The Transition to Air Breathing in Fishes: III. Effects of Body Size and Aquatic Hypoxia on the Aerial Gas Exchange of the Swamp Eel Synbranchus Marmoratus

Aerial respiration has evolved independently among many species of fishes which, as a result, possess a variety of anatomical, physiological, biochemical and behavioural specializations for aerial gas exchange (Carter & Beadle, 1931; Johansen, 1970; Graham, 1976; Kramer, Lindsey, Moodie & Stevens, 1978; Randall, Burggren, Farrell & Haswell, 1981). Despite a considerable diversity of air-breathing adaptations, fishes that breathe air can be separated into two groups, amphibious and aquatic air breathers (Graham, 1976). Amphibious air breathers rely on this respiratory mode during their frequent and routine terrestrial excursions, whereas aquatic air breathers, which for the most part remain confined to water, periodically surface and gulp air to obtain supplemental O₂ (Johansen, 1970; Graham, 1976).

Because they seldom leave water, aquatic air-breathing fishes lack specializations such as for aerial vision, terrestrial locomotion, desiccation resistance and nitrogen excretion usually present in amphibious fishes (Graham, 1976; Gordon, Ng & Yip, 1978; Iwata et al. 1981). Also, aquatic air breathers, unlike most amphibious forms, typically have separate anatomical structures and gas exchange surfaces for aerial and aquatic respiration (Johansen, 1970). The air-breathing organ (ABO) of these fishes is usually highly specialized and sequestered in the body away from direct contact with ambient water (Graham, 1976). This allows a fish to consume O₂ from the ABO while simultaneously ventilating its gills in water for CO₂ and N2 release, electrolyte and volume regulation and, depending upon aquatic conditions, some O₂ uptake (Johansen, 1970; Graham, 1983).

This paper investigates aspects of the aquatic aerial respiration of the swamp eel Synbranchus marmoratus one of only a few fish species that, owing to conditions in its natural habitat, has evolved the capability for both amphibious and aquatic aerial respiration (Liem, 1980; Heisler, 1982). The swamp eel ranges from southern Mexico to southern Brazil and is found in streams, rivers, lakes, ponds and swamps (Breder, 1927; Carter & Beadle, 1931; Luling, 1958; Rosen & Greenwood, 1976; Kramer et al. 1978). This species breathes air while inhabiting hypoxic water and during terrestrial excursions in search of prey and new habitats as well as when it is confined to its relatively dry mud burrow during the tropical dry season (Carter & Beadle, 1931; Bicudo& Johansen, 1979; Heisler, 1982). Synbranchus marmoratus is also one of the few species that utilize gills and adjacent buccopharyngeal epithelium for air breathing (Liem, 1980). This means that unlike other aquatic air-breathers this fish cannot ventilate its gills while it holds air in its ABO (Heisler, 1982).

The major objective of this research was to compare the swamp eel’s ability for aquatic aerial respiration with that of other species. Previous papers in this series (Graham & Baird, 1982; Graham, 1983) demonstrated that a suite of biochemical and physiological compensations for hypoxia are initiated by the onset of facultative air-breathing in the armoured catfishes Ancistrus chagresi and Hypostomus plecostomus. As a result of hypoxia acclimation, both these species were able to reduce their air-breathing frequency in hypoxic water (Graham & Baird, 1982). It was further shown that hypoxia-acclimated Ancistrus has an increased air-breathing efficiency and a heightened capability for aquatic respiration in hypoxia (Graham, 1983). Acclimation to hypoxia, however, did not change the threshold aquatic O₂ partial pressure that elicited air-breathing in these species (Graham & Baird, 1982). Gee (1980) also reported no effect of hypoxia acclimation on the air-breathing threshold of the mud minnow, Umbra limi.

In contrast to these findings Bicudo & Johansen (1979) reported that, following 6 weeks of air-breathing and acclimation to hypoxia, S. marmoratus from Brazil initiated air-breathing at a significantly higher (54·0 mmHg) than did normoxia-acclimated control fish (30·1 mmHg). Aquatic hypoxia also appears to have less of an effect on the gas exchange capacity of S. marmoratus. Both the total blood haemoglobin (Hb) concentration and the Hb-O₂ affinity of air-breathing Ancistrus and Hypostomus are elevated in hypoxic water (Graham, 1983 and unpublished data). Weber, Wood & Davis (1979) also reported that 4–7 days of air-breathing and hypoxia increased the Hb-O₂ affinity of two Brazilian armoured catfish species. However, this treatment did not affect either the total Hb or the Hb-O₂ affinity of 5. marmoratus.

These observations suggest that aspects of respiratory control and the response mechanisms to aquatic hypoxia, including air-breathing, in S. marmoratus are not the same as in other air-breathing fishes. To test this, we studied this species using specimens captured in Panama. Our first objectives were to confirm both the absence of a Hb concentration change and the presence of air-breathing threshold shifts, both of which have been previously reported for hypoxia-acclimated Brazilian S. marmoratus. We also determined the effect of body size on the air-breathing threshold and ABO volume and examined how and affect aerial and aquatic gas exchange. Our preliminary studies revealed differences between the air-breathing physiology of S. marmoratus from Panama and what has been reported for fish from Brazil. Thus an additional objective of this investigation was to determine if differences between South American and Panamanian populations could be attributable to geographic isolation and possibly different environmental selection pressures or to the experimental protocols of different investigators.

Swamp eels weighing between 0·5 and 900 g were collected by hand net and with live-baited minnow traps in the Burunga and Mandinga Rivers near Arraijan, Republic of Panama and transported by air to the Physiological Research Laboratory, Scripps Institution of Oceanography, La Jolla, California. Fish were maintained in dimly lighted aquaria (25–27 °C) on a natural photoperiod and fed beef liver or live goldfish at least once each week.

Groups of S. marmoratus were kept in hypoxic water and thus forced regularly to breathe air for up to 10 weeks. Light and temperature conditions and feeding regimen were the same as for control fish (above). The absence of aeration resulted in hypoxic conditions. Electric filters fitted with extensions on the in- and out-flow tubes were used to clean and mix water without aeration. Plastic sheeting draped snugly over the water surface of the aquarium and filter reservoir minimized the surface area for O₂ diffusion from air but provided space, along the edges of the tank, for fish to surface and gulp air. Water temperature and in the hypoxic tanks were monitored daily with a Yellow Springs Installent (Model 54) O₂ meter and probe.

Control and hypoxia-acclimated S. marmoratus were compared for blood Hb content and for their threshold. Contrasts were made between fish of similar body size. Blood comparisons were made using fish (125–450 g) that had been in hypoxia for 14–28 days, a time determined to be sufficient for an erythropoietic response to occur in other air-breathing fishes (Graham, 1983). Blood samples (0·2–1·0 ml), taken by cardiac puncture, were withdrawn into tuberculin syringes that were flushed with heparin and dried prior to use. To facilitate handling, fish were cooled in water at 10–12°C for 5–10min before blood sampling. Total Hb was estimated using the cyanmethaemoglobin procedure (Graham, 1983).

Air-breathing thresholds in progressive hypoxia (i.e. the at which fish initiate air-breathing) were compared in control and hypoxia-acclimated groups of S. marmoratus using procedures described by Graham & Baird (1982). Fish acclimated to hypoxia for 6–10 weeks were used in these tests in order to ensure comparability with Bicudo & Johansen (1979). Groups (TV = 5–6) of test fish (40–152g) were transferred to a 301 experimental aquarium (26±1°C) containing aerated water. The aquarium was back lighted and positioned behind a visual blind. Following a 12–24 h adjustment period, aeration was stopped and aquarium was slowly (about 0·7 mmHg min^-1) reduced by bubbling N₂ gas into the tank through an air stone. at the time of the first air breath by each test fish was recorded and at least three replicate threshold tests were made on consecutive days for each group. In the 24–30 h interval between replicate testing, control fish were maintained in aerated water. Hypoxia-acclimated fish were re-exposed to hypoxic water soon after transfer to the experimental tank and, in successive threshold tests, 6–10 h prior to testing. Neither previous threshold tests nor the treatment between tests affected subsequent threshold determinations.

Air-breathing threshold determinations were also carried out on control fish weighing between 6 and 850 g in order to learn the effects of body size.

Gas exchange measurements were made using an L-shaped Lucite respirometer submerged (except for the upper end of the vertical section) in a constant temperature (25·0 ±0·1 °C) water bath (Fig. 1). The chamber was darkened with black plastic and its uppermost section was covered by a dark cloth, dimly back lighted, and observed through a blind. Depending on fish body size, either 1·8 or 4·01 respirometers were used. Fish that had been starved 24 h were placed in the chamber and then allowed a 24 h adjustment period before experiments were begun. During this time filtered, aerated water was continually pumped through the respirometer from a reservoir also located in the water bath. When the respirometer was closed, water was recirculated and, because of fish respiration, decreased. In air-breathing experiments a fish was allowed access to air at the top of the vertical section (Fig. 1) and, once had dropped to air-breathing threshold level, the fish would periodically ascend, take a gulp of air, and then sink to a resting position with all or most of its body in the horizontal section.

Aerial gas exchange measurements were made on fish that had been air-breathing for 12–96 h. In preliminary experiments fish were required to take successive breaths from an atmosphere in which the O₂ partial pressure was continuously recorded (Fig. 1 inset). At the beginning of each test period a 75–100ml air phase was enclosed at the top of the vertical section by a rubber stopper. The YSI O₂ probe mounted in the stopper projected into the air phase. This was initially calibrated (25 °C) in air and N₂ gas, and a Scholander gas analyser was used to verify correspondence between the electrode reading and O₂ levels in the air phase. To ensure that air-phase O₂ remained close to saturation, the contents of this space were flushed completely with fresh air (using a syringe and Tygon hose connector) between every 3–4 air breaths. The vertical section was continuously observed so that the air release, stealthy ascent, and air gulp by the fish could be accurately recorded. The volume of an air breath was calculated from a measurement of the vertical displacement distance of the respirometer water surface following submergence of the fish with air in its ABO. Volume constants (ml mm^-1 displacement) were predetermined for each respirometer. Release of an air breath caused an abrupt step drop in the O₂ signal and was also confirmed by observation. The amount of O₂ utilized from the air breath was calculated from the magnitude of the change in the O₂ signal and the known air-breath and air-phase volumes. Utilization of O₂ could then be calculated and related to the duration of the air breath, or expressed as an instantaneous O₂ consumption rate ⁠. A correction factor for the rate of O₂ diffusion from the air phase into the hypoxic respirometer water was necessary and could be computed from the steady gradual decline in the air-phase O₂ record. Also, since S. marmoratus was observed routinely to exhale all gas from its ABO at the end of an air-breath, the calculated breath volume, when corrected for O₂ utilization, would indicate ABO volume and this could be related to fish body size.

Procedures described above did not permit separate analyses of each air breath for both CO₂ and O₂ content and necessitated both an indirect measurement of ABO volume and a correction for O₂ diffusion from air to water. To eliminate these limitations and to examine the effects of on aerial gas exchange, the respirometer was fitted with a Y-shaped side arm (Fig. 1) that allowed an air-breathing fish regular access to atmospheric air. Aided by favourable back lighting of the open water surface, fish in the respirometer quickly learned how to surface for an air breath. Also, since fish typically released air breaths from below the surface and always several minutes prior to surfacing for another gulp, expelled gas bubbles invariably ascended under the collection apparatus mounted over the vertical section of the tube. A gas collector similar to that described by Graham (1983) was fitted into the vertical section. This consisted of a funnel filled with hypoxic water and a graduated centrifuge tube that fitted tightly into the neck of the section (Fig. 1). A 20gauge blunt needle mounted flush with the inner apex of the tube permitted attachment of a stopcock and syring assembly. Using the large vertical syringe (Fig. 1) expelled gas was immediately (15 s) pulled into the tube for a volume determination, then raised to the level of the stopcocks and withdrawn sequentially into syringes 1 and 2. Hypoxic water contacted gas in the centrifuge tube, funnel, and in stopcock 1, but mercury was used to fill the dead spaces of syringes S₁ and S₂ and stopcock 2 (Fig. 1). Gas taken into syringe S₂ was analysed for O₂ and CO₂ content with the Scholander and Radiometer O₂ and CO₂ electrodes calibrated at 25 °C. The fish was observed continually to time both air breaths and the inter-air-breath intervals and to verify that all released gas was captured. Gas diffusion between water and an expelled breath necessitated rapid processing of a breath (Graham, 1983). A mean hourly aerial (STPD) was calculated from O₂ uptake and utilization data recorded for a series (usually covering 1 h or longer) of air breaths. Measurements of and were made at regular intervals during each test in order to examine their effects on breath duration and gas exchange. In some tests a 5 % CO₂:95 % N₂ gas mixture was initially bubbled into respirometer water to determine the combined effects of aquatic hypoxia and hypercapnia on aerial respiration.

Respirometer tests without an air phase were conducted to determine how progressive aquatic hypoxia affected the aquatic of S. marmoratus. Procedures (Graham, 1983) were as follows: the respirometer was closed and the rate of O₂ decline was monitored using the YSI electrode mounted in the flow of the system. Antibiotics (furacin and penicillin) were initially added to the reservoir to reduce background microbial respiration, and blank respiration corrections were made for each fish. Instantaneous estimates were made for each 10 mmHg range of ambient ⁠. Tests were repeated on successive days for all fish and runs were continued down to a of 15 mmHg or to the point where a fish repeatedly ascended the vertical section ‘to search for air’. The effect of on aquatic ventilation rate was determined by counting the ventilations of fish (N = 15) held in a clear respirometer tube positioned behind a blind and dimly back lighted.

At or above a of 120 mmHg, the mean aquatic of S. marmoratus (10–790 g) is 26·4 ± 5·4 ml kg^-1 h^-1 (x̄ ± 95 % confidence limits, N = 84, 25 °C). The of fish denied access to air steadily decreases in progressive hypoxia in the pattern of a metabolic O₂ conformer (Fig. 2 and Discussion). A least squares regression analysis reveals a significant correlation between mean aquatic (within each 10 mmHg increment of ⁠) and from 30 –150 mmHg (r = 0 ·90, N= 12, P <0 ·05), and the slope of the regression equation is significantly different from zero (0 ·179 ±0 ·061).

The aquatic ventilation rate of 15 S. marmoratus (76 –360g) is 27 ·9 ±3 ·1 ventilations min^-1 (x̄ +95% confidence limits, range 17 –39, N = 133, 25 °C). Larger fish tend to have lower rates but the negative correlation is not significant (P > 0 ·05). In normoxic water, S. marmoratus does not ventilate its gills continuously and periods of apnoea regularly occur (see Discussion and Heisler, 1982). Fish exposed to progressive hypoxia ventilate their gills both slightly faster and for a greater amount of time (i.e. more each hour) than in normoxia. However, mean ventilation rates did not change even with Pwo2 values as low as 5 –15 mmHg. Ventilation rates are also not different in control and hypoxia-acclimated S. marmoratus.

Air-breath duration and the O₂ content, and volume of expelled air breaths were determined for 17 S. marmoratus (58 –760 g). Over the range 0 –20 mmHg, average air-breath duration is 15 ·7 ±2 ·1 min (x̄ ± 95 % confidence limits, range 2 –53, N = 137, 25 °C). The inter-air-breath interval of this fish, that is the time between the release of an air breath and the gulping of the next one, averaged 15 ·1 ±2 ·7min (x̄ ±95% confidence limits, range 1 –42, N = 117). Although both mean air-breath duration and mean inter-breath interval are similar, no correlation exists between them (also see Heisler, 1982) nor does breath duration relate to body size. Further the inter-air-breath interval is not correlated with O₂ utilization on either the preceding or succeeding air breaths, with fish body size, or with ⁠. Decreasing ⁠, however, has a significant negative effect on air-breath duration (Fig. 3) with shorter but more frequent air breaths occurring in progressive hypoxia.

Mean aerial O₂ utilization correlates with air-breath duration (r= 0 ·80, P <0 ·05, N= 22); however, the relationship (Fig. 4) is not linear and the rate of O₂ uptake is much greater from short air breaths than from longer ones. This is similar to results obtained from fish with cannulated ABOs (Johansen, 1966; Bicudo & Johansen, 1979). Fig. 4 reveals that at mean air-breath duration (15 ·7min) O₂ utilization approaches 80%. In a few cases utilization is nearly 100% whereas in others it is relatively low. Although variability in O₂ utilization (and duration) was evident for individual fish, no relationship between utilization and body size was found.

The mean aerial of S. marmoratus (Fig. 2), estimated only for the time air breaths were held and thus not including the inter-breath interval, is 23 ·3 ± 2 ·2 ml kg^-1 h^-1 (x̄ ± 95 % confidence limits, range 6 ·9 –55 ·7, N = 137, 25 °C). This is not significantly different from aquatic in normoxic water (26 ·4ml kg^-1 h^-1) and is not significantly affected by which ranged from 0 ·5 to 51 ·4 mmHg in these tests (Fig. 2).

Each ABO volume estimate for S. marmoratus, either measured directly or calculated from a water displacement volume, was corrected for O₂ utilization and plotted against body weight (Fig. 5). Volumes determined for each fish were tightly grouped around the mean and no correlations were found between ABO volume and ⁠, air-breath duration, or inter-air-breath duration. The 95 % confidence limits around the exponent relating ABO volume to body weight (0 ·737 ± 0 ·088) in Fig. 5 do not overlap 1 ·0, indicating that the rate of increase in ABO volume occurs disproportionately with respect to body size in S. marmoratus.

Expelled air-breaths from five S. marmoratus (N =49, Table 3, Fig. 6) were examined to learn the effects of air-breath duration and on expired breath O₂ and CO₂ partial pressures Mean values for four groupings of air breaths, separated on the basis of relative duration (i.e. shorter or longer than mean duration) and conditions in the respirometer, are contrasted in a diagram (Fig. 6) and in Table 3. Fig. 6 reveals a correspondence between and and the expected inverse relationship for breath duration and Also shown are 13 mean breath points (typical duration 5 min, reported fuBrazilians, marmoratus by Bicudo & Johansen (1979). Because of their longer breath duration, breaths from fish in the present study have both a lower mean O₂ conten and a higher mean CO₂ content than those studied by Bicudo & Johansen (1979).

Table 3 summarizes data for aquatic O₂ and CO₂ conditions and for breath duration and both and in the four groups of data shown in Fig. 6. Moderately good agreement exists between each group’s duration and O₂ utilization relationships (Table 3) and that shown in Fig. 4. Depending upon both and the time an air breath was held, did not always exceed ⁠. Table 3 shows that is greater than all breaths released by fish in ambient values of 5 –10 mmHg irrespective of breath duration. In a higher range, is above in only 7 of 16 (44%) breaths held 8-12min (Table 3). For breaths held 14 –35 min, the number in which exceeds increases significantly (Chi square contingency = 6 ·8, P < 0 ·05) to 8 of 11 (73 %, Table 3). The aerial respiratory exchange ratio (RE, Table 3) of S. marmoratus is very low. This value was computed for each group using only positive values and each was corrected for both and the of inspired air prior to calculation.

Typically, fishes are metabolic O₂ regulators and thus can maintain down to a critical (Ultsch, Jackson & Moalli, 1981). Documented instances of metabolic O₂ conformity among lower vertebrates are rare and most data showing this response are considered equivocal (Ultsch et al. 1981). Our study shows that Panamanian S. marmoratus is a metabolic O₂ conformer when exposed to progressive aquatic hypoxia without access to air. The air-breathing threshold of this species varies directly with body size (see below) and, for the weight range of fish examined (6 –850 g), the percentage reduction in aquatic at threshold extends from 43 % (850 g) to 72% (6g) of mean aquatic measured above 120mmHg (Fig. 2). These reductions are greater than in the armoured catfish Ancistrus which regulated aquatic nearly down to its air-breathing threshold (33 mmHg, Graham, 1983). Additional studies (J. B. Graham & T. A. Baird, in preparation) have shown that the energetic cost of gill ventilation is higher in S. marmoratus than in other fishes. Metabolic O₂ conformity with hypoxia may therefore be an energy-saving mechanism utilized by this fish until, at a combination of and reduced relative determined by body size, air-breathing must be initiated (Fig. 2).

Our finding of O₂ conformity is different from that of Bicudo & Johansen (1979), who determined that Brazilian S. marmoratus regulate down to 30 –50 mmHg ⁠. Also in contrast are our observations that the gill ventilation rate of S. marmoratus remained constant in progressive hypoxia. Possible bases for these and other physiological differences between Panamanian and South American populations of S. marmoratus are discussed below.

The effect of body size on the air-breathing threshold of S. marmoratus may be due to factors such as the scaling of ABO (respiratory chamber) volume (Fig. 5) which Becomes relatively smaller in larger fish. Also important are the smaller body-surface area to volume ratio of larger fish which would limit cutaneous O₂ uptake, especially in hypoxia (Heisler, 1982), and the energetic cost of ventilation which is relatively high for S. marmoratus and increases with body size (J. B. Graham & T. A. Baird, in preparation).

Our experiments demonstrated no effect of hypoxia acclimation on the air-breathing threshold of S. marmoratus and thus do not verify the results of Bicudo & Johansen (1979) with Brazilian fish. These workers did not specify their procedures for threshold determination but reported that five fish (mean weight 123 ·8 g, range 76 –190) ‘chronically’ adapted to hypoxia (25 °C) for 6 weeks commenced air-breathing at a significantly higher than did control fish. Our tests were done using fish of similar size (Table 1) acclimated to hypoxia and regularly air-breathing for 6 –10 weeks. Thus unknown differences in experimental protocol probably account for the different results (see below). However, our finding of no effect of hypoxia on the airbreathing threshold of S. marmoratus is similar to results for other species (Gee, 1980; Graham & Baird, 1982) and suggests that similar mechanisms (i.e. ambient O₂ sensors, a reduced aquatic in hypoxia, and possibly limited cutaneous O₂ uptake) trigger air-breathing in this fish (Heisler, 1982).

Compared with air-breathing data for Brazilian fish (Bicudo & Johansen, 1979), S. marmoratus from Panama held air-breaths longer (15 ·7 vs 5 ·10 min) and had correspondingly higher average O₂ utilizations (80 vs 40 –50 %). Bicudo & Johansen (1979, p. 61) reported a 2 –3 ml ABO volume for a 150g fish but did not indicate sample size or the variability around this estimate. Since our Fig. 5 shows a 150 g fish to have about a 6 ml ABO volume, it would seem that the ABO of Brazilian fish is smaller. However, we calculated ABO volume from the body weight (74 –211g) and tidal volume data given by Bicudo & Johansen (1979, Table 2) and obtained values of 2 ·9 –10 ·8 ml which, when plotted with body weight, agree closely with our data in Fig. 5. The aerial RE of Brazilian fish was found to be about 0 ·1 by Bicudo & Johansen (1979) and the data points reported by them are in line (Fig. 6) with values in the present study. Bicudo & Johansen (1979) did not specify respirometer in their tests and slight differences between their RE estimate and our lower values (Table 3) may reflect the absence of a correction factor.

Analyses of expelled breaths indicate that aerial CO₂ release by S. marmoratus has no effect on O₂ uptake and is a passive process directly related to breath duration and The positive effect of necessitated use of a correction factor prior to calculation of aerial RE (see above and Results). This suggests that a steady state equilibrium condition for CO₂ exists between S. marmoratus and its ambient water and that this establishes, during the time air is held, a CO₂ diffusion gradient from fish tissue to gas in the ABO. Depending upon gradient steepness and breath duration, a net outward flux of CO₂ may occur. However, since little exhalant CO₂ is released aerially (Table 3), S. marmoratus must either accumulate respiratory CO₂ in its tissues during an air-breath and then flush this gas by aquatic ventilation during the inter-air- eath interval, or it must release CO₂ simultaneously while air-breathing, utilizing an equatic extra-branchial exchange surface such as the skin (Heisler, 1982).

Air-breath duration is affected by (Fig. 3) but the aerial of S. marmoratus remained independent of this factor (Fig. 2). The average durations of both the air breaths (15 ·7 min) and the inter-air-breath intervals (15 ·1 min) of this fish are much longer that has been observed for other species (Gee, 1976; Kramer & Graham, 1976; Graham & Baird, 1982). Because of its long inter-breath interval the air-breathing frequency (i.e. the number of complete breath and inter-breath cycles that occur hourly) of S. marmoratus is much less than for either Ancistrus or Hypostomus. By combining the linear equation for air-breath duration and determined for this fish (Fig. 3) with its mean inter-breath interval, we obtain an estimated air-breathing frequency of about 4 breaths h^-1 at a of 0 mmHg and 2 breaths h^-1 at 30 mmHg. These frequencies are similar to some observed by Bicudo & Johansen (1979, Fig. 5) but are much lower and change less with than those of hypoxia-acclimated Ancistrus (13 breaths h^-1 at 0 mmHg, vs 5 at 30 mmHg) and Hypostomus (13 at 0 mmHg vs 8 at 30 mmHg) (Graham & Baird, 1982). A negative relationship between air-breathing frequency and can be attributed to reductions in air-breathing effectiveness caused by the greater diffusive loss of aerially-obtained O₂ in more hypoxic water (Graham & Baird, 1982; Graham, 1983). This problem, common to many air-breathing fishes, stems from the ‘series’ circulation between ABO and gills and the need to ventilate gills in hypoxic water (for N₂ and CO₂ release and ion balance) which leads to transbranchial O₂ loss (Johansen, 1970; Johansen, Mangum & Lykkeboe, 1978). The above comparison shows a smaller effect of on the air-breathing frequency of Synbranchus than for either Ancistrus or Hypostomus, which is to be expected since the former holds air over its gills and cannot simultaneously ventilate water while air-breathing. Nevertheless, these calculations do suggest that some aerial O₂ is lost by S. marmoratus and this probably occurs through its scaleless skin. Studies in progress show that the skin of this fish is active in aquatic uptake and that some cutaneous O₂ loss could occur during air-breathing, particularly if a cutaneous pathway for CO₂ release is used (Heisler, 1982).

Our finding that hypoxia acclimation had no effect on the Hb concentration of S. marmoratus agrees with results obtained by Weber et al. (1979), who acclimated Brazilian fish to hypoxia for 4 –7 days. These workers also found that hypoxia exposure did not alter the Hb- O₂ affinity (indexed by erythrocyte phosphate-Hb ratio) of their study fish. Together these observations indicate that a prolonged transition period is not required to develop air-breathing proficiency in S. marmoratus which, regardless of its history of exposure to hypoxia, retains the physiological capacity to make a complete and abrupt transition to aerial respiration. This is unlike the gradual hypoxia-acclimation response of Ancistrus which is initially intolerant of severe hypoxia (Graham, 1983), but consistent with the natural respiratory requirements imposed on S. marmoratus by terrestrial excursions and by conditions that occur in some of its habitats at night when environmental O₂ demand can result in the rapid onset of severe hypoxia (Johansen, 1970; Kramer et al. 1978).

Compared to most other air-breathing fishes S. marmoratus has a high blood Hb content (Johansen, 1970; Johansen et al. 1978). Combining data from several studies (Lenfant&Johansen, 1972; Johanseneta/. 1978; Weber et al. 1979), we calculate that the South American population of S. marmoratus has a mean Hb of 13 ·8 ± 1 ·0 g% (x̄ ± 95 % confidence limits). This is significantly higher than the overall mean (combined hypoxia and control fish) of Panamanian S. marmoratus (11 ·1 ± l ·6g%) and may reflect genetic isolation and different environmental selection pressures in the two regions (see below). The high Hb of S. marmoratus probably plays a role in blood buffering (Johansenet al. 1978; Heisler, 1982) since this fish accumulates CO₂ (Table 3) in its tissues and often occurs in habitats that are acidic and subject to large thermal fluctuations (Kramer et al. 1978).

Respiratory physiology studies with S. marmoratus have been mostly done with fish from South America (Carter & Beadle, 1931; Johansen, 1966, 1970; Lenfant & Johansen, 1972; Johansen et al. 1978; Bicudo & Johansen, 1979; Weber et al. 1979; Heisler, 1982). Our study of Panamanian 5. marmoratus reveals several physiological differences between fish from these two areas. A second species of synbranchid eel, Ophistemon aenigmaticum, occurs sympatrically with S. marmoratus throughout Brazil and northern South America, but not in Panama (Rosen & Greenwood, 1976). Since S. marmoratus and Ophistemon are very similar in appearance the possibility cannot be discounted that some results reported for South American S. marmoratus may have actually been obtained from O. aenigmaticum. (Another species, 5. madierae occurs in eastern Bolivia and western Brazil, but physiological studies of fish from this region have probably not been conducted.) The ‘slow’ vs ‘fast’ response times for arterial and pH noted for specimens by Heisler (1982, Fig. 5) may be the result of studying different species.

Although additional studies are needed, differences between the aquatic respiratory patterns, air-breathing thresholds, and air-breath durations observed by Bicudo & Johansen (1979) and those reported in the present study seem to be caused by experimental procedures. Bicudo & Johansen (1979, Table 1) found the aquatic of their fish to vary following different exposure times in hypoxia. Also, these workers only acclimated fish in the respirometer for 4 –6 h prior to measurements. The mean of their control group (normoxia, 25 °C, Bicudo & Johansen, 1979, Table 1) was 39 ·8 ± 8 –0 ml kg^-1 h^-1 (x̄ ± 95 % confidence limits), which is significantly higher than our value for Panamanian fish at or above 120 mmHg (26 ·4 ± 5 ·4 ml kg^-1 h^-1, 25 °C, Fig. 2). Unsettled fish may have a higher ; this would affect O₂ storage and in turn alter metabolic responses to hypoxia, giving the appearance of regulation (Ultsch et al. 1981). Bicudo & Johansen did not specify the conditions under which they observed ventilation, and their graph (1979, Fig. 5) showing a steady rise in aquatic ventilation rate in increasing hypoxia presents no statistical information. The maximum rate they observed, about 32 ventilations min^-1, is slightly higher than our mean value of 28. J. B. Graham & T. A. Baird (in preparation) have observed that hyperoxia greatly Rduces aquatic ventilation in 5. marmoratus and in normoxia this fish uses a pattern intermittent aquatic branchial respiration in which, depending upon both body size and ⁠, periods of apnoea are regularly interspersed between intervals of aquatic? ventilation. We conclude that factors related to handling stress, the use of differently sized fish, and observation periods of insufficient duration to allow correction for regular periods of apnoea in high may have all contributed to the different findings for ventilation. Our preliminary studies demonstrated that shorter duration air breaths commonly occurred immediately after the induction of air-breathing in hypoxia. A rapid (1 h) onset of hypoxia also affected air-breathing and experiments conducted too soon after handling altered both air-breathing threshold and ⁠.

The higher Hb concentration of South American fish may reflect genetic differences that have evolved through isolation and different environmental selection pressures and could also be further investigated. Our literature review indicates that neither fish body-size differences nor methodological factors account for the recorded Hb differences. In addition, the proportionally shifted haematocrits in these two populations (Panama 40%, this study; South America 47%, above cited Hb references and Heisler, 1982) indicate the Hb differences between them are real. In contrast to Panama the broad expanse of the Amazon Basin may contribute to the regular (annual) occurrence of extreme seasonal conditions (e.g. the dry season) that are uniformly severe in all habitats occupied by S. marmoratus (Kramer et al. 1978 ; Heisler, 1982) in South America and this may have intensified selection for respiratory adaptations such as a higher Hb.

Partial support for this work was provided by the Smithsonian Institution, Smithsonian Tropical Research Institute, the Marine Biology Research Division of Scripps Institution, and PHSRRO7011. Major funding for this research was through NSF Grant DEB 79 –12235. Appreciation is expressed to D. Abel, G. de Alba, K. Dickson, I. Rubinoff and N. Smith for advice, comments on manuscript drafts, and other types of assistance. The gracious cooperation of the Department of Natural and Renewable Resources (RENARE) of the Republic of Panama is gratefully acknowledged.