A Peripheral Arterial O₂-Sensitive Pathway to the Respiratory Oscillator of the Shore Crab Carcinus Maenas

It is now well known that the ventilatory activity of aquatic animals is correlated mainly with oxygenation of the ambient medium (Dejours, 1981). In decapod crustaceans, water is propelled through each gill chamber by the action of a modified process of the second maxilla, called the scaphognathite. These processes act as bilateral suction pumps promoting a forward flow of water from inhalant apertures at the base of the walking legs to exhalant openings beside the mouthparts on either side of the animal. Hydrostatic pressure recordings in the branchial cavities allow measurements of scaphognathite beat frequencies and also provide quantitative information about the flow rate of water through the gills. Assuming constant resistance to flow, this rate is directly related to the negative pressure level in the actively ventilated gill chamber. Such observations have clearly shown that ventilation rate increases in hypoxic water, whereas the gill flow rate is depressed under hyperoxic exposure (see, for example, Dejours and Beekenkamp, 1977; Taylor, 1982; Cameron and Mangum, 1983; Jouve-Duhamel and Truchot, 1983).

The localization of the O₂-sensitive areas responsible for these responses, as well as the corresponding afferent neural pathways, has seldom been studied. A priori, a chemosensitivity to oxygen could derive from the site of respiratory rhythm generation within the central nervous system (CNS) itself or originate from peripheral areas, facing either the external water or the blood. A central O₂ sensitivity in Crustacea was first suggested by McMahon and Wilkens (1975) to explain a relatively long latency in the resumption of the ventilatory rhythm in deeply hypoxic lobsters upon re-exposure to aerated sea water. Wilkens et al. (1989) have recently addressed this question directly by examining the responses of the crab ventilatory central rhythm generator to changes in oxygen tension of saline perfusing the isolated thoracic ganglion. Although transient increases in ventilatory rates could be seen to follow steps from hyperoxic to hypoxic levels of (thereby suggesting a direct compensatory ability of the CNS), the long-term response was a decline in rhythm frequency. On this basis, Wilkens et al. (1989) concluded that peripheral O₂ receptors or higher control centres in the CNS are needed to sustain in vivo compensatory responses to changes in oxygen tension.

Direct arguments for peripheral chemosensitivity have come from observations on crayfish (Larimer, 1964; Massabuau et al. 1980; Massabuau and Burtin, 1984). By recording gill chamber hydrostatic pressure and blood O₂ tension in the pericardial cavity of the intact animal during and after transient application of hyperoxic water at the inhalant openings, Massabuau et al. (1980) showed that a ventilatory depression appeared before the hyperoxic blood had reached the heart and thence the central nervous system. This suggested that O₂-sensitive receptors, providing neurally mediated information to the centre, must be located peripherally in the gill region, probably in contact with the blood on its way to the heart. However, the afferent pathways for this or other peripheral O₂ chemosensory reflexes have yet to be identified.

In the present study, we have investigated the mechanism of O₂ sensitivity in the shore crab Carcinus maenas. Rhythmic ventilatory movements, consisting of alternate levation and depression of the blade of each scaphoghathite, are driven in crabs by bilateral respiratory oscillators located in the anterior part of the ventral thoracic ganglion. The activity of an oscillator can be readily monitored in vitro by recording its motor output in two different ventilatory nerves, depressor (DN) and levator (LN), with each containing 2–4 excitatory axons per scapho-gnathite muscle, of which there are five depressors and five levators (Young, 1975). In the completely isolated thoracic ganglion, perfused through the sternal artery (Simmers and Bush, 1980), the central O₂ responsiveness of the respiratory centre was examined by varying oxygen tension in the perfusing saline (see also Wilkens et al. 1989). Additionally, in a semi-isolated preparation of the ganglion, with intact ventilatory nerves and circulation to the mouthparts and scaphognathites, tests were made for the presence of O₂-responsive sensory inputs by selective deafferentation, interruption of blood flow and application of drugs known to be selectively active on O₂ chemoreceptors in vertebrates. Finally, we examined the possibility that O₂-sensitive chemosensory information could be conveyed to the respiratory centre from or via the cerebral ganglion.

Male shore crabs Carcinus maenas, 60–100 g wet mass, obtained from local fishermen, were kept in the laboratory in flowing sea water at seasonal temperatures for at least 1 week before use. They were fed twice weekly with mussels or fish meat. Experiments reported here were conducted on a total of 114 crabs.

Electromyographic recordings of scaphognathite activity were obtained using Teflon-coated wires implanted bilaterally into levator muscles of each scaphognathite through small holes drilled in the dorsal carapace (Fig. lAi) using procedures described by Rezer and Moulins (1983). The crab was placed on a bed of gravel in an aquarium with opaque walls (Fig. lAii). Oxygen partial pressure of the water was controlled by bubbling gas mixtures (air+N₂) or (O₂+N₂) obtained with mass flowmeters (Tylan). In some animals, the circumoesophageal connectives were surgically sectioned through a window opened in the dorsal carapace and then replaced and glued with cyanoacrylate adhesive. This operation caused loss of postural control but did not obviously affect animal viability for at least 1 or 2 days. The placement of muscle recording wires and transection of connectives were verified by dissection at the end of the experiment.

After autotomy of the chelae and walking legs, the dorsal carapace, the heart, viscera, gills and brain were removed. The thoracic ganglion, anterior nerve trunks and ventral arterial system to the mouthparts region were left intact (Fig. IBi; Simmers, 1978). The preparation was pinned down on Sylgard in a dish containing Carcinus saline. The saline composition was (mmol l⁻¹): Na⁺, 500; K⁺, 12; Ca²⁺, 12; Mg²⁺, 20; Cl⁻, 576, buffered with 10 mmol l⁻¹ Tris-maleate to pH 7.2–7.4 (Roberts and Bush, 1971). The thoracic ganglion and the ventro-anterior region were continuously perfused with the same saline at a flow rate of 2 ml min⁻¹ through a cannula inserted into the descending sternal artery. All in vitro experiments were performed at room temperature (15–20°C).

Starting from the semi-isolated preparation (Fig. 1Bi), the anterior nerve trunks were cut and the perfused thoracic ganglion was removed from the thorax and pinned on Sylgard in a small Petri dish containing saline (Simmers and Bush, 1983). Since the leg arteries were cut upstream of the autotomy plane, they were ligated to reduce leaks and improve perfusion of the ganglion.

In both isolated and semi-isolated preparations, oxygen partial pressure of the perfused saline was controlled with a gas exchanger placed in the perfusion line immediately before entry to the sternal artery (Fig. IBii). The gas exchanger was made from several metres of gas-permeable silicone tubing (Silastic) packed in a plastic syringe through which gas mixtures circulated at 11 min⁻¹. The gas mixtures at various values were prepared from compressed air, nitrogen or oxygen tanks using mass flowmeters (Tylan; precision ±1 %). Continuous measurements with an oxygen electrode showed that 99 % of the expected value was attained at the level of the sternal artery about 5 min after switching from one gas to another.

In some experiments on semi-isolated or isolated preparations, potassium cyanide (Merck), lobeline (Sigma) or almitrine bismesylate (Servier) was injected as a bolus into the perfusion line downstream of the gas exchanger. The injected doses were 0.5–50 μg kg⁻¹ for KCN, 0.5-5/zg kg^-1 for lobeline and 2–3 mg kg⁻¹ for almitrine in volumes varying between 10 and 100 μl. Dye injection showed that the transit time for the bolus to reach the anterior sternal artery was about 10 s.

Extracellular recordings of spontaneous motor output to a scaphognathite were obtained in both semi-isolated and isolated preparations with glass suction electrodes placed on the peripheral levator and/or depressor branches of the ventilatory nerve root (Fig. IBi). Activity was monitored en passant either from intact nerves of the semi-isolated preparation or from the central stumps of transected nerves in the isolated preparation. Signals were amplified and displayed with conventional equipment and permanent records were made with a Gould ES 1000 electrostatic recorder.

Ventilatory cycle frequency (f) was determined from measurements of the period (f=l/p Hz) between the onset of consecutive bursts either in chronically recorded scaphognathite muscles (e.g. Fig. 2A), or in the first active motoneurone per cycle in in vitro nerve recordings (e.g. Fig. 3A). In all in vivo experiments, the animal was allowed to recover from electrode implantation for at least 3 h before recordings and O₂ trials commenced.

For a given trial, ventilatory frequency analyses were made from the time (about 10 min in vivo and about 5 min in vitro) that the O₂ tension had reached its prescribed value in the gill chamber or the sternal artery. Three 30 s sequences were then measured over a recorded 5 min period, and the combined mean frequency value ± S.D. was plotted on a histogram as a function of the prevailing (expressed in kPa). In some cases (values quoted in text), data obtained under the same O₂ conditions in different experiments were pooled and the overall mean values ± S.E.M. were calculated.

Respiratory responses to changes in water oxygenation were studied in intact animals to establish reference values of the scaphognathite beat frequency for comparison with responses observed in isolated, more or less deafferented preparations. Electromyographic recordings demonstrated an inverse relationship between respiratory frequency and the oxygen partial pressure of the ambient medium. The data shown as an example in Fig. 2A, B were from an animal which

Respiratory responses to changes in water oxygenation were studied in intact animals to establish reference values of the scaphognathite beat frequency for comparison with responses observed in isolated, more or less deafferented preparations. Electromyographic recordings demonstrated an inverse relationship between respiratory frequency and the oxygen partial pressure of the ambient medium. The data shown as an example in Fig. 2A, B were from an animal which ventilated bilaterally throughout the whole range tested. In normoxic water ⁠, the frequency was about 2 Hz, increasing to 3–4.5 Hz in hypoxia and decreasing slightly in hyperoxia. In 10 similar experiments, mean scaphognathite frequency was 2 ± 0.3Hz in normoxia, 4.6 ± 0.45Hz in hypoxia at and 1.4 ± 0.25 Hz in hyperoxia at ⁠. At the latter values, moreover, crabs often displayed respiratory pauses occurring either unilaterally or bilaterally. These cycle frequency values obtained electrophysiologically are in full agreement with previous observations obtained by recording the hydrostatic pressure changes in the gill chamber due to the movements of the scaphognathite itself (Dejours and Beekenkamp, 1977; Jouve-Duhamel and Truchot, 1983).

To test whether the cerebral ganglion is involved in these oxygen-dependent respiratory responses, the circumoesophageal connectives, which link the brain to the thoracic ganglion, were severed in the whole animal. When successful (A=3), this operation suppressed the so-called ‘shadow reflex’, a transient arrest of scaphognathite activity upon a visual stimulation. Fig. 2C shows the results of one such experiment in which the scaphognathite frequency was recorded at various water levels, before and after severing the circumoesophageal connectives. Compensatory respiratory responses to changes in ambient oxygen level persisted in the operated animal (in this case for 3 days after surgery), indicating that the brain is not implicated in the expression of O₂ chemosensitivity.

In vitro preparations allowed application of an oxygen-related stimulus via the vascular route by perfusing the ventral arterial system through the sternal artery that normally carries haemolymph pumped downwards from the heart. After descending through the thoracic ganglion, and supplying capillaries to it, the sternal artery gives rise to segmental arteries radiating to the walking legs and chelae, and two major vessels that extend in an anteroposterior direction (Sandeman, 1967). The anterior artery runs forward medially, below the circumoesophageal connectives, eventually branching into smaller vessels that supply the scaphognathites and mouthparts (see Fig. 5A).

Initially, we used a semi-isolated preparation in which the ventilatory nerves to the scaphognathites and the vascular supply to the ventral anterior region were kept intact (see Materials and methods). Changes of in the perfusing saline were applied either abruptly by switching from hyperoxic to hypoxic conditions and vice versa (Fig. 3B), or progressively by graded steps from hyperoxia to hypoxia (Fig. 3A,C). Such preparations (N=53) perfused with hyperoxic saline exhibited continuous bilateral activity (14%), unilateral activity (21%) or bilateral ventilatory pauses (65%); a variability similar to that observed in intact hyperoxic animals (Duhamel-Jouve, 1982). By contrast, switching from hyperoxic to hypoxic saline elicited rhythmic bilateral activity in 92 % of these preparations with a cycle frequency that was higher (2.26 ± 0.2Hz) than in those preparations that expressed bilateral or unilateral activity in hyperoxic conditions (0.7 ± 0.1 Hz).

The increase in frequency was long lasting and showed no sign of temporal adaptation. Fig. 3B shows an example of this response and also demonstrates that it is fully reversible. When saline was decreased step by step from hyperoxia to hypoxia, the preparation also exhibited a progressive increase in respiratory frequency (Fig. 3A,C). Thus, respiratory responses qualitatively similar to those found in the whole animal when exposed to various water oxygen levels can be demonstrated in a semi-isolated preparation perfused with a saline at various levels. This indicates that an O₂ chemosensitivity was still present in the semiisolated preparation. As evident in Table 1, however, all absolute frequency values and the changes in frequency observed at different levels in semi isolated preparations were significantly lower than those recorded in the intact animal.

In the semi-isolated preparation, the perfusing saline at various levels has access first to the thoracic ganglion containing the respiratory centre and then to the ventro-anterior arterial system that supplies the mouthparts region (see Fig. IBi). The observed O₂ chemosensitivity could thus originate either centrally or peripherally in the perfused areas. The possible existence of O₂-sensitive structures in the respiratory centre was therefore tested on a perfused preparation of the completely isolated thoracic ganglion (see Materials and methods). As shown in Fig. 4A,B, the responses observed were completely different from those recorded from semi-isolated preparations. In contrast to the intact animal and semi-isolated preparations, all 11 isolated ganglia tested displayed robust bilateral rhythmic activity when perfused with hyperoxic (100 kPa) salines. Under these conditions, moreover, respiratory frequencies were relatively high (1.27 ± 0.2Hz), but they decreased steadily with saline ⁠, reaching mean values of 0.60 ± 0.07Hz in normoxia (21 kPa) and below 0.37 ± 0.05Hz in deep hypoxia (5.3 kPa). When perfusion of hyperoxic saline was resumed, respiratory frequency rose back to higher levels, with a marked transient rebound in some preparations. That a decrease in O₂ supply depresses the activity of the central oscillator in a graded manner is consistent with a metabolic effect, and indicates a lack of direct O₂ chemosensitivity of the oscillator itself or of areas located elsewhere in the thoracic ganglion. We therefore conclude that a peripheral O₂-sensitive area is located on the vascular side in the ventro-arterial region which is absent in the isolated preparation but is normally supplied with arterial haemolymph via the sternal artery. This notion is supported by the following experiments involving either interruption of saline flow to the anterior region or section of the ventilatory nerves which supply this region in semi-isolated preparations.

The mouthparts region of each side is innervated from the thoracic ganglion by several nerves of which three innervate the ipsilateral scaphognathite (Young, 1975): the levator nerve (LN) sending motor axons to all levator and one depressor (D2a) muscles and containing sensory fibres notably from the oval organ, a mechanoreceptor located in the scaphognathite blade (Pasztor, 1969); the depressor nerve (DN) containing motor axons to all depressor muscles except D2a; and the sensory nerve (SN) whose function is presently unknown. Semi-isolated preparations (N=18) in which all anterior nerves except the ventilatory nerve bundles were severed (Fig. 5A) exhibited O₂ responsiveness as evident from the first three changes in oxygen tension in the representative example of Fig. 5B. Subsequent bilateral transection of ventilatory nerves under hypoxic perfusion (N=12) elicited an immediate reduction in respiratory frequency. A switch to normoxic and then hyperoxic saline induced an increase in frequency, a response

The mouthparts region of each side is innervated from the thoracic ganglion by several nerves of which three innervate the ipsilateral scaphognathite (Young, 1975): the levator nerve (LN) sending motor axons to all levator and one depressor (D2a) muscles and containing sensory fibres notably from the oval organ, a mechanoreceptor located in the scaphognathite blade (Pasztor, 1969); the depressor nerve (DN) containing motor axons to all depressor muscles except D2a; and the sensory nerve (SN) whose function is presently unknown. Semi-isolated preparations (N=18) in which all anterior nerves except the ventilatory nerve bundles were severed (Fig. 5A) exhibited O₂ responsiveness as evident from the first three changes in oxygen tension in the representative example of Fig. 5B. Subsequent bilateral transection of ventilatory nerves under hypoxic perfusion (N=12) elicited an immediate reduction in respiratory frequency. A switch tq normoxic and then hyperoxic saline induced an increase in frequency, a response which was the opposite from that observed before nerve section but resembled the response recorded from the isolated thoracic ganglion (see Fig. 4). These effects of ventilatory nerve transection were also observed in preparations where anterior non-ventilatory nerves were left intact. When the ventilatory bundles were severed under hyperoxic perfusion (N=6), either an increase in frequency was observed or a bilateral rhythm appeared if the preparation was previously quiescent. In addition, reducing the saline after nerve section led to a decreased frequency. Thus, selective transection of the ventilatory nerves in semiisolated preparations suppresses the hyperventilatory response to hypoxic perfusion as well as the ventilatory depression observed with perfusion of hyperoxic saline. Moreover, in other semi-isolated preparations (N=3) where the ventilatory nerves remained intact, suppression of O₂ responsiveness was also observed upon interruption of flow to the anterior ventral region by ligating or severing the anterior sternal artery upstream of the bifurcations of the scaphognathite arteries (Fig. 5A,C). All these results argue in favour of a population of O₂-sensitive chemoreceptors being located peripherally within the arterial system of the mouthparts region, probably around the scaphognathites, and whose afferent information is conveyed to the centre via the ventilatory nerves.

Peripheral O₂ chemoreceptors located in carotid bodies of vertebrates are known to be reversibly stimulated by drugs such as sodium or potassium cyanide, lobeline or almitrine bismesylate. The effects of these agents on respiratory motor output were tested in our perfused crab preparations, under conditions where spontaneous ventilatory activity was minimal, i.e. in hyperoxia for the semiisolated preparation and moderate hypoxia for the isolated ganglion. Various doses contained in 10–100 μl of saline were injected into the perfusion line to the sternal artery. Injections of similar volumes of vehicle saline were without detectable effect. As seen in the recordings of Figs 6 and 7A for the semi-isolated preparation, a dose of 15 μlg kg⁻¹ potassium cyanide elicited a reversible increase in respiratory frequency. Under hyperoxic saline, this preparation was active on one side only. About lmin following KCN injection, the respiratory frequency increased on the active side while a tonic discharge developed on the other, previously silent, side, turning progressively into a rhythmic activity. After peaking at 6–8 min, the activity on both sides declined and the preparation returned to a quiescent state after 9–10 min. That the rhythm-generating centres remained viable after KCN treatment is evident in the bottom recording of Fig. 6 and the right side of Fig. 7A; about 20min after the original KCN injection, transection of the ventilatory nerves elicited resumption of strong bilateral rhythmicity that continued for as long as the now isolated ganglion remained exposed to hyperoxic saline. Similar responses to KCN were observed in 14 preparations and at doses as low as 1.5 μgk g⁻¹. As shown in Fig. 7B, potassium cyanide at doses between 0.5 and 50 μgk g⁻¹ had no detectable effect on ventilatory activity recorded from the completely isolated thoracic ganglion (N=6), demonstrating that the respiratory stimulation caused by the drug was originating at a peripheral location. Furthermore, administration of KCN after section of the ventilatory nerves in semi-isolated preparations (N=5) elicited no increase in respiratory frequency.

Similarly, lobeline injection at.doses varying from 0.5 to 5 μg kg⁻¹ elicited no responses in the isolated thoracic ganglion (N=6) but caused a reversible increase in respiratory frequency of semi-isolated preparations (N=3) perfused with hyperoxic Ringer (Fig. 8A). Almitrine bismesylate at doses between 2 and 3 mg kg⁻¹ was also stimulatory in semi-isolated preparations (N=8), the response appearing with a similar delay but being of shorter duration (Fig. 8B), and again having no detectable effect on the completely isolated ganglion (N= 10).

Although it is well known that ventilatory activity is strongly dependent on mbient oxygen levels in Crustacea, the location of the O₂-chemosensitive areas as well as the corresponding afferent pathways have not been clearly identified. In this context, in vitro perfused preparations such as those used in this study offer a number of distinct advantages over intact animals. First, the motor output from the respiratory oscillator can be directly and conveniently recorded from the ventilatory nerves, in the virtual absence of many sensory ‘disturbances’ present in the whole animal. Second, topical application of the oxygen stimulus is possible at known locations, particularly via the vascular route as performed in this study. Third, various degrees of deafferentation of the respiratory centre make it possible to analyse the contribution of spatially separate structures and pathways in the observed responses. Using this approach, we succeeded in locating an O₂-chemosensitive afferent pathway which most probably participates in the O₂-dependent ventilatory responses in the whole animal. Our results not only demonstrate the existence of a peripheral O₂ chemosensitivity within the arterial system of the anterior ventral region but, importantly, exclude possible locations within the central nervous system itself.

First, we found that the cerebral ganglion is not required for the expression of the respiratory responses to oxygen in the crab. This could be inferred from the observation that ventilatory frequency increased with decreasing of the perfusing saline in the semi-isolated preparation in which the cerebral ganglion was absent. We also found that surgical section of the circumoesophageal connectives left the response unaffected in the whole animal. This excludes the presence of an O₂ chemosensitivity on the anterior appendages, e.g. the antennae, which are innervated from the cerebral ganglia. External O₂-sensitive chemoreceptors located on the book gills and intercoxal membranes of Limulus polyphemus are apparently responsible for an immediate resumption of ventilatory movements when O₂ is re-admitted to the animal previously kept in anoxic water (Page, 1973; Crabtree and Page, 1974). Whether such an external O₂ chemosensitivity also exists in Carcinus maenas has not been assessed in this study, but in any case it could not arise from the anterior appendages.

Second, a decrease rather than an increase in respiratory frequency was observed when was reduced in the saline perfusing the completely isolated thoracic ganglion which contains the respiratory oscillator. In a recent study on the crab thoracic ganglion in vitro (Wilkens et al. 1989), a similar direct relationship between rhythm frequency and 0₂ tension was found to prevail after relatively long exposure times. These authors, however, also reported that a transient long-latency increase in ventilation frequency could accompany a step decrease in perfusing ⁠, and this was interpreted as a direct O₂-dependent metabolic effect on the ventilatory oscillator itself. Although such transient responses were not observed in our experimental conditions, it seems clear that neither the respiratory oscillator itself nor other structures in the ganglion are responsible for the sustained compensatory ventilatory responses observed in the whole animal. That an overall decline in respiratory frequency follows a decrease in the perfusing saline could most simply be explained by a metabolic depression resulting from reduced O₂ supply to the ganglionic tissue. In addition to the strong hyperventilation in hypoxic water, an abrupt drop in ventilatory activity is usually observed as very low ambient oxygen levels in the whole animal (Taylor, 1976; Massabuau and Burtin, 1984). This may be related to the metabolic depression of the central oscillator we observed in the isolated thoracic ganglion under hypoxia. However, Massabuau and Burtin (1984) showed in hypoxic crayfish that upon re-admission of oxygenated water, ventilatory activity increased before oxygenated blood reached the heart and thus the suboesophageal ganglion, which contains the respiratory centre, suggesting that ventilatory depression in deep hypoxia was also peripherally mediated.

Since experiments on the isolated ganglion demonstrated that the increase in respiratory frequency under hypoxic saline could not originate centrally, the responses observed on the semi-isolated preparation indicated that a peripheral O₂ chemosensitivity originated from within the arterial system at an as yet unknown location in the ventro-anterior region of the thorax. Suppression of the response by severing ventilatory nerves alone, as well as by interrupting saline flow with ligation of the anterior sternal artery, strongly argues in favour of a population of (O₂-sensitive chemoreceptors being located around the scaphognathites. The hypoxia-induced increase in respiratory frequency recorded in the semiisolated preparation was, however, much smaller than that observed in the whole animal. This could be interpreted in two ways. First, central depression under hypoxia may be antagonistic to O₂-responsive excitatory inputs from the periphery. Second, it is highly probable that only a fraction of the O₂-sensitive chemoreceptors present in the whole animal was stimulated in the semi-isolated preparation. Receptor structures in the gill region, such as those possibly existing in crayfish (Massabuau et al. 1980), may well be present in the crab. This location would strategically be better for early sensing of an oxygen change in the ambient medium. If such gill receptors exist in the crab, they would signal O₂ changes well before the hypoxic arterial blood could reach the scaphognathite region. The insensitive reflex pathway demonstrated in the present study may serve to complement such early-warning detection in the gills by providing a continuous (tonic) feedback of information to the respiratory oscillator about O₂ levels in the circulating blood, albeit at longer latency.

Several other aspects of our observations deserve comment. The O₂-chemo-sensitive pathway found in the semi-isolated preparation was strongly stimulated by agents known to be excitatory for vertebrate peripheral O₂ chemoreceptors (Heymans and Neil, 1958). Besides providing suggestive evidence for the existence of such chemoreceptors in the crab, our observations also indicate that basic mechanisms of O₂ chemosensitivity may be the same throughout the animal kingdom. In contrast, it is noteworthy that O₂ chemosensory inputs to the centre may be either excitatory or inhibitory, according to the prevailing conditions, as has been suggested for the lugworm Arenicola (Dejours and Toulmond, 1988). This is best seen in experiments involving section of the ventilatory nerves. When nerve section was performed under hypoxic perfusion, an immediate reduction in respiratory frequency was observed, indicating that the peripheral input was exciting the oscillator. Conversely, nerve section under hyperoxic perfusion elicited an increased activity, revealing that peripheral chemoreceptor input was depressing the oscillator under hyperoxia. Whether these contrasting responses could be effected by two different classes of O₂ chemoreceptors remains to be established.

We thank Dr J.-Ch. Massabuau (Strasbourg) and Dr H. Guenard (Bordeaux) for kind gifts of almitrine bismesylate. This study was supported, in part, by a Postgraduate Student Grant from the Moroccan Government to HZ.