '·· r. 2-250 shows PTC 1 corresponding to PTC 2 of Fig. 2-23D, demonstrating this kind of discontinuity. Note that this discontinuity appears even though the behaviour of the preparation is smoothly changing (Fig. 2-7); there is no real physical discontinuity. A second kind of discontinuity is the threshold-type behaviour shown in Figs. 2-16 and 2-17 and discussed in section {iii) above and in CHAPTER 3 below. I reiterate that this discontinuity of all-or-none depolarization appears to be real, and that I believe it arises from the stochastic nature of the behaviour of single channels in the cardiac membrane (see CHAPTER 3). A similar statement can probably be made with regard to all- or-none repolarization. A third kind of discontinuity is evident when the perturbed cycle passes through regions of phase space wherein lie the tightly-coiled trajectories of an unstable spiral point or unstable 2-96 Figure 2-25. First transient phase transition curve {PTC 1 ) for four pulse amplitudes: D. 24 nA. A. 5 nA B. 6.5 nA C. 8 nA The dashed lines in the upper right hand corners of all four panels are approximations to the data, necessary because the stimulus artifact obscures the action potential upstroke for stimuli delivered sufficiently late in the cycle. The dashed line in the middle of panel B indicates that the response is continuous in this region. The discontinuity in panel 0 is a Kawato-Suzuki type discontinuity, and is discussed in the text. in Fig. 2-230. PTC 2 for these data is shown The increased scatter in the data points of Fig. 2-23 is due to the inherent fluctuations in the interbeat interval of the aggregate. (Aggregate# 1: diameter= 114 2-97 ~m.) A .~ • •• , ..... B • • ••• 6.5 nA 5 nA 0 e---------------------------~ 0 0 0 0 .............. c D • .. • 8 nA 24 nA 0 0 0 0 small-amplitude limit cycle. In this case, there are apparent discontinuities in the perturbed cycle length function (T 1 /T 0 vs. \/T 0 ), since the action potential fires on one or other of the crests of the subthreshold oscillatory activity (Figs. 2-13, 2-14). Glass and Winfree (1984) have shown in a simple two-dimensional limit cycle model (motivated by the results of these experiments) that real discontinuities in the perturbed cycle length function can occur if a threshold is included in the model. It remains to be seen whether the gaps occurring in Fig. 2-13 where action potentials do not fire would occur (in the absence of noise) in a higher-dimensional ionic model of continuous differential equations that does .not possess a discontinuous threshold. A fourth kind of discontinuity appears when the system has three equilibrium points (Fig. 2-248). This occurs when the perturbed cycle pierces the stable separatrix associated with the saddle point. There is little experimental evidence for this behaviour (one aggregate), but it can be shown to occur in ionic models (Clay, Guevara, and Shrier, 1984). The influence of noise (arising from single channel activity) on this behaviour would presumably be to smear out the discontinuity. For example, the arbitrarily large values of T1 predicted in the model would not occur in experiments (Lecar and Nossal, 197la, 1971b). 2-98 .- c. Annihilation of Spontaneous Activity It appears to be impossible to abolish spontaneous activity in the aggregate with a 20 msec duration depolarizing pulse. Although the type 1- type 0 border can be crossed using a pulse of this duration, there is no guarantee that a null-space of full dimensionality, if present, will be encountered. Thus, it is conceivable that it might be possible to abolish spontaneous activity in the aggregate by using a pulse duration different from 20 msec; I have not investigated this possibility. However, ionic modelling ·indicates that faster-beating aggregates only have one equilibrium point (Clay, Guevara, and Shrier, 1984}, which is unstable (see CHAPTER 3}. 0 Ionic modelling of the sinoatrial node and Purkinje fibre also indicates that, without any external intervention, it is impossible to stop spontaneous activity with a single pulse of current (Guevara, unpublished). Thus, in these cases, the stable manifold of the equilibrium point (the phaseless set or null-space) has dimensionality less than the dimension N of the phase space of the equations. d. Cardiac Rotors Winfree (1982, 1983a, 1983b) has recently suggested that there may be a connection between the topology of phase resetting 0 2-99 of cardiac oscillatorsand the induction of paroxysmal tachycardia or fibrillation by premature stimulation. I will now give an example that I think illustrates Winfree•s main points. Consider the situation shown in Fig. 2-26, where one has a monolayer of spontaneously oscillating cells. Let us assume that these cells have only one equilibrium point, lying in the plateau range of potentials. Assume that at the point in time considered in Fig. 2-26, a wave of activation has just propagated uniformly across the sheet from right to left. The electrical activity in all cells along the same vertical line is synchronous, with the voltage of a cell at any horizontal coordinate being given by the corresponding point on the waveform shown at the top of the figure. Suppose that a premature stimulus is then applied at the location 0 s. Suppose further that the stimulus is sufficiently strong to cause type 0 phase resetting in cells sufficiently close to S (e.g. at location A). Cells sufficiently distant from S (e.g. at· location C) will only feel a small electrotonic influence, and will undergo type 1 phase resetting. By continuity, there will be a point B, somewhere between A and C, which will be on the border between type 1 and type 0 phase resetting; the perturbation takes the state-point of the cell located at B to the phaseless manifold of the equilibrium point. By continuity, there will be cells in a neighbourhood of B that will have all possible values of eventual phase. This sets up optimal conditions for a circulating helical wave of excitation - a cardiac reverberator or rotor - to come into 0 2-100 o Figure 2-26. Schematic diagram used in the text to illustrate Winfree's theory of fibrillation. The rectangle represents a monolayer of cardiac cells. The voltage of any cell in the monolayer can be found by projecting a vertical 1 ine up to the waveform at the top of the figure. See text for further description. 2-101 !.-... • / I ! 0 i '-- ! / u • CO • < being. c If the oscillating cells making up the planar medium can be stopped with a single pulse, it is not difficult to see that a spatial 11 black hole11 will be created, which will cover a small area of the plane (Winfree, 1982, 1983a, 1983b). may even grow in size (Winfree, 1983b). This black hole Note that the centre of the rotor will be located at a point some distance away from the point of stimulation. It is of some interest in this connection to note that in a review article on ventricular fibrillation, Wiggers ( 1940) wrote: 11 The fact that only strong stimuli, applied during the vulnerable phase, are followed by undulatory waves and fibrillation can be interpretated to signify that they exert some influence at a distance, which modifies conduction and permits reentry, or 0 ... . 11 The above argument of Wi nfree can be modified to account for the case of a sheet containing quiescent, but excitable, cells. This modification is essential if one wishes to make a connection with tachycardia and fibrillation, since these dysrhythmias generally occur in cardiac tissues that are taken to be quiescent. To carry out this modification, one simply has to replace the concept of phase with that of latency. The continuity argument needed is one based on the continuity of latency. By latency, I mean the time to the next maximum of voltage in a cell following delivery of a stimulus. By continuity of 1 atency, I mean that this time is a continuous function of the stimulation intensity. Note that continuity of latency is largely due to continuity in the function relating the voltage attained following a stimulus to the 0 2-102 strength of that stimulus. c That continuity has been demonstrated in the Hodgkin-Huxley equations for quiescent squid axon (Clay, 1977), and is to be expected in ionic models of cardiac tissue which possess only one equilibrium point (FitzHugh, 1955,1960}. The trypsin-dissociated ventricular aggregate does not show a continuous progression from type 1 to type 0 phase resetting as stimulus intensity is increased; the response is discontinuous in the neighbourhood of the type 1 - type 0 border (e.g. Fig. 2-16). As mentioned earlier, many cardiac tissues, both quiescent and spontaneously active, show this discontinuous all-or-nothing response to premature stimulation applied close to the end of their refractory period. This discontinuity is partly due to a competition between the depolarizing fast inward sodium current INa 0 and the repolarizing potass'ium current Ix (see CHAPTER 3). Thus, the continuity assumption crucial to Winfree's scheme is missing in the case of tissues demonstrating fast action potentials. However, in tissues with slow action potentials, the continuity condition necessary for the Wi nfree mechanism to take place may be present. t1anoeuvres which lead to slow action potentials {and slow conduction) in ventricular muscle uniformly decrease the ventricular fibrillation threshold; this fact fits in well with Winfree's hypothesis. 2-103 CHAPTER 3 THE IONIC BASIS OF SPONTANEOUS ACTIVITY AND OF PHASE RESETTING IN THE AGGREGATE: NUMERICAL INVESTIGATION OF A PARTIAL MODEL 0 11 The theory is a mess and seems inherently messy." John Guckenheimer, 1980b 0 INTRODUCTION In this chapter, I assemble an ionic model of spontaneous electrical activity in the embryonic chick ventricular heart cell aggregate and investigate its response to perturbation with brief duration current pulses. The model consists of a set of ordinary differential equations that describe the ionic currents that flow across the cell membrane. These equations are obtained from published analyses of data from voltage clamp experiments. There have been few models of spontaneously active cardiac tissue assembled from voltage clamp data: models for sinoatrial node (Yanagihara, Noma, and Irisawa, 1980; Irisawa and tJoma, 1982; Bristow and Clark, 1982) and for Purkinje fibre 0Joble, 1962; Gul'ko and Petrov, 1970; M::Allister, Noble, and Tsien, 1975; DiFrancesco and Noble, 1982a). The model of Beeler and Reuter (1977) for quiescent ventricular muscle can be made spontaneously active if a constant depolarizing bias current is applied. There appears to have been only one systematic study undertaken on the phase resetting of an ionic model of spontaneously oscillating cardiac cells (Bristow and Clark, 1982). The model investigated in this chapter is a partial one, in that it only simulates phase 3 of the action potential (the late, rapid phase of repolarization}, phase 4 (the diastolic or pacemaker potential), and phase 0 (the upstroke of the action potential). 3-1 This is due to the fact that only five currents are included in the model: the fast inward sodium current INq' the plateau current Ix, the pacemaker current IK , and the background currents INa b and 2 ' IK • Other currents known to exist in the aggregate, such as the l slow inward current Isi (Josephson and Sperelakis, 1982), are not included in the model, since a quantitative description of their properties is not presently available. In Section 2 below, I describe the individual currents and give their mathematical formulation. Section. 3 briefly outlines the method used to numerically integrate the system equations. Section 4 shows the spontaneous activity displayed by the model and its response to current pulses of brief duration. Finally, in Section 5, I relate the activity of the model to the experimentally 0 observed behaviour described in Chapter 1 and discuss the shortcomings of the model. 2. FORMULATION OF THE MODEL: THE INDIVIDUAL CURRENT COMPONENTS Throughout, voltages are in units of millivolts (mV), currents in nanoamperes ( nA), conductances in microsi ernens ( JJS), and times in seconds {sec). All conductances and currents assume an aggregate diameter of 200 Jlm. ·Inward currents are negative; outward currents are positive. c 3-2 The TTX-sensitive, fast inward sodium current INa is both time (t) and voltage (V) dependent, having both activation (m) and inactivation (h) variables ( 3-1) where gNa is the fully activated conductance and ENa is the Nernst equilibrium potential for the sodium ion. is thus a linear, gated channel. The fast sodium channel The activation and inactivation variables m and h each obey first-order kinetics m= a m(1-m) -em m ( 3-2a) The overdot denotes differentiation with respect to time. The rate coefficients am, sm' ah' and eh are given (in units of inverse seconds) by (Ebihara and Johnson, 1980) 0 an= 320 (V+47.13)/(1-exp{V+47.13)) I, ( 3-3a) Bm = 80 exp(-V/11) {3-3b) 3-3 ah c ( 3-3c) = 135 exp (-(V+80}/6.8) eh= 3.56 x 10 3 exp(0.079V) + 3.1 x 10s exp (0.35V). (3-3d) Figures 3-lA and 3-lB are plots of these activation and inactivation rate constants as functions of voltage. 2) m~ Equation (3- be rewritten • 1 (3-4a} m =-(m -m) T m QC) • '1 h =T- ( hoo -h) h (3-4b) ' where -rm( Th} is the time constant of activation (inactivation) and mQO (hQO) is the asymptotic or steady-state value of m{h) at a fixed potential ( 3- Sa) m QC) = a m'I( a mm +e ) ( 3-Sb} ( 3-Sc} ( 3-Sd} 3-4 Figure 3-1. Variables associated with the activation and inactivation of INa plotted as functions of voltage. Unmodified equations of Ebihara and Johnson (1980). 0 A. The rate constants of activation (am and 8m}. B. The rate constants of i nac ti vati on (ah and Sh}. c. The steady-state values of m and h (m and h ). D. The time constants of activation and inactivation CO (Tm and T 00 h). The scales in panels A and B are chosen to facilitate comparison with the corresponding curves shown in Ebihara and Johnson ( 1980). All of these curv.es are similar to those ~otted in Ebihara and Johnson (1980) with the exception of the m curve (see text). QO 3-5 1 6121121121. (sec-1) e 121121121. 121. 121 -3121.121 -6121. 121 5121121. 121 ( sec-1) 121. 121 B 25121.121 121. 121 -9121. 121 0 -6121.121 -4121.121 1. 121 • 5 121. 121 -1121121.0 • 1211211214 Tm (sec) -5121. 121 121. 0 • 1214 D • 1211211212 • 02 0 121. 121 0. 121 -10121.121 -50.0 121. 121 Th (sec) Figures 3-1C and 3-1D are plots of -r , 'h, m , and h as functions m 0 "" "" of voltage. The expressions for INa in eqn. (3-1) and for am, sm, and ah in eqn. (3-3} have the same functional form as those originally used to describe the fast inward sodium current in the giant axon of the squid (Hodgkin and Huxley, 1952). However, the maximal specific sodium conductance is much larger in nerve than in the aggregate (120 mS cm- 2 versus 23 mS cm- 2 ). This difference is in part responsible for the smaller maximal rate of rise of the upstroke in the aggregate. Even though the rate constants are larger in the aggregate at 37°C than in the squid axon at 15°C, the ratio Th/-rm is about the same in both preparations over a wide range of potentials (Ebihara and Johnson, 1980). 0 In the model for the aggregate, the h"" curve is shifted by about 12 mV in the hyperpolarizing direction relative to the h "" curve in the Hodgkin-Huxley model for the squid axon: there is little overlap of the m"" and h"" curves in the aggregate model (Fig. 3-1C) and hence theoretically 1ittle window current INa in the 00 pacemaker range of potentials (-100 mV ~V~ -60 mV). The window current INa (Attwell et al., 1979) is the steady state value of Oil INa at a potential V and is given by INa = m}hj'Na ( V-E~Ja). 00 Thus, INa would not be expected to play a significant role in the genesis of phase 4 depolarization. However, following addition of tetrodotoxin ( TTX), which specifically blocks INa' the spontaneous beat rate of an aggregate slows before action potential generation 0 3-6 c stops. This slowing is due to a prolongation of the duration of phase 4 {Colizza, Guevara, and Shrier, 1983: Fig. 6}. Thus, INa is experimentally implicated in controlling the duration of phase 4, if not its slope. The window current INa has recently been measured in the "" aggregate {Brochu, Shrier, and Clay unpublished}, and is much larger than that predicted in the pacemaker range of potentials using the equations given above. The equations for m listed above were obtained by voltage clamping at potentials more positive than -40 mV. Thus, these formulae are not necessarily applicable in the pacemaker range of potentials. Furthermore, there are inconsistencies in the paper of Ebihara and Johnson (1980) regarding the activation variable m. 0 Using their published equations for the rate constants am and sm (eqn. (3-3a,b) above), I obtain curves of am and am as functions of voltage very similar to their curves (my Fig. 3-lA, their Fig. 4C). However, using these same equations, one obtains a curve for m as a function of voltage "" (Fig. 3-lC) which .is not the same as the one shown in Fig. 4A of Ebihara and Johnson (1980). However, both curves fit the data points well for V more positive than -40 mV; they are only different for V more negative than about -40 mV, where there are anyway no experimental data points. Another problem arises if one carries out a numerical simulation of the membrane voltage using the equations given above for INa and the equations given below for the other currents. membrane potential slowly becomes more positive, but does not attain threshold: the upstroke of the action potential is not 3-7 The generated. 0 It is reasonable to suspect that this deficiency is associated with the aforesaid problems with the activation variable m. Clay has shown that inserting an extra factor of 0.1 multiplying the term (V+ 47.13) appearing in the exponential in the denominator of eqn. (3-3a) resolves this problem (Clay, Guevara, and Shrier, 1984). Moreover, the predicted window current is now close to the experimentally measured value (Shrier and Clay, unpublished). rm. Figure 3-2 shows the modified curves for am, m~. and However, these modified curves do not fit the data points of Ebihara and Johnson {1980) as well as the curves {Fig. 3-1) resulting from their original unmodified equations. In what follows, I nevertheless incorporate the extra factor of 0.1 into the model. 0 Equation (3-3) for the rate constants is taken from voltage cl amp data obtained from aggregates prepared from 11-day-ol d chick embryos {Ebihara and Johnson, 1980; Ebihara et al., 1980). This equation can be applied to 7-day-old aggregates (i.e. aggregates prepared from 7-day-old embryos} , since the maximal upstroke • velocity Vmax does not change appreciably from day 7 to day 11 (Ebihara et al., 1980; Clay and Shrier, 1981b; Colizza, Guevara, and Shrier, 1983). It is unlikely that there could be simultaneous compensating shifts in opposite directions in the maximal sodium conductance and in the kinetics between day 7 and day 11 that would • leave both Vmax and the potential at which it occurs (~ - 20 mV) unchanged. Ebihara and Johnson (1980} found a maximal specific sodium conductance of 23 mS cm- 2 in a 75 urn-diameter aggregate that had an 3-8 0 Figure 3-2. Variables associated with the activation and inactivation of INa plotted as functions of voltage. Equations of Ebihara and Johnson (1980}, with am modified as indicated in the text. A. The rate constants of activation (am and Sm}. B. The rate constants of inactivation ( ah and Bh}. Q c. The steady-state values of m and h (m and h ) • "" (10 D. The time constants of activation and inactivation (TmandTh). There is now a significant INa due to the overlap of 00 moo and hoo shown in panel 0 c. 3-9 1 6121121121. (se c-l) e 121121121. 121. 121 -6121. 121 5121121. 121 (sec-1) -3121. 121 121. 121 B 25121.121 121. 121 -8121. 121 0 -6121. 121 -4121. 121 " 1. 121 • 5 121. 121 -1121121. 121 • 1211211214 Tm (sec) -5121.121 121. 121 • 1214 D • 1211211212 • 1212 ( 121. 121 121. 121 -1121121. 121 -5121. 121 121. 121 Th sec ) active cell surface area of 1.47 x 10- 3 cm 2 • Thus, an aggregate of 0 diameter 200 lliD would have gNa = 641 l-!S in eqn·. (3-1), since membrane surface area is directly proportional to the cube of the aggregate diameter (Clay, DeFelice, and DeH~an, reversal potential for the sodium current, E~Ja, 1979). The was experimentally measured to be 29 mV (Ebihara and Johnson, 1980), which agrees well with 38 mV, the Nernst potential calculated for the sodium ion from the measured intracellular sodium concentration of 33.5 {McDonald and DeHaan, 1973). ~~ The discrepancy between the two values may be partly due to the different ages of aggregates and the different external potassium concentrations used in the two studies. In eqn. (3-1), I use ENa The fast sodium current 0 IN~ = 40 mV. has been described using voltage clamp techniques in several cardiac tissues including the sinoatrial node (Noma, Yanagihara, and Irisawa, 19t7), atrium (Rougier, Vassort, and Stampfli, 1968), Purkinje fibre (Dudel and Rildel, 1970; Colatsky and Tsien, 1979; Colatsky, 1980), ventricular muscle (Lee et al., 1979; Bodewei et al., 1982), and ventricular heart cell aggregates (Nathan and DeHaan, 1979; Ebihara and Johnson, 1980; Ebihara et al., 1980). INa is not necessary for spontaneous action potential generation in tissues with slow action potentials such as the sinoatrial node (Yamagishi and Sano, 1966) or the atrioventricular node {Zipes and Mendez, 1973); however, in most tissues with fast action potential s, pharmacological blockade 3-10 of Iua with TTX abolishes spontaneous activity. Thus, INa can be involved in generating the action potential upstroke, in maintaining the action potential duration (Attwell et al., 1979), or in contributing to spontaneous phase 4 depolarization. Application of a voltage-clamp step in the pacemaker potential range (-100 mV < V < -60 mV) yields a current waveform with an exponential time course. Such voltage clamp data has been interpreted (Shrier and Clay, 1980; Clay and Shrier, 198la, 1981b; Shrier and Clay, 1982; Clay and Shrier, 1983} as evidence for the existence of a time-dependent, potassium-ion pacemaker current IK 2 0 in the aggregate that is similar to that e~r1ier Purkinje fibre (e.g. Noble and Tsien, 1968). described in Like the potassium current in squid axon, IK activates but does not inactivate 2 (3-6) The activation variable is denoted by s, while fK (V) gives the 2 current flow through the fully activated (i.e. s=l) channels. Unlike the fast sodium channel in both squid axon and heart, and unlike the potassium channel in squid axon. the IK channel is not 2 ohmic: it rectifies in the inward direction, with the dependence of current on voltage given by the nonl inear function fK (V) 2 3-11 0 . where + p2 = 1/(1 {3-8a} + exp (-e(V-EK)/kT)) {3-8b) Pz { 3-8c) Yz-- • + Pz This formulation for the rectifying channel comes from a 0 si ngl e-fil e knock-on model o.f ion traffic through a channel which possesses a block1ng particle and which has 2 ion-selective sites (Hodgkin and Keynes, 1955; Clay and Shlesinger, 1977; Clay and Shrier, 1981a). The parameter 62 gives the probability that the blocking particle will be dislodged if struck by a permeant + potassium ion, p2 gives the probability that it will then enter the channel, t 2 is the average time between collisions, N2 is the total number of IK channels, EK is the equilibrium potential for the 2 potassium ion, e is the elementary charge, k is Boltzmann S 1 constant, and T is the temperature C'K). mv. 3-12 ForT= 35°C, kT/e = 26.5 The activation variable s obeys first order kinetics 0 ( 3-9) or s =-}- (s..,-s), ( 3-10) s with (3-lla} S 00 0 =a s/(a s + Bs ). ( 3-llb} The rate constants are given by as= 1.05 (V+57) I (1-exp(-0.2(V+57))} (3-12a} a5 (3-12b) = 0.095 exp (-0.075 (V+57)). Note that the rate constants as and as for the activation and deactivation of IK have the same functional form as those for the 2 potassium current IK in squid axon. Furthermore, the functional forms for as and a 5 are the same as those for am and am given 3-13 earlier in eqn. {3-3). 0 The constants in eqn. (3-12) were found by fitting eqns. (3-11) and (3-12) to the experimentally determined values of r S and sCO in 15 aggregates {Clay and Shrier, 198la). Although these 15 experiments were carried out at external potassium concentrations ranging from 1.3 mM to 4.8 mM, the bestfit through all the data (eqn. (3-12)) can be used, since the kinetics of IK do not appear to be a function of the external 2 potassium concentration (Clay and Shrier, 1981a). as, as, S 00 , and •s as functions of V. Figure 3-3 shows The curves for sco and rs agree well with the corresponding ones in Fig. 5 of Clay and Shrier (198la) obtained from eqns. (3-11) and {3-12). Voltage clamp steps in the pacemaker range of potentials produce instantaneous changes in both IK and the time-independent 2 0 background current Ibg due to changes in driving force. These changes are complicated by the rectification present in both currents. However, the contaminating effect of Ibg can be removed by computing the ratio of the time-dependent changes in current at the "on11 and "off.. of a voltage clamp pulse. Shrier (1981a), this "ratio analysi~ (Noble and Tsien, 1968) gives a value of 1090 nA for N2 et 2 - 1 in eqn. (3-7). in the model. According to Clay and I adopt this value The calculated Nernst potential for potassium at an external potassium concentration of 1.3 mM is -124 mV, assuming an intracellular potassium concentration of 146.1 mM (McDonald and DeHaan, 1973). While the reversal potential for IK has not been 2 experimentally obtained at 1.3 mM, it has been obtained at higher 3-14 0 Figure 3-3. Variables associated with IK plotted as functions of 2 voltage. 0 A. The rate constants of activation (as and B. The steady-state value of s {s }. c. The time constant of activation (T 5 ). CO 3-15 e5 ). 5121. 121 5. 121 A 0 121. 121 r21. 121 -1121121. 121 -5121. 121 121. 121 -5121.121 121. r2l -5121.121 r21. 121 1. 121 sex> •5 0 121. 121 -1121121. 121 2. 121 Ts c 1. 121 (sec) 121. 121 -1121121.121 potassium concentrations, and the experimental values agree very 0 well with the calculated Nernst values (Clay and Shrier, 1981a). thus use EK I = -124 mV in eqn. (3-8a). The formula for IK in eqn. (3-6) is similar to that 2 originally used to .describe the potassium current IK in squid axon (Hodgkin and Huxley, 1952). However, following a voltage clamp step in the aggregate, there is an exponential (and not a sigmoidal) change in current. Thus, the activation variable s in eqn. ( 3-6) is taken to the first power and not to the fourth power as in the Hodgkin-Huxley equations. Also note that the pacemaker current IK 2 is not responsible for the repolarization phase of the action potential, as is IK in nerve, since it is activated over a much more negative range of potentials. 0 constants for IK Not unexpectedly, the rate are about two orders of magnitude smaller than 2 those for IK in nerve. The IK current described by eqns. {3-6) to (3-12) is also 2 similar to that originally described in adult Purkinje fibre (Deck and Trautwein, 1964; Vassale, 1966; Dudel et al., 1967; Noble and Tsien, 1968; Peper and Trautwein, 1969; Hauswirth, Noble, and Tsien, 1972a; McAllister, Noble, and Tsien, 1975). a maximum value of about 1.3 sec at V ~ -80 mv However, Ts has in the aggregate (Fig. 3-3C}, as contrasted with a maximum value of about 2.3 sec at V :'::! -80 mV in Purkinje fibre. This may have to do with the fact that the duration of phase 4 is much longer in Purkinje fibre than in the aggregate, since IK is intimately tied to the generation 2 0 3-16 of diastolic depolarization (see RESULTS below). c Currents similar to IK apparently do not exist in adult atrial (Brown, Clark, and 2 Noble, 1976a, 1976b) or ventricular (Beeler and Reuter, 1970, 1977) muscle, in atrial aggregates (Shrier and Clay, 1982) or in the sinoatrial node (Noma and Irisawa, 1976). IK also gradually 2 disappears with increasing age in ventricular aggregates fabricated from tissue taken between day 7 and day 12 of development (Clay and Shrier, 198lb). Thus, while IK is not found in quiescent tissues, 2 it is also not necessarily found in spontaneously oscillating tissues. Finally, within the last three years, evidence has accumulated suggesting that the depolarization-activated IK does 2 not exist in Purkinje fibre, and is to be replaced by a hyperpolarization-activated inward current If carried by both 0 sodium and potassium ions (see DISCUSSION). (ii1) Ibg Voltage clamp steps in the pacemaker range of potentials produce an instantaneous change in current followed by a time-dependent change with an exponential time course due to activation or deactivation of IK • Part of this instantaneous 2 change in current is due to a change in the drivi.ng force for potassium ions through the IK channel, another part derives from 2 the rectification of the IK channel, and a residual part is 2 thought to be a result of alteration in the level of background 3-17 current fiow through time-independent (but voltage-dependent) 0 channels. Once IK is determined as outlined in the previous 2 ' section, its steady-state contribution, given by s (V)fK (V), can 00 2 be subtracted from the total current fiowing in the steady state during a voltage clamp step to a given voltage. In this manner, the residual current (i.e. the background current Ibg) can be determined. Partly because the reversal potential of the extracted current component is about -50 mV, it has been modelled (McAllister, Noble, and Tsien, 1975; Clay and Shrier, 1981a) as the sum of a linear inward sodium current (INa,b) and an inwardlyrectifying outward current carried mostly by potassium ions (IK ) l (3-13) 0 where (3-14) and (3-15a) where (3-lSb) and 3-18 ( 3-lSc) 0 with + p3 = 1/(l+exp(-e(V-E 3 )/kT)), (3-16a) (3-16b} { 3-16c) Figure 3-4 shows plots of INa,b' IK , 1 IK , 3 IK,._' and Ibg as functions of potential. 0 This description of the rectifying IK channel is of the same 3 fonn as that earlier used to describe the fully-activated, inwardly rectifying IK channel, but assumes four ion-selective sites 2 . instead of two, since the range of voltage over which rectification occurs is narrower for Ibg than for IK • The additional term IK '+ 2 must be added to IK since Ibg' although inwardly rectifying in the 3 pacemaker range of potentials, becomes progressively more outward as V increases to values more positive than about 4). Thus, the model used for IK ~70 mV (Fig. 3- in the aggregate is similar to 1 that used in models of Purkinje fibre (tt:All ister, Noble, and Tsien, 1975) and ventricular muscle (Beeler and Reuter, 1977), with the exception that the functional form of the equation used to model the rectifying IK component is that of Clay and 0 3 3-19 0 Figure 3-4. The background current Ibg' its components IK and 1 0 I~la,b ( Ibg = IK IK and IK 3 4 ( IK 1 1 + INa,b}, and the IK = IK 3 subcomponents + IK ) as functions of 4 voltage. 0 1 3-20 0 3121.121 I (nA) 121. 121 JNa,b 0 -3!2! .. 121 -1121121.!2! -75.121 v(mV) -5121 .. 121 Shlesinger {1977) and not that of Adrian (1969). 0 By using the IK subtraction technique outlined above, Clay 2 and Shrier {1981a) found E3 = -95 mV, E4 = -40 ~Na,b = 0.202 mV, and s3 ~S, = 0.63. §K 4 = 0.516 ~s, The value given in the caption of Fig. 11 of Clay and Shrier (198la) for N3 et 3 incorrect (Clay, personal communication). -1 is For reasons given in section (v) below, I set N3 et31 = 303 nA and change §K from 0.516 4 ~S to 0.9 ~S. As stated before, ENa = 40 mV. Preliminary voltage clamp data obtained from steps to voltages more positive than -65 mV reveals that there is a gated 0 channel that is largely responsible for the later repolarization phase of the action potential in the aggregate (Shrier and Clay, 1982). Since the fully-activated current is independent of voltage over a wide voltage range and since the current displays a single exponential time course in response to a voltage clamp step, it can be modelled by the equation ( 3-17) where I~ is the fully-activated current and the activation variable x has first order kinetics 3-21 0 • -a (1-X) + B X. X= X X 0 ( 3-18) The rate constants for the opening and closing of the gate are given by ax = 0.04 (V-10) I (1-exp(-0.1(V-10))) Bx ~ (3-19a) (3-19b) 0.01 exp(-0.1(V-10)}. These equations are the same as those employed by Shrier and Clay (1982) to describe Ix in 12-day-old aggregates, with the sole exception that (V-10) occurs in both expressions instead of (V+10). 0 This modification better describes the 7-day-old aggregate, since there is a shift in the kinetics of Ix with development (Clay, personal communication). The functional form of eqn. (3-17) is not similar to that originally used to describe the pl atea1J currents Ix and Ix in Purkinje fibre (Noble and Tsien, 1969a, 1969b; l 2 McAllister, Noble, and Tsien, 1975). However, ax and Bx have the same functional form as those of the corresponding rate constants for m and s given earlier, and were obtained by fitting the expression ( 3-20) 0 3-22 to the time constants obtained from voltage clamp experiments in 0 7-day-ol d aggregates ( Shri er and Cl ay, unpublished). In eqn • . (3-17), I~= 100 nA (Shrier and Clay, 1982}. sx, Tx and x~ Fig. 3-5 shows ax• plotted as functions of potential. Note that x and ~ Tx are very small in the pacemaker range of potentials relative to s~ (v) and Ts respectively (Fig. 3-3B,C). Current-Voltage Characteristics As mentioned earlier, the two parameters N3 et 3 - 1 and gK were 4 modified from their published values (Clay and Shrier, l981a); this was done in order to make the steady-state current-voltage (IV} characteristic curve for ITTX = ITOT- INa (Fig. 3-6A) reasonably ~ 0 close to the experimentally-measured one shown in Fig. 2 of Clay and Shrier {1981a). The difference between the two curves for !TOT and ITTX in Fjg. 3-6A is the window current INa , which has a <.10 peak value of -25.0 nA at v· =- 47.7 mV. Using the unmodified equations of Ebihara and Johnson {1980), INa has a peak value of 00 -3.7 nA at V = -40.2 mV. Since s~l throughout diastolic depolarization (see RESULTS below), Fig. 3-68 shows the effect on the IV characteristic curves of setting s = 1 at all potential_s. Note that ITOT is very small (( 5 nA in magnitude) for V ( -65 mV, and that INa (the difference 00 between ITOT and ITTX) makes a significant contribution to ITOT for c 3-23 0 Figure 3-5. Variables associated with Ix plotted as functions of voltage. A. The rate constants of activation {ax and B. The steady-state value of x (x=). 0 c. The time constant of activation (Tx}. 3-24 ~\). • 5 5121121.121 A 0 • 25 121. 121 121. 121 -1121121.121 1. 121 -5121. 121 121. 121 -5121. 121 121. 121 8 0 121. 121 -1121121. 121 1121. 121 Tx c 5. 121 (sec) 0 121. 121 -1121121.121 -5121.121 0 Figure 3-6. A. Steady state current-voltage (IV) characteristics for IK 2 , Ibg' ITOT ITTX = ITOT - INa· = IK 2 + Ibg + INa + Ix, and The contribution of Ix to ITOT is negligible (( 0.15 nA) over this range of potentials. The total current has a local minimum at V = -48.5 mV, attaining a most negative value of -20.6 nA. B. Steady state IV characteristics for IK ' Ibg' 2 =1 ITTX' and ITOT with s curve for fK (V) 2 = IK at all poten~ials. The (s=l) is a bell-shaped 2 function of potential, attaining a maximum value of about 26 nA at V = -106.5 mV. 0 3-25 3fZJ.I2l 0 I (nA) A 12l.l2l -3121.121 -1121121.121 -75.12l -512l. l2l V(mv) 3121.12l I (nA) S= 1 12l.l2l -3121.121 -100.121 -75.121 V{mv) 0 -5121.121 V ) -65 mV. 0 3. METHODS The total current flowing through the membrane of an aggregate in this partial model is given by (3-21) where Iappl represents any current injected into the aggregate from an external source. Considering the cell membrane to be a leaky capacitor, application of the equation of state for a capacitor and Kirchoff's first law leads to the equation 0 • 1 .. V = - !' 1TOP (3-22) where C is the membrane capacitance. aggregate, C = 0.023 ~F (Clay~ For a 200 ~m-diameter DeFelice, and DeHaan, 1979). Thus, the partial ionic model of the aggregate is formally a fivedimensi anal system of coupled· ordinary differential equations . V= i 0 (V,m,h,s,x) (3-23a) (3-23b) c 3-26 .= f 0 h 2 (V,h} ( 3-23c) ( 3-23d) ( 3-23d) where the f; are nonlinear functions obtained from eqns. (3-1) to ( 3-22) above. To carry out the simulation, eqn. (3-23) was numerically integrated in double precision {16.3-16.6 significant decimal digits) using a fixed time-step implementation of the algorithm due to Rush and Larsen {1978). 0 The program used was modified from one kindly supplied by David Clapham. was 50 ~sec in most instances; a step size of 10 used on occasion. s = 1, x = The iteration step size employed ~sec or 1 ~sec Initial conditions of V = -70 mV, m = 0, h was = 1, 1 were set, and the integration carried out until the threshold voltage {arbitrarily taken as V = -55 mV) was attained on the upstroke of the action potential. The initial conditions approximate the values expected for a complete limit cycle model once the steady state of oscillation is reached (e.g. McAllister, Noble, and Tsien, 1975}. 3-27 see 4. RESULTS (i) Analysis of Spontaneous Activity 0 Figure 3-7A shows the transmembrane voltage computed during unperturbed activity in the model. There is an initial rapid · repolarization from the starting voltage of -70 mV to an -97.3 mV. ~1DP of This is followed by a slower diastolic depolarization to Thus, the intrinsic interbeat interval is 0.524 sec, threshold. assuming that the time from up stroke to -70 mV on the repol ari zi ng 1 imb of the action potential is 0.150 sec. Figure 3-7B shows the total ionic current flowing through the membrane. The initial rapidly declining outward current, which is responsible for the repolarization from -70 mV to t1JP, is due to a fast deactivation of Ix (Fig. 3-7C,G). Th~ time constant -rx is very small during this part of the simulation(..-; 0.04 sec). During. diastolic depolarization, both INa {Fig. 3-7E) and Ix (Fig. 3-7C) are effectively zero. The magnitudes of IK and Ibg decline in a 2 roughly parallel fashion, thus maintaining a small net inward current IK + Ibg that is responsible for diastolic depolarization 2 (Fig. 3-70). In the final stages of phase 4, the total current rapidly increases in the inward direction (Fig. 3-7B) due to activation of INa (Fig. 3-7E,J). Throughout most of the period of diastolic depolarization, the activation variables for IK continually declines (since 2 s 0 < S00 } , attaining a minimum value of about 0.7 (Fig. 3-7H). This deactivation contributes to the shutting down of IK shown in Fig. 2 3-28 0 Figure 3-7. Spontaneous unperturbed activity of several variables in the model as functions of time. starts at t = o. F. V G. x 0 H. s ,s 00 I. T s J. m,h. Throughout this simulation, 'm< .13 msec and 'h < 28.2 msec. Voltages are in mV, currents in nA, and times and time constants in sec. size 0 The simulation = 50 ]..lSec. 3-29 Integration step sa. a 0 sa. a A V -zs.a V -zs. a -1aa. a -1aa. a 1aa. a 1. 11 8 ITOT a. a X -1aa. a taa. a lx 0 a 1. Gl sa. m • !!I ra. ra sra. a. a 11 2. 11 m. m 1. 121 -sa.11 m. ra G .s 121. c F 121. 121 1. 121 E J h INa -sa. 11 .s -1aa.a L 0 a. a • :Z!!I t .!!I ra. a • :zs t .!I 3-70. 0 However, a large part of the decline in IK is due to the 2 ongoing inward rectification that takes place in the I!( channel '2 (Fig. 3-68) as the membrane depolarizes from MOP to threshold. Late in phase 4, both IK and Ibg approach zero, as does their sum 2 IK + Ibg· Thus, in this model, the properties of IK and Ibg are 2 2 such that the sum IK + Ibg just suffices to bring the membrane 2 into the voltage range for activation of INa· Both IK and Ibg are 2 fully turned off by the time of the action potential upstroke. However, by the time of the next MOP, both currents are back to their maximal amplitudes (with s ~ 1, s~ ~ 0, Ts ~~sec), ready to decline together once again in the subsequent phase of diastolic depolarization. 0 (11) The Effect of a Current Pulse on the Voltage Waveform Figure 3-8 (left panel) shows the effect on the transmembrane potential difference of delivering a 20 msec duration, 20 nA amplitude current pulse at various phases of the spontaneous cycle. A 20 nA current pulse in a·200 ~m-diameter aggregate theoretically has the same effect as a 2. 5 nA pulse ·r n a 100 urn diameter aggregate, since membrane surface area increases with the cube of aggregate diameter (Cl ay, DeFel ice, and DeHaan, 1979). simulation starts at t The = 0. A pulse delivered before t = 6 msec produces a prolongation in interbeat interval; the same pulse delivered later than t = 6 msec produces an abbreviation of interbeat interval. 0 Late in the cycle, the latency from the stimulus to upstroke decreases with increasing coupling interval 3-30 Figure 3~8. Effect of delivery of a 20 msec duration depolarizing current pulse on the spontaneous ac ti vi ty of the model. The amplitude of the current pulse is 20 nA in the left panel, 60 nA in the right panel. The top trace in each panel shows unperturbed activity; the numbers to the left of the other traces indicate the 0 time at which the current pulse is injected. point t = 0 corresponds to the start of each . simulation. aggregate. 0 The The model is for a 200 JJm-diameter Integration step size = 50 11sec. 3-31 20 nA 0 -·~· 60nA [ -2~ a _::~~~~ •~ O(msec)L , , 20 40 60 0 80 100 150 200 300 I 1 ) -lam. ~ ..... . f ~ ..... . f 1.5 J u-.--,-, f 1 I , , , 1.0 I o 1 Le... ,, .. 2.0 f . ~ 2.5 J ... , .. l .L .. .. , L I ..... . .~ J [ -··~···· • 25 t(sec) .5 ~,, 3.0 1 Ill O.O(msec)l, ' 0.5 m. m 0 (mV)L • . o .~ , ' f f f '~' .-/ •~! t'\-:' f ~ ' I / .. ~.. r __.-/ ~=·~~· b:. ' ' '~ ' ~ r ~:11~'' b.. ~ ... , 3.5 V. ........ . 4.0 f, V' ' ra. a I I I I • 25 t{sec) I I I I • 5 tc. 0 Due to the increase in membrane slope resistance with increasing diastolic time, the change in voltage that occurs during the current pulse increases with increasing tc. Increase in the current pulse amplitude to 60 nA (7.5 nA in the model of a 100 ~m-diameter aggregate) results in larger maximal prolongation and abbreviation, and a less gradual transition from prolongation to abbreviation of cycle length as tc is increased (Fig. 3-8, right panel). The range of tc over which the transition from maximal prolongation to maximal shortening takes place is about 2 msec wide at this amplitude, as compared with about 70 msec at a pulse amplitude of 20 nA (Fig. 3-8, left panel). The transition from prolongation of cycle length to abbreviation of cycle length is continuous at an amplitude of 20 nA. 0 Further increase to 80 nA (10 nA in a 100 to a transition that occurs within 1 ~sec ~m aggregate) leads or less (Fig. 3-9A). Figure 3-98 shows the trajectories in the current-voltage {IV) phase plane. Thus, during the current pulse, both trajectories follow an almost identical route; however, soon after the stimulus is turned off, the two trajectories go their separate ways. Figure 3-10 summarizes the effect on interbeat interval of a current pulse, showing plots of the normalized perturbed interbeat T1 /T 0 as a function of the normalized coupling interval tc/T 0 for three different current amplitudes. Since the time from the upstroke of the action potential to -70 mV on the repolarizing limb of the action potential is assumed to be 0.150 sec, T0 = 0.150 + 0.374 = 0.524 sec. 3-32 Q Figure 3-9. A. Effect on the voltage wavefonn produced by an 80 nA amplitude, 20 msec duration current pulse. For the trace labelled I, the current pulse is turned on at t = 0.770 labelled II, at t = msec; for the trace 0.771 msec. Thus, a change in tc of 1 JJSec can produce remarkably different behaviour at this stimulus intensity. B. Trajectories in the current-voltage phase plane for the two trials shown in panel A above. The point labelled a is the·starting point (V= -70 mV}. 0 The trajectory moves from b to c when the current pulse is switched on, from c to d during the current pulse, and from d toe when the pulse is switched off. From e, the trajectories evolve in two quite different directions: trajectory I (pulse applied at t = 0. 770 msec), trajectory II (pulse applied at t = 0.771 msec}. The two trajectories overlie each other from a toe on a diagram of this seal e. The simulations in both panel A and panel B were terminated at V = -40 mV. 3-33 Integration step size 0 A -25. 121 V (m V) -1121121.121 0. 121 • 25 . 5 t(s ec) B 10121. 0 lror (nA) 1 121. 0 -11210.0 -11210. 121 -75. l2l V(mV) -512l. l2l 0 Figure 3-10. Phase-resetting data plotted for three different pulse amplitudes: nA ( 0 ). with T0 10 nA ( 0 ) , 20 nA ( 6 ) , and 80 T1 /T 0 is plotted as a function of t/T = 0.524 sec. T1 is found by adding 0 , 0.150 sec to the time computed for threshold to be attained, since the simulation is started with V= -70 mV. Similarly,\ is computed by add.ing 0.150 sec to the time t at which the current pulse is injected. Points are computed every 50 msec for tc ) 250 msec, and every 10 msec for 150 msec " \ <: 250 msec. In addition, for the 20 nA curve, points are computed every 2 msec for 150 msec tc ..: 160 msec, in order <; to illustrate the continuity of that curve. Integration step size = 50 ~sec. 3-34 0 0 ~ 1 ------- ----·---- ----------- -- ---- --- - ---------- --.- 0000 0 00 ~ ----------- .... 0 0 0 D. 0 0 D. 0 0 0 0 0 o. 0 0 0 10 20 nA 80 1 .o (111) Analysis of the Response to Current Pulse Perturbation Figure 3-8 showed the effect of delivering a 20 nA amplitude, 20 msec duration current pulse at several coupling intervals. effect of a pulse delivered at t The = 0 msec (i.e. tc = 150 msec) is to prolong the i nterbeat interval beyond control (Fig. 3-llA}. Figure 3-llB shows that during the current pulse, !TOT is more inward than in control (Fig. 3-7B), so that V falls more slowly In fact, during the latter half of the pulse, the than in control. current is inward and the membrane depolarizes. Since the final phase of repolarization proceeds more slowly, and since rx is very small during this time (tx = 5 msec), Ix is maintained at a higher outward level for a longer period of time (Fig. 3-llC) than in control (Fig. 3-7C). Thus, when the current pulse is switched off, Ix is still non-zero (Fig. 3-llB) in contrast to the control case ( Fi g • 3- 7B) • The sum IK + Ibg is also outward immediately after the 2 current pulse ends (Fig. 3-110) in contrast to control (Fig. 3-70), so that ITOT is outward (Fig. 3-llB). This is in part due to a slight decrease in the rate at which s declines (Fig. 3-llH), which in turn is due to the doubling in r 5 from about 0.5 sec to 1.0 sec produced by the perturbation (Fig. 3-7!, Fig. 3-lli). However, the fact that IK + Ibg is outward is primarily a consequence of the 2 fact that the change in membrane voltage produced by the current pulse causes a bigger decrease in the amplitude of Ibg than in IK 2 0 3-35 0 Figure 3-11. The time course of several variables in the model in response to a 20 nA amplitude, 20 msec duration depolarizing current pulse delivered at t =0 msec. The arrows at the bottom of panels A and F indicate the time during which current is being injected. A. V B. ITOT C. Ix F. V G. H. s, X D. IKz' Ibg' IK +Ibg E. INa r. Ts J. m,h. 2 s~ L . Voltages are in mV, currents in nA, and times and time constants in sec. 0 Integration step size 3-36 = 50 ~sec. Sill. Ill 0 5111. Gl A V V -zs.121 -1111(11. Ill 1111121. Ill ITOT 0 t 1. Ill B X Ill. Ill G • !5 Ill. Ill -lalll. a lx -25.111 -1111111.111 t F c laa. Ill !1111. a • !I a. 111 Ill. Gl sa. a 1. Gl 2. 11.1 D Ts Ill. Gl 1. 11.1 a. -sa. Ill Ill. Ill 111 1. Ill E INa -!1111. 121 J • 5 m -lllla. 13 0 a. Ill. 13 111 • zs t • 5 Ill. liiJ . ;zs t .s {Fig. 3-110). This is because of the different degree of rectification present in the two currents (Fig. 3-68}. The effect of the current pulse is thus to reset the activity backwards in time: the latter half of each curve in Fig. 3-11 is virtually superimposable with the final part of the corresponding curve in Fig. 3-7. Figure 3-12 shows the effect of delivering a 20 nA amplitude, 20 msec duration current pulse at a coupling interval of 170 msec. This leads to a shortening of the interbeat interval {Fig. 3-12A). The effect of the current pulse is basically to charge the membrane capacitance in a linear fashion (Fig. 3-12A), since the stimulus current is much larger than the ionic membrane current {Fig. 312B). The perturbation has little or no effect on Ix (Fig. 3-12C) or its activation variable x (Fig. 3-12G). At the end of the current pulse, the sum IK + 1bg is not 2 much different from the control value at that time, since IK and 2 Ibg change by about equal amounts during the current pulse·(Fig. 3120}. Thus dV/dt is about the same in both instances {Figs. 3-7A and 3-12A). However, -r 5 (Fig. 3-121) has been reset to a higher value {about 1.3 sec) than in control (about 0.6 sec). As the membrane depol arfzes, -r 5 (Fig. 3-12!) proceeds along the falling limb of its bell-shaped curve (Fig. 3-3C) instead of along the ascending limb as in control (Fig. 3-71). Thus, immediately after the end of the current pulse, s falls more slowly than in control, aided by the fact that s~ is higher at the end of the current pulse than in control (Fig. 3-121, Fig. 3-7!, eqn. (3-10)). 0 3-37 0 Figure 3-12. The time course of several variables in the model in response to a 20 nA amplitude, 20 msec duration depolarizing current pulse delivered at t = 20 msec. The arrows at the bottom of panels A and F indicate the time during which current is being injected. 0 A. V B. 1TOT . C. IX F. V G. X H. s, s D. IK ' I bg ' I K2 +I bg E. IN a 2 CO I. T s J. m, h. Vol tages are in mV, currents in nA, and times and time constants in sec. ]..I 0 Integration step size = 50 Sec. 3-38 5111.11.1 0 V S0.1lJ A V -2S.IlJ -lliJS. 11.1 1011.1. Ill lror t 0 t 1. 11.1 B X -lllJ0.11.1 lx -2s.0 -111.111.1. 11.1 11.1. llJ lllJS.II F .s Ill. 11.1 c G 1.'------ 1. 11.1 SS.IlJ .s a. a 11.1. 0 38.11.1 2. 11 Ts 11. 11 I 1. 11.1 lbg -as. 11 111.11.1 11. s 1. E s .s INo -ss. s m a. s -1011. 11 0 a. a • 2!5 t .!5 s. s • 2!5 t .s At first sight, one would think that resetti n~· 0 ; to a higher value should produce an increase in outward current and so lead to a prolongation and not an abbreviation of the interbeat interval. However, once again the rectification properties of the IK and Ibg 2 channels appear to be more important than the time-dependent behaviour of the IK channel. The effect of the pulse is to 2 advance the activity: most of the traces in Fig. 3-12 for the time following the delivery of the current pulse are virtually superimposable with the final parts of the corresponding traces during unperturbed activity (Fig. 3-7). 5. DISCUSSION (f) Spontaneous Activity Figure 3-7 shows that diastolic depolarization is not simply due to a decrease in the outward time-dependent pacemaker current IK , but is in fact due to the simultaneous decrease in the 2 magnitudes of two large (in comparison to ITOT) oppositely directed currents, IK and Ibg· A similar situation occurs in the MNT model 2 for Purkinje fibre (Me All i ster, Noble, and Tsi en, 1975}. Furthermore, one cannot say that the decrease in IK is due to a 2 progressive deactivation of that current (i.e. decrease in the activation variables) as diastolic depolarization proceeds. In fact, IK continues to fall in late diastole after the minimum of s 2 0 3-39 is attained, when s is actually increasing (Fig. 3-70,H). 0 In th'is model, s falls from a maximum of about 1.0 at MOP to a minimum of about 0.7, as it does in the MNT model. the aggregate, IK However, in the model of falls by a factor of about 7 during this time 2 (Fig. 3-70}. Thus, it is the rectification properties of IK and 2 Ibg that play a major role in detennining the evolution of the transmembrane voltage during the phase of diastolic depolarization in this model of the aggregate. model, where the fall in IK This is in contrast to the MNT is largely due to and occurs in a 2 fashion parallel with the deactivation of IK (McAllister, Noble, 2 and Tsien, 1975). The magnitudes of IK 2 , Ibg' and IK 2 + Ibg decline as pacemaker depolarization progresses (Fig. 3-70). 0 In fact, the sum IK + Ibg is close to zero towards the final part of phase 4 {Fig. 2 3-70) • Thus, changes in any of the parameters contra11 i ng IK or. 2 Ibg that lead to an increase in IK 2 or decrease in Ibg of only a few nanoamperes will abolish spontaneous phase 4 depolarization. For example, increase of gK in eqn. {3-lSc) from 0.9 ~s to 1.2 ~s 4 abolishes spontaneous activity. Injection of a constant hyperpolarizing current of 4.3 nA {i.e. Iappl = 4.3 nA in eqn. (321)) also stops activity. Injection of a constant hyperpolarizing current of this magnitude will also stop spontaneous action potential generation in the aggregate (Guevara, unpublished). 3-40 0 As pacemaker depolarization enters its final phase, the currents Ibg and IK turn themselves off, since their sum IK 2 + 2 Ibg serves to depolarize the membrane increasingly into the region of rectification of both channels (Fig. 3-6B). As the sum Ibg + IK 2 approaches zero, the depolarization process is taken over by INa (Fig. 3-7E,J}. Thus, as in the MNT model, the final part of pacemaker depolarization is due to the window current INa • This CO can be seen in experiments on aggregates in which TTX is added to the bathing medium: there is a slowing of beat rate due to prolongation of the duration of phase 4 (Colizza, Guevara, and Shrier, 1983), presumably due to a decrease in the window current. Decrease of gNa in eqn. (3-1) to 44 llS abolished spontaneous activity. Thus, in surrmary, it appears that the three currents INa, IK , 2 and Ibg control the duration of phase 4 in the aggregate. the window current Isi However, may also play a role, since there is also a CO slowing in the beat rate of an aggregate when the slow inward channel blocker 0600 is applied, and since the window current Isi "" is approximately ~ of the window current INa in the MNT model of 01) Purkinje fibre in the pacemaker range of potential s. The arrangement in which diastolic depolarization, even in physiological preparations such as Purkinje fibre and the 0 3-41 sinoatrial node, appears to be due to the decline in magnitude of 0 two large currents does not seem to be a very economical one. An alternate mechanism would be to have, for example, a single much smaller outward current which declines in value throughout phase 4. This would decrease the number of channe1 s needed in the membrane, as well as significantly decrease the load on the Na+-K+ pump. One theoretical reason for having two 1arge currents has to do with control of spontaneous activity. Many target organs in the body (e.g. the heart} have a reciprocal sympathetic-parasympathetic innervation with high levels of resting tone in both branches of the autonomic system. Alteration in the state of the system is accomplished by simultaneously decreasing the level of activity in one branch and increasing it in the other. I suggest that there are many advantages to this mode of control, analogous to the advantages seen when differential input rather than single-ended input is used in electronic amplifiers. The two large currents that flow during diastolic depolarization in the aggregate may be analogous in some way to the two high resting levels of neural tone. A neurotransmitter or circulating hormone may have simultaneous effects on more than one current. For example, acetylcholine both increases the outward potassium current and decreases the inward sodium-calcium current in the sinoatrial node. These two simultaneous changes work in concert to slow the spontaneous beat rate; there is however a "differential amplifier.. c 3-42 kind of effects in that the "non-inverting" or "excitatory" input is being decreased, and the,. inverting" or u inhibitoryu input increased. There has recently been a re-interpretation of the results of the original voltage clamp experiments carried out in the pacemaker potential range in Purkinje fibre. These results were originally thought to indicate the presence of a depolarization-activated outward current IK carried by potassium ions (Dudel and Trautweins 2 1958; Trautwein and Kassebaum, 1961; Dudel et al., 1967; 1969). Deck and Trautwein, 1964; Noble and Tsien, 1968; Peper and Trautwein, Recent work has indicated that this time-dependent pacemaker current may actually be a hyperpol ari zati on-activated inward current that is carried by both potassium and sodium ions DiFrancesco and Noble, 1982a, (DiFrancesco, 1980, 1981a, 198lb; 1982b}. This current is termed If, and is in sone respects similar to the hyperpolarization-activated inward'current If (also called ih) found in the sinoatrial node (Brown, DiFrancesco, and Noble, 1979; DiFrancesco and Ojeda, 1980; Yanagihara and Irisawa, 1980; Irisawa and Noma, 1982). Arguments can be made both for and against the reinterpretation of IK in terms of If in Purkinje fibre (e.g. Cohen, Falk, and 2 Kline, 1982). However, algebraic computation (DiFrancesco, 198la) and numerical simulation (DiFrancesco and Noble, 1982b) suggest that as far as the voltage and total current waveforms are concerned, it does not matter whether the IK. or the If description is used. 2 Thus, reformulation of the description of IK given above 2 0 3-43 in eqns. (3-6) to {3-12) in terms of If should not change the 0 phase-resetting behaviour of this model of the aggregate. things presumably changed are the deta~ls The only of how the currents IK 2 (or If) and Ibg are involved. The mechanism of spontaneous activity in the model of the aggregate is quite different fr·om the mechanism demonstrable in two similar models of the sinoatrial node (Yanagihara, Noma, and Irisawa, 1980; Irisawa and Noma, 1982). Three currents change significantly during the pacemaker depolarization of the sinoatrial node: the current is, the slow inward current, which is largely responsible for the action potential upstroke; the current it' which is a time-independent leakage current of unknown origin; and the current iK' which is a potassium-ion current responsible for the repolarization phase of the action potential. During pacemaker depolarization in these sinoatrial node models, there is a progressive movement of is in a more inward direction, and a progressive movement of ii in a more outward direction. There is a1 so a small decline in the magnitude of \, which remains outward throughout phase 4. Note that ih (:If) is not involved in generating spontaneous phase 4 depolarization. An alternative, less-physiologically based ionic model of the sinoatrial node has a pacemaker potential that is generated by an IK type of mechanism 2 similar to that involved in the pacemaker phase of Purkinje fibre or of the aggregate (Bristow and Clark, 1982). Periodic action potential generation can be elicited in models 0 3-44 of quiescent ventricular myocardium by inj ecti nq a constant 0 depolarizing current (Beeler and Reuter, 1977). Since this bias current is much 1 arger than the sum of the membrane-generated currents during diastole, there is a linear phase 4 depolarization until threshold is attained. Thus, the mechanism is similar to that involved in action potential generation in the Hodgkin-Huxley model for squid axon subjected to a constant depolarizing bias current. Another case in which spontaneous phase 4 depolarization can take place without the primary involvement of a time-dependent current is in an ionic model of atrial aggregates {Shrier and Cl ay, 1982). Atrial aggregates fabricated from 10- to 14-day-old embryos do not demonstrate a time-dependent change in current fall owing a 0 voltage clamp step made in the pacemaker range of potentials. Thus, IK or If is not present; there is only a background current 2 apparent. Yet, there is a slow pacemaker depolarization to threshold; the rate at which this depolarization proceeds is largely controlled by the membrane time constant {Shrier and Clay, 1982). Thus, there are several different arrangements of ionic currents that can 1ead to periodic activity in cardiac cells. The mechanisms can be quite different from system to system as illustrated above. However, the periodic activity is presumably due to the presence of a 1 imit cycle in the phase space of the set of differential equations describing the system. 0 3-45 Ionic modelling indi'cates· that this limit cycle can arise from one or more 0 bifurcations as a parameter in the system is changed (Guevara, unpublished). In the latter case, a sequence of complex periodic and aperiodic phenomena are seen, which are analogous to those seen experimentally. (ii) Comparison of the Phase-Resetting Behaviour of the Model with Experiment The phase-resetting response of the ionic model shown in Figs. 3-8 and 3-9A and summarized in Fig. 3-10 is similar to the experimental response detailed earlier in CHAPTER 2 in the fall owing ways. (i) The response is biphasic: an early depolarizing pulse prolongs the cycle length; a late depolarizing pulse shortens the cycle length. (ii) As stimulus intensity increases, the maximum possible prolongation and abbreviation increase, the neutral point at which T1/T 0 equals unity moves to a small er coupling interval, and the transition from prolongation to shortening becomes more abrupt. (iii) The transition from prolongation to shortening can occur with a change in coupling interval of less than 1 0 3-46 msec at a sufficiently high pulse amplitude (e.g. 0 compare Fig. 3-9A with Fig. 2-16 or Fig~ 2-17). Moreover, the transition occurs at about the time of MOP in both instances. (iv) The current amplitude needed to elicit a certain behaviour in the model is in the same range as that seen experimentally, keeping in mind that the model is for a 200 ~m-diameter aggregate, and that aggregate membrane area increases as the cube of the aggregate diameter. The most notable discrepancy between the experimental and modelling results is at lower pulse amplitudes. Experimentally, prolongation is produced at small amplitudes when the slope of the pacemaker potential is reset to a value less than in control {e.g. Fig. 2-5, middle panel). The model does not respond to a small current pulse with a change in the slope of diastolic depolarization (Fig. 3-8, left panel}. that if ~ight It can be however shown changes in some of the parameters in the model are made (which leave the steady-state IV and the currents IK and Ibg 2 within the range of experimentally determined values}, there results a more linear pacemaker potential that does tend to respond to a small current pulse with a change in its slope (Clay, Guevara, and Shrier, 1984}. However, even with these modifications, c prolongations of at most 10% are produced. 3-47 This does not agree· with the experimental results, where prolongations 0 {i.e. 1.2 ( T1/T 0 ( of~ 20-30% 1.3) can be routinely obtained in all aggregates (Fig. 2-10A,B). Another discrepancy is that the long delays shown in Figs. 211 to 2-14 are not reproduced by the model. However, this is not unexpected, since long delays were only seen in more slowly beating aggregates, and the model investigated in this chapter is one for a faster-beating preparation with an intrinsic period of 524 msec. It can be shown that oscillatory activity in the pacemaker range of potentials in response to a current pulse can be produced in a model differing slightly from the one developed in this chapter (Clay, Guevara, and Shrier, 1984). This involves shifting the IV characteristic of ITOT shown in Fig. 3-6A in the hyperpolarizing direction so as to sl 0~1 the beat rate; the modification creates an equilibrium point lying in the pacemaker range of potentials. ~1oreover, since this equilibrium point is unstable, spontaneous . activity cannot be annihilated with a single current pulse. 11 This Saddle-node bifurcation also results in the creation of an 11 additional equilibrium point which is a saddle-point {Fig. 2-24). (iii} Ionic Mechanisms Involved in Phase Resetting The above analysis of a partial model shows that the response to a current pulse is very complex, with ,the membrane voltage and the various currents feeding back on one other in a very complicated manner. However, I will now summarize what I see the c 3-48 main mechanisms of phase resetting to be ·as revealed by study of c ionic models of the aggregate. For a very small current amplitude, essentially only IK and 2 Ibg are involved in the phase-resetting behaviour. because: This is {i} the pulse amplitude is too small in comparison to Ix to affect events during the repolarizing limb of the action potential ; and ( i i} the pulse amplitude is too small to bring the membrane potential into the voltage range for activation of INa except for times very late in the cycle. There is a resetting of the activity of IK and Ibg forward or backward in time, producing 2 abbreviation or prolongation of cycle length respectively (Figs. 311, 3-12). This behaviour hinges upon the rectifying properties of the IK and Ibg channels. 2 0 The maximum prolongation and abbreviation of cycle length seen will be quite small at this stimulus intensity. At the other extreme, for a very large current pulse, essentially only INa and Ix are involved in the phase resetting behaviour. If such a pulse is delivered after r~DP occurs, it will depolarize the membrane into the range of activation of INa; the membrane will immediately come to threshold and there will be an abrupt shortening of the cycle length. If the pulse is delivered not too much earlier than the time of occurrence of MOP, the same kind of behaviour will be seen. If however the pulse is delivered even earlier, Ix will still be sufficiently 1arge so as to nullify most of the depolarizing effect of the pulse; repolarization will continue at a reduced rate, producing an increase in action 3-49 potential duration, a more positive t>10P, and .an increased cycle 1 ength. At intermediate current amplitudes, all four currents come into play. For example, a pulse delivered at about the time of MOP (when Ix is quite small) can affect Ix, .rK 2 and Ibg· The same pulse delivered later in the cycle can affect IK and Ibg' as well 2 as INa to a lesser extent. (iv) Discontinuity in the Phase-Resetting Response There is a coupling interval somewhere just before r1DP occurs where a large current pulse will take the membrane voltage very close to the threshold voltage. 0 Due to the very fast rate at which INa becomes activated in this range of potential s {Fig. 3-2A) and the relatively large value of the conductance gNa' one of two behaviours will result (Fig. 3-9A}. The behaviour is very delicate: for some time immediately following termination of the current pulse, the difference in potential between traces I and II in Fig. 3-9A is only 7 ~v. I have not investigated whether the response shown in Fig. 39A is continuous; that is, whether delivery of pulses with 150.770 msec ~ tc ~ 150.771 msec would produce action potentials with upstrokes falling in between those of the two traces shown in Fig. 3-9A. However, in the Hodgkin-Huxley model for squid axon, it has been shown that a response presumably equivalent to that shown in c 3-50 Fig. 0 3-9A is in fact continuous. In that case, the voltage arrived at following a stimulus must be changed in increments of 1 femtovolt in order to reveal the continuity (Clay, 1977). kind of "quasi-threshold phenomena~ This (FitzHugh, 1955) cannot be easily investigated numerically on a computer that has at most only 16-17 significant decimal digits. The equilibrium point lying in the plateau range of potential s may have some sort of saddl e-1 ike geometry associated with it. If such be the case, then there might indeed be a true all-or-none phenomenon in this model {FitzHuqh 1955, 1960). While the response shown in Fig. 3-9A is probably continuous, in real life the situation is somewhat different (Figs. 2-16, 2.- 17). We know that the Hodgkin-Huxley approach to the description of ionic currents {formulation as a system of continuous ordinary differential equations) is incorrect. A more precise description of ionic currents is a stochastic one in which currents arise from the random openings and closings of a population of single channels (e.g. Clay and DeFelice, 1983). A single ionic channel passes a current of roughly 2 pA and opens for about 5 msec. A single opening and closing of this channel thus transfers a charge of 10 fC across the membrane. For a 200 capacitance of 0.023 this results in a change in the membrane potential of ~F, about~ ~V. ~m-diameter aggregate with a Thus, the difference in the two responses shown in Fig. 2-9A can be produced if 14 channels open in one case during the pulse and do not open in the other case. Intermediate responses would only be seen if fewer than 14 channels change their 0 3-51 activity during the current pulse. 0 Since many thousands of channels open and close during the current pulse, this eventuality is unlikely. Even if less than these 14 critical channels open or close during the pulse so that the voltage at the end of the pulse changes by less than 7 uV, there is no guarantee that the intermediate responses theoretically predicted in continuous models will be found. Thus, only the two behaviours shown in Fig. 3-9A will be seen experimentally. the s~ngle This situation will be reinforced if channels do not operate independently of each other; that is, if there is some form of cooperative phenomenon in the membrane {Changeux et al., 1967; Carnay and Tasaki, 1971). The perturbed cycle lengths of traces I and II in Fig. differ by about 0.8 of the spontaneous cycle length. 0 3~9 Thus, type 0 phase resetting is not present, unless a further phase shift difference of 0.2 can be produced during the post-stimulus cycles. This eventuality cannot be tested, since the partial model does not generate repetitive activity. In summary, combined experimental and modelling work indicates that the action potential in the spontaneously beating aggregate at 35°C is indeed all-or-none, provided that stimulation is carried out in a narrow range of potentials near MOP. (v) Suggestions for Future Ionic Modelling Work The partial model for the electrical activity of the aggregate 0 3-52 developed and investigated in this chapter is a preliminary one in many respects. currents. There appear to be problems with all of the These problems were revealed by comparing the predicted with the experimental phase-resetting response, and incidentally illustrate the value of using the phase-resetting behaviour of an ionic model as a test of its validity. (i) INa: The issues are as follows: The problem with rm was mentioned earlier. The modification made to am is not entirely satisfactory, since it produces a poorer fit to the 'm data than the original equation of Ebihara and Johnson (1980). Perhaps modifications to Sm, am, or an should be made instead. 0 ( ii) IK , Ibg: The model does not account very well for the 2 response to a small amplitude pulse which leaves the membrane potential in the pacemaker range of potential s. In particular, prol ongati ons of cycle length of more than about 10% are not produced. This probably indicates a problem with IK or Ibg (or both). 2 (iii) Ix: The formulation of Ix may need to be modified, si nee recent work shows that Ix in the atrial aggregate actually has two components, I xl and I x2 , which are activated in two different ranges of potential c and Shrier, unpublished}. 3-53 (Cl~y The major deficiency of the model is that, being an incomplete 0 model, it does not produce cyclic activity. At a minimum, I s.~ would have to be added to generate realistic cyclic activity. In such a full model, graded action potentials (Fig. 2-7) would presumably be produced in response to an early high-amplitude depolarizing stimulus. Furthermore, the change in the form of the PTC could be investigated as a function of stimulus strength. example, it ~ould For qe possible to see if the model would produce the sequence {type 1 (monotonic) +type 1 (non-monotonic) +type a} in PTC~ 23). as the amplitude of a depolarizing pulse is increased (Fig. 2This sequence is seen in the MNT model for Purkinje fibre which, being a complete limit cycle model, generates cyclic activity (Guevara, unpublished: see CHAPTER 6}. There has been only one systematic investigation published to date concerning the phase-resetting properties of an ionic model of cardiac tissue (Bristow and Clark, 1982). Although the ionic mechanisms of spontaneous activity and thus of phase-resetting may be different in different cardiac oscillators, the topological properties of the phase-resetting should be similar. Only systematic investigation of several different ionic models can test this hypothesis. 3-54 0 0 0 ,......,.. '-"'