Gas exchange, storage and transport in voluntarily diving Xenopus laevis

Earlier studies on enforced diving in amphibians showed that O₂ stores were rapidly depleted and that O₂ uptake after a dive was greater than pre-dive levels for quite long periods (Jones, 1967, 1972). Thus, contributions from anaerobic sources appeared to be necessary to support the energetic demands of the tissues during prolonged periods of submergence. More recent studies on freely diving Xenopus laevis have shown that voluntary dives of 30 min duration can be supported entirely by oxidative metabolism. The substantial metabolic acidoses noted in the earlier studies were probably caused by increased levels of activity associated with the forced dives (Boutilier & Shelton, 1986b). However, very little is known about the way in which O₂ stores in different parts of the body are utilized and regulated during prolonged voluntary dives or how those stores are renewed when the dives come to an end.

There is considerable evidence that blood flow to the lungs of Xenopus is reduced in the periods between lung ventilations (Shelton, 1970), as it is to the lungs of other amphibians and reptiles (Johansen, Lenfant & Hansen, 1970) or to the air-breathing organs of many fish (Johansen et al. 1968a; Johansen, Lenfant, Schmidt-Nielsen & Petersen, 1968b). The functional significance of these variations in blood flow may be to reduce the cardiac output during a dive to conserve energy, to regulate the flow of O₂ from lung to blood as the stores are used (Shelton, 1976) or to recharge the blood stores as quickly and effectively as possible when the animal breathes (Shelton, 1985). The present experiments were undertaken to see whether measurements of the rates of depletion and renewal of the lung O₂ store would give some indication of the mechanisms involved as well as offering possible explanations of the extremely variable durations of voluntary dives seen in Xenopus (Boutilier, 1984). Measurements of lung CO₂ were also made in order to assess the significance of changes in this gas. The relative importance of storage and skin exchange in dealing with the CO₂ produced during a dive was of interest in view of the steady states in acid-base balance seen after long dives (Boutilier & Shelton, 1986b).

A new method, using a closed extracorporeal loop with and electrodes, was developed in order to make continuous measurements of partial pressures in the lungs of undisturbed Xenopus making voluntary dives. A simple data-processing method was used to make corrections for response delays associated with the loop and electrode characteristics. More conventional sampling techniques were used to examine the equilibria between blood and gases in the lung.

Twelve adult female Xenopus (110–150g) were obtained from a commercial supplier and maintained in the laboratory at 25°C.

Before cannulating the lungs, a soft piece of tubing was inserted into the glottis of an anaesthetized animal (0·06% solution of MS-222) and the lungs were inflated until their outline was clearly visible along the body margin. The apex of one lung was approached by a dorsal incision in the body wall and a small hole made in the lung tip. The lung was kept inflated by the tracheal tubing whilst a catheter was inserted through the hole and advanced well into the central lung cavity. The inserted portion of the catheter was a 1·7-cm PP200 cuff which had been forced-fitted and heat-sealed onto the end of a piece of polypropylene tubing of known length and diameter (see below). The cuff had been heat-flared at both ends and 10 wedges were cut from its wall to decrease the chances of any fluid or tissue blockage during gas withdrawal. Tissue at the apex of the lung was looped with thread and fastened tightly against the catheter on each side of the flared rim at the cuff join. The other lung was cannulated in an identical fashion. Cannulations caused negligible blood loss and the catheters remained patent and flowing for up to 7 days. Subsequent dissection confirmed the catheters’ positions and showed that little tissue damage or oedema had resulted. All animals remained healthy for the duration of the experiments.

Following surgery, animals were placed in a thermostatted water bath (25°C) which consisted of an inverted cone (in the water bath) and a cylindrical breathing chamber positioned at the water-air interface (see Boutilier, 1984, for illustration). The water was continuously aerated (water = 140–150 Torr) and the air chamber (15 ml) was flushed with humidified gases (25 °C) at a constant flow rate of 300 ml min⁻¹, which provided for complete gas renewal every 3 s. Gases flowing out of the breathing chamber passed through a pneumotachograph screen on either side of which pressures were detected by a differential pressure transducer (Hewlett-Packard Model 270). The amplified output (HP-8801 A carrier preamplifier) from the transducer was recorded directly on a Devices chart recorder and stored on an Instrumentation tape recorder (Racal Store 4).

After 5–10 min of artificial lung ventilation following surgery, animals began to dive and surface in a normal voluntary fashion. Breathing movements were detected by the pneumotachograph. The lung cannulae were then connected to a peristaltic pump (Gilson Minipuls 2) and the extrapulmonary circulation of gases was begun (Fig. 1). During transit, the gases passed through serially connected O₂ and CO₂ electrodes whose output could be continuously monitored. In the course of pilot experiments, it was decided that lung gas should be circulated at a flow rate of 1 ml min⁻¹. This rate of flow had no apparent effect on the animals’ behaviour, and ventilation patterns were similar to those described earlier (Boutilier, 1984). It was important to begin the extrapulmonary circulation as soon as possible following surgery so that the animals would get used to the experimental arrangement. Animals were given at least 24 h to recover from the surgery before data collection began.

Lung gas and levels were measured using Radiometer electrodes mounted in thermostatted cuvettes (25 ± 0·5°C) and coupled with Radiometer PHM-71 or PHM-72 display meters. Linear outputs from each electrode were continuously displayed on a dual pen chart recorder (model CR652, JJ Instruments, Southampton). Electrodes were calibrated with precision gas mixtures (Wösthoff pump, Bochum) which were humidified at the experimental temperature. Periodic calibrations were carried out during the experiments, without disrupting the lung gas circulation, by diverting the flow and bypassing the electrodes. The temporal response characteristics of the loop were also determined. To do this, gases were drawn at 1 ml min⁻¹ from an equilibration chamber containing 4% CO₂/96% N₂ into the input tube of the loop. After the electrode outputs had stabilized, the loop was subjected to a step change in gas tensions by switching the input tube to a second equilibration chamber containing a 1 % CO₂/99% air mixture. Step changes in gas concentrations were also performed in the opposite direction. The response curves from each electrode output were recorded twice daily and subsequently used to compensate for the time delays in the extracorporeal loop. The cannulae making up the extracorporeal loops were always prepared with tubing of identical bore and length, and were often retrieved from one animal for use in the next. The geometric delays of the individual electrodes (i.e. the time between the input step change and the first appearance of an electrode response in either the first or second measurement chamber) were always constant at a flow of 1ml min⁻¹. Because of the writing differential between the pens on the flat bed recorder, the first electrode in the series was connected to the pen advanced further in time on the chart paper. The pens were adjusted so that their writing difference equalled the geometric time difference between the first and second electrode. The actual transit time between lung gas and each measurement chamber was confirmed at the end of each experiment by drawing water (at 1 ml min⁻¹) through the excised lung cannulae. The average time delays from cannula tip to each measurement chamber were identical to those determined during ‘empirical’ gas calibrations using the same dimension tubing. Simultaneous pneumotachograph measurements of breathing movements were appropriately adjusted in time by positioning the ventilation information onto the chart recording of lung gas tensions. Some of the tape-recorded ventilation records were played back at speeds that facilitated precise matching of ventilation and lung gas information.

A typical output trace from a electrode in the extracorporeal loop (curve A), following an air-N₂ step change at the lung input cannula (curve B), is shown in Fig. 2. After a geometric delay of 26 s, due to passage of the gas up the tube from input to electrode, the response begins and proceeds at a progressively decreasing rate to the new equilibrium position. A number of factors are certainly involved in such a time-dependent response, and include the mixing of gases as they flow in the catheter and through the electrode chambers, the diffusion rates through electrode membranes, and the electrochemical events within the electrode (Hahn, Davis & Albery, 1975). The overall response is, therefore, likely to be the result of the interaction of several factors, but for practical purposes it was possible to treat the curves from both and electrodes as single exponentials. Plots of log (P_t–P_∞/P₀–P_∞) against time [where P₀ and P_∞ are the partial pressure readings at the beginning (0) and end (∞) of a full-range response to a step change, and P_t is the reading after time t] were practically linear over the whole range (regression coefficients; 0·97–0·99). Consequently, time constants (τ) for the electrode response could be determined by measuring the time taken for the readings to reach 1/e of the total response to a step change. Corrections to be added to a reading at time t could then be calculated:

An example of the and electrode outputs is shown in Fig. 3, the records showing 90-min periods of uninterrupted lung gas circulation. Also shown are the simultaneously recorded breathing movements which have been adjusted in time for the geometric delay of the loop (26 s). Electrode output data as in Fig. 3 were digitized using an Apple graphics tablet coupled with an Apple II microcomputer (Apple Computer Inc., California). Placement of an electronic pen on the tablet face defined a discrete numerical pair in the total x–y array in the tablet. The computer was programmed to receive successive pen placements at a constant time increment of 10 s, corresponding to 1 mm of chart record. Coordinates were stored in a threedimensional array (⁠⁠, and time) and thereafter processed according to a programmed version of equation 1 above. Software was developed in order to redraw the corrected data in analogue format and to plot the coordinates in the form of versus diagrams. Accuracy of data entry on the graphics tablet was periodically checked by re-entering the same sections of chart record. Deviations of the retrieved data never exceeded ±0·4Torr, well within the overall experimental error.

These experiments were carried out at 25 °C on 21 male and female Xenopus laevis (82–134g). The femoral artery and lung (N = 10 animals) or left atrium and lung (N = 11 animals) were chronically cannulated so that systemic arterial blood or pulmonary venous blood could be sampled at the same time as lung gases. The techniques of lung cannulation (see above) and femoral catheterization (Boutilier, 1984) have already been described. In order to cannulate the left atrium, a mid-line ventral incision was made through the pectoral girdle to expose the anterior region of the heart. A 2–4mm incision was made in the pericardium overlying the left atrium. The atria of Xenopus, unlike those of ranid or bufonid anurans, have several fingerlike projections. The longest of these fingers was pulled through the pericardial incision and clamped off with polypropylene-coated artery clamps, isolating a blind end about 4 mm in length. Catheters were prepared by heat flaring the tip of a length of PP 50 tubing and cutting wedges out of the cannula wall just behind the heat flare over a 2·5 mm length. This was found to decrease the likelihood of blockage due either to blood clot formation or to drawing of the thin auricular wall over the holes during blood sampling. The cannula tip then was inserted through a small hole cut in the apex of the isolated portion of auricle and loosely tied in place by trapping a continuous piece of the auricular tissue with a loop of thread. With the clamp removed, the catheter was advanced into the central region of the auricle and then anchored firmly in place with a purse-string suture. The pericardium and chest cavity were carefully sutured and the cannula was fixed with thread at several points along its exit. Cannulations remained patent for several days with negligible blood loss.

Arterial and measurements were made (25 ±0·5°C) by allowing blood to flow through the femoral cannula into Radiometer electrode cuvettes connected in series. The blood displaced a 2 % CO₂/98 % air mixture into a syringe barrel connected to the outlet of the electrode pair. Pulmonary venous blood from the left atrium (⁠ and ⁠) was sampled in a similar fashion except that a small negative pressure had to be applied from the outlet syringe. After measurement, using Radiometer PHM-72 display meters, blood was reinfused into the animal by pushing the CO₂/air gas mixture back into the cuvettes from the syringe. The gas mixture was then immediately displaced once again with a 1-ml sample of lung gas, which had been drawn into a Hamilton gas-tight syringe over the period of time when blood was flowing into the electrode pair. Electrodes were calibrated frequently, with humidified gas mixtures from a Wösthoff pump.

Examples of the original recordings of partial pressures in alveolar gas, together with lung ventilation, are shown in Fig. 3 for undisturbed and freely diving Xenopus. One to two days of recordings were taken whilst the animals were kept in airequilibrated conditions (Fig. 3A). Subsequently, the toads were exposed to elevated levels of CO₂ (Fig. 3B) delivered at the skin by bubbling a 1% CO₂/99% air mixture into the water. However, the blowhole at which the animals ventilated their lungs was maintained at normocapnie levels. Measurements of water levels showed that these remained high throughout all stages of the latter experiments (>140Torr), whereas water ranged from 7 to 8Torr.

Oscillations in and corrected for electrode response delays, are shown in Fig. 4A,B from the first and second days of voluntary diving after recovery. Within the period shown for the first day, seven complete diving-emergence sequences were analysed, with dives ranging from 8 to 17 min in duration. The analysis shows that even a single lung ventilation is sufficient to raise the to high levels. In Fig. 4A, for example, a single ventilation caused a change from 38 to 120Torr. Compared to the oscillations in (20-150Torr), levels, whether for a single ventilation or a series of breaths, always cycled over a comparatively small range (12–19 Torr). levels fell rapidly as soon as the animal began to breathe. Upon submergence, the CO₂ levels increased over the first few minutes of the dive and, thereafter, remained in a steady state. In contrast, fell progressively throughout the dive so that oxygen uptake was eventually occurring with little or no substitution of CO₂, the net effect of which must be a decline in lung volume. The records of Fig. 4B reveal some changes in the breathing behaviour on the second day of the experiment, the animal having remained in the apparatus overnight. Though the cyclical patterns of were essentially the same as those reported earlier, the depletion of was far more gradual during dives on the second day and, despite the longer periods of submergence, did not reach such low levels. Differences of this sort were not necessarily confined to experiments on different days; they could occur within hours of each other.

The most rapid depletion rates of lung gas O₂ stores during voluntary diving occurred when the animals were seen to be spontaneously active (Fig. 5A). Activity also brought about much larger and rather more variable changes in (Fig. 5 A) and the dives tended to end when levels were still relatively high. The broad range of depletion rates that were observed are illustrated in Fig. 6. Although there can be little doubt that a strong correlation exists between the duration of the dive and the rate at which falls, the relationships are far from being simple and it was never easy to predict when a dive would end, particularly in active or disturbed animals.

The lung gas data obtained in the loop experiments can be plotted (Figs 5, 7, 8) on a diagram (Fenn, Rahn & Otis, 1946), a presentation which has been used extensively in the analysis of unimodal gas exchange. In a unimodal system in which the partial pressure changes are simple functions of the total volumes of gas exchanged, alveolar and expired gas points fall on a straight line whose slope represents the metabolic respiratory quotient (RQ) and whose origin on the abscissa is the inspired ⁠. Ventilation changes cause the alveolar and expired gas points to depart from the line because of the complicating effects of O₂ and CO₂ stores in the body, but when steady states are restored, these points return to the RQ line. In bimodally breathing animals, the analytical use of such a diagram is beset with problems since the lung gases are exchanged in a ratio which is not the same as the metabolic RQ, the bulk of the CO₂ being removed at the aquatic exchanger. In addition, the exchange ratio changes continuously as these animals are all intermittent breathers and steady states are almost never seen. Finally, the fact that the lung decreases in volume during a period of breath-holding will result in a distortion of the relationship between partial pressure changes and the volumes of gas exchanged. Nevertheless, the diagram is a convenient way of presenting partial pressure relationships in bimodal animals, even though the significance of the lung R lines is different from that in a unimodal system.

As Figs 5B and 8 show, the lung R values are about 0·8 in resting animals during a period of lung ventilation, but they fall very rapidly as a voluntary dive proceeds and comes into steady state. At this stage, the instantaneous values for lung R, as shown by the slope of the experimental line, fall to zero; in fact the actual gas exchange ratio may be slightly negative because the lung volume is falling and CO₂ must be absorbed in order to keep constant. Even in undisturbed animals, the values can vary over about 3–4 Torr at any value during different dives, as the shaded area in Fig. 5B shows. The relationships change during spontaneous activity (Fig. 5B), with significantly higher levels being found at any value than in resting animals. Partial pressures in the alveolar gas of animals that were prevented from surfacing for 30 min, after diving voluntarily (Fig. 7), also showed marked differences from those of freely diving animals at rest. Experimental dives of this sort led to a considerable rise in and fall in soon after the animals had attempted to surface and breathe. The increased activity associated with such dives leads to a lactacidosis and an increased production of CO₂ through acidification of the blood (Boutilier & Shelton, 1986b), this being the proximate cause of the upward trend in lung R values in the latter stages of the dive (region B, Fig. 7). During the first breathing period after surfacing from such an enforced dive, the animals would increase their ventilation by some 10-fold over that seen before the dive (Boutilier & Shelton, 1986b). The levels of lung R at this stage became much higher than the metabolic RQ (region C, Fig. 7). Increasing the of systemic venous blood, through exposure of the animal to hypercapnic water, also led to considerable hyperventilation after a dive as well as to high lung R values (Fig. 8). Aquatic hypercapnia, like activity, caused an increase in throughout the period of submergence and usually brought the dive to an end when levels were still comparatively high (Figs 3, 8).

The data shown in Fig. 9 were collected at various stages during breath-holding and lung ventilation cycles inXenopus that were diving voluntarily. In these graphs, gas tensions in femoral (A,C) or left atrial (B,D) blood are plotted against those of simultaneously determined lung gas samples. Regression lines for each set of data can be used to estimate representative gradients from gas to blood at any or values.

The gradient from lung to pulmonary venous blood averaged 9 Torr and changed little over the full range of fluctuation. The variation in the data was quite low (Fig. 9B). The gradient in the reverse direction was also fairly constant at 2 Torr. These gradients are both significantly larger than the alveolar arterial differences found in the mammalian lung and suggest inefficiencies in the amphibian gas exchanger (a high R_diff in Fig. 10). They may be due to longer diffusion distances from alveoli to arteries (Weibel, 1972), anatomical shunts between pulmonary artery and vein (Smith & Rapson, 1977) or inhomogeneities of perfusion causing and differences to occur in different regions of the lung.

In general, the levels in the femoral artery (Fig. 9A) were lower than those in the left atrium, though the differences were less marked at low values. Variation in the data was much greater than that in the pulmonary venous blood. Both of these differences can be attributed to the variable shunting of blood from the left atrium with some of that from the right (SI in Fig. 10) as the two streams pass through the undivided ventricle. The greater difference between and at high values does not necessarily indicate that the shunt is larger at these levels. Because of variations in the slope of the dissociation curve, small quantities of venous blood have substantial effects on when added to fully saturated blood from the pulmonary vein but much smaller effects as that blood becomes progressively less saturated during a dive. However, the changes in blood flow rates through the heart, evidence for which is reasonably convincing (Shelton, 1970), suggest that there must be changes in shunt pattern during a dive.

During periods of voluntary diving, the lungs ot Xenopus are important sources of stored O₂, the rate of use being clearly related to activity levels and dive duration (Figs 4, 5, 6). In an earlier paper, which suggested that anaerobiosis was unimportant in free dives (Boutilier & Shelton, 1986b), calculations were presented to show that undisturbed Xenopus (100 g, temperature 20 –25 °C) could remain submerged for 23–28 min by using O₂ stored in lungs and blood for its cell metabolism. The data in the present paper allow a more detailed analysis of store use. Though the rates of O₂ uptake from the unventilated lung vary considerably, two main patterns emerge. During prolonged dives, falls slowly and regularly, and in some cases the rate of decline increases in the later stages of the dive (Figs 3, 4, 6). In comparison, during shorter dives declines exponentially, falling at high rates in the early stages and at progressively lower rates as the dive proceeds (Figs 4, 5, 6). There are many variations of pattern, however, and there can also be sudden changes in the rate of depletion during a dive (Fig. 6). Factors additional to the metabolic rate of the cells seem to be important in determining the way in which the lung store is used.

An idealized model of the O₂ exchange, storage (Fig. 10A) and transport system (Fig. 10B) of Xenopus (100g at 20 –25°C) makes it easier to examine these factors. Oxygen input into the animal from the environment goes on through the lung at an average rate of 55 μl min⁻¹, though the actual rate will be determined by lung ventilation and perfusion (QP) and will vary between 0 μl min⁻¹ during a dive and about 500 μl min⁻¹ during recovery. Input also occurs through the skin at 15μl min⁻¹, a rate which is assumed to stay constant in the model throughout the dive. The resting O₂ consumption of the cells is set at a constant 70 μl min⁻¹ (all values derived from Emilio & Shelton, 1974). The main O₂ stores are found in the lungs (total volume 8 ml, Boutilier & Shelton, 1986b) and the blood (total volume 13 ml, Emilio & Shelton, 1980), the tissue stores being small and unimportant. The blood volume has been subdivided into arterial (3 ml) and venous (10 ml) components, based on the distribution found in mammals; the venous component has been further subdivided into pulmonary veins (1ml), cutaneous veins (0·5 ml) and systemic veins (8·5 ml). These are all estimates, but they are reasonably realistic and do not seriously affect the principles underlying the model’s operation.

It is also clear from the model that many other possibilities exist in the relative rates of use of the different stores during a dive and that these possibilities are determined largely by changes in blood flow (Fig. 10B). Selective pulmonary vasoconstriction (RP increase) would cause Q̇p to fall more than Q̇s and increase shunt 1, thus conserving-O₂ stores in lung and pulmonary vein (Shelton, 1970, 1976). Systemic and venous blood stores would then fall more rapidly than shown in the model, enhancing cutaneous gas exchange, and eventually causing the lung stores to be used at progressively increasing rates in the later stages of a dive (Fig. 6). In a highly active animal, on the other hand, cardiac output and blood flow to lungs and body almost certainly continue at a high level (Emilio & Shelton, 1972), so that arterial — venous differences would not be increased and values in all parts of the storage system would move closer together than shown in the model. The rate of decrease would then depend on the total body capacitance for O₂ with roughly similar values in all stores. As the model shows, this total capacitance increases progressively as falls, to a maximum at about 28 Torr. would, therefore, decline rapidly at first in an active animal, but at progressively decreasing rates as the maximum capacitance value of the total store was approached (Fig. 6).

The elimination and storage of CO₂ is more difficult to model with any degree of certainty because the gas behaves quite differently from O₂. The skin is extremely effective in removing CO₂; in total, more than 80% of the gas produced in metabolism can leave by that route in Xenopus (Emilio & Shelton, 1980). In addition, the CO₂ capacitances of blood and tissues are very large and not accurately quantified. Since the CO₂ molecule can hydrate and then dissociate into H⁺ and HCO₃⁻ ions, any chemical group that can act as an H⁺ acceptor is a potential CO₂ store. Other complications include the large Haldane effect in Xenopus blood (Boutilier & Shelton, 1986a), the low blood buffer slope in vivo (Boutilier, 1981), suggesting a large transfer of HCO₃⁻ to the extravascular compartment as CO₂ rises during a dive, and the importance of intracellular stores.

The overall effect of these differences is that the changes during a dive are very much smaller than those of ⁠, not merely in alveolar gas but also in blood, as Fig. 9 illustrates, and in cells and tissues. Ultimately, in the later stages of a dive, steady states occur in as the stores reach equilibrium and all the CO₂ produced by the resting metabolism leaves via the skin. These results are consistent with our observations of little change in acid-base status of arterial blood during the later stages of a voluntary dive (Boutilier & Shelton, 1986b). The gas exchange ratio of the stores and the instantaneous gas exchange ratio of the lung reach zero at these times (Figs 5,7).

Activity in the diving animal causes levels to increase (Figs 5, 7). Even so, the instantaneous gas exchange ratio usually falls to zero towards the end of a dive, presumably when the increased gradient from blood to environment causes the CO2 loss to balance the increased production. Activity during a forced dive causes even higher levels of (Fig. 7) because of the addition to the blood of protons produced during anaerobic metabolism (Boutilier & Shelton, 1986b). When such dives (with increased levels of CO₂ in body stores) come to an end, the very considerable hyperventilation is accompanied by lung R values that are much higher than the metabolic RQ (Fig. 7). Exposure of animals to hypercapnic water has a similar effect (Fig. 8), and even in resting animals high lung R values can be seen during ventilation (Fig. 5B). They are attributable to another difference in the O₂ and CO₂ stores, namely that the rates of adjustment are different. Thus the hyperventilation and increased cardiac output at the end of a dive enable the O₂ stores to be renewed and reach equilibrium rapidly, whereas the CO₂ stores continue to run down for some time. This is because the former occur mainly in lungs and blood, whereas the latter are distributed throughout blood, tissues and cells. Differential rates of store adjustment during a dive are not so obvious, because the rates of change in the stored gases are much lower than they are when the animal surfaces.

The role of the lungs in eliminating CO₂ has been examined in some recent studies (Gottlieb & Jackson, 1976; Jackson & Braun, 1979; Boutilier, McDonald & Toews, 1980; Boutilier, 1984) and it has been suggested that these organs are important in the control of CO₂ levels in the body, despite the fact that the skin eliminates the major part of the gas produced by metabolism. The present analysis shows that the lungs are important in eliminating the CO₂ stored during a period of breath-holding, but that they are not of major importance in forming part of that store themselves. Lung ventilation lowers levels in the body, which causes the amount of CO₂ lost through the skin to fall. During a breathing period, therefore, the CO₂ stores are run down and some CO₂ produced by the cells is directly eliminated by the lungs. During a breathing pause, however, the CO₂ stores build up and the amount of CO₂ lost through the skin increases. By fine adjustment of levels and of CO₂ stores, systems controlling lung ventilation can determine both the levels of CO₂ in the body and the relative amounts of the gas removed through the two exchangers.

Since tidal volume changes rather little in Xenopus, the most important variable in the regulation of ventilation is the breathing frequency (Boutilier, 1984). Commonly, breaths are grouped into bursts, between which there are long periods of breathholding, though other patterns do occur (Shelton & Boutilier, 1982; Boutilier, 1984). The present experiments suggest that falling O₂ levels (either concentrations or partial pressures) in lungs and blood may be the important stimulus to take a breath or begin a breathing burst. Voluntary dives do not appear to go below the point of maximum total capacitance in the O₂ stores (approx. 28Torr), though breathing is often triggered at higher levels than this with considerable variation in the triggering level. The rate of use of the O₂ stores is probably a factor of some significance because dives end at higher values of when the decline is rapid than they do when it is slow. There is no evidence that the or levels are directly involved in triggering breathing because they, and the pH of arterial blood (Boutilier & Shelton, 1986b), remain constant during the later stages of a dive (Figs 4, 5B). On the other hand, elevated CO₂ values, as seen when the animals are active or in hypercapnic water, are associated with more frequent breathing, bursts being triggered at relatively high values. There is obviously some interaction between the two gases; it may be that elevated CO₂ levels affect the sensitivity of a predominantly O2-mediated response. It is well known that hypercapnia increases the ventilatory response to in mammals (Cunningham, 1974; O’Regan & Majcherczyk, 1982).

Since different regions of the O₂ store can run down at different rates, some knowledge of receptor location is crucial in understanding the total system. At the moment, there is no direct evidence as to receptor sites or function. The main O2 stores are in the lungs and systemic venous blood (Fig. 10). They could be regulated with reference to a single site receptor in, say, the arterial blood but only if information were also available about heart output and the levels of blood flow or of peripheral resistance (Rp, Rs and Rc) in the major divisions of the circulation. This would be a very indirect means of regulation with such labile storage and transport systems. More widespread receptor sites would clearly make for simpler control of the stores and work is currently going on to try to locate them.

It is equally difficult to say with certainty where the information that brings breathing to an end can originate. Ventilation stops when and are within certain broad limits (Figs 3, 4), but again there is a marked lack of precision in the gas levels at which animals will end a breathing burst and dive. Hyperventilation following a dive continues for some time after high levels of and have been achieved, particularly when the animals have been active or exposed to hypercapnic water (Figs 7, 8). This suggests that breathing may continue until stored CO₂ has been removed from the cells and blood and that in the blood and lungs, or pH in the blood, may play a more important role than in ending a breathing burst. Again there is almost no information on the location and characteristics of such CO₂ receptors (see, however, Milsom & Jones, 1977; Kuhlmann & Fedde, 1979).

What is clear from all these studies on Xenopus is that the breathing-diving behaviour and the associated oscillations of respiratory gases in the body are much more regular in undisturbed animals than they are in those that are disturbed or active. Brain centres higher than the medulla must play an important part in regulating the diving behaviour and can clearly override information from the central and peripheral sites responsible for the control of breathing. It is this involvement of higher centres that confers much of the unpredictability on the system.

These studies were supported by grants from the University of East Anglia, the SRC and the NSERC, to all of whom we express our thanks. RGB was in receipt of a Commonwealth Scholarship. We are grateful to Dr Peter Croghan for his help with computer programming.