Multifunctional rapidly adapting mechanosensitive enteric neurons (RAMEN) in the myenteric plexus of the guinea pig ileum

Abstract

An important feature of the enteric nervous system (ENS) is its capability to respond to mechanical stimulation which, as currently suggested for the guinea-pig ileum, is encoded by specialized intrinsic primary afferent neurons (IPANs). We used von Frey hairs or intraganglionic volume injections to mimic ganglion deformation as observed in freely contracting preparations. Using fast voltage-sensitive dye imaging we identified rapidly adapting mechanosensitive enteric neurons (RAMEN, 25% of all neurons) in the myenteric plexus of the guinea pig ileum. RAMEN responded with phasic spike discharge to dynamic changes during ganglion deformation. This response was reproducible and increased with increasing forces. Deformation-evoked spike discharge was not changed by synaptic blockade with hexamethonium, ω-conotoxin or low Ca²⁺/high Mg²⁺, defunctionalization of extrinsic afferents with capsaicin or muscle paralysis with nifedipine, suggesting direct activation of RAMEN. All RAMEN received hexamethonium-sensitive fast EPSPs, which were blocked by ω-conotoxin and low Ca²⁺/high Mg²⁺. Seventy-two per cent of RAMEN were cholinergic, 22% nitrergic, and 44% were calbindin and NeuN negative, markers used to identify IPANs. Mechanosensitivity was observed in 31% and 47% of retrogradely traced interneurons and motor neurons, respectively. RAMEN belong to a new population of mechanosensitive neurons which differ from IPANs. We provided evidence for multifunctionality of RAMEN which may fulfil sensory, integrative and motor functions. In light of previously identified mechanosensitive neuron populations, mechanosensitivity appears to be a property of many more enteric neurons than generally assumed. The findings call for a revision of current concepts on sensory transmission within the ENS.

Introduction

The enteric nervous system (ENS) can regulate reflex pathways that control gut motility and mucosal transport function independently from the central nervous system (CNS), hence its alias ‘the second brain’ (Gershon, 1999). This intriguing capability is, among others, due to coding of mechanical stimuli by enteric neurons. Such neurons have been referred to as ‘sensory’ neurons despite the fact that this terminology may be misleading as they do not encode information in the same way that vagal or spinal sensory neurons encode mechanosensory information by specialized endings in the gut wall (see Blackshaw et al. 2007). Using extracellular electrodes, mechanosensitive neurons have been described in the myenteric plexus which responded to probing the tissue with the tip of a glass electrode or a 20 μm platinum wire (Wood & Mayer, 1974; Mayer & Wood, 1975). Some of these neurons behaved like slowly adapting units showing tonic spike discharge, others like rapidly adapting units responding with phasic spike discharge at the onset of stimulation. These studies provided evidence for the existence of mechanosensitive enteric neurons but did not reveal detailed electrophysiological behaviour, other possible functions of these neurons or their neurochemical coding.

One of the best studied concepts on sensory pathways in the ENS derives from studies in the myenteric plexus of the guinea-pig ileum and suggests the existence of specialized intrinsic primary afferent neurons (IPANs). According to this concept, IPANs respond to mechanical stimulation with spike discharge (Furness et al. 1998; Kunze & Furness, 1999). IPANs have Dogiel type II morphology (cells with multiple long processes), are cholinergic and in addition immunoreactive for the calcium-binding protein calbindin (Calb) and/or NeuN (Furness et al. 1998; Van Nassauw et al. 2005). Electrophysiologically, they are AH neurons which display a long-lasting afterspike hyperpolarization and receive slow EPSPs but rarely fast EPSPs (Kunze & Furness, 1999). Activation of IPANs by sustained tissue stretch depends on smooth muscle tension because it is abolished by muscle-paralysing drugs (Kunze & Furness, 1999).

The IPAN concept has been recently challenged (Spencer & Smith, 2004; Wood, 2004; Smith et al. 2007; Blackshaw et al. 2007; Wood, 2008; Schemann & Mazzuoli, 2009). Studies performed in the guinea pig distal colon found that unipolar Dogiel type I neurons that behaved like tonically spiking S neurons were activated during sustained colonic distension whereas the putative IPANs did not respond (Spencer & Smith, 2004; Smith et al. 2007). According to their morphology and axonal projections those neurons were termed mechanosensitive interneurons to indicate that they fulfil more than one function.

Some limitations of previous studies investigating mechanosensitivity in the ENS include lack of information on reproducibility of responses, certain bias when mechanosensitivity has only been studied in a particular neuron population, poorly defined stimulus modalities, rather descriptive analysis of response patterns or recording of late responses only after onset of tissue stretch. Some of the limitations were due to technical difficulties as most studies involved conventional electrophysiology using sharp electrodes (Smith et al. 2007).

We attempted to identify mechanosensitive neurons that responded to rapid ganglion deformation similar to what we observed in freely moving preparations from the ileum. Furthermore, we aimed to provide evidence for the concept that mechanosensitive neurons in the ENS are in fact multifunctional (Wood, 1970; Schemann & Schaaf, 1995; Pan & Gershon, 2000; Blackshaw et al. 2007; Smith et al. 2007). In order to validate this concept, we identified mechanosensitive neurons in the myenteric plexus of the guinea pig ileum, the preparation which served as the basis for the IPAN concept. We studied neuronal responses to mechanical stimuli that mimic changes during contractile activity rather than using sustained stretch with a fast neuroimaging technique based on the use of a voltage-sensitive dye. The voltage-sensitive dye imaging has the advantage of simultaneously recording from all neurons in the field of view.

Methods

Tissue samples

Male guinea pigs (Dunkin Hartley, Harlan GmbH, Borchen, Germany) were killed by cervical dislocation followed by exsanguination (approved by the local animal ethical committee and according to the German guidelines for animal protection and animal welfare). The ileum was quickly removed and further dissected in carbogen-aerated (95% O₂, 5% CO₂; pH = 7.40) Krebs solution containing (in mm): 117 NaCl, 4.7 KCl, 1.2 MgCl₂.6H₂O, 1.2 NaH₂PO₄, 25 NaHCO₃, 2.5 CaCl₂.2H₂O, 11 glucose. The mucosa, submucosa and circular muscle were removed in order to obtain longitudinal muscle–myenteric plexus preparations. In some preparations the circular muscle layer was kept intact in order to record contractions in intact mucosa-free tissues. The preparations (5 mm × 10 mm) were pinned onto a silicone ring that was placed in a recording chamber continuously perfused with 37°C carbogen-aerated Krebs solution with a rate of perfusion of 11 ml min⁻¹. The average amount of stretch above slack was 58 ± 20% in the circular and 18 ± 8% in the longitudinal direction.

Organotypic tissue culture and retrograde tracing of circular muscle motor neurons and interneurons

The organotypic culture method has been previously described in detail (Neunlist & Schemann, 1997). After removal from the animal and several washes, a 4 cm × 2 cm segment of ileum was pinned onto a Sylgard-covered Petri dish and the mucosa and submucosa were removed. To retrogradely label neurons projecting to the circular muscle a DiI (1,1′-didodecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate; Molecular Probes, Eugene, OR, USA)-coated glass bead (diameter 50–100 μm) was placed onto the circular muscle between ganglia. Criteria for specificity of labelling of circular muscle motor neurons have been previously published (Michel et al. 2000). In other preparations the circular muscle layer was removed and the DiI glass bead was placed directly onto a single myenteric ganglion in order to label interneurons (Reiche et al. 2000). Following the bead application, 30 ml of culture medium (Dulbecco's modified Eagle's–Ham's F-12 medium; Sigma, Deisenhofen, Germany) containing 1 μm nifedipine, 10% heat-inactivated horse serum (CC Pro, Karlsruhe, Germany), 1% antibiotic antimycotic (CC Pro) and 40 μg ml⁻¹ gentamicin (Sigma). The tissue was maintained in a humidified incubator at 37°C and equilibrated with 5% CO₂ in air for periods of 48–120 h.

Multisite optical recording technique with voltage-sensitive dye

We used a multi-site optical recording technique (MSORT) as previously described in detail to detect signals from the fluorescent potentiometric dye 1-(3-sulfanatopropyl)-4-[β-[2-(di-n-octylamino)-6-naphthyl]vinyl] pyridinium betaine (Di-8-ANEPPS) which is incorporated into the cell membrane (Neunlist et al. 1999; Michel et al. 2005; Schemann et al. 2005). Individual ganglia were stained with Di-8-ANEPPS by local pressure application through a microejection pipette loaded with 20 μm Di-8-ANEPPS dissolved in DMSO and Pluronic F-127 containing Krebs solution. The dye staining did not change the electrophysiological properties of the nerve cells (Neunlist et al. 1999). The chamber containing the preparation was mounted onto an epifluorescence Olympus IX 71 microscope (Olympus, Hamburg, Germany) equipped with a 75 W xenon arc lamp (Optosource, Cairn Research Ltd, Faversham, UK). Illumination of the preparation was achieved by a software-operated shutter (Uniblitz D122, Vincent Associates, New York, USA). Di-8-ANEPPS-stained neurons were visualized with a ×40 objective (UAPO/340, NA = 1.4, Olympus, Hamburg, Germany) using a fluorescence filter cube consisting of a 545 ± 15 nm excitation interference filter, a 565 nm dichroic mirror and a 580 nm barrier filter (AHF Analysentechnik, Tübingen, Germany). The fluorescence images were acquired and processed by the Neuroplex 8.3 software (RedShirtImaging, Decatur, GA, USA). This set-up allowed us to measure relative changes in the fluorescence (ΔF/F), which is linearly related to changes in the membrane potential (Neunlist et al. 1999). The fluorescence changes were detected with a cooled charge-coupled device (CCD) camera made of 80 pixels × 80 pixels (RedshirtImaging). Optical signals were processed with a computer; the frame rate was 1–2 kHz which enables the detection of action potentials. With a ×40 objective the spatial resolution of the CCD system was ∼24 μm² per pixel. Individual cells can be identified since the dye incorporates into the membrane revealing the outline of individual cell bodies. The overlay of signals and ganglion image allowed the analysis of the responses from single neurons (Michel et al. 2005).

To evoke fast EPSPs via electrical stimulation of interganglionic fibre tracts a Teflon-coated platinum electrode (diameter 20–25 μm) connected to a constant-voltage stimulator was used (600 μs pulse duration with amplitudes ranging from 1 to 8 V which were suprathreshold).

Since any fluorescent dye recording will eventually cause dye bleaching and/or phototoxicity illumination time was a crucial factor. On the one hand it had to be kept to a minimum but also had to be long enough to reveal representative and reproducible responses of neurons. It turned out that recordings with durations of 1255–2000 ms for the mechanical stimulations and 628 ms for the electrical stimulations yielded reliable and reproducible responses. Nevertheless, in some ganglia we recorded for up to 5000 ms to detect possible late onset responses.

Behaviour of ganglia in freely contracting tissues

We studied deformations of Di-8-ANEPPS-labelled ganglia during muscle movement in flat sheet preparations consisting of longitudinal, circular muscle and myenteric plexus. The movements of the tissue and the dye-labelled ganglia were recorded with a CCD camera (LU165M, Lumenera Co., Ottawa, ON, Canada) and the imaging software AMCap 9.10 (Noël Danjou) and Infinity Capture 4.0.2 (Lumenera Co.).

Mechanical stimulation of ganglia and neurons

Two techniques were applied to deform ganglia and ganglion cells in order to mimic deformations observed in freely contracting tissue. Firstly, von Frey hairs were used to apply a defined force onto the ganglion. Secondly, we used intraganglionic injections of small volumes of Krebs solution.

As von Frey hairs we used whiskers (exerting forces between 0.30 and 2.70 mN) from a cat which were connected to a motorized micromanipulator (DC-3K, Märzhauser, Wetzlar, Germany) to control advancement of the hair. Hairs were first positioned just above the ganglion and then advanced to create ganglion deformation.

For intraganglionic volume injections we inserted a micropipette into a fibre tract close to the site where it entered the ganglion. The micropipette was filled with the same oxygenated and buffered Krebs solution that was used for superfusing the preparation (identical composition, pH and osmolarity). Volumes were injected by a pressure-controlled picospritzer (Parker Hannifin Co., Cleveland, OH, USA). As it is not possible to impale the fibre tract achieving a tight seal we used pressure-ejections ranging from 34 to 138 kPa. The pressure was adjusted to values that led to deformation of the entire ganglion as visually inspected.

Pharmacology

For pharmacological studies the following substances were added to the Krebs solution perfusing the tissue: 1 μm nifedipine (Sigma, Schnelldorf, Germany) and 0.02 μmω-conotoxin GVIA (Alomone Labs, Jerusalem, Israel) were perfused for 20 min while 200 μm hexamethonium bromide (Sigma) and 10 μm capsaicin (Sigma) were perfused for 30 min. In some experiments 20 min perfusion of the entire tissue with a low Ca²⁺/high Mg²⁺ Krebs solution, containing (in mm) 98 NaCl, 4.7 KCl, 16 MgCl₂.6H₂O, 1.2 NaH₂PO₄, 25 NaHCO₃, 0.25 CaCl₂.H₂O, 11 glucose, was used.

Immunohistochemistry

In order to study the neurochemical code of the mechanosensitive neurons immunohistochemistry was performed. Tissue specimens were fixed overnight at room temperature in a solution containing 4% paraformaldehyde and 0.2% picric acid in 0.1 mol l⁻¹ phosphate buffer and then washed (3 × 10 min) in phosphate buffer. The samples were permeabilized in NaHCO₃/Na₂CO₃ (pH 8.6)/NaN₃ (0.1%)/glycerol at 50% for 30 min, 80% for 30 min and 100% for 120 min followed by several washes in phosphate buffer. Whole mount preparations were first incubated in PBS/NaN₃ (0.1%)/horse serum (4%) for 1 h at room temperature followed by 24 h and 3 h incubation with the primary and secondary antibody, respectively. As primary antibodies we used rabbit anti-calbindin (Calb; 1:1000; Chemicon, Limburg, Germany), mouse anti-NeuN (1:500; Chemicon), goat anti-choline acetyltransferase (ChAT; 1:100; Chemicon) or rabbit anti-nitric oxide synthase (NOS; 1:500; Alexis, Grünberg, Germany). Besides specifications provided by the suppliers, specificity of primary antibodies were previously published for rabbit anti-Calb (Reiche et al. 1999), mouse anti-NeuN (Van Nassauw et al. 2005), goat anti-ChAT (Li & Furness, 1998; Pfannkuche et al. 1998), and rabbit anti-NOS (Kummer et al. 1992).

As secondary antibodies we used donkey anti-rabbit conjugated to carbocyanin (CY2; 1:200; Dianova, Hamburg, Germany), donkey anti-mouse conjugated to 7-amino-4-indodicarbocyanin (CY5; 1:500; Dianova), donkey anti-goat conjugated to carbocyanin (CY2; 1:200; Dianova) or donkey anti-mouse conjugated to biotin (1:50; Dianova). Biotin-conjugated secondary antibody was visualized using streptavidin conjugated with aminomethylcoumarin acetate (AMCA; 1:50; Dianova).

Finally, specimens were washed in PBS, mounted on poly-l-lysine-coated slides and coverslipped with a solution of PBS (pH 7.0)/NaN₃ (0.1%) containing 65% glycerol. The preparations were examined with an epifluorescence microscope (Olympus, Japan), equipped with appropriate filter blocks. Pictures were acquired with a video camera connected to a computer and controlled by Scion image software (Scion Corp., Frederick, MD, USA). Frame integration and contrast enhancement were employed for image processing.

Data analysis and statistics

We counted the number of dye-labelled neurons in the field of view and analysed the number of mechanosensitive neurons per ganglion and the frequency of action potentials. For signal analysis we used Neuroplex 8.3 (RedShirtImaging), Igor Pro 6.03 (Wavemetrics Inc., Lake Oswego, OR, USA) and Image J 1.32j (Wayne Rasband, National Institutes of Health, USA) software. The statistical analyses were performed with SigmaStat 3.1 (Systat Software Inc., Erkrath, Germany) and SigmaPlot 9.0 (Systat Software Inc.). All data are presented as mean ± standard deviation or, when not normally distributed, as median value together with the 25% and 75% quartiles. Differences in the spike frequency between several control stimulations were performed with a rank sum test. To test the differences in action potential frequency before and after drugs application we used one-way analysis of variance on ranks or repeated measures analysis of variance on ranks. Differences were considered significant when P was < 0.05.

Results

Behaviour of ganglia in freely contracting tissues

Ganglion deformation during tissue contractions was observed in 30 ganglia. During muscle activity ganglia became briefly distorted resembling the twisting movements of seaweed floating in the water (Fig. 1A; Supplemental Movie 1, available online only). We made no attempts to analyse the deformation in detail as the only reason to perform these experiments was to verify ganglion deformation in a freely contracting tissue.

Figure 1.

A, deformation of a Di-8-ANEPPS-stained ganglion during contractile activity of a flat sheet preparation from the ileum. The two frames were taken from a movie which can be accessed online as Supplemental material (Movie 1) and demonstrate that the lower part of the ganglion is distorted during muscle movements. B, neuronal deformation during intraganglionic volume injection. The four frames were taken from Movie 3 (online Supplemental material). Ganglion deformation is evident during volume injection. This is paralleled by deformation of individual neurons as illustrated for the neuron encircled by white lines. The response of this neuron to volume injection is shown in the left trace below the frames; signals have been filtered with a Butterworth filter (low pass 200 Hz, high pass 15 Hz). The bar below the trace indicates the onset and end of the volume injection. The very first deflections are mechanically induced artefacts (arrows) due to the volume injection. After that the neuron fires 4 action potentials. The trace on the right shows the response of the same neuron to electrical stimulation of interganglionic fibre tracts. The first action potential is a compound action potential reflecting axonal spike discharge (see Schemann et al. 2002). This is then followed by a fast EPSP. C, deformation during von Frey hair stimulation. The four frames were taken from Movie 2 (online Supplemental material). The white dotted circle indicates the stimulation site where the von Frey hair touched the ganglion. The von Frey hair deformed a relatively small area of the ganglion while distant areas of the ganglion showed no noticeable deformation. The white arrows mark two neurons: neuron 1 is located within the deformed area whereas neuron 2 is located in the non-deformed area. Below are the responses of the two neurons to von Frey hair stimulation; signals have been filtered with a Butterworth filter (low pass 150 Hz, high pass 25 Hz). The duration of the stimulation is indicated by the bar below the trace; the von Frey hair stimulation remained for the entire recording period. Neuron 1 fired 4 action potentials; the first deflection is the mechanical artefact (marked by an arrow). In contrast neuron 2 did not show any artefact and no action potential discharge. D, response of a neuron to two deformation stimuli (intraganglionic volume injection) which were applied 2 s apart. The neuron responded with action potential discharge after both stimulations (6 spikes with the first and 4 spikes with the second stimulation); signals have been filtered with a Butterworth filter (low pass 250 Hz, high pass 13 Hz). The bars below the traces indicate onset and end of the volume injection. The first deflection after the mechanical stimulus is the mechanical artefact (marked by arrows). The neuron in E (left trace) responded with 5 spikes to mechanical stimulation with the von Frey hair; the duration of the stimulation is indicated by the bar below the trace. The part highlighted in grey is shown on an expanded time scale in the trace on the right side in order to illustrate the fast onset of spike discharge; the first spike occurred 6 ms after stimulus onset. Signals have been filtered with a Butterworth filter (high pass 9.3 Hz).

Identification of rapidly adapting mechanosensitive enteric neurons (RAMEN) with von Frey hairs

With the von Frey hair technique recordings were performed in 52 ganglia from 16 guinea pigs. The mean number of nerve cells in the field of view was 33 ± 9. The von Frey hair was deforming a relatively small area of the ganglion while distant areas of the ganglion showed no noticeable deformation (Fig. 1C; online Supplemental Movie 2). Von Frey hair stimulation with 0.30 and 0.70 mN forces evoked no neural response (10 ganglia, 4 guinea pigs). Increased forces of 1, 1.6, 2 and 2.7 mN caused stronger deformations associated with discharge of one or more action potentials in neurons located within the deformed area. The number of responding neurons and the action potential discharge was directly related to the stimulus strength (Fig. 2A). Both the number of responding neurons as well as the spike frequency gradually increased from 12 ± 8% to 21 ± 13% and 1 Hz to 2.0 Hz, respectively.

Figure 2.

A, spike discharge in mechanosensitive neurons gradually increased with increasing forces of the von Frey hairs (from 0.7 to 2.7 mN). Asterisks mark significant increase of spike discharge compared to a 0.7 mN von Frey hair. Number of neurons are given in parentheses. Data were obtained from different ganglia probed with different von Frey hair forces. Recordings are from a mixed population of neurons. B, spike discharge increased with increased durations of intraganglionic volume injections; increasing the pulse duration from 200 to 400 ms significantly enhanced the spike discharge (significance marked by asterisks). Number of neurons are given in parentheses. Data were obtained from the same neurons which all responded to both 200 and 400 ms injections. C and D, reproducibility of the responses to mechanical stimulation by intraganglionic volume injections. Four consecutive stimulations (c1–c4) were applied every 15 min in 13 ganglia and responses from 85 neurons were analysed. The injection pipette remained in the same position. C shows that the number of responding neurons remained unchanged. D illustrates that the spike discharge of the 85 neurons remained stable along the four injection stimuli.

The von Frey hair evoked an initial fast deformation followed by a sustained deformation till the hair was retracted after the end of the recording period. The response of mechanosensitive neurons consisted of an immediate rapidly adapting spike discharge which occurred only during the initial phase of deformation, hence they are referred to as rapidly adapting mechanosensitive enteric neurons (RAMEN). Touching and deforming the ganglion with the von Frey hair always caused mechanical artifacts which can be seen as deflections in the traces. Those deflections had particular shapes and time courses which allowed them to be easily distinguished from action potentials. In some ganglia the mechanical artifact was small enough to conclude that the first spikes occurred 2–4 ms after the onset of the stimulus (Fig. 1E). The spike discharge lasted on average 813 ± 527 ms. The time of occurrence of the last action potential in the volley was 1258 ± 526 ms after the onset of the stimulus. The latest ever recorded spike occurred 1968 ms after the stimulus.

The von Frey hair technique had a major disadvantage. It was impossible to deform the exact same area of the ganglion twice as with every retraction the von Frey hair was slightly displaced. At the same time there was a slight displacement of the ganglion. Even with the high precision motorized manipulator it was not possible to reposition the von Frey hair onto the same spot. This made reproducibility tests and further pharmacological investigations of the response impossible.

Identification of RAMEN after intraganglionic volume injection

With the intraganglionic volume injection it was possible to achieve rapid reproducible deformations of the entire ganglion (Fig. 1B; online Supplemental Movie 3). Despite short pressure pulses of 200 ms or 400 ms, the deformation was sustained for 160 ± 90 s. During this time the volume was redistributed in the ganglionic network and the ganglion regained its original shape. Thereby, it was possible to record both the response to the early dynamic and late sustained deformation.

Deformation in response to intraganglionic volume injection was performed in 90 ganglia from 50 guinea pigs. The mean number of nerve cell bodies in the field of view was 35 ± 10. Deformation evoked a spike discharge in 25 ± 11% of neurons (range 3–71%). The response pattern was identical to the one seen with the von Frey hairs and consisted of rapidly adapting spike discharge. About 30% (107 neurons) of the RAMEN fired one action potential only. All other neurons fired a volley of action potentials ranging from 2 to 28. As with the von Frey hair stimulation, intraganglionic volume injection also caused mechanical artifacts which can be seen as deflections in the traces. Those deflections had particular shapes and time courses which allowed them to be distinguished from action potentials. In some ganglia the mechanical artifact was small enough to conclude that the first spikes occurred 3–5 ms after the onset of the stimulus. The spike discharge lasted on average 302 ± 231 ms. The time of occurrence of the last action potential in the volley was 447 ± 231 ms after the beginning of the stimulus. The median value of the action potential frequency in the responding neurons was 2.0 (1.0/4.0) Hz. In six ganglia recordings of up to 5 s after the injection were performed. All responding neurons fired action potentials within 1000 ms after the stimulus. Therefore, in all remaining experiments recording periods of a maximum of 2000 ms were used. In three ganglia from two guinea pigs we performed 1.8 s-long recordings 1 and 2 min after the onset of the volume injection. In no case have we observed any ongoing or late onset spike discharge.

In 13 ganglia from 12 guinea-pigs we found a high reproducibility of the responses to 4 mechanical stimulations 10–15 min apart. The number of responding cells per ganglion and the spike discharge remained stable (Fig. 2C and D). In 5 ganglia we tested reproducibility with 11 stimulations given every 15–20 min: neither the number of responding neurons nor the action potential discharge showed significant variations (data not shown).

In 10 ganglia from 6 guinea pigs we performed experiments with different durations of injection which would lead to increased volume injection and deformation. Increasing the pulse duration from 200 to 400 ms increased the percentage of responding neurons per ganglion from 16 ± 11% to 22 ± 12%. Moreover, the spike discharge significantly increased from 0.9 to 2.4 Hz (Fig. 2B).

In four ganglia two deformation stimuli were applied within a 2 s interval and revealed similar responses although a decrease in the number of responding neurons (31 ± 5%vs. 22 ± 11%) and in the spike frequency 1 (0.4/2.0) Hz versus 0.3 (0.3/1.3) Hz was observed (Fig. 1D). Importantly, these experiments demonstrated responses to two fast consecutive mechanical stimuli. The observed decrease in spike discharge was most likely due to the fact that the ganglion was still deformed in response to the first injection because it took several minutes to redistribute the injected volume and to regain original ganglion shape. Thus, the additional rapid deformation during the second injection is occurring at a time when the ganglion is still deformed. The degree of dynamic changes in ganglion shape was therefore smaller than during the first injection. These findings support the relevance of dynamic changes for the responsiveness of RAMEN.

Occurrence of fast EPSPs in RAMEN

Without exception we recorded in all mechanosensitive neurons fast EPSPs (Fig. 1B). Their amplitude remained unchanged after the mechanical stimulation (5 ganglia, 3 guinea pigs): the median amplitude expressed in percentage ΔF/F was 0.8 (0.4/1.0) before and after the mechanical stimulation. All fast EPSPs were strongly suppressed by hexamethonium (4 ganglia from 4 guinea pigs): the median amplitude was 0.9 (0.7/1.0)%ΔF/F before and 0.04 (0.0/0.1)%ΔF/F after hexamethonium application. The fast EPSPs were fully abolished by ω-conotoxin (5 ganglia, 5 guinea pigs; 1.4 (1.1/1.9)%ΔF/F vs. 0.0%ΔF/F). Likewise, fast EPSPs were blocked in the presence of low Ca²⁺/high Mg²⁺ Krebs solution (8 ganglia from 4 guinea pigs; 1.1 (0.0/1.7)%ΔF/F vs. 0.0%ΔF/F).

Pharmacology of mechanosensitive responses in RAMEN

The percentage of responding neurons remained unchanged during nifedipine perfusion (13 ± 9%vs. 12 ± 10%; 8 ganglia, 8 guinea pigs). Likewise, the action potential discharge remained unchanged (Fig. 3).

Figure 3.

Responses of mechanosensitive neurons to intraganglionic volume injections are not changed by nifedipine, hexamethonium, capsaicin or ω-conotoxin and increased in low Ca²⁺/high Mg²⁺ Krebs solution. Spike discharge during control stimulations (Ctrl) are shown as white whisker plots, spike discharge after treatment as grey whisker plots. Note that mechanical stimulation in low Ca²⁺/high Mg²⁺ evoked a significant increase in spike discharge which remained at a higher level after washout and reperfusion of normal Krebs solution; see text for further explanation. Number of neurons are given in parentheses. Significant increase is indicated by asterisks.

The percentage of mechanosensitive neurons remained unchanged in the presence of hexamethonium (20 ± 8%vs. 20 ± 9%; 10 ganglia, 9 guinea pigs). The action potential discharge also remained unchanged (Fig. 3).

The percentage of mechanosensitive neurons was not changed in the presence of ω-conotoxin which fully blocked fast EPSPs (18 ± 8%vs. 16 ± 5%; 7 ganglia, 7 guinea pigs). The action potential discharge also remained unchanged (Fig. 3).

In eight ganglia from four guinea pigs we were able to analyse the effect of low Ca²⁺/high Mg²⁺ Krebs solution to block synaptic release on deformation-evoked responses. Lowering Ca²⁺ had in some ganglia detrimental effects on dye recordings as the signal disappeared only a few seconds after the low Ca²⁺/high Mg²⁺ solution entered the bath. These negative effects were irreversible for that particular ganglion. This may have been caused by changes in membrane properties associated with low Ca²⁺ concentrations which then interfere with the dye incorporation into the membrane. We therefore only analysed data from those experiments where we recorded signals before and after washout of the low Ca²⁺/high Mg²⁺ solution. The percentage of mechanosensitive neurons did not significantly change in the presence of the low Ca²⁺/high Mg²⁺ solution (30 ± 8%vs. 25 ± 8%). The spike discharge significantly increased from 2.7 (1.8/4.4) Hz to 4.4 (1.8/8.9) Hz and remained at 4.0 (2.7/9.8) Hz even after washout (Fig. 3). This increase in spike discharge did not appear to be due to synaptic blockade by low Ca²⁺/high Mg²⁺ solution as we did not observe an increased spike discharge with hexamethonium or ω-conotoxin. Low Ca²⁺/high Mg²⁺ solution elevated spike discharge in AH neurons (Grafe et al. 1980) as well as in S-like neurons (Schemann & Wood, 1989). This effect very likely contributed to the deformation-evoked increase in spike discharge in the presence of low Ca²⁺/high Mg²⁺ solution. A detailed analysis revealed different response patterns in the presence of low Ca²⁺/high Mg²⁺ solution and after washout. In 33% of RAMEN we observed an increased spike discharge to mechanical deformation during perfusion of low Ca²⁺/high Mg²⁺ solution which returned to control levels after washout. In 25% of RAMEN the spike discharge increased during perfusion of low Ca²⁺/high Mg²⁺ solution and remained at a high or even higher level after washout. The behaviour of these neurons was mainly responsible for the, on average, increased spike discharge during perfusion of low Ca²⁺/high Mg²⁺ solution and after washout. In 18% of RAMEN the response remained constant before, during and after washout of low Ca²⁺/high Mg²⁺ solution. In 15% of RAMEN the response to mechanical deformation decreased in low Ca²⁺/high Mg²⁺ solution and recovered to control values after washout. We have not observed a decreased spike discharge in the presence of ω-conotoxin. This would suggest that the effect of low Ca²⁺/high Mg²⁺ solution is not due to blockade of synaptic transmission because fast EPSPs were equally abolished in low Ca²⁺/high Mg²⁺ solution and ω-conotoxin. The most likely reason for the decreased spike discharge in some of the neurons in the presence of low Ca²⁺/high Mg²⁺ solution is the strong Ca²⁺ component of action potentials in AH neurons (Hirst & Spence, 1973; Hirst et al. 1974).

Defunctionalization of extrinsic primary afferents nerves by capsaicin (Weber et al. 2001) did not change the percentage of mechanosensitive neurons (23 ± 7%vs. 19 ± 5%; 6 ganglia, 6 guinea pigs). The frequency of action potentials also remained stable (Fig. 3).

Neurochemical coding of RAMEN

In 11 ganglia from 6 guinea pigs we studied the neurochemical code of RAMEN. While 56% of mechanosensitive neurons were immunoreactive for NeuN, 29% were immunoreactive for Calb. Double staining revealed that 29% of mechanosensitive neurons were immunoreactive for both NeuN and Calb, whereas 27% were immunoreactive for NeuN only (Fig. 4A). With 44% the highest percentage of mechanosensitive neurons was neither Calb nor NeuN immunoreactive. It is noteworthy, that the Calb-immunoreactive RAMEN fired action potentials in response to deformation at a significantly lower frequency than the Calb-negative RAMEN (1.9 ± 0.9 Hz vs. 5.1 ± 6.4 Hz).

Figure 4.

A, neurochemical code of mechanosensitive neurons. The image on the left illustrates a Di-8-ANEPPS-stained ganglion. The orange dots mark the neurons which responded to intraganglionic volume injection. Orange dots with white circles mark neurons which are neither Calb nor NeuN immunoreactive. Orange dot with green circle marks a neuron which is Calb and NeuN immunoreactive and orange dot with blue circle marks a neuron which is NeuN immunoreactive. Calb and NeuN staining of the ganglion is shown in the central and right images, respectively. B, responses in DiI-traced circular muscle motor neuron and interneuron. The arrows mark the locations of the DiI-traced neurons. Images were taken after recordings with Di-8-ANEPPS which explains the background staining. The traces below show the response of the circular muscle motor neuron (left trace) and the interneuron (right trace) to intraganglionic volume injection. In both traces strong mechanical artifacts are evident at the beginning of the injections (marked by arrows). These are followed by 3 spikes in the motor neuron and 5 spikes in the interneuron. The signals have been filtered with a Butterworth filter (low pass 300 Hz, high pass 20 Hz). Bars below traces indicate onset and end of intraganglionic volume injection.

In 13 ganglia from 4 guinea pigs, ChAT and NOS immunohistochemistry revealed that 72% of mechanosensitive neurons were immunoreactive for ChAT, 22% were immunoreactive for NOS and the remaining 6% were negative for both markers.

Multifunctionality of RAMEN

Retrograde labelling with the tracer DiI can be used to identify enteric neurons with defined functions (see Brookes, 2001). To provide evidence for multifunctionality of RAMEN we recorded deformation-induced action potential discharge in DiI-traced interneurons or motor neurons. The response pattern was very similar to that which we observed in freshly dissected preparations.

Intraganglionic volume injections evoked action potentials in 31 ± 33% of DiI-labelled interneurons (14 ganglia, 4 guinea pigs) at a median frequency of 2.8 (1.9/4.5) Hz. The spike discharge lasted 309 ± 277 ms. The time of occurrence of the last action potential in the volley was 455 ± 295 ms after the onset of the stimulus. The last action potential ever recorded was 1097 ms after the beginning of the stimulus.

In 19 ganglia from 8 guinea pigs intraganglionic injections evoked spike discharge in 47 ± 36% of the DiI-labelled circular muscle motor neurons at a median frequency of 4.7 (2.8/7.5) Hz. Spike discharge lasted 467 ± 317 ms. The time of occurrence of the last action potential in the volley was 606 ± 295 ms after the onset of the stimulus. The last action potential ever recorded was 1118 ms after the beginning of the stimulus.

The number of neurons which were sufficiently filled with DiI to reveal their detailed morphology was too small to make firm conclusions. However, shapes of mechanosensitive DiI-traced interneurons suggested either large, oval cell bodies with multiple long processes resembling Dogiel type II neurons or small cell bodies with numerous short processes resembling Dogiel type I neurons (Fig. 4B and C). Whenever possible to analyse, circular muscle motor neurons appeared to have Dogiel type I morphology.

Discussion

This study revealed three main findings. Firstly, we identified mechanosensitive myenteric neurons in the guinea pig ileum which fired action potentials in response to ganglion deformation. Secondly, the response pattern of mechanosensitive neurons suggests that they behave like rapidly adapting mechanosensors that respond to dynamic changes. We therefore suggest referring to these neurons as RAMEN (rapidly adapting mechanosensitive enteric neurons), a terminology that best describes their response pattern. Thirdly, we provided evidence that most, if not all, RAMEN are multifunctional, serving sensory, integrative and/or motor functions.

The non-invasive voltage-sensitive dye imaging had the benefit of simultaneous recordings from all neurons in the field of view. Signals were acquired from all neurons independent of their size. This allowed us to quantify the proportion of mechanosensitive neurons without any bias which often occurs with intracellular electrodes because of the difficulties in achieving stable recordings in small-sized neurons. Despite these advantages voltage-sensitive dye imaging has limitations. At first, it is not possible to continuously record for several minutes as long recording periods compromise the signal to noise ratio due to dye bleaching. Despite this limitation we are confident that we did not miss late onset responses, at least in those mechanosensitive neurons identified in the present study, for the following reasons. Firstly, we were able to record for up to 5 s without seeing late onset spike discharge. Secondly, in some ganglia we performed several recordings (1.8 s) 1 and 2 min after the onset of the mechanical stimulation and never observed any spike discharge. The mechanical stimulation caused slight movements of the ganglion and therefore also of neuron cell bodies. Any movement will cause signals as pixels detect changes in fluorescence. We had to filter our signals to reduce the amplitude of those low frequency signals caused by mechanical artifacts. This prevented identification of slowly developing depolarization before or during action potential discharge.

Both stimulation techniques, the von Frey hair and intraganglionic volume injection, produced almost identical spike discharge patterns. Using the von Frey hair technique only neurons located close to the site of ganglion deformation responded. The intraganglionic volume injection evoked a more reliable deformation of the entire ganglion and was not restricted to a small region. Nevertheless, the volume was never equally distributed throughout the ganglion which resulted in variable degrees of deformation in different areas of a ganglion. The reason for the preferred volume flow is unknown but may be due to intraganglionic differences in cell density and/or variable density of extracellular matrix, both of which are likely to produce different resistances to flow. We therefore cannot rule out that the proportion of RAMEN found in this study (25%) is an underestimation because it was not possible to achieve equal volume flow within the ganglion.

The response of RAMEN to mechanical stimulation is reproducible suggesting that this is a physiological mechanism and not caused by mechanical damage of the neuron. This is also supported by the findings that we could stimulate and record from RAMEN with no significant changes in the response pattern for over 3 h. Other evidence that the ganglion was not damaged came from the immunohistochemical data which revealed normal structure of the ganglion and neurons. Last but not least, the fast EPSP amplitude remained stable after mechanical stimulation.

The action potential discharge of RAMEN never lasted longer than 2 s although the ganglion stayed deformed for much longer periods. It is noteworthy that RAMEN fired action potentials during the initial phase of deformation but not during sustained deformation. This was not due to desensitization or other mechanisms that inhibited spike discharge during the sustained phase of deformation because spike discharge in RAMEN occurred in response to two deformation stimuli applied within 2 s. Thus we conclude that RAMEN respond to dynamic changes only. All experiments were performed in stretched preparations (58% above slack in circular and 18% in longitudinal direction). We cannot rule out that this tissue stretch may have some influence on the ability of RAMEN to respond to sustained deformation. However, this is unlikely because maintained tissue stretch evoked ongoing spike discharge in IPANs in the ileum (Kunze et al. 1998) as well as in mechanosensitive interneurons in the colon (Spencer & Smith, 2004).

When using brief mechanical stimulations Kunze et al. (2000) also recorded from IPANs a pattern of action potential discharge very similar to the one we found. In line with these and our results IPANs in the mouse small intestine also responded to mechanical deformation of myenteric ganglia with von Frey hairs (Mao et al. 2006). This response remained after synaptic blockade and the discharge pattern would suggest activation of rapidly adapting mechanosensors (Mao et al. 2006). However, RAMEN differed from IPANs in various aspects as reported in previous studies (Kunze et al. 1998, 1999, 2000; Mao et al. 2006). Firstly, all RAMEN received fast EPSPs which were absent in IPANs. Secondly, not all RAMEN are Calb or NeuN immunoreactive. Thirdly, RAMEN did not fire action potentials during sustained deformation. Mechanosensitive interneurons in the guinea-pig colon received fast EPSPs (Spencer & Smith, 2004). However, behaviour of RAMEN differed from mechanosensitive interneurons in two main aspects. Firstly, RAMEN did not respond to sustained deformation; secondly, spike discharge in RAMEN rapidly adapted.

Stretch- or deformation-sensitive neurons within the myenteric plexus have been previously described using extracellular recording techniques (Wood, 1970; Ohkawa & Prosser, 1972; Wood & Mayer, 1974; Mayer & Wood, 1975). In these experiments, briefly probing the tissue with the tip of a glass electrode or a platinum wire evoked repeatable responses in myenteric neurons. Some of these neurons responded with tonic discharge and others with phasic spike discharge that stopped during a sustained stimulus with constant intensity; the latter was also observed in RAMEN. The stimulus modalities were different between our and previous studies which makes direct comparisons difficult, and it may be that some of the RAMEN are a subpopulation of previously identified mechanosensitive enteric neurons. In this respect it is noteworthy that some RAMEN were Calb positive and may be identical to those IPANs that responded to brief von Frey hair stimulation (Kunze et al. 2000). However, an important difference remains because all mechanosensitive neurons identified in our study received fast EPSPs including the Calb-positive RAMEN. Mechanosensitivity appears to be a property of numerous enteric neurons that respond to different stimulus modalities with a specific discharge pattern. Moreover, mechanosensitive neurons belong to different neuron classes, at least when considering the classification schemes used so far based on morphology, neurochemical coding, electrophysiological properties or projection patterns (see Brookes, 2001).

The conclusion that RAMEN responded directly to mechanical stimulation is based on several findings. Firstly, the responses were not changed after blockade of synaptic transmission with hexamethonium, ω-conotoxin or low Ca²⁺/high Mg²⁺ suggesting a direct stimulation of mechanosensitive neurons. These data agree with previous findings that responses in mechanosensitive IPANs to probing with von Frey hairs were not diminished after synaptic blockade (Kunze et al. 2000). They also agree with the finding that blockade of the synaptic transmission with low Ca²⁺/high Mg²⁺ did not affect responses in mechanosensitive interneurons (Spencer & Smith, 2004). Secondly, neither the number of responding neurons nor the action potential frequency were changed after long-term capsaicin treatment of the tissue which resulted in defunctionalization of TRPV1-expressing extrinsic afferents (Weber et al. 2001). In addition, perfusion of low Ca²⁺/high Mg²⁺ which would also be expected to block release from visceral afferents, did not suppress responses in RAMEN. Moreover, ω-conotoxin, which effectively suppressed activity in peripheral endings of primary sensory neurons supplying the guinea pig ureter (Maggi & Giuliani, 1991), did not inhibit responses in RAMEN. These findings indicated that peripheral endings of extrinsic visceral afferents did not contribute to the response of RAMEN. Thirdly, mechanosensitivity of RAMEN was independent of muscle tone which is in contrast to findings in IPANs which were activated by sustained tissue stretch (Kunze et al. 1999). The response of IPANs to sustained tissue stretch depended on the opening of stretch-activated channels in the muscle followed by muscle contraction that then led to activation of IPANs (Kunze et al. 1999). All the above argue against involvement of fast or slow synaptic transmission in deformation-evoked spike discharge of RAMEN.

The latency between the stimulus onset and the appearance of the first spike in RAMEN is in the range reported for activation of mechanosensitive channels (less than 5 ms; Christensen & Corey, 2007) or intraganglionic laminar vagal endings (around 6 ms; Zagorodnyuk et al. 2003). Future studies have to address the molecular identity of putative mechanosensitive channels in myenteric RAMEN. It remains open whether non-neuronal structures are additionally involved in transducing the mechanical stimulus. Synaptic activation can be excluded as hexamethonium, ω-conotoxin and low Ca²⁺/high Mg²⁺ did not inhibit spike discharge. Unspecific activation of synapses by deformation is also unlikely because one would expect to see hexamethonium-sensitive responses due to release of acetylcholine, the main transmitter released by enteric synapses. This leaves glia or changes in the extracellular matrix as possible candidates that may be involved in mechanotransduction. This, however, remains speculative as mechanosensitive channels or receptors involved in mechanotransduction need to be identified in future studies.

The results with von Frey hair suggested that the threshold to evoke a response in RAMEN is around 1.0 mN. This value agrees well with the threshold of activation of stretch-sensitive extrinsic afferents which start to respond to mechanical stimulation between 0.9 and 2.0 mN (Lynn et al. 2003). Moreover, in the experiments with intraganglionic volume injection we found a significant difference in the percentage of responding neurons and in the spike discharge after increasing the injection time from 200 ms to 400 ms. This suggests that the threshold of putative mechanosensitive channels is not reached in all RAMEN with stimulation pulses of 200 ms. Future studies have to investigate whether the processes, the soma, or both, encode mechanical stimuli. At least in IPANs, processes appear to be activated whereas soma activity is inhibited by mechanical stimulation (Kunze et al. 2000). In our study it was not possible to distinguish between the contributions of the different mechanosensitive structures to the response as we very likely deformed both soma and neuronal processes. Based on our observation in freely contracting guinea-pig ileum it appears that contractile activity will very likely deform processes as well as somata. Our results suggest that the overall response of RAMEN is excitatory independent of whether the deformation stimulus is encoded by the process or the soma.

It is obvious that mechanosensitive neurons in the ENS have region-specific properties and respond to different stimulus modalities. Some respond to sustained tissue stretch while RAMEN appear to only respond to dynamic changes during deformation. What may be the functional relevance of rapidly adapting mechanosensitive enteric neurons receiving fast synaptic input? The behaviour of RAMEN may be suitable to respond to changes during phasic muscle activity which is most common in the small intestine. The rapid adaptation in RAMEN may be very important in the gut where neurons have to rapidly integrate a variety of different stimuli in order to adjust their output and to appropriately modulate muscle activity. This is relevant to gut behaviour as circumferential distension of the small intestine evoked peristalsis in the presence of an inert meal while the same distension is associated with segmentation when nutrients are present (see Blackshaw et al. 2007). The activity of RAMEN is highly modulated by synaptic input. Thus, there is a low stimulus fidelity which allows setting the gain in a sensory network. The modulation of the response to a stimulus via fast synaptic input allows the enteric circuits to rapidly accommodate to changes in the microenvironment and thereby to appropriately determine gut behaviour. Our results suggest that interneurons and even motorneurons exhibit mechanosensitivity. As such the reflexes that determine peristalsis are modulated directly by the forces that they regulate because many neurons respond to the stimulus that they directly control. The functional consequences for peristalsis of this type of neuronal behaviour remain to be determined. The low stimulus fidelity together with the ability to modulate responsiveness of mechanosensitive neurons may be one way to maintain integrative functions and to prevent the gut from becoming ‘epileptic’. It remains a challenge to incorporate multifunctionality of mechanosensitive enteric neurons into the established hard-wired circuitry of peristaltic reflex activity. It needs to be considered that it is currently unknown whether RAMEN have any role in initiating enteric reflex circuits which control muscle activity at sites distant from the stimulus location or rather reflect the ability of the system to locally adjust mechanosensitivity of a particular neuronal network.

RAMEN have no unique morphology or neurochemical code and belong to functionally different classes. The experiments performed on retrogradely traced neurons revealed mechanosensitivity of some of the interneurons and of even more circular muscle motor neurons. These experiments have to be expanded to also study mechanosensitivity of longitudinal muscle motor neurons or neurons projecting to the mucosa. Nevertheless, our findings provided strong evidence for the concept that RAMEN are multifunctional. It seems from this and previous studies that the functional allocation of enteric neurons, in particular those with mechanosensitive properties, requires some revision. The identification of RAMEN supports the existence of multifunctional enteric neurons which may fulfil sensory, integrative and motor functions. Multifunctionality in the ENS may pave the way for novel concepts on sensory transmission in the gut.

Glossary

Abbreviations

ChAT

choline acetyltransferase

ENS

enteric nervous system

IPANs

intrinsic primary afferent neurons

NOS

nitric oxide synthase

RAMEN

rapidly adapting mechanosensitive enteric neurons