Arterial blood gas levels and cardiovascular function during varying environmental conditions in a mudskipper, Periophthalmodon schlosseri

Mudskippers (Gobiidae: Oxudercinae) actively emerge from water and spend part of their life on the surface of intertidal mudflats, although the degree of terrestriality varies widely among species. The terrestrial activities of mudskippers include foraging, courtship and territorial defence (Clayton, 1993; Murdy, 1986). Many mudskippers construct burrows in the soft substratum of the mudflat and use them for reproduction and for refuge (Atkinson and Taylor, 1991; Brillet, 1976; Clayton, 1993). Because of the highly reducing environment of the mudflat (Scholander et al., 1955), mudskipper burrows are usually filled with nearly anoxic water (El-Ziady et al., 1979; Garey, 1962; Gordon et al., 1978). Thus, mudskippers routinely experience a wide range of environmental conditions from total emersion to severe aquatic hypoxia.

To elucidate the physiological bases of the adaptation of mudskippers to the amphibious mode of life, the gas exchange of these fishes in water and air has been extensively studied. Two recent review. s have compiled data on aerial and aquatic O₂ uptake rates for mudskippers (Graham, 1997; Martin and Bridges, 1999). These data indicate that mudskippers can generally maintain a constant irrespective of the respiratory medium (air or wate. r). Fewer studies have measured CO₂ elimination rates in mudskippers, but the available data suggest that, as with is similar in air and in water. Steeger and Bridges (1995) determined the respiratory exchange ratio to be 0.77 in air and 0.81 in water for Periophthalmus barbarus. Schöttle (1931) reported a ratio of 0.72–0.78 for three species of Periophthalmus in air, but she did not measure aquatic gas-exchange rates. Other marine intertidal fishes are able to release CO₂ into air at a similar rate as into water (Graham, 1997; Martin and Bridges, 1999). This is in contrast to freshwater air-breathing fishes, which have very low RER values during air exposure (Martin, 1995).

In contrast to the amount of information available on gas exchange of mudskippers, little is known about the blood gas levels and cardiovascular physiology of these fishes. This lack of information stems mainly from the difficulty of working with tiny mudskippers, which usually weigh from a few grams to 50 g. Their small size precludes the application of physiological techniques such as chronic cannulation of blood vessels and placement of blood flow probes. Electrocardiograms have been recorded for a few species of mudskipper (Aguilar and Graham, 1995; Garey, 1962; Gordon et al., 1969), and one study used a pulsed Doppler technique to measure blood flow for Periophthalmus barbarus (Bridges et al., 1995).

Periophthalmodon schlosseri, which inhabits intertidal mudflats of Southeast Asia (Murdy, 1989), is an especially suitable subject for physiological investigations of mudskippers because of its exceptionally large size. The largest specimen of P. schlosseri used in this study weighed 225 g, and all weighed over 100 g. Thus, adult fish of this species easily allow chronic cannulation for repetitive blood sampling and blood pressure measurement. Apart from its large size, P. schlosseri is outstanding among mudskippers in that it is so well adapted for air breathing that aquatic confinement may be stressful. Two recent studies of P. schlosseri have demonstrated that th.e aquatic of P. schlosseri was lower than the aerial (Kok et al., 1998; Takeda et al., 1999) and, in addition, Takeda et al. (1999) showed that the fish was unable to repay an O₂ debt in water but that it could in air. The reduced capacity for aquatic respiration is probably a result of its specialized gill morphology. P. schlosseri has branched gill filaments, thick gill rods and fused secondary lamellae, all of which appear to be ill-adapted for aquatic gas exchange (Low et al., 1988).

Recent studies of P. schlosseri from Malaysia have provided extensive data regarding respiratory gas tensions in its environment. At low tide, P. schlosseri spends long periods on the exposed mudflat surface (Kok et al., 1998) or immerses its body in the hypoxic water pools of its burrow with its head out of water to breathe air (Ishimatsu et al., 1998). At high tides, most fish follow the tide, swimming along the water line with their snout and eyes above the water surface (Kok et al., 1998; Murdy, 1986). However, P. schlosseri guarding a burrow stays within the burrow for several hours during high tides (A. Ishimatsu, T. Takeda, Y. Tsuhako and Khoo, K. H., unpublished observations). The burrows of P. schlosseri are filled with nearly anoxic water, but contain gas whose ranges from less than 1 kPa to as high as 19.9 kPa, and varies from 0.4 to 11.8 kPa. The fish is likely to use the gas stored in the burrow as an O₂ reservoir during burrow confinement (Ishimatsu et al., 1998). Thus, in its natural habitat, P. schlosseri probably experiences a wide range of ambient O₂ and CO₂ levels.

The purpose of the present study was to examine changes in arterial blood gas levels, blood pressure and heart rate in chronically cannulated P. schlosseri during total emersion, during aquatic hypoxia with access to air and during prolonged forced submersion to obtain basic information on the physiological performance of the species under conditions that the fish may encounter in its natural habitat.

Most fish (Periophthalmodon schlosseri (Pallas)) used in this study were collected on the mudflats of Penang, Malaysia. Some of the fish used in 1995 (air exposure experiment: blood gas determinations, see below) were collected in a tributary of the Selangor River near Kuala Selangor, Malaysia. At both these locations, air temperature ranged from 28 to 31 °C, whereas water temperature ranged from 28 to 36 °C. The dissolved O₂ concentration of burrow water varied from almost zero to approximately 80 % of that of air-equilibrated water at the surface, but was less than 3 % below 50 cm (Ishimatsu et al., 1998). The O₂ saturation of open water ranged from 30 to 80 %. The salinity of the water sampled from P. schlosseri burrows varied from 15 to 27 ‰ except for those in the tributary of the Selangor River, where salinity was only 1–2 ‰. In 1995, fish were transported to the laboratory at Universiti Sains Malaysia, where the air exposure experiment (blood gas determinations, see below) was conducted. Fish were kept individually in glass aquaria, partially filled with 50 % sea water (17 ‰), for at least 1 week before use. They were fed daily on chopped shrimps, but starved for 2 days before use. Both air and water temperatures were stable at approximately 29 °C. In 1997 (aquatic hypoxia and forced submersion experiments) and 1998 (air exposure experiment: blood pressure and ammonia determinations), individuals were shipped by air freight to the Marine Research Institute in Nomozaki, Japan. Acclimation conditions were the same as in 1995, except that the glass aquaria were placed in water baths maintained at 30 °C. Overall mean body mass was 174±6.6 g (±1 S.E.M.; range 124–225 g; N=24).

Fish were anaesthetized by immersion in a 50 % solution of tricaine methane sulphonate in sea water (0.6 g l⁻¹; MS-222, Sigma) until opercular movements stopped. During surgery, the gills were irrigated with a solution of tricaine (0.3 g l⁻¹) aerated with pure O₂. A short incision (1.5 cm) was made on the ventral body surface, somewhat to the left of the midline and posterior to the left pectoral fin. The mesentery was blunt-dissected to expose a branch of the coeliaco-mesenteric artery alongside the spleen. The vessel was occlusively cannulated using polyethylene tubing (PE-50; Clay Adams) with a tapered tip. Whenever possible, the cannula was advanced so that the tip lay within the lumen of the coeliaco-mesenteric artery. The cannula was filled with heparinized (50 i.u. ml⁻¹) saline. The cannula was then exteriorized through a hole made beside the cut, and the incision was closed using interrupted stitches. The incision was lightly powdered with oxytetracycline hydrochloride (Wako, Japan).

After surgery, fish were placed individually in plastic fish chambers (10 cm ×28 cm ×8.5 cm, height). The water depth in the chambers was maintained at 3 cm, allowing the fish access to both air and water. The aerial phases of the chambers were constantly ventilated with humidified ambient air at the experimental temperature, and the water phases of the chambers were supplied with 50 % sea water at a rate of 500–800 ml min⁻¹. The experimental arrangement consisted of two fish chambers, the water bath in which the fish chambers were placed, a 30 l water reservoir and two aeration columns. The water was recirculated and air-saturated by vigorous bubbling through the aeration columns. The chambers were covered with black plastic to minimize visual disturbance, and the fish were allowed to recover for 24 h. Haematocrit values (Hct) were repeatedly measured during the recovery period after surgery, and only fish with a stable Hct were used for experiments. Fish were killed by an overdose of MS-222 after experiments.

Seven fish (body mass 205±5.3 g) were used for blood gas measurements, and another group of five fish (172±20.1 g) for blood pressure and ammonia measurements in this experiment. One control blood sample was taken while the fish breathed bimodally. After clearing the cannula of saline, blood samples for and pH determinations (0.4 ml) were withdrawn into 1 ml tuberculin glass syringes to which short pieces of 22 gauge needle shaft were directly glued to minimize dead space volumes. The dead spaces (volume approximately 15 μl) were filled with concentrated sodium heparin solution (1000 i.u. ml⁻¹) before sampling. Immediately after sampling, the syringes were sealed by heating short pieces of PE-50 tubing attached to the needle shafts. Two haematocrit tubes were filled for determinations of plasma total CO₂ and Hct. Blood samples used for and pH determinations were reinfused into the fish after measurements.

After the the control measurement had been completed, the chambers were drained without disturbing the fish. Blood samples were taken 1, 3 and 5 h after emersion began. After 6 h of emersion, the supply of sea water to the chambers was restarted, and recovery samples were taken 2 h later. Experimental temperature was 29 °C.

Blood pressure, heart rate and plasma ammonia concentration ([NH₃]_p) were measured using a different group of mudskippers because the equipment needed for these measurements was unavailable when the blood gas study was conducted. Samples for blood pH and [NH₃]_p were taken using the same sampling schedule as above. Blood pressure and heart rate were measured by connecting the cannulae to pressure transducers.

Six fish were used in this experiment (body mass 164±6.5 g). After the control measurements had been completed, the water in the apparatus was made hypoxic by vigorously bubbling it through the aeration columns with pure nitrogen until water was below 0.9 kPa. Water was continuously monitored by an O₂ electrode inserted into the fish chambers. The nitrogen flow rate used was initially 15 l min⁻¹, but was reduced to 5 l min⁻¹ once water O₂ saturation declined below 10 %. The aerial phases of the fish chambers were constantly ventilated with room air, as during the control period, such that aerial was not affected by hypoxic water within the chambers. Blood gas, blood pressure and heart rate determinations were made 1, 3 and 5 h after the onset of aquatic hypoxia. After 6 h of hypoxia, the water was made normoxic, and the final samples were taken after 2 h of recovery. Fish had free access to ambient air throughout an experiment. Experimental temperature was 30 °C.

Six fish were used in this experiment (body mass 148±8.9 g). After the control measurements had been completed, the fish chambers were completely submerged in the water bath. Within a few minutes of submersion, the fish released air from the buccopharyngeal cavity. The gas was withdrawn from the chamber by suction, to prevent rebreathing. Blood samples were taken after 1.5, 5 and 12 h of submersion. After 12 h of submersion, the chambers were removed from the water to allow the resumption of bimodal breathing. The final blood samples were taken after 12 h of recovery. Blood pressure and heart rate were measured at 1.5, 2.5, 5 and 12 h of submersion and after 1 min, 30 min and 12 h of reimmersion. Water O₂ saturation was maintained at over 90 % throughout the experiment. Experimental temperature was 30 °C.

The blood and pH were determined, respectively, using an E101 O₂ electrode and E301/E351 pH electrodes connected to a meter (Cameron Instrument Co., Texas, USA). The and pH electrodes were calibrated with humidified nitrogen gas and room air, and standard buffers (IL Test Buffers: Instrumentation Laboratory, Massachusetts, USA), respectively. Blood O₂ content and plasma were measured using an Oxycon (Cameron Instrument Co.) and a Capni-Con 5 (Cameron Instrument Co.), respectively. The partial pressure of CO₂ (⁠⁠) was calculated from pH and using the solubility coefficient of CO₂ in plasma (Boutilier et al., 1984) and pK ′ (Siggaard-Andersen, 1976). The Hct value was determined using a microcentrifuge (10 700 g, 5 min). [NH₃]_p was measured using an enzymatic assay kit (Determiner NH₃, Kyowa-Medex, Tokyo, Japan). Water was monitored using an E101 O₂ electrode connected to an oxygen meter. Blood pressure was measured by connecting the cannulae to pressure transducers whose output was amplified by a polygraph and recorded with a four-channel recorder. The pressure measurement system was calibrated against a static column of water. Zero pressure level was established by placing the meniscus of the water column at the level of the water surface in the fish chambers and in the water bath during bimodal breathing and during submersion, respectively, or at the estimated level of the fish heart during air exposure.

Haemoglobin O₂ saturation was calculated by dividing the measured O₂ content of each sample by the estimated O₂ capacity at the Hct of the sample, after subtracting physically dissolved O₂ from the measured O₂ content, and is presented as a percentage. To estimate O₂ capacities at different Hct levels, a regression line between O₂ capacity and Hct was established. O₂ capacity was determined by measuring the O₂ content of blood after 15 min of equilibration with room air at 30 °C. Dissolved O₂ was subtracted from the O₂ content using the O₂ solubility for human plasma (10.49 μmol l⁻¹ kPa⁻¹ at 30 °C; Boutilier et al., 1984). The relationship between O₂ capacity (mmol l⁻¹) and Hct (%) was: O₂ capacity = −0.309+0.180Hct (r²=0.899, P<0.0005; N=11).

Statistical comparisons were made by repeated-measures analysis of variance (ANOVA). Pairwise comparisons with control values were made using Dunnett’s test. Values are expressed as means ± S.E.M. wherever possible.

Table 1 summarizes baseline values of blood gas level, blood pressure and heart rate calculated from the control data from all experiments. When undisturbed in normoxic water with free access to air, the fish usually breathed air every 3–5 min, although the interbreath interval varied among fish as well as within an individual fish. During breath-holding, the opercula were kept inflated, and both the mouth and the gill openings were closed. No aquatic ventilation was observed under the control conditions.

Arterial was unaffected by air exposure (Fig. 1A). Arterial pH during air exposure was significantly higher than the control value and remained higher at 2 h of recovery (Fig. 1B). failed to show significant changes throughout the experiment (Fig. 1C). Arterial was slightly, but significantly, decreased after 1 h of air exposure and at 2 h of recovery (Fig. 1D). Neither arterial blood pressure nor heart rate was significantly affected during air exposure or recovery (Fig. 2A,B). Heart rate differed greatly among individuals, which may have obscured possible changes, but inspection of individual plots also revealed no consistent trend in heart rate changes during air exposure or recovery. Arterial was not measured in this experiment. Air exposure did not appreciably influence the frequency of air ventilation.

Mean plasma ammonia concentration had increased approximately sixfold by the end of the air exposure and remained higher than the control level even at 2 h of recovery (Table 2).

Arterial values did not show any significant changes during aquatic hypoxia or recovery (Fig. 3A). Arterial did not change in response to aquatic hypoxia, but was significantly reduced by approximately 0.5 mmol l⁻¹ compared with the control value (2.3±0.3 mmol l⁻¹) at 2 h of recovery (data not shown). was calculated to be 53±7 % under control conditions and remained stable throughout the experiment (data not shown). Neither arterial pH (Fig. 3B) nor (Fig. 3D) changed in response to aquatic hypoxia, whereas showed significant decreases at 5 h of aquatic hypoxia and at 2 h of recovery (Fig. 3C).

Blood pressure did not change significantly in response to aquatic hypoxia (Fig. 4A). Heart rate was significantly higher than the control value at 3 h of aquatic hypoxia and at 2 h of recovery (Fig. 4B). Aquatic hypoxia exerted no appreciable effects on aerial ventilation rate.

The most dramatic changes were observed in this experiment. Blood ⁠, and were all significantly decreased at 1.5 h of submersion and remained low until 12 h of submersion (Fig. 5A–C). At 12 h of recovery, and were not significantly different from the respective control values, whereas remained significantly lower than the control value. Blood pH did not change from the control value during submersion, but became significantly higher than the control value at 12 h of recovery (Fig. 5D). and were significantly depressed during submersion, but both had recovered by 12 h of recovery (Fig. 5E,F).

Blood pressure was significantly lowered at 1.5 and 2.5 h of submersion, but then transiently increased at 5 h, such that the 5 h submersion value was not significantly different from the control value (Fig. 6A). The blood pressure was again significantly reduced at 12 h of submersion, and did not recover until 12 h post-submersion. Heart rate was significantly decreased in response to submersion (Fig. 6B). A full development of bradycardia took place over 3–12 min after the onset of submersion, and heart rate then remained stable at the low level except for some fluctuations due to underwater activity. When the fish took its first air breath upon re-emersion, heart rate increased immediately and became significantly higher than the control value. The heart rate at 12 h of recovery was not significantly different from the control value.

In response to forced submersion, fish immediately started vigorous gill ventilatory movements that persisted at rates of 40–50 min⁻¹ throughout the submersion period.

Hct decreased gradually and significantly during both air exposure and aquatic hypoxia (Fig. 7A). In contrast, the Hct during the submersion experiment showed no significant change from the control value throughout the experiment (Fig. 7B).

The present study provides the first in vivo blood gas and pressure data on a chronically cannulated mudskipper. The low mean arterial of P. schlosseri (Table 1) under control conditions can be partially attributed to the vascular architecture of the species. The cardiovascular system of mudskippers deviates little from the typical teleost design in which the branchial and systemic vascular beds are connected in series. There is almost no modification towards separation of O₂-rich blood (from multiple accessory gas-exchange surfaces; i.e. skin and buccopharyngeal and opercular mucosae) and O₂-poor blood (from the systemic capillary beds) (Graham, 1997; Schöttle, 1931). Therefore, the two types of blood will probably be mixed in the central veins and heart, and the resulting mixed blood is delivered to the systemic capillary beds as well as to the accessory gas-exchange epithelia. Presumably, the oxygenation status of the blood would be improved only marginally by passing through the gills because of the degenerative structure of the gills. The gill filaments are short, often bifurcated and sparsely arranged on the gill arch (Low et al., 1988). The secondary lamellae are covered by a thick epithelium and often fused (Schöttle, 1931). The low arterial is disadvantageous in terms of O₂ diffusion to the systemic tissues but it would enhance aerial O₂ uptake through the accessory gas-exchange surfaces perfused by the arterial blood. Moreover, it would retard O₂ loss when P. schlosseri stays in the often severely hypoxic water of mudflats.

Recently, Kok et al. (1998) reported blood gas levels and heart rate of P. schlosseri and another mudskipper Boleophthalmus boddarti. The blood value reported for P. schlosseri is slightly higher than our value, and the and pH values reflect more acidic conditions than in our study. These differences could be due to their sampling technique, i.e. severing the tail of anaesthetized fish for blood sampling. Pelster et al. (1988) reported blood gas and blood pressure levels of an intertidal rockpool blenny Blennius pholis chronically cannulated in the ventral aorta. This fish is a strictly aquatic breather in normoxic water and emerges only when water declines (Bridges, 1988). The ventral aortic blood and were reported to be 3.73 kPa and 0.34–0.39 kPa, respectively, during immersion (Pelster et al., 1988).

Air exposure did not cause a respiratory acidosis in P. schlosseri, as has been observed for most aquatic air-breathing fishes when they were subjected to air exposure (Graham, 1997). The blood pH of P. schlosseri increased slightly but significantly during air exposure and remained elevated after 2 h of recovery (Fig. 1B); however, the blood changed little (Fig. 1D). [NH₃]_p increased during air exposure, with higher rates of increase during the first 3 h. [NH₃]_p increased at 100 μmol l⁻¹ h⁻¹ for the first hour and at 90 μmol l⁻¹ h⁻¹ for the next 2 h, but the rate of increase was only 15 μmol l⁻¹ h⁻¹ after 3–5 h (Table 2). Upon reimmersion, [NH₃]_p decreased at 140 μmol l⁻¹ h⁻¹ during the first 2 h. The rise in blood ammonia level could have partly contributed to the development of metabolic alkalosis. Kok et al. (1998) also reported an alkalosis in air-exposed P. schlosseri, but they did not detect significant changes in blood ammonia level after 6 h of air exposure. A much higher [NH₃]_p (540 μmol l⁻¹) was reported by Ip et al. (1993) for P. schlosseri that had been kept in aquaria in which the fish were free to be in or out of water for 24 h at 25 °C, and the value increased to 920 μmol l⁻¹ after 24 h of air exposure. Peng et al. (1998) demonstrated that P. schlosseri has a greater capacity to detoxify ammonia by converting it to free amino acids than a more aquatic mudskipper, Boleophthalmus boddarti. The decreased rate of retention of plasma ammonia observed in the present study indicates that this detoxification mechanism may have been triggered by elevated [NH₃]_p. It should be noted that the observed [NH₃]_p, even after 24 h of air exposure, is still within the ‘normal’ range reported for resting fishes (Wood, 1993).

Both blood pressure and heart rate were maintained at the control levels during air exposure (Fig. 2A,B). Combined with the stable blood gas levels during air exposure (Fig. 1), these findings suggest that tissue perfusion was maintained as well. Thus, the cardiovascular system of P. schlosseri is probably able to function efficiently, in spite of the influence of gravity. One of the primary reasons for the failure of fish cardiovascular systems on land is that the gills collapse, increasing vascular resistance. The gills of P. schlosseri could probably retain their configuration on land because of the specialized morphology described above. Blood pressure and heart rate remained unchanged during the transition from submersion to emersion in B. pholis (Pelster et al., 1988).

P. schlosseri showed no significant changes in blood in response to severe aquatic hypoxia (Fig. 3A). These results indicate that P. schlosseri did not experience a net O₂ loss when exposed to hypoxic water. Other aquatic air-breathing fishes are known to lose O₂ across the gills in hypoxic water (Graham, 1997). P. schlosseri seldom ventilates the gills as long as air is accessible, and contact between hypoxic water and the respiratory surfaces is therefore minimized, reducing O₂ loss through the rudimentary gills and vascular lining of the buccal and pharyngeal cavities. Also, the gradient between the blood perfusing the gills and the ambient hypoxic water should have been small, because the of sinus venosus blood must have been lower than the arterial ⁠.

Another potential route for O₂ loss in severely hypoxic water is through the skin. Cutaneous respiration in. mudskippers has been estimated to be 40–80 % of the total (Tamura et al., 1976; Teal and Carey, 1967). Preliminary data suggested that 50 % of is through the skin in P. schlosseri confined in normoxic water as well as in air (T. Takeda, unpublished observations). It remains to be determined whether active regulation of capillary recruitment/derecruitment occurs in P. schlosseri in response to aquatic hypoxia as a means of controlling cutaneous gas exchange (Feder and Burggren, 1985).

P. schlosseri showed no hypoxic bradycardia (Fig. 4B) and no hypertension (Fig. 4A), as is known for a wide variety of strictly aquatic fishes (Fritsche and Nilsson, 1993). Rather, the heart rate tended to increase in response to aquatic hypoxia. The peripheral O₂ receptors are believed to occur in the first gill arch and to play an important role in triggering hypoxic bradycardia in fishes. The highly atrophied morphology of P. schlosseri gills suggests that the branchial O₂ receptors may have been lost in this species with diminishing reliance on water as the respiratory medium.

The precipitous fall of blood O₂ levels after forced submersion (Fig. 5A–C) demonstrated that the blood gas-transport system of P. schlosseri is poorly adapted for aquatic O2 extraction. This conclusion is further supported by its lower in water than in air (approximately 2–2.5 μmol g⁻¹ h⁻¹ during submersion compared with 3–3.5 μmol g⁻¹ h⁻¹ in air; Takeda et al., 1999). The reduced gills of the fish appear not to allow operation of the dual (buccal and opercular) pump system necessary for the effective countercurrent flow through fish gills, and this may have limited aquatic O₂ uptake in spite of the enhanced ventilatory activity. The reductions in blood and during submersion (Fig. 5E,F) are probably due to the increased gill ventilatory activity, in addition to the much higher solubility of CO₂ in water. The poor aquatic respiratory capacity agrees with our field observation that P. schlosseri seldom submerges for extended periods, and confirms the importance of the burrow gas pocket for respiration.

The pronounced bradycardia in P. schlosseri during submersion observed in this study (Fig. 6B) confirmed an earlier finding of submersion bradycardia in a congeneric species P. australis (=freycineti; Murdy, 1989) by Garey (1962) as well as recent results in P. schlosseri reported by Kok et al. (1998). In Garey’s study, heart rate decreased from 108 beats min⁻¹ to 46 beats min⁻¹ after 2 min of submersion and recovered to presubmersion level within 12 s after emersion. Similarly, Kok et al. (1998) demonstrated a gradual cardiac slowing in response to submersion and a more rapid increase in heart rate upon emersion. Although rapid changes in heart rate at the onset of submersion could be analyzed for only three individuals in our study, the analysis did demonstrate that submersion bradycardia did not develop rapidly to its full extent. In one fish, it took more than 10 min for the heart rate to stabilise at a lower level after submersion. In sharp contrast, heart rate increased abruptly in all individuals when the fish took its first air breath after emersion. Heart rate remained low before the first air breath even though the fish was partially emerged.

In P. schlosseri, bradycardia developed during submersion when ventilation was strongly stimulated. This is different from the situation in diving birds and mammals in which bradycardia is associated with apnoea (Butler, 1982). In spite of the enhanced ventilation rate, blood O₂ levels decreased significantly (Fig. 5A–C). Therefore, chemoreceptor stimulation by low blood appears to be important in developing bradycardia during submersion in this fish. Either blood or pH should be excluded as a candidate in evoking bradycardia because the former actually decreased and the latter did not change during submersion. It should be noted that the continuous water ventilatory movements seen during submersion were different from the more intermittent aerial ventilatory pattern of the fish, which consists of a single breath interposed by breath-holding for 3–5 min. The rapid tachycardia upon emersion is similar to that reported for birds and mammals in that it does not occur until the first breath is taken. The significantly lower blood pressure in spite of the tachycardia may indicate vasodilation, as is known to occur in birds and mammals (Butler, 1982; Butler and Jones, 1997).

The tendency for Hct to rise during submersion, although not statistically significant (Fig. 7B), is probably a physiological compensation for the reduced blood O₂ levels. A similar rise in Hct has been reported for water-breathing teleosts during aquatic hypoxia and is induced by catecholamine-stimulated release of erythrocytes from the spleen or by a reduction of plasma volume due to water movement to surrounding tissues (Jensen et al., 1993).

The present results demonstrate that P. schlosseri can maintain its oxygen homeostasis as long as individuals can breathe air, irrespective of the presence or absence of water, or its oxygenation level. However, the fish suffers from a severe hypoxaemia when access to air is denied, even in well-aerated water. A comparative study on less-terrestrial species, such as Boleophthalmus spp., will be useful for understanding how mudskippers are adapted to mudflat environments and what differential selection pressures have been exerted to generate a spectrum of mudskipper species possessing different degrees of terrestriality.

This study was supported by the Monbusho Grant-in-Aid for International Scientific Research (08045057) and by NSF IBN96-04699. N.M.A. was supported by an NSF-Monbusho Summer-in-Japan Fellowship during her stay in Japan. We thank Dr Richard Rosenblatt for reading the manuscript. We also thank Mr Keng Kuan, ‘the Old Man’, for his help.