Potential clinical relevance of the ‘little brain’ on the mammalian heart


  • J. A. Armour

    1. Centre de recherche, Hôpital du Sacré-Cœur de Montréal and Department of Pharmacology, Faculty of Medicine, Université de Montréal, Montréal, Québec, Canada
    Search for more papers by this author

  • This article is based on the Carl Ludwig Distinguished Lecture of the American Physiological Societies, Neural Control & Autonomic Regulation Section that was delivered at the Annual Experimental Biology Congress in Washington, DC, on 29th April 2007.

Corresponding author J. A. Armour: Hôpital du Sacré-Coeur de Montréal, Research Center, 5400 Gouin Boulevard West, Montreal, QC H4J 1C5, Canada. Email: drewarmour@hotmail.com


It is hypothetized that the heart possesses a nervous system intrinsic to it that represents the final relay station for the co-ordination of regional cardiac indices. This ‘little brain’ on the heart is comprised of spatially distributed sensory (afferent), interconnecting (local circuit) and motor (adrenergic and cholinergic efferent) neurones that communicate with others in intrathoracic extracardiac ganglia, all under the tonic influence of central neuronal command and circulating catecholamines. Neurones residing from the level of the heart to the insular cortex form temporally dependent reflexes that control overlapping, spatially determined cardiac indices. The emergent properties that most of its components display depend primarily on sensory transduction of the cardiovascular milieu. It is further hypothesized that the stochastic nature of such neuronal interactions represents a stabilizing feature that matches cardiac output to normal corporal blood flow demands. Thus, with regard to cardiac disease states, one must consider not only cardiac myocyte dysfunction but also the fact that components within this neuroaxis may interact abnormally to alter myocyte function. This review emphasizes the stochastic behaviour displayed by most peripheral cardiac neurones, which appears to be a consequence of their predominant cardiac chemosensory inputs, as well as their complex functional interconnectivity. Despite our limited understanding of the whole, current data indicate that the emergent properties displayed by most neurones comprising the cardiac neuroaxis will have to be taken into consideration when contemplating the targeting of its individual components if predictable, long-term therapeutic benefits are to accrue.

Aldous Huxley. Grey Eminence: A Study in Religion & Politics. The MacMillan Company of Canada, Toronto, 1941.

‘The insolubility of a problem has never deterred men and women from confidently propounding solutions. The method adopted is always the same: that of over-simplification. Thus, all but the immediate antecedents of the event under consideration are ignored. At the same time, all embarrassing complexities are mentally abolished. The event is thus made to seem simple enough to admit explanation in terms of a very few “causes”, perhaps even only one. Not unnaturally the results are disappointing.’

There is mounting clinical interest in the autonomic nervous system, particularly with respect to its role in the genesis of cardiac arrhythmias (Armour et al. 1972, 2005; Schwartz et al. 1978; Hirose et al. 2002; Chen et al. 2006; Patterson et al. 2005; Po et al. 2006; Scherlag & Po, 2006). With regard to heart failure, a neurocardiological perspective has also gained relevance of late, subsequent to the observation that β-adrenoceptor blockade or modification of the renin–angiotensin system retards its progression (Cohn, 1990; Dell'Italia & Ardell, 2004). Although the cardiac neuroaxis has been the subject of enquiry for the better part of a century (Howell, 1915), scant attention continues to be paid in the clinical world to its complex anatomy (Chiou et al. 1997) and function (Mittal & Lerman, 2002).

Recently, it has been promulgated that the intrinsic cardiac nervous system comprises ‘three fat pads’ containing two atrial and one ventricular ganglionated plexus (Chen et al. 2006). The term ‘fat pad’ only serves to engender confusion, since fat covers the base of adult human hearts. Another recent presumption concerns the fact that somata in each intrinsic cardiac ganglionated plexus only regulate adjacent tissues (Chen et al. 2006). Many of the concepts presented in this review were derived from in situ experimentation on larger mammals (canine, feline, guinea-pig and porcine models), owing to the relatively numerous populations of neurones in their peripheral autonomic ganglia that permit determination of how neurones in various intrathoracic ganglia interact with one another in situ. For instance, these models have permitted the study of how neurones in each major intrinsic cardiac ganglionated plexus communicate with one another. Indeed, concepts derived from large animal models have helped to clarify the presupposition that neurones in the right atrial ganglionated plexus control solely the sino-atrial node while those in the inferior vena caval–inferior atrial ganglionated plexus control only the atrio-ventricular node (cf. Mittal & Lerman, 2002). Although such functional and anatomical assumptions permit the contemplation of ablating a single intrinsic cardiac ganglionated plexus to modify a specific cardiac index (Scherlag & Po, 2006), in vivo data depict a primarily stochastic hierarchy (Randall et al. 2003; Waldman et al. 2006) that defies such simplistic assumptions. For instance, despite mounting evidence to the contrary (Moravec & Moravec, 1987; Steele et al. 1994; Horackova et al. 1999), the notion still exists that the intrinsic cardiac nervous system is comprised solely of cholinergic efferent postganglionic neuronal somata (Mittal & Lerman, 2002; Chen et al. 2006). Thus, if therapy is being developed on the basis of simplistic concepts such as accentuated antagonisms (Wilber & Morton, 2002), clinical uncertainties or errors are bound to arise (Carlson et al. 1992; Cummings et al. 2004; Oh et al. 2006).

It is the purpose of this review to present an overview of current information concerning the anatomy of peripheral autonomic neurones that regulate the heart. This forms the necessary basis for study of the putative interactions that occur among peripheral (Armour, 1991; Ardell, 1994; Randall et al. 1996) and central neurones (Andresen et al. 2004; Foreman et al. 2004) in the maintenance of adequate cardiac output. That intrinsic cardiac neurones are under the tonic influence of circulating hormones (Ardell, 2004) is an issue that will not be addressed in a meaningful manner here.

Intrinsic and extrinsic cardiac control At the outset it should be stated that this review focuses on extrinsic factors that are superimposed, so to speak, on those intrinsic to cardiomyocyte control.

Intrinsic factors. The law of the vertebrate heart was devised in the late 19th century from isolated cold-blooded vertebrate hearts (Frank, 1959) and later extended by Patterson, Piper and Starling to isolated mammalian hearts (Patterson et al. 1914). This time-honoured concept implies that the adequacy of cardiac output relies primarily on the degree of cardiomyocyte diastolic stretch; the Frank–Starling hypothesis. These authors identified the fact that ventricular chamber stroke volume is a function of regional cardiac muscle stretch elicited during diastole. This index is determined to a large extent by a chamber's returning venous blood. In the explanted heart, increase in venous return to a chamber induces greater diastolic muscle stretch such that the force of the subsequent contraction increases to expel the greater volume of blood returning to the right (Wiggers, 1914) or left ventricle (Patterson et al. 1914).

Extrinsic factors. Superimposed on such cardiomyocyte behaviour is the influence by circulating hormones and chemicals released directly into the substance of the heart by autonomic efferent postganglionic nerve terminals. It is relevant to recall that the normal heart is encased in a constraining fibrous pericardial sac that not only prevents organ displacement but also restricts beat-to-beat ventricular chamber diastolic distension. Given the constraint imposed by an anatomical feature that minimizes short-term ventricular diastolic distension in situ, extrinsic factors become relevant when considering how cardiac output is matched to changing body blood flow demands in different physiological states.

The reciprocal thesis of cardiac control The cardiac nervous system has long been conceived in terms of its two major outflow branches that exert reciprocal control over cardiac indices under the sole influence of central neuronal command regulating sympathetic (adrenergic) and parasympathetic (cholinergic) efferent preganglionic neuronal activity (Kollai & Koizumi, 1979; Levy & Martin, 1979). Increased cardiac sympathetic efferent neuronal tone increases cardiac chronotropism, dromotropism and inotropism (Brodde & Zerkowski, 1994); the reverse holds for the effects exerted by medullary (parasympathetic) efferent preganglionic neurones (Levy & Martin, 1979; Blomquist et al. 1987). In such a scenario, neurones in intrathoracic extracardiac (solely sympathetic; Hillarp, 1960; Skok, 1973) and intrinsic cardiac ganglia (solely parasympathetic; Levy & Martin, 1979) act as simple efferent relay stations that function in a reciprocal manner determined by central neurones (Skok, 1973; Gebber et al. 1996; Lewis et al. 2000). When inhibitory (parasympathetic efferent) motor neurones are most active, the activity generated by augmentor (sympathetic efferent) ones becomes suppressed, and vice versa (Langley, 1921; Kuntz, 1934; Kollai & Koizumi, 1979; Levy & Martin, 1979). An ‘accelerator and brake’ thesis dependent solely upon central neuronal command, although applicable in extreme situations, may be in need of revision in most physiological states that confront one over a lifetime.

Autonomic neuronal control also resides with collections of neurones on and adjacent to the organs that they subserve Collections of neurones in ganglia associated with abdominal organs are involved in regulating such organs and are not totally subservient to central neuronal command (Cooke, 1989). Recent evidence indicates that intrathoracic extracardiac and intrinsic cardiac neurones comprise a local distributive network that processes both centripetal and centrifugal information in cardiac control. During stable cardiac states, the various components that make up the peripheral cardiac neuronal hierarchy apparently do not transfer synaptic information concerning organ function to central neurones; hence, a general lack of awareness of normal cardiac status. At the same time, their integrative function is under the tonic influence of brainstem and spinal cord neurones (Andresen et al. 2004; Foreman et al. 2004), as well as circulating hormones (Ardell, 2004).

An overview of the anatomy of these peripheral cardiac neurones is presented first as a foundation with which to discuss what is known about their putative interactions. Given our evolving understanding of the stochastic nature of these interactions, it is now apparent that a simple reciprocal analogy is no longer adequate as a foundation on which to devise strategies that target the cardiac nervous system during the evolution of heart disease.

Anatomy of peripheral cardiac neurones

Cardiac afferent neurones The function of motor neurones innervating the substance of the heart is dependent to a considerable extent on the capacity of afferent neurones located in intrathoracic, nodose and dorsal root ganglia to transduce the cardiovascular milieu. Unipolar neurones associated with cardiac sensory neurites have been identified by anatomical means throughout both the nodose ganglia (Hopkins & Armour, 1989) and the dorsal root ganglia (Brown, 1967) from the C7 to T4 levels of the spinal cord (Vance & Bowker, 1983; Hopkins & Armour, 1989). They have also been identified in intrathoracic extracardiac (Armour, 1986b; Bosnjak & Kampine, 1989; Horackova et al. 1996) and intrinsic cardiac ganglia (Ardell et al. 1991; Yuan et al. 1994; Armour et al. 1997; Horackova et al. 1999; Arora et al. 2003). Such neurones display multiple chemical markers.

Cardiac efferent neurones Cardiac motor neurones, when activated, influence heart rate and atrio-ventricular nodal conduction, as well as atrial and ventricular inotropism (Levy & Martin, 1979; Randall et al. 1996).

Parasympathetic efferent neurones. Cardiac parasympathetic efferent preganglionic neuronal somata have been identified by both anatomical (Hopkins & Armour, 1984) and functional means (McAllen & Spyer, 1976) to be located primarily, but by no means exclusively, in the ventral lateral region of the nucleus ambiguus. Lesser numbers are located in the dorsal motor nucleus and the intermediate zone between these two medullary nuclei (Hopkins & Armour, 1984; Cheng et al. 1999). It has been proposed that cardiac motor neurones in the dorsal motor nucleus may be primarily concerned with regulating cardiac inotropism, while those in the nucleus ambiguus may be primarily concerned with heart rate (Gatti et al. 1995). These somata project axons to parasympathetic efferent postganglionic neurones located throughout the various atrial or ventricular ganglionated plexuses (Plecha et al. 1988).

Sympathetic efferent neurones. Spinal cord sympathetic efferent preganglionic neurones project axons via the T1–T5 rami to synapse with sympathetic efferent postganglionic cardiac neurones (Norris et al. 1977) that are located throughout both the superior and middle cervical ganglia, as well as the cranial poles of stellate ganglia. They are also located in mediastinal ganglia that lie adjacent to the heart (Hopkins & Armour, 1984), as well as each major intrinsic cardiac ganglionated plexus (Moravec & Moravec, 1987; Forsgren et al. 1990; Horackova et al. 1999).

Sympathetic postganglionic somata in each intrinsic cardiac ganglionated plexus project axons to widespread regions of the heart (Hopkins et al. 1984). Some adrenergic somata even project two axons via different cardiopulmonary nerves to innervate divergent cardiac regions. Such an anatomical arrangement ensures that the somata of sympathetic efferent postganglionic neurones located in one intrathoracic ganglionic locus influence widely divergent cardiac regions. The density of the adrenergic efferent neurites associated with these somata varies considerably throughout the ventricles. Their local anatomical ‘density’ does not reflect the capacity of local neurones to enhance regional inotropism. For instance, the capacity of adrenergic postganglionic somata to enhance ventricular inotropism is significantly greater in the ventricular outflow tracts and papillary muscles despite the fact that the density of their adrenergic efferent postganglionic neurites is similar to that found in other ventricular regions (Randall et al. 1996). This presents a problem if one is attempting to equate local tissue density of such neurites to the capacity of their associated somata to influence regional cardiac contractility.

Intrathoracic local circuit neurones Recent data indicate that many neuronal somata in individual intrathoracic ganglia, including those on the heart, project axons only to other neurones within the same ganglion. Other somata project axons to neurones in different intrathoracic ganglia (Armour, 1991), while others project axons to central neurones (Cheng et al. 1997). Many intrathoracic local circuit neurones receive inputs from extrathoracic sources as well (Armour, 1986a; Armour & Janes, 1988). Some of these intrathoracic, relatively large-diameter somata (i.e. ∼30 μm) form rosettes within their respective ganglion (Darvesh et al. 1987; Yuan et al. 1994; Armour et al. 1997), with their dendrites interconnecting within the ganglionic centre. Such data imply the presence of local information processing (Armour, 1991).

Normal cardiac control

Neurones involved in cardiac regulation are located from the level of the insular cortex (Oppenheimer & Hopkins, 1994) to the heart (Armour, 1991; Ardell, 1994). The intrathoracic component of this neuronal hierarchy has long been thought to act as a simple efferent relay station (Hillarp, 1960; Skok, 1973; Gebber et al. 1996; Lewis et al. 2000). As such, intrathoracic ganglia are represented as monosynaptic efferent relay stations that process centrifugal (efferent) inputs to the heart. Sensory information transduced by cardiovascular afferent neurones in nodose and dorsal root ganglia (Brown, 1967) is processed by the central nervous system to influence sympathetic postganglionic neurones in paravertebral ganglia via its spinal cord neuronal component, while parasympathetic postganglionic neurones in the target organ ganglia receive inputs from medullary neurones (Kuntz, 1934). It was long conceived that these two populations of cardiac motor neurones functioned in a reciprocal fashion (Levy & Martin, 1979).

Intrinsic cardiac cholinergic efferent postganglionic neurones receive direct synaptic inputs from medullary preganglionic neurones that are located primarily in the nucleus ambiguus, with lesser numbers being located in the dorsal motor nucleus and medullary regions in between these two nuclei (McAllen & Spyer, 1976; Hopkins & Armour, 1984; Gatti et al. 1995). The cardiac cholinergic postganglionic neuronal population is, in fact, quite limited with respect to all the neurones within the intrinsic cardiac nervous system (Armour, 1991). Sympathetic efferent postganglionic neurones involved in cardiac regulation receive inputs from caudal cervical and cranial thoracic spinal cord preganglionic neurones (Norris et al. 1977). The former are located in stellate, middle cervical, superior cervical and mediastinal ganglia, as well as in intrinsic cardiac ganglia (Armour, 1991). When maximally activated these sympathetic efferent neurones increase cardiac chronotropism, dromotropism and inotropism, while decreasing left ventricular chamber end-diastolic volume (Burwash et al. 1993). Parasympathetic efferent postganglionic neurones, when activated, do the opposite, including the suppression of ventricular inotropism (Randall et al. 1996). Furthermore, recent evidence indicates that many of the neurones in intrathoracic ganglia, including those on the heart (Randall et al. 2003), are in constant communication with one another (Armour et al. 1998) such that reciprocal antagonism between cardiac cholinergic and adrenergic postganglionic neurones may be the exception rather than the rule (Paton et al. 2005). Thus, the various populations of intrathoracic neurones, constantly interacting with one another and with central neurones (Lowie & Spyer, 1990; Andresen et al. 2004; Foreman et al. 2004), form temporally dependent reflexes that control spatially determined overlapping cardiac regions (Fig. 1).

Figure 1.

Proposed model for the cardiac neuronal hierarchy that emphasizes its intrathoracic components
The regional cardiac mechanosensory and chemosensory milieu is transduced by afferent neuronal somata located not only in nodose and dorsal root ganglia but also in intrathoracic intrinsic and extrinsic cardiac ganglia. This information engenders intrathoracic, as well as central (medullary and spinal cord) reflexes. The lower right-hand box indicates that circulating catecholamines influence cardiomyocytes not only directly but also indirectly via intrinsic cardiac adrenergic neurones.

The intrinsic cardiac nervous system Recent evidence indicates that the intrinsic cardiac afferent neurones transduce the local mechanical and chemical milieu of the heart, as well as that of major intrathoracic vessels (Ardell et al. 1991; Armour et al. 1997; Cheng et al. 1997; Horackova et al. 1999), to other neurones in their own ganglion, as well as to those in other intrinsic cardiac (Ardell et al. 1991) and intrathoracic extracardiac ganglia (Armour, 1986a). Recent evidence indicates that the synaptic interactions that occur among intrinsic and intrathoracic extracardiac neurones utilize a variety of neurochemicals (Armour, 1991). Peripheral neurones interact with central neurones (Yuan et al. 1993; Cheng et al. 1997) and therefore involve the entire cardiac neuronal hierarchy (Armour & Kember, 2004). The sensory neurones in each major intrinsic cardiac ganglionated plexus transduce primarily, but by no means exclusively, the chemical milieu of regions throughout the heart (Thompson et al. 2000; Armour & Kember, 2004). It has been proposed that such sensory inputs account for the generally stochastic behaviour displayed by many atrial and ventricular neurones (Waldman et al. 2006).

Collectively, populations of intrathoracic neurones display memory characteristics relative to cardiovascular events transduced during each cardiac cycle to influence efferent neuronal outputs to the heart during the same and the few subsequent cardiac cycles (Armour, 1976). As a consequence, cardiac perturbations initiated during a single arrhythmic beat, for instance, may affect efferent neuronal outputs for the next few cardiac cycles (Armour, 1976). Since intrathoracic afferent neurones influence local circuit neurones in intrinsic cardiac and intrathoracic extracardiac ganglia (Armour et al. 1998), in addition to central neurones (Yuan et al. 1994), redundant control is exerted over regional cardiac indices (Fig. 2). This may be why regional cardiac control remains relatively unaffected when the function of one neuronal population becomes compromised (Oh et al. 2006; Scanavacca et al. 2006).

Figure 2.

The influence that nicotine-sensitive neurons located in major atrial and ventricular cardiac ganglionated plexuses exert on select canine cardiac indices
Note that atrial tachydysrhythmias were induced by nicotine-sensitive neuronal somata within the dorsal and left atrial as well as cranial ventricular ganglionated plexuses. Abbreviations: RAGP, right atrial ganglionated plexus; DAGP, dorsal atrial ganglionated plexus; LAGP, left atrial ganglionated plexus; IVC-IAGP, inferior vena cava-inferior atrial ganglionated plexus; RVGP, right ventricular ganglionated plexus; VSVGP, ventral septal ventricular ganglionated plexus; CMVGP, cranial medial ventricular ganglionated plexus.

In order to clarify this fundamental concept, neurochemicals have been applied in limited quantities adjacent to neuronal somata within each major canine atrial or ventricular ganglionated plexus. In such a manner, local somata, not adjacent axons of passage, can be modified (Butler et al. 1990; Huang et al. 1994). Electrical current delivered to such loci activates not only adjacent somata and/or dendrites but also adjacent afferent and efferent axons of passage to produce different results from those elicited when solely somata and dendrites are activated in a locus (Butler et al. 1990). Presumably that is why focal electrical stimuli delivered to right atrial ganglionated plexus loci produce results that have been interpreted as indicating that sino-atrial nodal control resides primarily with neural elements in that ganglionated plexus (Ardell & Randall, 1986; Furukawa et al. 1990). Employing such methodology, it has also been presumed that atrio-ventricular nodal control resides with neurones that are solely located in the inferior atrial–inferior vena caval ganglionated plexus (Gatti et al. 1995; Zhuang et al. 2002). In fact, neuronal somata responsible for modulating either index reside in each major intrinsic cardiac ganglionated plexus (Fig. 2).

As a result of these functional anatomical realities, rapid ventricular rates accompanying atrial tachydysrhythmias can be slowed by inducing atrio-ventriculat nodal blockade subsequent to delivering electrical current to cholinergic efferent postganglionic neuronal somata and axons located in a number of ganglionated plexuses (Ali et al. 1990; Fig. 2), not solely the inferior vena caval–inferior atrial one (Zhuang et al. 2002). The same holds true for sino-atrial nodal control (Butler et al. 1990; Furukawa et al. 1990).

Intrinsic cardiac and intrathoracic extracardiac neuronal interactions It has been proposed that short-latency reflexes involved in modulating cardiac indices involve neuronal somata located adjacent to cardiomyocytes in intrinsic cardiac and mediastinal ganglia (Armour, 1976). Such short-latency (40 ms) intrathoracic reflexes apparently influence cardiac indices throughout each cardiac cycle. Longer-latency intrathoracic reflexes involving intrathoracic neurones modulate cardiac efferent neurones for a few cardiac cycles after the initiating event (Armour, 1986b). These various intrathoracic autonomic reflexes apparently involve local neuronal circuitry that collectively displays memory (Armour, 1991) such that cardiac sensory information generated during one cardiac cycle modulates cardiac efferent neurones not only during that cardiac cycle but also subsequent ones. Perhaps such polysynaptic reflex loops permit amplification of the efferent neuronal inputs to the heart, as well as the time-dependent effects depicted above relating to multiple cardiac cycles.

Interactions among peripheral and central neurones in cardiac regulation Some intrinsic cardiac and intrathoracic extracardiac neurones are under the tonic influence of spinal cord neurones (Foreman et al. 2004). Other intrinsic cardiac neurones are under the tonic influence of medullary neurones (Armour, 1991; Ardell, 1994). Furthermore, there are intrinsic cardiac local circuit neurones that apparently receive inputs from both of these central neuronal populations (Armour, 1991). It appears that the cardiac neuronal hierarchy is organized to provide the flexibility necessary for beat-to-beat co-ordination of regional cardiac indices via short-latency (intrinsic cardiac ganglia), medium-latency (intrathoracic) and relatively long-latency feedback (spinal cord and brain; Armour, 1976; Kollai & Koizumi, 1979; Katchenova et al. 1996). The relatively long-latency reflexes involving central neurones determine the tonic status of the peripheral nervous system. Descending inputs to the intrinsic cardiac neurones derived from medullary (Andresen et al. 2004) and spinal cord preganglionic neurones (Foreman et al. 2004) directly influence a limited population of intrinsic cardiac motor neurones, while at the same time influencing many local circuit neurones in an indirect manner (Armour, 1991). Medullary and spinal cord neurones also interact with one another, under the influence of neurones to the level of the insular cortex (Chechetto, 2004). Dorsal root ganglion cardiovascular afferent neurones initiate spinal cord reflexes (Foreman et al. 2004) while nodose ganglion ones (Katchenova et al. 1996) initiate brainstem-derived reflexes (Andresen et al. 2004). Such information exchange among neurones can also be indirectly influenced by sensory inputs from receptors located in other body regions (Armour, 1991). The shortest latencies of such centrally derived cardio-cardiac reflexes range from 125 to 350 ms (Armour, 1976), most affecting change over longer periods of time.

The relevance of these tonic central neuronal inputs to the intrathoracic nervous system is illustrated by the fact that many intrathoracic neurones, including those on the heart, no longer generate spontaneous activity in physiological states when their connections to central neurones are severed (Armour, 1991). In contrast, there are some intrathoracic sympathetic reflexes that receive tonic suppressor inputs from spinal cord motor neurones such that they become more prominent when disconnected from their central inputs (Armour, 1986a).

Some cardiac sympathetic efferent neurones generate activity that is phase related to cardiac dynamics reflective of cardiac or thoracic aorta mechanosensory inputs to the spinal cord (Armour, 1976). Some cardiac parasympathetic efferent preganglionic neurones also generate cardiac cycle-related phasic activity (Armour, 1976; Kollai & Koizumi, 1979). In fact, the phasic nature of such cholinergic efferent neurones is known to modulate heart rate more than if these inputs are stochastic (Levy & Martin, 1979). That most sympathetic and parasympathetic efferent postganglionic neurones generate stochastic activity presumably reflects the stochastic nature of sensory inputs derived from alterations in the cardiac chemical milieu (Armour, 1991). Although the details of efferent neuronal control remain elusive, it appears that the multi-synaptic interactions present within the intrinsic cardiac nervous system may engender a finer degree of regional co-ordination of cardiac indices than would be possible if control was solely dependent upon sensitizing afferent sensory neurites or efferent nerve terminals.

A relatively limited population of intrathoracic extracardiac and intrinsic cardiac local circuit neurones displays respiratory-related activity (Armour, 1986a, 1991). The activity displayed by this population appears to be dependent on two input sources: (i) right ventricular outflow tract mechanosensory neurites that transduce right ventricular outflow tract dynamic alterations secondary to changing pulmonary arterial resistance throughout each respiratory cycle; and (ii) inputs derived from pulmonary tissue mechanosensory neurites that transduce regional pulmonary tissue mechanics. Although the function of these respiratory-related inputs remains unknown, one may speculate that they supply some intrathoracic neurones, including those on the heart, with a capacity to simultaneously compute cardiac and respiratory status in order to balance cardiac efferent neuronal output to the demands of each organ.

Neurones throughout this hierarchy are in constant communication with one another via a host of neurochemicals to engender multilatency cardiovascular reflexes that appear to be to a considerable extent anatomically determined (Armour, 1976). The short-latency reflexes (i.e. 20–40 ms) depend to a considerable extent upon afferent neuronal transduction of the cardiac and great thoracic vascular mechanical milieu, modulating regional cardiac indices during each cardiac cycle. Intermediate-latency reflexes (i.e. 100–200 ms) apparently influence cardiac indices over a few cardiac cycles, while longer-latency ones may be responsible for longer-term effects such as the sustained hypertension that can be elicited during prolonged enhancement of cardiovascular sensory inputs to the medulla (Armour & Pace, 1982).

Implications of a primarily stochastic cardiac control system

Normal states The foregoing discussion highlights some of the interactions that are known to exist among peripheral and central neurones involved in cardiac regulation. The fact that limited intrinsic cardiac neuronal populations receive solely cardiac mechanosensory inputs results in short-term (beat-to-beat) control being confined to a relatively limited population of neurones (Waldman et al. 2006). The heart's chemical milieu is transduced over longer time scales by intrinsic and extracardiac intrathoracic neurones (Armour & Kember, 2004) that interact with central neurones (Malliani, 1982) to generate longer-latency reflexes that influence cardiac indices over multiple cardiac cycles (Waldman et al. 2006).

Widespread, spatially distributed cardiac indices can be modified by somata in one intrinsic cardiac ganglionated plexus, as demonstrated by activating local somata chemically without the aberration of activating local axons of passage as occurs with local current injection (Fig. 2). Neuronal somata in each major atrial or ventricular ganglionated plexus influence not only adjacent tissues, as is so frequently promulgated (Gatti et al. 1995; Chiou et al. 1997; Dickerson et al. 1998; Chen et al. 2006), but indices in all four chambers (Yuan et al. 1994). Thus, when considering the ablation of somata and/or nerves in one ganglionated plexus one cannot expect a predictably consistent response among different subjects. Rather, this redundant target organ control system appears to be so constructed as to assure unperturbed control when the function of one part becomes compromised.

Such functional redundancy may engender confusing results when local components within the hierarchy are targeted therapeutically (Cummings et al. 2004; Oh et al. 2006). Firstly, that is because neuronal somata in each major atrial or ventricular ganglionated plexus receive sensory inputs from widely distributed cardiac regions (Waldman et al. 2006). Secondly, each intrinsic cardiac neuronal population is in constant communication with the others (Thompson et al. 2000; Randall et al. 2003). Thirdly, many intrinsic cardiac local circuit neurones interact constantly with intrathoracic extracardiac (Armour et al. 1998) and central neurones (Yuan et al. 1994; Waldman et al. 2006). Induction of bradycardia by delivering electrical current to select loci within one intrinsic cardiac ganglionated plexus can be due to the activation not only of adjacent cardiac cholinergic efferent neuronal somata but also of their axons, as well as adjacent centrally projecting afferent axons or somata. Thus, ablation of one intrinsic cardiac ganglionated plexus may not produce detectable, long-term changes in cardiac indices. Fourthly, fat located at the base of adult human hearts is a continuum, obviating any easy delineation of the 10 major groupings of ganglionated plexuses associated with the human heart (Yuan et al. 1994). Such anatomy is certainly not consistent with the identification of a ‘third fat pad’ (Chiou et al. 1997), one adjacent to the aorta that can easily be confused with mediastinal ganglia (Armour, 1991). All of which is to state that a fuller understanding of the functional anatomy of this target organ nervous system is necessary before it can be targeted therapeutically such that results so derived have predictable long-term benefits.

Disease states There is growing evidence that the intrinsic cardiac nervous system can become directly involved in cardiac pathology (Hopkins et al. 2000). Increases in cardiac sensory inputs, such as occur in the presence of regional ventricular ischaemia, can result in the excessive activation of intrinsic cardiac second-order neurones, as well as neurones throughout the hierarchy. For instance, the flooding of the central nervous system by such novel information may influence cerebral neurones to such a degree that an individual becomes aware of altered cardiac status (Chechetto, 2004).

Cardiac arrhythmias. Cardiac arrhythmias can arise when the insular cortex is activated excessively (Chechetto, 2004). At the other extreme of this hierarchy, excessive activation of select elements within the intrinsic cardiac nervous system results in the genesis of atrial (Armour, 1976; 2005) or ventricular arrhythmias (Schwartz et al. 1978; Cardinal et al. 2004; Carlson et al. 1992). In accord with that, it has been promulgated that the ablation of neuronal elements in restricted cardiac ganglionated plexuses induces long-term suppression of atrial arrhythmias (Pappone et al. 2004; Scanavacca et al. 2006; Takahashi et al. 2006). It has also been proposed that excessive ‘vagal’ input to this nervous system facilitates the induction of atrial fibrillation secondary to programmed atrial stimulation (Scherlag & Po, 2006; Takahashi et al. 2006).

Perhaps more relevant to this discussion is the finding that atrial (Armour, 1991) and ventricular arrhythmias (Huang et al. 1994) can be elicited de novo when select components within each major intrinsic cardiac ganglionated plexus become excessively activated (Fig. 2). Since atrial and ventricular neurones are in constant communication with one another (Randall et al. 2003), as well as with more centrally located neurones (Armour et al. 1998), ablation of only one population may not obtund atrial arrhythmia formation (Mittal & Lerman, 2002). In fact, ablation therapy that targets limited components of the intrinsic cardiac nervous system is unlikely to produce consistent results, particularly when predicated on bradycardia induction (Pappone et al. 2004). The latter point is relevant because bradycardia is not a sine qua non for neuronal genesis of atrial arrhythmias (Cardinal, 2004). In contrast, modification of one population will influence others such that over time excessive activation of the whole becomes less likely. This may result in a reduced capacity for arrhythmia induction.

As an aside, the ‘virtual’ absence of sympathetic efferent postganglionic neurites in ventricular tissues of chemically sympathectomized canine preparations, as assessed histologically, proved not to reflect the capacity of sympathetic efferent neurones to enhance ventricular inotropism (elicited by direct nerve stimulation). In fact, sympathetic efferent postganglionic neuronal sprouting in ventricular tissues is difficult to quantify given the vast regional differences found throughout normal ventricles.

Decentralization of intrinsic cardiac neurones. It should be recalled that disruption of axons at the base of the heart does not ‘denervate’ the heart, since such a procedure only removes all inputs to the intrinsic cardiac nervous system (Ardell et al. 1991; Murphy et al. 1994). The fact that the intrinsic cardiac nervous system can function independent of central neuronal inputs (Ardell et al. 1991; Murphy et al. 1994) is another reason why attempts to ablate its select components may produce varied results (Carlson et al. 1992; Chiou et al. 1997; Hirose et al. 2002). Autonomic neuronal somata express considerable capacity to sprout neurites that innervate cardiomyocytes, whether studied in culture (Horackova et al. 1996) or in situ (Murphy et al. 2000).

Since the human intrathoracic extrinsic and intrinsic cardiac nervous systems are anatomically similar to those of large mammals, data derived from the canine (Yuan et al. 1994) and porcine models (Arora et al. 2003) may be applicable to man (Arora et al. 2001b). For instance, the transplanted canine heart offers a unique opportunity to study the capacity of the intrinsic cardiac nervous system to modulate cardiac indices, whether following acute cardiac transplantation (Murphy et al. 1994) or 1 year after cardiac auto-transplantation (Murphy et al. 2000). In both states, intrinsic cardiac neurones appear to help maintain cardiac status. After a few months, the intrinsic cardiac nervous system becomes partly re-innervated by intrathoracic extracardiac sympathetic efferent neurones; parasympathetic preganglionic neuronal re-innervation requires a longer time post-auto-transplantation (Murphy et al. 2000). Thus, if immunocompatibility exists between a donor heart and its host, the situation arises in which intrinsic cardiac neurones of a donor heart may receive inputs from recipient central neurones. Whatever degree of re-innervation occurs, the target organ's nervous system appears to be important in maintaining the integrity of cardiac function post-transplantation.

Angina of cardiac origin. The subject of the symptomatology of cardiac disease has a long and worthy history, starting with the observations of Heberdeen in the 1770s (Heberdeen, 1772). Many of the cardiac sensory neurites associated with nodose and dorsal root ganglion afferent neurones transduce adenosine (Armour, 1991; Huang et al. 1996), a chemical known to be released in increasing quantities by the ischaemic myocardium (Rubio et al. 1969). The relevance of adenosine in the genesis of cardiac pain became evident when Christer Sylvén and colleagues administered adenosine into the blood stream of diseased coronary arteries; the symptoms so engendered mimick those which the patients experienced during effort (Sylvén, 2004). In aggrement with that finding, dorsal root ganglion purine-sensitive afferent neurones are known to play an important role in the genesis of limb pain (Sylven, 2004). Furthermore, ventricular sensory neurites of dorsal root ganglion neurones become non-responsive to local ischaemia following adenosine receptor blockade (Huang et al. 1995b).

Spinal cord stimulation. Therapeutic approaches have been proposed that selectively target the intrinsic cardiac nervous system. One approach includes spinal cord stimulation therapy. Central neurones influence both intrathoracic extracardiac reflexes (Armour, 1976) and intrinsic cardiac neurones (Foreman et al. 2004). In fact, enhancement of spinal cord inputs to the intrinsic cardiac nervous system stabilizes its excessive activation when transducing regional ventricular ischaemia (Foreman et al. 2004). Enhanced sensory inputs to second-order intrinsic cardiac neurones may result in select populations becoming excessively activated (Arora et al. 2003), thereby increasing the likelihood of atrial arrhythmia formation (Cardinal, 2004). Indeed, enhanced spinal cord inputs to the target organ nervous system stabilize it in the presence of ventricular ischaemia (Foreman et al. 2004), perhaps to obtund its capacity to elicit atrial arrhythmias (Armour et al. 2005).

Pharmacology. Cardiac arrhythmias can arise when microlitre quantities of neurochemicals, such as β-adrenoceptor agonists or angiotensin II, are administered adjacent to select populations of intrinsic cardiac neurones (Huang et al. 1994; Horackova & Armour, 1997). β-Adrenoceptors or angiotensin II receptors associated with populations of these neurones influence cardiac motor neurones (Armour, 1991) to affect cardiac control (Farrell et al. 2001). Such modification also appears to influence the downward course of heart failure induced in animal models (Dell'Italia & Ardell, 2004). Pharmacological strategies currently employed in the management of cardiac arrhythmias (Cardinal, 2004) or heart failure (Dell'Italia & Ardell, 2004) may in fact also target select elements of the neuroaxis to support cardiomyocyte function indirectly.


Given the current paucity of detailed information concerning the anatomy and transduction capabilities of neurones located throughout the cardiac neuronal hierarchy, any prognostication as to the effectiveness of targeting its select elements therapeutically remains speculative (Mittal & Lerman, 2002). Recent evidence indicates that some benefit may accrue when elements within the intrinsic cardiac nervous system are ablated in the suppression of atrial arrhythmia (Scanavacca et al. 2006; Takahashi et al. 2006). Its relevance to heart failure therapy is just becoming apparent (Dell'Italia & Ardell, 2004).

The overriding thesis presented in this review is that the interactions that occur among its various cell stations are primarily stochastic in nature owing to the dominance of inputs derived from an ever-changing cardiac chemical milieu. Interactions engendered among its cell stations appear to be highly optimized to tolerate physiological perturbations. Given the fact that the capacity of its intrinsic cardiac components to influence cardiac indices deteriorates rapidly when studied with a Langendorff preparation (Kresh & Armour, 1997), understanding its function appears to require systems analyses based on data derived from observations in situ.

It is additionally hypothesized that populations of neurones throughout this hierarchy, when acting in isolation, may affect the smoothing function that its intrinsic cardiac component normally performs in the co-ordination of regional cardiac indices (Armour, 1991; Ardell, 1994). Indeed, linkage malfunction resulting from excessive inputs derived from the transduction of regional myocardial ischaemia can result in excessive activation of its local circuit neuronal population (Arora & Armour, 2003) such that atrial arrhythmias may arise (Armour et al. 1972, 2005; Cardinal, 2004; Foreman et al. 2004). Thus, it may not be surprising that preliminary data indicate that suppression of excessive activation of the intrinsic cardiac nervous system reduces its propensity to induce atrial arrhythmias (Foreman et al. 2004; Cardinal et al. 2006).

The fundamental message of this review is that any capacity to target select components within the cardiac neuroaxis therapeutically requires a more comprehensive understanding of the role that each plays in the co-ordination of regional cardiac indices. Given its primarily stochastic nature, it is difficult to predict the efficacy of attempts to influence individual components within the hierarchy without solid in situ experimental evidence. For instance, how can the creation of multiple transmural canals in the left ventricular wall by local laser application induce therapeutic benefit if it is assumed that such an intervention crates regional denervation (Bridges et al. 2007) when in vivo testing demonstrates that such does not occur (Hirsch et al. 1999)? Rather, this form of therapy apparently obtunds intrinsic cardiac local circuit neuronal function over time (Arora et al. 2001a), emphasizing the fact that reliable data in this field ultimately depend on in situ experimentation.



The author gratefully acknowledges the financial support of the Canadian Institutes of Health Research.