Evidence for glutamatergic mechanisms in the vagal sensory pathway initiating cardiorespiratory reflexes in the shorthorn sculpin Myoxocephalus scorpius

Without exception, vertebrates, in order to maintain homeostasis, regulate blood pressure and arterial gas concentrations by a variety of feedback loops,such as those that originate in peripheral chemoreceptors and baroreceptors. These receptors send afferent information into the central nervous system,which results in modulation of the activity in neurones generating respiratory rhythms and those controlling the cardiovascular system. While substantial knowledge exists regarding the central integration of activity in cardiorespiratory sensory afferents in mammals(Spyer, 1990; Talman, 1997; Van Giersbergen et al., 1992),nothing is known in fish, a separate, somewhat more primitive, group of vertebrates.

In mammals, the nucleus of the solitary tract (NTS) is the primary synaptic relay in the brainstem, where afferent information from visceral receptors is integrated. The NTS has been subdivided into different subnuclei based on its cytoarchitecture and the afferent and efferent connections of the neurons within it. The medial and lateral commissural subnucleus of the NTS has been shown to be the primary site of termination of cardiovascular afferent fibres,receiving inputs from carotid chemoreceptors, arterial baroreceptors and cardiopulmonary receptors (Loewy,1990; Van Giersbergen et al.,1992). Glutamate, an excitatory amino acid (EAA), is the strongest candidate for the neurotransmitter released by these afferents(Ohtake et al., 1998; Talman, 1997).

Information regarding the location of sensory areas in the medulla important for control of the cardiorespiratory system in fish is sparse, and information about the nature of their neurotransmitters and receptors is essentially lacking. It is documented that the gills are a major site for chemo- and baroreceptors, with their afferent nerves travelling in cranial nerves IX and X (Burleson et al.,1992). In the medulla, the areas of termination of afferent sensory fibres (Xs) are located dorsally and laterally above the sulcus limitans of His, whereas the motor area (Xm) is located ventral and lateral to the sulcus (Meek and Nieuwenhuys,1998). Although in most teleosts a clear NTS is absent, the visceral sensory area forms a continuous column dorso-laterally on either side of the 4th ventricle in the medulla, into which viscerosensory fibres of nerves VII, IX and X terminate in a rostrocaudally ordered fashion(Meek and Nieuwenhuys,1998).

Recently, it has been shown that EAAs are the neurotransmitters in taste pathways in goldfish (Carassius auratus; Smeraski et al., 1998), and immunohistochemistry has shown that glutamate is present in the nodose ganglia and vagal afferents in the shorthorn sculpin Myoxocephalus scorpius(J. Turesson and L. Sundin, manuscript submitted). Taken together, these results suggest that EAAs might also be the neurotransmitters in the general visceral sensory pathways conveying information from chemo- and baroreceptors via vagal and glossopharyngeal nerves.

A first step on the way to establish if glutamate is a functional neurotransmitter in the central processing of baroreceptor and oxygen chemoreceptor information in fish is to determine whether addition of glutamate into the vagal portion of the visceral sensory column elicits cardiorespiratory responses similar to the reflexes activated by stimulation of peripheral chemo- and baroreceptors. Therefore, the primary aim of this paper was to examine whether microinjection of glutamate, sometimes followed by appropriate antagonists, into different sites of the vagal sensory area(the terminal field of vagal afferent fibres characterised as the NTS in mammals) activates cardiorespiratory responses that mimic chemo- and baroreceptor reflexes. If clear responses were obtained, then glutamate could perhaps be used as a `mapping tool'. Accordingly, a second aim was to elucidate whether there was a distinguishable separation of areas in which different responses were elicited that might reflect a topographical arrangement of the central projection of receptor afferents. As knowledge of the central projections of the vagal afferent and efferent fibres in the medulla of the shorthorn sculpin is a prerequisite for reasonably accurate microinjections into the Xs, the initial aim of this study was to locate the Xs and Xm columns in this species, using a neuroanatomical technique.

Shorthorn sculpins Myoxocephalus scorpius L. weighing 172±18 g were caught on the Swedish West Coast by a local fisherman. They were given at least 3 days to recover from the effect of capture in holding tanks at 10°C and normal day/night length. All animal experiments have been approved by the local ethical committee in Gothenburg (No. 299/99).

On the day of surgery, the fish were anaesthetized in seawater containing 100 mg l^-1 MS 222 (ethyl m-amino benzoate; Sigma; 10°C) until breathing movements ceased. They were transferred to a surgical table where the gills were continuously irrigated with cooled, recycled water containing anaesthetic (40-50 mg l^-1 MS 222, 10°C).

The nodose ganglion was located by tracing the exposed branchial nerves centrally. Exposure was via a small incision (approximately 1 cm)made in the epithelium at the dorsal end of the 4th gill arch where it meets the roof of the opercular cavity, the operculum having been reflected forward. Using a 25 μl Hamilton syringe equipped with a 27-gauge hypodermic needle,5-10 μl of Fast Blue (Sigma), as a 2% solution in polyethylene glycol, was injected through the nerve sheath into the ganglion. When visual observation confirmed that the ganglion had turned yellowish in appearance, the needle was withdrawn and the puncture was closed with tissue glue. The incision was sutured and the fish was tagged, then returned to holding tanks for 7-10 days to allow axonal transport (orthograde and retrograde) of the tracer into the projections of the vagus, in the medulla. Each fish was then sacrificed by an overdose of MS 222 and heparin (0.2 ml, 5000 IU) injected into the caudal vein. The fish were exanguinated by perfusion with physiological saline (0.9%NaCl) using a ventral aortic cannula connected to a peristaltic pump. After 10-15 min, when the gills had turned white, the saline was switched to 4%formaldehyde solution and the fish were perfused for a further 15 min. The brain was then carefully dissected from the skull and placed in 4%formaldehyde in 0.1 mol l^-1 phosphate-buffered saline (PBS; pH 7.3)for at least 4-5 h at 4°C. Each brain was then rinsed for 30 min in PBS and stored in PBS containing 30% sucrose as a cryoprotectant. Finally, it was quick-frozen in isopentane cooled in liquid nitrogen and mounted on the stage of a cryostat. Serial, transverse sections, 20 μm thick, were cut,transferred directly to gelatine-coated glass slides and left to airdry overnight. The sections were coverslipped with glycerol mounting media and viewed under a fluorescence microscope (BX60, Olympus) connected to a digital video camera. Pictures were frozen on a TV monitor and captured by computer using the Micro Image software (Micro Image, Gothenburg, Sweden). To visualize the general histology of the labelled sections, some were stained for Nissl substance.

The day before the experiment, the third afferent branchial artery on the left side was cannulated (PE 50 tipped with a PE 10) according to the procedures described for Atlantic cod(Axelsson and Fritsche, 1994). This cannula was used to measure ventral aortic blood pressure(P_VA) and heart rate (f_H) and for the administration of drugs. Measurements of ventilation frequency(f_V) and amplitude (V_AMP) were made using impedance probes, fastened with suture thread stitched through each operculum.

On the day of the experiment, the fish was again anaesthetised (100mg l^-1 MS 222) and lowered into a plastic box placed between the steel bars of a modified stereotaxic frame (model SN-2N; Narishige Instruments,Tokyo, Japan). It was fixed in position with eye bars and also, initially,with a mouthpiece through which re-circulating respiratory water (40-50 mg l^-1 MS 222) flowed. As the animal started to breath spontaneously,the mouthpiece was withdrawn to deliver water approximately 2 cm in front of the snout. Using a dremel tool and vacuum suction, the skull was carefully opened (incision approximately 1.5 cm long) to expose the whole length of the medulla from the middle portion of the cerebellum to the first pair of the spinal nerves. The fish rested in a horizontal position on a height-adjustable platform inside the box. A standpipe controlled the water level, which was adjusted to cover the gills yet allowed the medulla to be uncovered.

Drugs were delivered into specific locations in the medulla from a single-barrel microinjection pipette (tip size 10-15 μm). Movements of the pipette were controlled by a micromanipulator (SM15 equipped with a base SM-15M). Injection volumes of 40-100 nl were delivered over a period of ≤1 s by applying a pulse of pressurized N₂ using a pressure injector(model PLI-100; Harvard Medical Systems, Holliston, MA, USA). The volume of drug delivery was controlled by changing the injection pressure, and the actual volume of the injection was determined by viewing the movement of the fluid meniscus in the barrel of the pipette, which was of known internal diameter, using a microscope (×50 magnification) equipped with a calibrated eyepiece micrometer.

The cannula was connected to a pressure transducer, the signal was amplified (4Champ; Somedic AB, Sollentuna, Sweden) and the leads from the impedance probes were connected to an impedance converter (model 2991; UFI,Morro Bay, CA, USA). The cardiorespiratory variables were continuously recorded to paper (recorder model 3701, LR 8100; Yokogawa, Tokyo, Japan), and the data were collected online, via data-acquisition software(Labview version 5.0; National Instruments, Solna, Sweden) onto a computer. Sampling frequency was 20 Hz, and mean values were subsequently created at 10s intervals. From the pulsed blood pressure and ventilation signals, fH and fV were derived using a Labview-based calculation program. The injection signal from the PLI-100 pressure injector was also sampled, which allowed exact timing of the injection with the cardiorespiratory responses.

In three fish, efforts were made to use a decerebrate and spinalectomized preparation to avoid any potential influence of anaesthesia on central reflex mechanisms. However, that approach was abandoned because these fish bled substantially, displayed low ventral aortic blood pressures (0.9-1.3 kPa) and never started spontaneous breathing. Instead of decerebration, light anaesthesia (40-50 mg l^-1 MS 222) was used, as it permits a smaller hole in the skull to be made and leaves the spinal cord intact, maintaining sympathetic outflow to the vessels. This approach significantly improved the blood pressure (2.0-3.8 kPa) in the animals, who now also started to breathe spontaneously. These blood pressures are comparable with those in an unanaesthetised and free-swimming sculpin(Fritsche, 1990).

Preliminary trials using 0.1 mmoll^-1 and 1.0 mmoll^-1of glutamate were employed to determine a concentration that would give clear and distinctive responses. The concentration (10 mmoll^-1) and volume range (40-100 nl) chosen are comparable with those commonly used for microinjections into the medulla of rats(Canesin et al., 2000; Dhruva et al., 1998; Le Galloudec et al.,1989).

The general protocol for the experiments was as follows. When stable cardiorespiratory parameters were established, usually around 40-60 min after securing the fish in the stereotaxic frame, unilateral microinjections of glutamate were made sequentially into different discrete areas of the Xs column along the medulla. When a response was elicited, the following injection was postponed (between 10 min and 60 min) until stable parameters were again established. The microinjection pipette was advanced through the sensory area (as determined by the nerve tracing) in steps of 0.1 mm down to a depth of 0.3 mm along a rostrocaudal direction from 2.0 mm rostral to -1.0 mm caudal of obex, in steps of 0.5 mm. Lateral coordinates applied were 0.3 mm,0.5 mm and 0.7 mm lateral to the midline. Each animal was subjected to 20-50 different injections, although not every coordinate received an injection in each animal and the order of injection sites varied among animals. When a clear and concise response was elicited, the pipette was raised, rinsed and vacuum loaded with the vehicle (0.9% NaCl) then lowered again to the same depth, and a control injection with the same or a larger injection volume was performed. To control for tachyphylaxis and possible mechanical damage of an injection, repetitive injections of glutamate in exactly the same area at 10 min intervals were performed in at least one site in each animal, and glutamate was sometimes re-injected into the area where a previous vehicle control had been carried out.

In five animals, at the end of the above-described general protocol, a site that had previously elicited a distinct bradycardia was again injected before an intra-arterial injection of the agonist atropine (1 mg kg^-1). After 20 min, the agonist injection was repeated at the same site.

For experiments using the N-methyl-D-aspartate (NMDA) receptor antagonist MK-801, the general protocol was as follows. Having obtained a cardiac response to unilateral microinjection of 40-100 nl of glutamate, the pipette was raised, rinsed then vacuum loaded with the antagonist (3 mg ml^-1), which was injected (40-100 nl) at the same depth. The agonist was then reloaded and the injection repeated. The time between the application of the antagonist and the agonist was 5-10 min. To further control for the specificity of the blockade, the pipette was lowered 0.1 mm beyond or moved 0.5 mm in a sagittal direction from the MK-801 saturated area, and the agonist injection was repeated.

Monosodium L-glutamate, dizocilpine (5R,10S)-(+)-5-methyl-10,11-dihydro-5H-dibenzo [a,d]cyclohepten-5, 10-imine hydrogen maleate (MK-801) and atropine were obtained from Sigma and dissolved in 0.9% NaCl.

Comparison of the cardiovascular effects before and after glutamate was made using a paired t-test. The same test was used for the comparison of the cardiovascular effects of glutamate before versus after MK-801, and before versus after atropine. Differences were considered significant at P<0.05. All values are means ± S.E.M.

A time period of 7-10 days (at 10°C) was sufficient to complete ortho-and retrograde axonal transport of Fast Blue from the nodose ganglia to the medulla (approximately 0.5-1.0 cm) in M. scorpius. In four animals,both the sensory (Xs) and the motor (Xm) columns contained the tracer, whereas in one animal only the motor column stained blue. The sensory nucleus and the motor nucleus of the Xth cranial nerve form two continuous columns parallel with and dorso-lateral to the fourth ventricle (4V) along the length of the medulla (Figs 1, 2). Even though both the motor and sensory columns were heavily marked with fluorescent tracer, the columns were clearly separated along their lengths(Fig. 1A). The sensory column is positioned slightly rostral to the motor column, and the total length of the two columns that fluoresce is 2500-2800 μm in each fish(Fig. 2).

Retrograde labelling with Fast Blue also identified cells in the nucleus ambiguus (NA) of the vagal motor column. This nucleus consists of a relatively small number of neurones located ventro-laterally with respect to the Xm and separated from it by a tract of nerve fibres(Fig. 1D).

The anterior ends of the two columns are located approximately 1.5-2.0 mm rostral of obex, and the posterior end of both columns stretches to 1 mm caudal of obex. At this caudal extremity, commissural fibres(Fig. 1B) cross above the central canal, to constitute the commissural nucleus of the Xs, while fibres crossing beneath are of motor origin (Fig. 1C).

Motoneurons, probably belonging to the nucleus ambiguus, in a ventrolateral position to the motor column were observed from obex to 0.9 mm rostral of obex. Observations of Fast Blue-filled neurons in this position were not common, probably resulting from incomplete staining and a vague and disperse nucleus in this species.

As the dorsal sensory and the ventral motor columns are located rather close to each other, adjacent at depths of 0.4-0.5 mm from the dorsal surface of the medulla, the results reported here are restricted to injections made down to 0.3 mm.

To control for non-specific pressure and volume effects of the injections,the vehicle (0.9% NaCl) alone was delivered into the same sites (equal or larger volume) where glutamate had previously elicited a response. The vehicle produced small insignificant blood pressure increases only in one fish. In addition, two fish displayed bradycardia, a concomitant blood pressure decrease and a short apnea when accidentally large volumes (150-300 nl) were injected. When smaller injection volumes were applied at the same sites no responses were evoked.

Repeated injections of glutamate into the same site at 10 min intervals did not decrease the responsiveness of the animal. Hence, the repetition of a glutamate injection into a vehicle-applied site always produced a response.

Dependent on the injection site in the Xs, glutamate elicited decreases in heart rate (f_H) and either increases or decreases in ventral aortic blood pressure (P_VA), ventilation frequency(f_V) and amplitude (V_AMP). A tachycardia was never observed. Occasionally, an injection elicited a transient apnea. The coordinates and responses for each injection in all animals are summarised in Fig. 3. Sometimes, an injection elicited a response in a single cardiorespiratory parameter (Fig. 4) or, at other times, in two or three or all of them (Figs 5, 6). Blood pressure and the ventilatory responses could be biphasic, often reflecting the relationship between f_H and P_VA on the one hand and fV and V_AMP on the other hand. In clear cases,such as a bradycardia-induced drop in P_VA(Fig. 7), this depressor event was excluded from the summarised data in Fig. 3. Only depressor responses that were apparently independent of a change in f_H are included (e.g. Fig. 5). The convention adopted with respect to ventilatory parameters is that reciprocal changes in frequency and amplitude are recorded as the appropriate excitatory response. Thus, an increase in amplitude that led to a decreased frequency or a marked increase in frequency that resulted in reduced amplitude have been recorded as an increase in either variable.

Some salient features of these seemingly complex patterns of response summarised in Fig. 3 are:

1. Caudal of obex: no responses to injection in the midline but a bradycardia and reduction in P_VA, fV and V_AMP following injection into more lateral sites, 0.5 mm caudal of obex.

2. At obex: a bradycardia plus specific pressor/depressor responses, including a definitive depressor area 0.3 mm lateral to obex. Some sites induced increases in fV.

3. Rostral (0.5 mm) of obex: bradycardia plus pressor/depressor effects;increases in fV and V_AMP.

4. Rostral (1.0 mm) of obex: bradycardia; pressor responses but no depressor responses rostral of this level; both fV and V_AMPincreased or decreased.

5. Rostral (1.5 mm) of obex: bradycardia; pressor responses; only increases in V_AMP, increases or decreases in fV.

6. Rostral (2.0 mm) of obex: bradycardia; pressor responses; increases in V_AMP, independent decreases in fV.

In summary, certain main features emerge with regard to each recorded variable: a bradycardia and specific pressor responses were induced by injections at most reactive sites, both caudal and rostral of obex; depressor responses were obtained at and immediately (0.5 mm) caudal or rostral of obex; fV was increased by injection into some sites at and up to 1.5 mm rostral of obex, while a decrease in fV accompanied injection into sites just caudal (0.5 mm) and 2.0 mm rostral of obex; V_AMP was increased by injection into most areas rostral of obex and decreased by injections just caudal (0.5 mm) and rostral (1.0 mm) of obex.

It is clear from these data that, while there is some evidence for rostrocaudal distribution of projections from specific receptor-mediated responses, a reflex bradycardia is induced by injection of glutamate into most sites either side of obex. The induced bradycardia sometimes resulted in cardiac arrest, in one case for up to 4 min(Fig. 7).

Along the fourth ventricle at the medial (0.3 mm lateral) injection sites,glutamate always induced a bradycardia that sometimes was very marked, in one extreme case without a heart beat for up to 4 min(Fig. 7). In five animals, the non-competitive antagonist of NMDA receptors, MK-801, was injected into bradycardia-inductive sites and the glutamate injection was repeated. MK-801 abolished the rapidly glutamate-induced bradycardia (Figs 7, 8). The bradycardia was also blocked by a systemic injection of atropine(Fig. 9).

The neuroanatomical study revealed that the sculpin had vagal sensory (Xs)and motor (Xm) columns similar in all respects to those described in other teleost fish (Burleson et al.,1992; Meek and Nieuwenhuys,1998). As such, it is not new information but does serve the initial role of identifying injection sites within the Xs in this species. The present study is to be extended by further use of fluorescent markers to study the detailed topography of the vagal supply to specific target organs, such as the heart and branchial arches, including both the location of cell bodies and their processes. This will enable central injection and eventual recording from these specific sites to identify areas integrating cardiorespiratory reflexes and generating central interactions. In the present study, we show that glutamate injected into the vagal sensory column in the dorso-lateral medulla in fish elicits several cardiorespiratory responses, demonstrating that it may be an important neurotransmitter released by the afferents of baro- and chemoreceptors, as has been suggested for mammals(Schaffar et al., 1997; Sykes et al., 1997; Talman et al., 1980). The most ubiquitous response obtained was a bradycardia, which can be explained by the fact that this is a component of both the baroreceptor and chemoreceptor reflex responses in fish (Taylor et al.,1999). However, these separate reflexes may be distinguished by their accompanying changes in physiological variables, with the baroreflex leading to a reduction in blood pressure and the chemoreflex leading to an increase in both blood pressure and ventilation. Examination of these responses reveals some evidence of a topographic separation of the projections from these different reflexogenic areas. Our mapping showed that the cardiorespiratory responses characteristic of a chemoreflex (bradycardia,increased blood pressure, ventilation frequency and amplitude; Fritsche and Nilsson, 1993) are located rostral to obex, whereas responses typical of a baroreflex (a bradycardia accompanied by a decrease in blood pressure; Lutz and Wyman, 1932) are located at the level of obex or just caudal to it, i.e. in the commissural segment.

Interestingly, the specific location of the terminal field within the NTS is crucial for the production of respiratory or cardiovascular reflexes in mammals (Dhruva et al., 1998; Marchenko and Sapru, 2000). In mammals, both chemo- and baroafferents terminate in the commissural nucleus of the NTS (Loewy, 1990; Van Giersbergen et al., 1992). Within this restricted area, the baroreflexogenic field is located rostral to the chemoreflexogenic field (Dhruva et al.,1998). With the finding of a depressor area lateral to obex at the beginning of the commissural segment, our results seem similar to the location of depressor areas in the commissural nucleus in mammals. However, caudal to the depressor area, we found no evidence for a chemoreflexogenic zone. In fact, most of the injections in this region produced no responses at all. Instead, responses simulating chemoreflexes were elicited rostral of obex. This is consistent with the fact that most peripheral chemoreceptors have been described as being located on the gills of fish. The gill arches are innervated sequentially by the IXth glossopharyngeal nerve then the first four branches of the vagus. The fifth branch innervates the viscera, including the heart (Burleson et al., 1992; Taylor et al., 1999). Thus,chemoreceptor afferents will travel in the more rostral projections into the Xs from the branchial branches of the vagus nerve, which terminate rostral of obex (Taylor, 1992). Caudal of obex, at the commissural nucleus, the afferents of the most caudal root terminate. Thus, this structure only receives sensory information from visceral afferents rather than from the gill arches(Kanwal and Caprio, 1987; Lazar et al., 1992; Morita and Finger, 1987). The finding of a specific depressor site 0.3 mm lateral to obex substantiates that the barostatic reflex in fish, implicating changes in the resistance of the vessels, may project through the area innervating the heart(Taylor, 1992).

In addition to the distribution of reaction patterns simulating a specific reflex, injection of glutamate could sometimes elicit a unitary response such as an increase in respiratory amplitude or a decrease in blood pressure. This suggests specific areas in the Xs for reflex control of each cardiorespiratory variable. Identification of these `single' responses may have been prejudiced by the extracellular injection technique. Although different response patterns could be obtained with a pipette movement of just 0.1 mm, the spread of the injection solution will probably cause stimulation of many neighbouring neurons. With smaller injection volumes and even smaller steps during mapping,a better picture of this single response topography may evolve. Nevertheless,this single response topography may accord with the physiological evidence for more than one population of oxygen receptors in fish, which elicit different cardiorespiratory parameters dependent on their peripheral location (specific gill arch or extrabranchial) or orientation (monitoring respiratory water or blood oxygen levels) (Burleson and Smatresk, 1990; Smatresk et al., 1986; Sundin et al., 1999, 2000). If glutamate is, as in mammals (Schaffar et al.,1997; Sykes et al.,1997; Talman et al.,1980), the neurotransmitter released by the afferents of baro- and chemoreceptors in fish, there should be glutamate receptors on target neurons binding the EAA. Indeed, our results show that the non-competitive NMDA receptor antagonist MK-801 effectively blocked the glutamate-induced bradycardia. Similarly, NMDA receptors mediate a glutamate-induced bradycardia in rats (Canesin et al., 2000; Colombari et al., 1997). In addition, the data following systemic injection of atropine show that the bradycardic responses produced by microinjection of glutamate along the 0.3 mm lateral coordinates are due to parasympathetic neurotransmission, so that a glutamatergic mechanism for chemo- and baroreflex activation in fish seems likely. This is borne out by the demonstration that NMDA receptors in the NTS are involved in the bradycardic element of both the chemoreflex(Haibara et al., 1995) and the baroreflex in rats (Canesin et al.,2000).

In conclusion, glutamate applied to different areas in the Xs of the sculpin evoked responses simulating chemo- and baroreflexes. There was some evidence for a topographic separation of these two areas with a chemoreflexogenic zone rostral to a baroreflexogenic zone. The ubiquitous,glutamate-induced bradycardia depended on NMDA receptors in the sensory pathway and was of muscarinic cholinergic, and therefore vagal, origin. Evidence has thus been presented that glutamate may have been present as a key neurotransmitter in the reflex control of the cardiorespiratory system from early in vertebrate evolution. Thus, this work may provide a first step in establishing the fundamental central mechanisms for the processing of chemo-and baroreceptor signals in all vertebrates.

This work was supported by grants from the Swedish Research Council (VR),Magnus Bergvalls Stiftelse, The Royal Swedish Academy of Sciences and Stiftelsen Lars Hiertas Minne.