Dr Paul P Bertrand, Department of Physiology, University of Melbourne, Victoria 3010, Australia. Email:


1. The enteric nervous system (ENS) is present in the wall of the gastrointestinal tract and contains all the functional classes of neuron required for complete reflex arcs. One of the most important and intriguing classes of neuron is that responsive to sensory stimuli: sensory neurons with cell bodies intrinsic to the ENS.

2. These neurons have three outstanding and interrelated features: (i) reciprocal connections with each other; (ii) a slow excitatory post-synaptic potential (EPSP) resulting from high-speed firing in other sensory neurons; and (iii) a large after-hyperpolarizing potential (AHP) at the soma. Slow EPSP depolarize the cell body, generate action potentials (APs) and reduce the AHP. Conversely, the AHP limits the firing rate and, hence, reduces transmission of slow EPSP.

3. Processing of sensory information starts at the input terminals as different patterns of APs depending on the sensory modality and recent sensory history. At the soma, the ability to fire APs and, hence, drive outputs is also strongly determined by the recent firing history of the neuron (through the AHP) and network activity (through the slow EPSP). Positive feedback within the population of intrinsic sensory neurons means that the network is able to drive outputs well beyond the duration of the stimuli that triggered them.

4. Thus, sensory input and subsequent reflex generation are integrated over several hierarchical levels within the network on intrinsic sensory neurons.

List of abbreviations

After hyperpolarization


After-hyperpolarizing potential


Action potential


Enteric nervous system (collectively refers to neurons with cell bodies present in the wall of the intestine)


Excitatory post-synaptic potential


Inhibitory post-synaptic potential


The functions of the gastrointestinal (GI) tract are controlled by an ever-changing and complex interplay between local nerves, muscle and glands, central innervation and hormonal control.1–5 This adaptive control is important because the GI tract is presented with a bewildering array of nutritive, bulk or toxic substances to which it must respond appropriately.6–9 The GI tract can adapt to its local environment, in part because of the complex local innervation located in the wall of the intestine: the enteric nervous system (ENS). The ENS contains the cell bodies of motor neurons, interneurons and, importantly, sensory neurons.10–12 These intrinsic sensory neurons (or intrinsic primary afferent neurons; IPAN13,14) have three outstanding properties. The first is the reciprocal synaptic connections between intrinsic sensory neurons. The second important feature is their prominent after-hyperpolarizing potential (AHP) that, following a somatic action potential (AP), lowers the excitability of the cell for up to 20 s. Finally, in the population of intrinsic sensory neurons, synaptic transmission results in long-lasting, slow excitatory post-synaptic potentials (EPSP). As well as generating APs at the cell body, slow EPSP (long-duration synaptic potentials initiated by G-protein-coupled receptors) reduce the AHP and, so, provide strong positive feedback within the network (the term ‘network’ is used throughout the present review to refer to a population of interconnected neurons, either real or represented by a computer model).

The focus of the present review is on recent developments in our understanding of these features in the function of the circuitry of the ENS. Intrinsic sensory innervation will be covered first, followed by a detailed examination of the properties of intrinsic sensory neurons from the guinea-pig myenteric plexus, with particular emphasis on the AHP, the slow EPSP and their interactions. Finally, the multiple hierarchical levels of sensory processing, from sensory terminals to networks, will be covered, highlighting the three important properties of the intrinsic sensory neurons.

In the latter part of this review, we have placed an emphasis on computational studies of the ENS, in particular network aspects of enteric sensory neurons. Computational studies are used routinely in parallel with experimental studies in the central nervous system (e.g. see the volume containing Durstewitz et al.15), but are still rare in the study of the ENS. One of the aims of computational studies is to ‘explain’ how complex responses in an organ or neural network arise from the interactions between the simple components of the system. Predictions from these models are limited by the quality of the data and the extent to which the data can be generalized to other systems. Further, the goals of the investigators involved can strongly influence the types of assumptions and simplifications made. Thus, we believe computer simulations are best used as a tool to test predictions from experimental studies and to form new predictions to be tested experimentally.


Sensory innervation of the intestine

The intestine has two separate but equally important forms of sensory innervation. The first, and better known, is the primary afferent nerve terminals arising from cell bodies in the dorsal root ganglia and the nodose ganglia; these fibres run predominately in the spinal and vagal tracts, respectively (Fig. 1). These are the afferents that give rise to conscience sensations and pain arising from the intestine. Because the cell bodies are located outside of the intestine, they are referred to as extrinsic sensory neurons of the intestine (or extrinsic primary afferent neurons). The extrinsic sensory neurons also have an efferent role and seem to innervate neurons in both the myenteric and submucous plexes, although their exact targets are unknown.

Figure 1.

Sensory innervation of the intestine. A diagram showing a cross-section of the wall of the gut with the connections from the extrinsic sensory neurons (left) with cell bodies in the nodose and dorsal root ganglia and the intrinsic sensory neurons (right) with cell bodies located in the submucosal plexus (SMP) and myenteric plexus (MP). Both sets of sensory neurons have terminals in the MP and SMP with a variety of target neurons (not shown). The intrinsic sensory neurons are multipolar and their terminals project to the SMP, MP and the mucosal epithelium, where they come into contact with the enteroendocrine cells. For simplicity, the mucosal process is shown coming from the cell body; in many cases, these projections seem to come from a branch of a myenteric projection. All the projections of the intrinsic sensory neurons can conduct action potentials and all are specialized for the release of neurotransmitter. The intrinsic sensory neurons innervate most functional classes of enteric neuron. They do not appear to innervate the muscle or blood vessels, but they do appear to innervate the enterocytes. The extrinsic neurons have release sites in the MP, SMP and subepithelial space. They also have specialized terminal structures that are the site for the transduction of mechanical stimuli (not shown; see Zagorodnyuk et al.95). EPI, epithelium; CM, circular muscle; LM, longitudinal muscle.

This distinction would be superfluous were there no other sensory neurons from which to distinguish them. In fact, the intestine does contain the cell bodies of sensory neurons that project to the mucosa and to other enteric ganglia13 (Fig. 1). The intrinsic sensory neurons are also (perhaps more correctly) called intrinsic primary afferent neurons (IPAN; for a discussion of nomenclature, see Furness et al.13,14). The intrinsic sensory neurons do not give rise to conscience sensation, but they are in the direct, minute-to-minute control of intestinal function and seem to function simultaneously as sensory neurons and, possibly, as interneurons.

Sub-types of intrinsic sensory neurons

Populations of intrinsic sensory neurons may exist that respond to different modalities, such as stretch or other mechanical stimuli, or to the luminal chemical content. There are clear differences between the neurons in electrophysiological properties, morphology, location and chemical code but, to date, a clear correlation with their sensory properties has not been demonstrated. For the purposes of the present review, the intrinsic sensory neurons will be treated as a single population with overlapping properties.

Basic properties of intrinsic sensory neurons

Intrinsic sensory neurons have only been identified directly in the guinea-pig ileum,13 although their presence has been inferred by experiments in many other regions and species. Electrophysiological recordings have been made from chemosensory neurons in the myenteric plexus that are sensitive to acid at a low pH (1–3),16 acid at a more physiological pH (3–5) to basic pH (9–11) and to the short-chain fatty acid acetate at a neutral pH (7).17 A large proportion of these chemosensory neurons is also sensitive to the sensory mediators 5-hydroxytryptamine (5-HT)18,19 and/or ATP.20 Studies using a less direct approach have shown that sensory neurons sensitive to glucose or to distortion of the mucosal epithelium are in the submucosal plexus and that their activation is mediated by 5-HT.21 Electrophysiological studies have shown that myenteric sensory neurons are not sensitive to distortion of small areas (< 1 mm2) of mucosa,16 but are sensitive when larger areas are stimulated.17 The neurons sensitive to tension in the wall of the intestine are also located in the myenteric plexus;22 their firing is inhibited by smooth muscle paralysis with L-type calcium channel blockers.23 The myenteric terminals of these intrinsic sensory neurons respond to mechanical probing with a burst of APs.24

Most electrophysiology experiments that have identified the intrinsic sensory neurons directly (such as those above) have used L-type calcium channel blockers to paralyse the smooth muscle. This compound also reduces the release of sensory mediators from the mucosal enteroendocrine cells (for review, see Racké and Schwörer25) and, thus, probably leads to an underestimation of their sensitivity. The use of these compounds is a major stumbling block and will be a challenge for future studies to overcome.

Morphology and projection pattern of intrinsic sensory neurons

Intrinsic sensory neurons in the guinea-pig ileum have Dogiel type II morphology;13 most regions of the intestine in most species have some neurons with this morphology. They account for 25–35% of all neurons in the guinea-pig ileum.10,26 Dogiel type II neurons are large, smooth bodied and multipolar, with projections to the mucosa and between the myenteric and submucosal plexes (Fig. 2a). Approximately half have two projections, whereas the remainder have three or more projections.27 Dogiel type II neurons provide excitatory innervation to many cells within the same and circumferentially located ganglia and to virtually all functional classes of neuron.

Figure 2.

Morphology of an intrinsic sensory neuron and a sensory map for acetate. (a) A photomicrograph showing an intrinsic sensory neuron in the myenteric plexus (inset: a larger-scale view (2× magnification) of the cell body and associated fibres). The neuron had AH-type electrophysiology (i.e. a neuron with a long-lasting after-hyperpolarizing potential following the action potential) and Dogiel type II morphology. The neuron (black) was filled with biocytin during electrophysiological recordings and processed later to reveal the cell shape and projections. The dotted lines represent the boundaries of the ganglia and fibre tracts in the myenteric plexus. The arrow to the left indicates the direction of the intact mucosa. (b) The upper traces are voltage traces from a different intrinsic sensory neuron in the myenteric plexus. Intact mucosa was nearby. The lower diagram shows a schematic representation of the positions of the cell body (right; SN, sensory neuron) in the myenteric plexus (MP) and of the intact mucosa (left). The points represent areas of mucosa that were stimulated with acetate (100 mmol/L, pH 7). The filled circle represents an area that evoked a response from the neuron, whereas open circles represent areas that did not respond to application of acetate.

As mentioned in the Introduction, an important feature of the intrinsic sensory neurons is that they also innervate many other intrinsic sensory neurons.28 The projections are a few millimetres in both directions longitudinally and > 10 mm around, which is almost the full circumference of the intestine.27 This means that there are closed loops of excitatory transmission within the population of sensory neurons (even when the circumference has been cut open; Figs 2a,3a). Ten per cent of these neurons also have a long projection to anally positioned ganglia located up to 10 mm away,29 although the targets of these projections remain to be identified. All the projections of Dogiel type II neurons can conduct AP30 and all have transmitter release sites.31 In the mucosa, the projections of Dogiel type II neurons innervate the full length of the villi.31 The anatomical extent of the mucosal projection has been estimated as 1 mm2 using retrograde dye techniques32 and > 6 mm2 using electrical stimulation of the mucosa;33 this has been correlated with the functional sensory field, which was one-quarter of the area33 (Fig. 2b).

Figure 3.

Three key properties of intrinsic sensory neurons of the intestine. (a) Schematic drawing of the interconnections between sensory neurons. The grey neuron represents a point of reference. It is a true multipolar neuron with branching. This neuron communicates with other sensory neurons via slow excitatory post-synaptic potentials (EPSP; see (c)). (b) A voltage trace from an intrinsic sensory neuron. A single electrical stimulus was applied to a fibre tract (indicated by the open arrow) containing a projection of this neuron evoking an invading action potention. The cell body of the sensory neuron has a long-lasting spike after-hyperpolarization, the AHP. These AHP can reduce the excitability of the cell body substantially and are due to an opening of a calcium-activated potassium channel. (c) The upper traces is a voltage trace from an intrinsic sensory neuron. A train of electrical stimuli was applied to a fibre tract (the start is indicated by the open arrow; 20 Hz for 1 s). Downward deflections are due to negative current pulses (100 pA; lower trace) used to monitor input resistance. Note that during the peak of the slow EPSP, there is a large increase in the size of the hyperpolarization and, thus, in the input resistance of the neuron consistent with this depolarization being mainly due to a closure of potassium channels.

In guinea-pig ileum and colon, some non-Dogiel type II neurons with single axons have been postulated to be tension or length sensitive, respectively.22,34 However, because these neurons are not multipolar, it is unclear how they convey sensory information to other neurons. We could speculate that these uniaxonal neurons have a branched structure, like the Dogiel type II neurons, that allows axon reflexes to play some role in the propagation of sensory information.

Electrophysiological properties of intrinsic sensory neurons

The after-hyperpolarizing potential

In the guinea-pig ileum, intrinsic sensory neurons are classified electrophysiologically as ‘AH’; that is, they have a long-lasting afterhyperpolarizing potential (AHP) following the AP.35,36 A large AHP greatly reduces the excitability of the sensory neuron cell body (Fig. 3b). Intrinsic sensory neurons in other regions or species may not have a large AHP.37 The basis of the AHP is calcium entry during the AP, which opens a calcium-activated potassium channel35,38 that probably has an intermediate conductance.39 This potassium channel is regulated by several second messenger systems, including protein kinase A40–42 and protein kinase C.41,43

Synaptic transmission

Synaptic transmission in the ENS can be placed into well-defined classes depending on whether the duration of the synaptic event is fast, slow4 or intermediate.44 Fast synaptic events are mediated by ligand-gated ion channels and typically last from 10 to 30 msec. Slow events are mediated by G-protein-coupled receptors and have distinctly longer time-courses, lasting from minutes45 to hours.46 Intermediate-duration events are also mediated by G-protein-coupled receptors and last 0.5–2 s and include an excitatory post-synaptic potential (EPSP) and an inhibitory post-synaptic potential (IPSP), both found in the submucous plexus. One idea is that intermediate events are sustained by a direct G-protein to ion channel interaction.

Fast synaptic input to intrinsic sensory neurons

Fast synaptic input is only infrequently observed in intrinsic sensory neurons. However, intrinsic sensory neurons possess many receptors that may, under the right conditions, generate a fast EPSP (a short-duration synaptic potential mediated by ligand-gated ion channels), such as nicotinic acetylcholine, 5-HT3, GABAA and, recently, P2X receptors20,47–50 (also, see below).

Slow synaptic input to intrinsic sensory neurons

As mentioned in the Introduction, intrinsic sensory neurons have prominent slow excitatory post-synaptic potentials (i.e. slow EPSP). These events typically require a brief train of stimuli (3–10 pulses, 5–30 Hz) to evoke and cause membrane depolarizations of 3–30 mV. Slow EPSP can reach threshold for AP generation and so are true synaptic potentials in addition to being modulatory events. One source of these slow EPSP is from other intrinsic sensory neurons;51 thus, as a group, the intrinsic sensory neurons can be predicted to function as interneurons and form a self-reinforcing network (Fig. 3a). Because they have predominately circumferentially orientated projections, reflexes are coordinated around the full circumference of the region involved52,53 and we speculate that this coordination may underlie the initiation or propagation of some reflexes.54

The slow EPSP is primarily mediated by substance P, or a related tachykinin released from other sensory neurons, that acts at neurokinin receptors (NK3 and possibly NK1).55–57 The sensory neurons also receive descending projections from a class of interneuron that releases 5-HT,3,58 which may act through 5-HT1P59 or 5-HT7 receptors (P Bertrand and E Thomas, unpubl. data, 2004).60 In addition, the intrinsic sensory neurons have many G-protein-coupled receptors that can produce a slow EPSP-like response.40,41,61 In particular, there are receptors for the sensory mediators cholecystokinin and calcitonin gene-related peptide40 and for inflammatory mediators.62–64

There are also slow inhibitory post-synaptic potentials (IPSP) in the intrinsic sensory neurons.65,66 These have been observed in response to electrical stimulation of circumferential fibre tracts,67 indicating that transmission of IPSP could come from other sensory neurons. The transmitters, receptors and quantitative properties of IPSP have yet to be elucidated.

Long-term changes in intrinsic sensory neuron properties

The properties of the intrinsic sensory neurons are not fixed, but vary over time. For example, the AHP reduces the excitability of the sensory neuron cell body. However, the AHP can itself be reduced during a slow EPSP (Fig. 3c). The AHP is reduced for long periods following prolonged stimulation or during inflammation.46,68–70 In addition to changes in resting properties, the amplitude and frequency of fast EPSP are increased in sensory neurons from inflamed tissue.69,70 Controlling these changes in excitability is a potentially important therapeutic target.71

What the intrinsic sensory neuron does

The ENS has a large repertoire of motor programmes that control intestinal movements. These include segmentation and mixing patterns that are used during nutrient absorption, numerous forms of peristalsis (including reverse peristalsis) to move the chyme to different regions of the intestine for different kinds of processing, emesis for removal of noxious content and several interdigestive motor patterns. The ENS engages a motor programme based on stimuli such as the tension or length of the muscle, compression of mucosal epithelium, nutrients in the lumen, hormonal levels, extrinsic input etc. Some of this information may be conflicting; for example, a bolus provides a mechanical stimulus that evokes peristalsis, whereas the nutrient content of the bolus provides a chemical stimulus that evokes segmentation. We suggest that the population of sensory neurons combines multimodal input in order for a coordinated and appropriate reflex to be produced.


Significant sensory processing seems to occur at multiple levels in the population of intrinsic sensory neurons: at sensory terminals, at the cell body and at the level of the network. The amount of processing at each level will depend on where stimuli occur along the length of intestine and on the type and duration of stimulation.

In a sensory neuron, input consists of the number, frequency and/or pattern of AP discharge coming from the sensory (input/afferent) terminals. For chemosensory terminals, this pattern of APs is probably the result of an interaction between ligand-gated ion channels on the terminals20 and sensory mediators released from sensory cells (e.g. enteroendocrine cells), whereas for mechanosensory terminals this pattern seems to be determined by the coupling of mechanosensitive ion channels in the terminal to other structural elements in the tissue.23 Sensory output consists of the APs travelling out the output/efferent processes and, via synaptic transmission, directly evoking AP discharge or indirectly modulating ongoing activity in second-order neurons (other sensory neurons, interneurons or motor neurons). In the dorsal root ganglia, the extrinsic sensory neurons innervating the intestine (and other organs) are pseudo-unipolar; thus, sensory output to the central nervous system follows directly from sensory input. For the intrinsic sensory neurons in the intestine, the situation is more complex. They are true multipolar cells, so the majority of the APs must cross the cell body to reach other efferent projections (though a minority may bypass the cell body via an axon branch). In addition, the second-order neurons they synapse with are often other sensory neurons. Thus, for the intrinsic sensory neurons, the integration of sensory information can occur at the input nerve terminals, at the cell body and by recruitment of other sensory neurons at the network level.

The nerve terminal integrates sensory input

The intrinsic sensory nerve terminals and associated structures are the first step in the process of sensory processing. An example of the importance of this can be seen in monosynaptic local reflexes from mucosal stimulation of the intrinsic sensory neurons to motor neurons. The burst of fast EPSP recorded in these motor neurons has a time-course that is similar to the burst of APs recorded from the intrinsic sensory neuron.17 Similarly, when the output of descending reflexes (distension) is compared with the local reflexes (electrical stimulation of the mucosa), the inhibitory junction potentials in the circular muscle have a similar time-course. This suggests to us that, for some very simple reflexes, the final shape of the output can be determined at the level of the input nerve terminal of the sensory neuron.

Mucosal nerve terminals preferentially fire bursts of APs

The intrinsic sensory neurons all have a projection that terminates in the subepithelial space near the base of the enterocytes in the mucosal epithelium (see Fig. 1). These terminals are responsible for detecting chemical and mechanical stimuli in the lumen. Experiments designed to probe their properties have found that a variety of stimuli activate these terminals, but that the terminals respond with a stereotyped burst of APs. For example, in approximately 70% of the intrinsic sensory neurons with an intact projection to the mucosa, there was spontaneous, ongoing bursting.17,72,73 These bursts of APs were deduced to be coming from the mucosal projections and invading the cell body.73 Similarly, when chemical stimuli were applied to the mucosa, these same neurons responded with high-frequency bursts of APs (20–50 Hz, three to 10 APs; e.g. see Fig. 2b). Transient and prolonged stimuli give a similar pattern of response that has led us, and others, to speculate that the bursts are generated by a process of sensory transduction.6,8 The sensory cells in the epithelium (e.g. ATP-containing enterocytes or the 5-HT-containing enterochromaffin cells) could detect luminal stimuli and release discrete packets of sensory mediators. These mediators could then act on ligand-gated ion channels on the mucosal nerve terminal to evoke a discrete burst of APs.18,20 Other mechanisms may also contribute to the generation of these bursts, such as specialized voltage- or ion-activated ion channels on the nerve terminal.73

Mucosal nerve terminals can amplify input

Bursts of APs originating at mucosal terminals occur both when a stimulus is applied to the mucosa and when there is ongoing spontaneous activity.17,72,73 In addition, some of these neurons appear to participate in a positive feedback loop with elements in the mucosa.73 For example, in approximately 40% of the intrinsic sensory neurons with an intact projection to the mucosa, an AP initiated by a brief somatic depolarization was followed by a burst of APs. The APs arose from a flat membrane potential and were not sensitive to somatic hyperpolarization73 (Fig. 4). If the triggering AP was prevented by hyperpolarization, then the following APs were prevented too. This indicates that the burst of APs was triggered by APs in the processes rather than in the soma. Using a combination of somatic and mucosal stimulation, it was shown that the somatic AP must travel along the mucosal projection towards the terminal in order to evoke the subsequent burst.73 In addition, the burst itself travels via the same projection back to the cell body. Taken together, this suggests that APs travelling from non-mucosal projections may invade the mucosal projection and elicit a burst of APs. This burst then travels back to the cell body and out to the other projections. Thus, the mucosal terminal and associated structure may, under some conditions, act as an amplifier where a single AP can be amplified into a burst of APs.

Figure 4.

The intrinsic sensory nerve terminal integrates stimuli. Inset: diagram indicating that complex interactions are occurring at the nerve terminal. (a) Schematic diagram showing how input from non-mucosal projections can cross the cell body of the intrinsic sensory neuron and invade the mucosal projection. At the mucosal terminal, feedback between epithelial elements (here represented as an enteroendocrine cell) can lead to a burst of action potentials (APs) travelling back to the cell body. In effect, the single AP has been amplified by the nerve terminal processes. (b) Voltage traces from a single intrinsic sensory neuron. The upper trace shows that a single AP was elicited at the cell body by a brief (3 msec) depolariaing current pulse passed through the recording electrode (at the filled bar). The APs and the current pulse occur together. This single AP would be expected to travel out all the projections of the multipolar neuron (unless blocked by an after-hyperpolarizing potential in the soma). Following the single evoked AP is a short burst of APs. These APs were not initiated at the soma. Note the hyperpolarized potential at which they are generated. The middle traces shows, for the same data, the first derivative of voltage with respect to time. Each of the later APs has an inflection on the upstroke, indicating they are generated by an AP invading the soma from a distal projection. The lower trace shows that when the membrane potential was lowered to −80 mV, a large depolarizing pulse can still generate an AP that travels out the projections. However, the invading APs are now revealed as proximal process potentials, or incoming APs that failed to elicit somatic APs. The traces in (b) have been adapted from Bertrand.73 EC cell, enterochromaffin cell; SN, sensory neuron.

Output nerve terminals and other properties

The intrinsic sensory neurons have many interesting properties, more than just the core properties that are the focus of the present review. For example, the terminals of the intrinsic sensory neurons that are in the mucosa or the myenteric plexus are likely to have afferent (input) and efferent (output) properties. Thus, at any given moment, they may function to sense a stimulus (e.g. nutrients in the lumen or tension in the wall of the intestine) or to release transmitter (e.g. to excite enterocytes or to excite neighbouring neurons). We suggest that a single terminal may perform these tasks independently or, more likely, the terminal performs these tasks simultaneously.

In addition, enteric nerve terminals are well-endowed with G-protein-coupled receptors that can either increase or decrease the efficacy of synaptic transmission. Although not shown directly, we suggest that the terminals of the intrinsic sensory neurons possess one or more of the following receptors. The inhibitory G-protein-coupled receptors appear to be most important (i.e. presynaptic inhibition), the most prominent of which is the muscarinic M2 receptor.74 Some other important receptors are α2-adrenoceptors, µ-opioid receptors, 5-HT1A receptors, galanin receptors,75 adenosine A1 receptors76 and GABAB receptors.77 Finally, clinically important 5-HT4 receptors seem to enhance release and are the targets for a class or prokinetics (e.g. cisapride).78,79

The cell body integrates information

In the guinea-pig ileum, it is thought that the cell body of the intrinsic sensory neuron is of prime importance for the integration of sensory information. Wood4 suggested that the AHP acts as a gate for APs travelling across the cell body. At rest, the gate is open and an AP may cross the cell body. Following the AP, the generation of an AHP would close the gate and block any subsequent APs from crossing the soma (Fig. 5). This gating process has been confirmed experimentally; when an incoming AP was blocked or reduced to a proximal process potential at the soma, it did not cross over to other projections and propagate along them.80 In addition to the AHP, Kunze et al.24 have described a hyperpolarization of the soma that was induced by gentle probing. If this hyperpolarization was generated by normal intestinal movements, then we would predict that it would also participate in somatic gating. The important role of the AHP and the experimental accessibility of the cell body mean that the excitability of the soma has become the main focus of research into the generation of aberrant motor patterns in inflammatory disease states.71

Figure 5.

An after-hyperpolarizing potential (AHP) at the intrinsic sensory neuron soma can block action potentials (APs) traversing the body. Inset: diagram indicating how integration can occur at the cell body. (a) Voltage traces taken from a single intrinsic sensory neuron from the myenteric plexus. Acetate (100 mmol/L, pH 7) was applied to a section of intact mucosa close to the cell body. The upper traces shows that, under control conditions, the cell body is poorly excitable (hyperpolarized) and the AHP blocks the majority of APs. The lower trace shows that, following prolonged stimulation (in this case, a rebound excitation following washout of a low calcium solution), the cell body is more excitable (depolarized). Importantly, the AHP no longer prevents the APs from crossing the cell body and invading other projections. (b) Schematic diagram representing the gate in the cell body turning on and off. For simplicity, the mucosal process is shown coming from the cell body. 1. A chemical stimulant evokes a train of APs at the terminal that travel to the cell body. 2. When the cell body has a functional AHP (grey) only a few of the APs pass through the soma. 3. When the cell body is excited (white) the AHP does not block APs from crossing the soma. RMP, resting membrane potential; EC cell, enterochromaffin cell; SN, sensory neuron.

Bursts of APs bypass the gate at the cell body of the intrinsic sensory neuron

The generation of the AHP is complex; there is an approximate 60 msec delay between an AP and the start of the AHP.36,39 During this period, APs can still be generated at the soma. Thus, a short burst of APs could follow the first AP and cross the soma without being blocked by the AHP. This has been seen experimentally where the frequency (20–50 Hz) and number of APs (three to 10 APs) in the bursts from mucosal stimuli (described above) are such that most APs evoked somatic APs and, thus, would be expected to cross the cell body to another process. Thus, bursts of APs from distal processes may bypass gating by the AHP at the intrinsic sensory neuron cell body (Fig. 5).

These observations suggest that the AHP may have a smaller role in controlling excitability than supposed previously. This idea has been expressed by Cornelissen et al.,81 who found that, under control conditions, multipolar Dogiel type II neurons in the pig rarely have a long-lasting AHP (but see below).

What if the AHP in response to a single AP is small or suppressed completely? One idea is that high-frequency bursts from mucosal projections may reactivate the gate through build up of intracellular calcium that drives the AHP.82 Thus, we predict that the cell body may be able to continue to act as a gate when the AHP is reduced, such as during a slow EPSP, inflammation or after prolonged stimulation.45,46,69

Other properties of the intrinsic sensory neurons

An intriguing possibility is that sensory neuron morphology may contribute to the way these neurons process stimuli. For example, in addition to being multipolar, these neurons are highly branched (see Fig. 2a). Branch points allow sensory information to propagate without crossing the body of the sensory neurons (i.e. axon reflexes). From data in the literature, we predict that from one-third to one-half of the traffic may avoid the gate at the soma.27

The multipolar/branched projection pattern also seems prone to collision of APs. For example, we predict that if tension and mucosal compression were applied simultaneously, as a bolus in the lumen would, then APs from projections in the myenteric plexus should, at some point, collide with APs from the mucosal projection. Axon reflexes and collision of APs are properties that remain ripe for future electrophysiological or computer modelling investigations.

The network integrates sensory input

The population of intrinsic sensory neurons form an interconnected network. For long-lasting stimuli, or for stimuli that are strong, the number and frequency of APs generated in the intrinsic sensory neurons will generate slow EPSP in other sensory neurons. This means that the network of sensory neurons becomes significantly involved in processing stimuli and driving motor or secretory behaviour over the time-course of the slow EPSP (minutes) and potentially much longer.

The network of intrinsic sensory neurons

The properties of the larger network of sensory neurons are poorly understood and it is not known whether this network may have ‘emergent’ behaviours not intuitively predicted from the properties of individual neurons. Computational studies using computer models of intrinsic sensory neurons can be used to develop hypotheses about how networks of these neurons may behave. At the very least, the networks, as a whole, must have mechanisms to control firing. In addition, to encode stimuli, they must allow graded responses to differing levels of sensory input. Reciprocal connections and strong excitation between sensory neurons suggests there may be a great deal of positive feedback within these networks. Computer simulation studies agree with this and have predicted that, without some form of balancing inhibition, positive feedback is so strong that even small stimuli will lead to runaway firing.53,83 There are a number of candidate mechanisms that may, individually or in combination, control the firing of the intrinsic sensory neurons at the network level. Two of these mechanisms are the AHP and the IPSP.

The AHP provides activity dependent negative feedback

The AHP is an obvious candidate mechanism for negative feedback within the network of intrinsic sensory neurons. However, as noted, not all regions or species have intrinsic sensory neurons with a large AHP. Further, excited intrinsic sensory neurons may not express an AHP. Recent studies using computer simulations have found that even a small AHP may play a role in controlling excitation at the network level53 (Fig. 6). At the other extreme, AHP as large as those in the guinea-pig ileum may provide too much inhibition. Again, computer simulation studies suggest that networks of intrinsic sensory neurons with unsuppressed AHP have very limited firing rates, even for strong sensory input. When the network has such a limited input/output relationship, it may not be able to drive reflexes, especially over periods of time outlasting the initial stimulus, and, so, is not likely to be biologically useful. Slow EPSP cause a reduction in the size of the AHP45 through a second messenger-mediated phosphorylation of the potassium channels underlying the AHP.40,41,84 When this interaction is included in computer simulations of intrinsic sensory neuron networks, the network is able to maintain a firing rate that is graded with the strength of sensory input and it is able to track changes in the strength of the input.53,83,85

Figure 6.

The network of intrinsic sensory neurons has emergent properties. Inset: diagram indicating that integration is occurring at the level of the network. (a) Graph showing overall firing of a network of intrinsic sensory neurons modelled on a computer. When neurons lack an after-hyperpolarizing potential (AHP), even small stimuli cause the network to fire at maximum rate and encoding of the original stimulus intensity is lost. When each action potential (AP) leads to a full sized AHP (unsuppressed AHP), network firing rates are highly constrained and, again, the original stimulus intensity is not encoded. Finally, if the AHP is only suppressed by the slow excitatory post-synaptic potential (EPSP), then the balance between positive and negative feedback allows the network to give graded responses to inputs. Note the logarithmic scale for both axes. (b) A voltage trace from an intrinsic sensory neuron during electrical stimulation of a fibre tract (at the black bar). The slow EPSP evoked APs and suppressed the AHP. (c) Computer-generated voltage traces from an intrinsic sensory neuron showing a slow EPSP evoked by a discrete stimulus and evoked during prolonged network activity (as graphed in (a)).

The IPSP and negative feedback

Another form of negative feedback within these networks could be provided by IPSP. These have been observed in intrinsic sensory neurons65,66 in response to circumferential stimulation,67 indicating that they probably come from other sensory neurons. However, this is yet to be confirmed and the transmitters, receptors and quantitative properties of IPSP have yet to be elucidated. Computer simulation shows that, by interacting with slow EPSP, IPSP have a regulatory effect on intrinsic sensory neuron network firing similar to that of the AHP. Interestingly, because IPSP are likely to have significantly shorter durations than slow EPSP, under some circumstances an IPSP may be able generate rebound bursts of APs.83 These bursts are likely to be chaotic and we predict that they would be observed as synaptically driven erratic bursting. Such firing patterns were seen frequently when using extracellular recording techniques (for a review, see Wood86).

Other forms of negative feedback

Computer simulations studies show that the negative feedback provided by either the AHP or the IPSP has a short latency that allows the intrinsic sensory neuron network to rapidly reach a stable firing rate. If there is a large difference between the rates of onset of positive and negative feedback, we predict that more complex firing patterns are possible, such as propagating waves of activity87 (see below) and local oscillations (E Thomas, unpubl. obs. 2003). An example of such a slow negative feedback mechanism is activity dependent loss of synaptic efficacy (e.g. through receptor internalization88). However, whether these processes cause negative feedback in the way predicted is, at this stage, speculative.

How intestinal reflexes may work

The three main properties of the intrinsic sensory neurons can account for much of the observed behaviour of the intestine. Below are three examples of how these properties may be combined to integrate stimuli and produce simple intestinal reflexes.

Peristalsis and accommodation

Rapid distension of the intestinal wall causes a polarized reflex (muscle relaxation anal to the stimulus, contraction oral) and, in the whole organ, initiation of peristalsis.89 Conversely, a slow distension leads to an accommodation of muscle tension and allows for increased luminal content.90 A maintained tension predominately causes low-frequency firing of AP22,23 rather than firing in high-frequency bursts.20,73 Thus, APs are generated in the network of intrinsic sensory neurons infrequently and we would predict that the overall firing rate of the network will be limited by the AHP. This may preferentially activate inhibitory pathways, leading to the accommodation reflex. A fast distension may recruit more sensory neurons and/or generate a higher-frequency input of APs in the network and allow it to activate an excitatory reflex, but we would predict that the output of the network would still be limited by the AHP.

Enhancement of reflexes

It is clear that distension-evoked reflexes can be enhanced by other stimuli. For example, Yuan et al.91 showed that a maintained mucosal compression caused an enhancement of the distension-evoked reflex. Smith et al.92 showed the same, but used the application of acid to the mucosa to enhance the reflex. Mucosal compression or application of acid causes a burst of APs in the intrinsic sensory neurons. These bursts could trigger slow EPSP in the network and change the way it transduces stimuli. The number and frequency of APs (20–50 Hz, three to 10 APs) in bursts of APs from the mucosa are ideally suited to evoke a slow EPSP.66,85,93 We would then predict that the interconnectedness of the intrinsic sensory neurons would ensure that the burst-evoked slow EPSP causes positive feedback and a change to an excitatory motor pattern.51,53

Switching between motor programmes

Suppression of the AHP may, in part, be the key to switching between different motor programmes. When the AHP is suppressed, long periods of high firing are possible. One example of this is phase III of the migrating motor complex. The migrating motor complex is a motor programme that occurs in the fasted state, lasts hours and is enacted over the entire small intestine. It culminates in phase III, which is a slowly moving region of strong contractile activity that starts in the duodenum and travels to the terminal ileum. Computer simulation studies have predicted that if the AHP is suppressed sufficiently, then local regions of the network can enter a high firing rate state. This firing excites neighbouring regions and, so, a wave of activity is generated that propagates with a speed similar to that of the phase III.87

Much work remains to be done in order to understand reflexes in the intestine and the intrinsic sensory neuron network that underlies much of its behaviour. In particular, more studies are needed like that of Brookes et al.,94 where the rate of distension and force needed to evoke a reflex are carefully controlled. Under these conditions, the state of the intrinsic sensory neuron network can be more easily modelled and more definitive experiments can be performed.