Adaptations to a Terrestrial Existence by the Robber Crab Birgus Latro: II. in vivo Respiratory Gas Exchange and TRANSPORT

Birgus latro L. is an obligate air-breather, normally entering water only to drink or release eggs. The respiratory apparatus consists of paired lungs, formed by the evaginated lining of the branchiostegites, and 14 pairs of rather small gills (Semper, 1878; Harms, 1932). The lungs are ventilated by the anteriorly placed scaphognathites which draw air forwards through the lung. However, the efficiency of ventilation and the length of diffusion paths within the lung are not known. The epithelium and cuticle of the lining of the lung are thin and suited to a gas-exchange function (haemolymph-gas diffusion distance 0·5–1·2μm) (Storch & Welsch, 1984). In contrast, the gills seem unsuited to gas exchange; their surface area is small and their structure, the nature of their attachment and their location in a pocket of the branchiostegite preclude their effective ventilation with air (Semper, 1878; Cameron, 1981a).

In the field, Birgus is slow-moving, and often remains immobile for long periods. The crabs may travel long distances in search of food, however, and on Christmas Island (Indian Ocean) appear to be semi-nomadic, moving through the rainforest in search of fruiting trees. They must also travel to the sea during the spawning season. Rapid locomotion is seen only on disturbance, but even then the crab soon backs into a crevice or climbs a tree pausing as soon as it reaches safety. On behavioural grounds it is apparent that the species does not sustain high levels of activity but is capable of sustained slow or intermittent activity. This is backed up by physiological evidence indicating that Birgus cannot sustain high levels of aerobic activity (Smatresk & Cameron, 1981). The crabs do, however, show ventilatory adaptations to exercise and can rapidly correct respiratory acidosis engendered by exercise.

The roles of the gills and lungs in gas exchange and acid-base balance in landcrabs are poorly understood and there is little direct evidence concerning their relative contributions. In Birgus, the role of the gills was investigated indirectly by studying the effects of gill ablation (Harms, 1932; Smatresk & Cameron, 1981). Ablation resulted in a small increase in ventilation rate and a reduced ability to cope with respiratory acidosis after exercise. At least in the short term, therefore, the gills appear to perform only a minor role in gas exchange and acid-base balance. Either the main site of gas exchange is normally the lung or the lung is capable of compensating for the gills.

In this study, the roles of the gills and lungs in gas exchange and acid-base balance of Birgus were examined by sampling pre-and postpulmonary and pre-and postbranchial haemolymph. This allowed more direct measurement of the contributions of each organ and avoided the problems of trauma and compensation by other organs following gill removal. Additionally, we investigated the effects of prolonged exercise on acid-base balance, production of lactic acid and its contribution to metabolic activity. In vivo data for oxygen content and partial pressure of the haemolymph have been combined with in vitro data for haemolymph (Morris et al. 1988) to determine the role of the pigment in oxygen carriage.

Specimens of Birgus latro were collected from Christmas Island (Indian Ocean) and maintained at the University of New South Wales as described previously (Morris et al. 1988). For 24 h prior to sampling the animals were isolated without food and water in plastic fish-boxes at 25°C.

The partial pressures of oxygen and carbon dioxide (⁠ and ⁠) in Birgus haemolymph samples were measured using the electrode/cuvette of a BMS3 blood microsystem (Radiometer, Copenhagen, Denmark) thermostatted at 25°C. Calibration was by humidified gas mixtures delivered by Wösthoff gas-mixing pumps (Type SA 18, Wösthoff, Bochum, FRG). was displayed on the expanded scale of a PHM 72 meter (Radiometer) and on a Strathkelvin oxygen meter (Strathkelvin, Glasgow, UK). Haemolymph pH was measured using the G299a capillary electrode of the BMS 3 and displayed on a PHM 72 (Radiometer).

The oxygen content of the haemolymph samples was measured by the method of Tucker (1967) according to Bridges et al. (1979) using 5 or 10 μl samples. The carbon dioxide content was determined by the method of Cameron (1971) employing 10 and 15mmoll⁻¹ NaHCO₃ standards.

[L-lactate] in the samples was determined using the Boehringer test kit (catalogue no. 139084, Boehringer Mannheim GmbH, Mannheim, FRG) and [urate] using the Sigma test kit (no. 685, Sigma Chemical Co., St Louis, USA). Concentrations of calcium and copper were determined with a Varian AA175 atomic absorption spectrophotometer as described previously (Sparkes & Greenaway, 1984).

In the first experimental treatment arterial and venous samples were taken from quiescent and exercised Birgus latro (400–500 g). Exercise was for a 30-min period in a humidified 25°C constant-temperature room and was induced by repeatedly startling the animal into an escape response. Most specimens became refractory towards the end of this period. Samples were taken in 1 ml glass syringes the dead space of which had been previously filled with saline matching the osmotic pressure of the haemolymph and equilibrated with CO₂ (7·4 mmHg rest and 14·8 mmHg exercised). The samples were stored on ice for a maximum of 15 min and the final 30% of the haemolymph was discarded.

Arterial haemolymph was sampled directly from the pericardial cavity via a small hole in the carapace dorsal and posterior to the heart and 2–3 mm to one side of the anterior-posterior axis. This hole was drilled at least 24 h prior to the experiment and did not penetrate the hypodermis. After sampling the hole was plugged with silicone grease.

Venous samples were withdrawn from the sinus at the base of the second walking leg which receives haemolymph returning from the leg and from the ventral sinus. It is possible, as with other decapods, that the sample does not represent fully mixed venous haemolymph. Each animal was used once only and the sampling period was <30 s. The number of animals in each group was never less than 10. In addition to measurement of and in arterial and venous haemolymph the concentrations of L-lactate, Ca, Cu and urate were also determined in this series. Differences between exercised and rested parameters were tested using the t-test and differences between arterial and venous data sets using a paired-sample t-test.

In this series the rested and exercised treatments were repeated but, in addition to arterial and venous haemolymph, samples of haemolymph returning from the lungs were taken from the pulmonary veins. The positions of these vessels were determined by examining casts of the vascular system of Birgus (C. A. Farrelly & P. Greenaway, unpublished data). In the crabs used, these veins were approximately 3 mm in diameter and were sampled via a second hole drilled in the dorsal carapace. Otherwise, sampling was as for the pericardial site and in general as described for series A. Large samples could be obtained from the exercised animals as the duration of the sampling period was less crucial. In rested animals only small samples could be obtained quickly and, as a consequence, the partial pressures of oxygen and carbon dioxide were not determined. Concentrations of Ca, Cu, L-lactate and urate were determined in arterial samples. Pulmonary – arterial and pulmonary – venous differences were tested using the t-test for paired samples.

In this series, afferent and efferent branchial haemolymph was sampled. The gills of Birgus are housed in a semiflexible pocket formed from the ventral margin of the carapace (Harms, 1932). The posterior margin is quite flexible and it was possible to move this aside to expose the last five gills (nos 10–14). Haemolymph leaving the gill lamellae passes into large efferent sinuses on the underside of the gill which in turn feed the branchiopericardial vessels running to the pericardial cavity. Efferent branchial haemolyph was taken from these sinuses using a 25 μl gas-tight Hamilton syringe with an exchangeable 26 gauge hypodermic needle with low dead space which had been prefilled with saline. Haemolymph was sampled directly from the efferent sinuses of gill 10 with a success rate >90%. In the few cases where a sample could not be obtained within 2 min the animal was discarded for the purpose of the experiment. When samples of 12–20μl were obtained 5 μl subsamples were used for the determination of and ⁠. When sample sizes >20μ1 were obtained 10μl was used in each case. In the short period between sampling and analysis, mixing between the saline in the dead space and haemolymph was restricted to a narrow boundary layer contained within the residual blood in the syringe.

A sample of afferent branchial haemolymph was taken immediately after the efferent sample (within 10 s). This was drawn from the venous sinus supplying the branchial circulation of that limb. Large samples (250 μl) could be taken and this allowed the determination of pH in addition to and ⁠. The [Cu] was not determined in series C.

Owing to the nature of the sampling method used, animals in series C could not be considered quiescent but nor were they exercised. This group are accordingly referred to as disturbed animals throughout this report. Differences between pre-and postbranchial samples were tested using the t-test for paired samples.

Arterial oxygen partial pressure

Venous oxygen partial pressure

Arterial carbon dioxide partial pressure

Venous carbon dioxide partial pressure

Oxygen content

Carbon dioxide content

Arterial oxygen content

Venous oxygen content

Arterial carbon dioxide content

Venous carbon dioxide content

PHa Arterial pH

PHv Venous pH

Prebranchial afferent supply to gill 10

Postbranchial efferent supply from gill 10

Pulmonary efferent supply from the branchiostegal lung

The measurements made in series A are summarized in Table 1. In resting Birgus there were significant arterial – venous (a–v) differences in the and of the circulating haemolymph. Exercise changed the absolute values of measured parameters but the a–v differences were maintained. There was a significant a–v difference in the pH of the haemolymph of exercised crabs. Interestingly, exercise resulted in a significant decrease in but was unaffected. Conversely, decreased without a decrease in and increased significantly as a result of exercise, whereas and declined significantly. This change in the carbonic acid equilibrium was reflected by an exceptional decrease in haemolymph pH. Both pHa and pHv fell by approximately 0·7 pH units during the 30 min exercise period. A major part of this acidosis was metabolic in origin as the concentration of circulating lactate increased by more than 27 mmol l⁻¹. There was an increase in the mean [Cu] of circulating haemolymph from 2·63 to 3·24 mmol l⁻¹ but variability among individuals meant that the difference was not significant. The concentration of urate in the haemolymph of resting crabs was 0·23 mmol l⁻¹ in these animals and the small increase with exercise was not significant. An increase in [Ca] of approximately 1 mmol l⁻¹ was observed in both arterial and venous haemolymph after 30 min of exercise. Using combined data for arterial and venous [Ca] this increase was significant (0·05 P<0·02).

In series B the respiratory and acid-base parameters of haemolymph from the pulmonary vein (with the exception of ⁠) approximated those of the pericardial rather than venous haemolymph (Fig. 1). As in series A, exercise dramatically altered the absolute values of all measured parameters but the relationship between the three samples persisted with the changes in the afferent pulmonary haemolymph largely mirroring those in the arterial samples. Maximal of the pulmonary haemolymph in an exercised animal was 100·3 mmHg.

In an attempt to understand the role of the lung in gas exchange, pulmonary – arterial and pulmonary – venous differences in pH and gas content were calculated and tested for significance (Table 2). Arterial haemolymph from rested crabs contained significantly less oxygen than haemolymph returning from the lung, but considerably more than the venous sample. This trend was even clearer in exercised animals (Table 2), indicating that significant oxygen loading occurred in the lung. The situation for CO₂ was more complex. In rested Birgus there was no significant decrease in during passage from the venous system through the lung but, nevertheless, arterial haemolymph contained significantly less CO₂ than venous haemolymph, indicating either that CO₂ was excreted elsewhere or that an increase in pH had occurred. The measured pH differences were significant, but not in a direction that would support the titration of HCO₃⁻. In the exercised crabs there was no significant difference in the °f arterial and pulmonary haemolymph. However, the of postpulmonary haemolymph was greater than that of pericardial haemolymph. Despite these data some CO₂ appeared to be lost at the lung as both partial pressure and content decreased, the latter significantly (Table 2). The small but significant concomitant increase in pH could, however, largely account for decreased at a constant The levels of Ca, Cu and L-lactate in the haemolymph all increased significantly during exercise, but the concentration of urate was variable among animals and no significant trend could be detected.

The results of the investigation of changes in haemolymph gases during passage through the branchial circuit (series C) are summarized in Table 3. There was no significant increase in the of the haemolymph during passage through the gill, but there was a highly significant loss of CO₂ as the mean decreased by 2·30mmol I⁻¹. The mean prebranchial pH (7·465) was lower than the venous pH of rested Birgus. Assuming that all haemolymph pH changes during gill passage were due to the loss of CO₂ it was possible, using a relationship determined in vitro for (Morris et al. 1988), to extrapolate to a postbranchial pH. This value (7-595) was similarly lower than the pHa measured in resting Birgus. The extrapolated values for post-and prebranchial were determined in similar fashion by inserting in vitro values together with the determined into the Henderson-Hasselbalch equation. This allowed estimation of values where none were measured (see Table 3).

The mean pre-and postbranchial values were derived as indicated in Fig. 2. Arterial values approximated a mean value for of 11 mmol l⁻¹ and therefore the equilibrium curves were normalized to this value. The curves were constructed by abstracting the appropriate data for pre-and postbranchial pH from Morris et al. (1988). The mean values were then plotted with these curves to supply the appropriate ⁠. Also indicated in Fig. 2 is the range of ⁠, values measured and these data reflect that in 19% of the animals there was a decrease in during passage through gill 10.

The concentration of Cu was used as a measure of haemocyanin concentration. Measurements of oxygen capacity in air-equilibrated haemolymph and [Cu] showed, however, that [Cu] overestimates Hc-O₂max by approximately 19%. Therefore, haemocyanin-oxygen equilibria were calculated assuming that [haemocyanin] = 0·41[Cu] (Fig. 3). The arterial and venous data from rested crabs (Fig. 3A) and arterial, venous and pulmonary data from exercised crabs (Fig. 3B) were also described by equilibrium curves calculated from the previously determined in vitro relationship (Morris et al. 1988). The relative importance of dissolved compared with haemocyanin-bound oxygen was evaluated by normalizing 100% saturation as 1·1 mmol l⁻¹ and calculating the dissolved fraction over the appropriate range (Fig. 3A,B).

In resting Birgus, the haemocyanin of arterial, and therefore pulmonary, haemolymph remained saturated, and at the highest recorded dissolved O₂ made up 10% of the total haemolymph load. A fall from a mean arterial pressure of 44 mmHg to a venous tension of 10 mmHg reduced dissolved O₂ by 0·05 mmol l⁻¹ and reduced haemocyanin saturation from 98 to 61%. Thus a mean total of 0·46 mmol l⁻¹ O₂ was extracted from the haemolymph by the tissues during each circuit of the haemolymph, with approximately 11% of this coming from the dissolved fraction. Haemolymph passing through the lung of resting Birgus was completely reoxygenated, often to levels above that of the arterial haemolymph, before returning to the heart (Fig. 1).

A different situation was observed in Birgus that had been exercised to near exhaustion (Fig. 3B). The large decrease in pH markedly right-shifted the equilibrium curve, reducing haemocyanin oxygen affinity. However, haemocyanin in the haemolymph leaving the lungs was still 93% saturated, although in arterial haemolymph saturation fell to 79%. Despite maintenance of close to, or even greater than, pre-exercise levels, venous saturation fell to a mean value of 25%. Under these conditions the a–v difference in the dissolved fraction was reduced to 0·02mmoll⁻¹, whereas the He–O₂ difference averaged 0·70mmoll⁻¹ providing a total a–v difference of 0·72 mmol l⁻¹.

Using the and pulmonary O₂ contents (0·92, 0·29 and 1·08 mmol l⁻¹, respectively) further analysis of lung and gill function was possible. Assuming that efferent pulmonary and branchial supplies were the only return to the heart, and bearing in mind that no branchial uptake of O₂ occurred, the dilution of pulmonary haemolymph by efferent branchial haemolymph could be determined on a percentage basis. Accordingly, it was calculated that 79% of the return to the heart came from the pulmonary circuit and 21% from the branchial circuit, a ratio of 3·8:1 in favour of the lung.

To analyse further the changes in acid-base balance that occurred in exercising Birgus (Fig. 4), mean values calculated from experimental series B and C were fitted to diagram. The similarity of arterial, venous and pulmonary parameters in the rested crabs was evident from the plotted mean values. A similar situation was seen for exercised crabs. Nevertheless, significant differences in and pH did exist within each of these groups (Table 2).

The large difference between the values for resting and exercised crabs was due to a complex change in acid-base balance of mixed respiratory and metabolic origin. The significant acidosis had been partially compensated by the end of 30 min of exercise. Despite the greater variation in values from exercised Birgus, the apparent titration down the nonbicarbonate buffer line from venous to pulmonary conditions was significant (Table 2). Thus, up to 0·8 mmol I⁻¹ CO₂ was lost from the haemolymph of active crabs during each lung transit. In quiescent animals, the CO₂ lost via the lungs appeared negligible. Comparison of this value with the amount lost during gill transit (2·3 mmol l⁻¹) must take into consideration the greater relative flow through the lung calculated above. A value of 0·5 mmol l⁻¹ CO₂ lost by all routes from the haemolymph of quiescent crabs per circuit was calculated accordingly. The available data for pre-and postbranchial haemolymph were fitted to the same diagram and indicated that the perturbations of acid-base balance resulting from disturbance could be accounted for by an increase in titrating pH along the nonbicarbonate buffer line.

The roles of lactate, H⁺ and CO₂ generation in determining the acid-base balance of exercising Birgus were analysed in detail (Fig. 5). In this analysis, measured in vivo values were used to predict the expected pH; (a) assuming a change in alone (b) assuming a stoichiometric release of lactate and H⁺ into the extracellular compartment and (c) taking into account changes in both measured [L-lactate] and ⁠. In Fig-5 the difference ac is the depression in pH expected given the increase in L-lactate, ab is the actual change in pH and be is the difference due to changes in ⁠. The predicted end-point position, B₃, was below the measured value which implies that the haemolymph was losing and/or gaining H⁺ and HCO₃⁻ at differential rates, thus raising pH above that predicted. The observed [L-lactate] in the haemolymph could more than account for the very low pH in exercised animals assuming stoichiometric release of protons.

The lungs of Birgus latro were clearly effective in oxygen uptake as the haemocyanin in pulmonary haemolymph was saturated in both resting and exercised animals and high levels were attained. In contrast, no significant contribution to oxygen uptake was discernible by gill 10. This situation obtained in the gills generally as pericardial haemolymph (mixed pulmonary and branchial) had a lower oxygen content than pulmonary haemolymph, indicating a dilution of oxygen-rich pulmonary haemolymph with low-oxygen branchial return.

Available evidence suggested that elimination of CO₂ was apportioned rather differently. Some loss of CO₂ across the lungs of Birgus must have occurred by simple diffusion in all crabs, as the lungs are permeable to gases (viz high uptake of O₂) and a partial pressure gradient for CO₂ existed. Such loss was obviously negligible in resting animals. However, the higher levels in exercised animals and reported hyperventilation with exercise (Cameron & Mecklenburg, 1973; Smatresk & Cameron, 1981) may have enhanced diffusional loss enough to explain the measured loss (0·8 mmol I⁻¹) by this route following exercise. Carbonic anhydrase has been identified in the lipid (membrane-bound) fraction of lung epithelial cells and may have catalysed CO₂ formation from the bicarbonate pool in the haemolymph (S. Morris & P. Greenaway, unpublished data).

Significant diffusional loss of CO₂ across the gills by passive means was unlikely, given the measured gradient and the negligible inward movement of O₂. Carbonic anhydrase, however, was present in both lipid (membrane-bound) and cytoplasmic fractions of the gills of Birgus (S. Morris & P. Greenaway, unpublished data) and may have been responsible for the observed elimination of CO₂.

The lungs contained only membrane-bound enzyme at a lower activity (S. Morris & P. Greenaway, unpublished data).

Anaerobiosis does not produce CO₂ although the accompanying acidosis may transiently increase loss of CO₂ previously produced. High metabolic activity in Birgus will not, therefore, be accompanied by a proportionate increase in CO₂ output. Repayment of the oxygen debt incurred will require elimination of CO₂ but full recovery requires several hours (Smatresk & Cameron, 1981). Use of anaerobic metabolism during periods of high metabolic activity will minimize the rate of CO₂ elimination required and may be valuable in an animal where branchial gas exchange is demonstrably limiting.

The significance of gills and lungs in the elimination of CO₂ depends on haemolymph flow to these organs as well as the fractional removal achieved. When both these factors were considered it was apparent that the lungs and gills eliminated approximately equal amounts of CO₂ in exercised crabs (provided that there was no increase in fractional removal by the gills following exercise).

Cardiac output is related to the oxygen-carrying capacity of the haemolymph and, using available data (McMahon & Wilkens, 1983), the cardiac output of Birgus was estimated at 45 ml kg⁻¹ min⁻¹. Using this value and our calculated value for removal of CO₂ from the haemolymph of resting Birgus (0·5 mmol l⁻¹), the rate of CO₂ output was calculated to be 22·5μmolkg⁻¹ min⁻¹. This compares favourably with the measured by Cameron & Mecklenburg (1973), 24μmol kg⁻¹ min⁻¹, and yields a value for the respiratory exchange ratio (R) of 0·94.

Exercise involved a very high level of anaerobic metabolism which resulted in a 100-fold increase in circulating levels of L-lactate to a mean level of 25 mmol I⁻¹ and in excess of 40 mmol I⁻¹ in one individual. These values are much higher than reported for other crustaceans and more than twice that found previously for Birgus following a 5-min period of exercise on a treadmill (Smatresk & Cameron, 1981). The leg muscles of Birgus possess a high activity of lactate dehydrogenase which can turn over up to 0·01% of muscle mass min⁻¹ as L-lactate (S. Morris & P. Greenaway, unpublished data).

The 30-min period of exercise caused an extremely large mixed acidosis of the haemolymph despite the relatively high buffering capacity of the haemolymph proteins. Measured levels of L-lactate more than accounted for this change, and it was evident that significant compensation occurred during the exercise period. In disturbed animals the observed acid-base changes were explicable as a respiratory acidosis.

Compensation for metabolic acidosis may have been achieved by several mechanisms. First, there was an increase in haemolymph [Ca] of 1–2 mmol l⁻¹ during exercise. Assuming that this was drawn from exoskeletal CaCO₃, the resultant formation of HCO₃⁻ would remove 2H⁺ per Ca²⁺ and this alone could account for the observed compensation of the acidosis excess. A second mechanism of compensation involving branchial exchange of ions (H⁺/Na⁺ and HCO₃⁻/C1⁻) is theoretically possible. Several species of landcrabs have been shown to release isosmotic urine into the branchial chambers where salt resorption occurs across the gills (Wolcott & Wolcott, 1982, 1984, 1985). Estimations of urinary salt output from data in the literature (Gross, 1964; Cameron, 1981b; Kormanik & Harris, 1981) suggest that a maximum of 22mmolkg⁻¹ day⁻¹ of NaCl is available for ionic exchange. Substantial resorption of NaCl from the urine certainly occurs in the branchial chambers (P. Greenaway & S. Morris, unpublished data) and this would allow excretion of protons into the urine. Clearly this mechanism too could account for the compensation seen, but at present must be speculative.

A third possible mechanism of compensation is that differential efflux of H⁺ and lactate or H⁺ and HCO₃⁻may occur from the metabolizing muscle tissue. Boutilier et al. (1986) have reported preferential release of H⁺ from Xenopus muscle when the pH of the extracellular fluid was high, switching to preferential release of lactate as pH fell during exercise. From the evidence (Fig. 5), it is possible that the latter is occurring in exercised Birgus with retained H⁺ being buffered intracellularly, and again this could explain much of the apparent compensation.

In a previous study of acid-base balance in Birgus (Smatresk & Cameron, 1981), an appreciable portion of the metabolic acidosis immediately after exercise could not be explained by the measured [L-lactate] and it was suggested that some other unidentified metabolite accounted for the remainder of the acidosis. In this study, protons accompanying L-lactate alone could account for all the observed acidosis and no other organic metabolites need to be considered. It is possible that differential release of H⁺ and lactate from muscle cells, discussed above, may be responsible for the lack of stoichiometry in haemolymph H⁺ and lactate loads observed in the previous study.

The oxygen content of arterial haemolymph of Birgus (1·1 mmol l⁻¹) is at the top of the range for crustaceans (McMahon & Burggren, 1988). Transport of oxygen in simple solution was always small (<0·2mmoll⁻¹) and the high oxygen capacity was conferred by a high concentration of haemocyanin.

Utilization of circulating oxygen increased from 0·46 mmol l⁻¹ (11% dissolved) to 0·72 mmol l⁻¹ (2·8% dissolved) following exercise, an increase of 57%, and oxyhaemocyanin was clearly the major carrier of oxygen in the haemolymph. Increased delivery in exercise was achieved by maintaining near saturation of arterial haemocyanin whilst the large (for a terrestrial crab) Bohr shift (Morris et al. 1988) depressed (despite a small rise in from 10 to 13mmHg).

Although the in vitro Bohr shift was large between pH8 and pH 7·3, below this it decreased sharply (Morris et al. 1988). This prevented extreme right-shifting of the equilibrium curve which would have compromised oxygenation at the lungs and thus reduced and oxygen delivery. Increasing temperature decreased oxygen affinity in Birgus (Morris et al. 1988) and it is clear from the in vitro and in vivo studies that temperatures above 25 °C in exercised crabs would compromise oxygen loading during exercise. In the resting state, however, a temperature increase reducing affinity could potentially increase oxygen delivery to the tissues. Thus the increased metabolic demand for oxygen with rising temperature could be satisfied in resting animals but would exacerbate problems of oxygen delivery during exercise.

The haemocyanin of Birgus was largely insensitive to [L-lactate] (Morris et al. 1988) and there may be several reasons for this. First, potentiation of haemocyanin by L-lactate would marginally increase arterial oxygen saturation but would necessitate very low levels to release adequate oxygen to the tissues. Resultant reduction of the gradient from haemolymph to cells would then restrict diffusional delivery of oxygen. Second, the ability of Birgus to increase ventilation of the lungs by >500% (Smatresk & Cameron, 1981) would effectively maintain arterial saturation during exercise. Thus potentiation by lactate would have no real selective advantage and, indeed, could be disadvantageous. This may provide an explanation for the insensitivity to modulators seen in the respiratory pigments of air-breathing crabs generally. Modulation of the pigment was chiefly by pH and thus was a function of acidosis.

The of exercised crabs (12–13 mmHg, Table 1; Fig. 1) was high considering the high level of anaerobic metabolism and oxygen debt, suggesting that oxygen supply to active tissues was diffusion-limited. Oxygen delivery per circuit was increased (57%) during exercise and further increase would be enabled by the reported increase in heart rate with exercise (Smatresk & Cameron, 1981). Even so, delivery was totally inadequate for sustained activity. The crab clearly has a large anaerobic capacity for violent exercise (defensive leg flicking and rearing) but this is normally sustained for only a short period during which the animal seeks a safe refuge where the large oxygen debt can be repaid.

We should like to thank Caroline Farrelly for her help on numerous occasions throughout this investigation and also David Hair for technical assistance. SM and BRM thank Professor T. Dawson, Dr A. M. Beal and other members of the School of Biological Science, University of NSW for their assistance and hospitality. Financial support was provided by ARGS grant A18616299 to PG; a NATO overseas fellowship (NERC, UK) and an AHFMR research allowance to SM; and NSERC grants A5762, T767O and IC-0265 to BRM.