Respiratory Physiology of the Lake Magadi Tilapia (Oreochromis alcalicus grahami), a Fish Adapted to a Hot, Alkaline, and Frequently Hypoxic Environment

The tilapia Oreochromis alcalicus grahami is a unique ureotelic teleost, the only fish that lives in the alkaline hotsprings of Lake Magadi, Kenya. Physical conditions and fish behavior were monitored in the Fish Springs Lagoon area, a site where the tilapia were particularly abundant. Water Po2 and temperature fluctuated more or less in parallel in a diurnal cycle from less than 20 Torr and less than 25° C at night to greater than 40° Torr and 38° C during the day, whereas pH remained constant at approximately 9.8. Field laboratory tests demonstrated that routine Ṁo2 (under normoxia) increased greatly from 27° C to 36° C (Q10 = 62) but then stabilized at a very high level (~34.5 μmol g−1 h−1) up to the lethal temperature (~ 42.5° C), a pattern that was adaptive to the natural diurnal regime. The Po2 threshold for survival during acute exposure (≤ 1 h) was approximately 16 Torr. Ṁo2 from water was well maintained down to a Po₂ of 60 Torr, below which it declined. Under such hypoxic conditions, the fish performed supplementary surface breathing when allowed access to air. Both the better oxygenated surface layer and air bubbles were inspired, resulting in significant uptake of O2. The Po2 threshold for surface breathing was 1.8-fold higher at 37.5° C than at 31° C. Surface breathing and voluntary entry of fish into air were observed in the field. The blood O2 dissociation curve at 30°-32° C was h3perbolic, with a high afinity (P50 = 6 Torr), low cooperativity (Hill coeficient = 1.18), and no Bohr effect over the extracellular pH range 8.2-8.6.


Introduction
The tilapia Oreochromis alcalicus grahami is the only fish species known to exist in the highly alkaline pools fed by geothermal springs at Lake Magadi in the Rift Valley of Kenya (Coe 1966). This fish thrives in a hostile environment (water of pH 9.6-10.4, osmolality = 525 mOsmol kg-', Cco2 = 180 mmol L-1, temperature 05 420C), with intense predation by birds. Coe (1966) and Trewevas (1983) have described the physical properties of the lake and much of the basic biology of 0. alcalicus grahami, while Reite et al. (1974) have documented its resistance to high pH and high temperature. Earlier studies have investigated various aspects of its ionoregulatory physiology (Leatherland et al. 1974 Rather less is known about the respiratory physiology of 0. alcalicus grahami. Blood pH's, both extracellular and intracellular, are high relative to the predictions of standard pH versus temperature relations in fish (Cameron 1984), reflecting low Pco2 and high HC03 levels Johnston et al. 1983;Wood et al. 1989Wood et al. , 1994. The only study on 02 transport in this species used a dialyzed hemolysate of whole blood rather than intact red cells ). At 350C, 02 affinity of this stripped hemoglobin was high (Ps0 = 4 Torr) and there was no Bohr effect over what is the probable physiological pH range (7.5-8.5). Despite the high environmental temperature, the hemoglobin does not exhibit unusual thermostability ). Maina (1990) reported that respiratory surface area of the gills was unusually high and diffusion distance low, presumably adaptations to hypoxia and/or high metabolic rates.
The present study investigates several aspects of the respiratory physiology of this unique species. Our first objective was to characterize the temperature and 02 conditions faced by 0. alcalicus grahami in its natural environment and to relate this to the fish's behavior. The Fish Springs Lagoon area (see Coe 1966), employed as the collection site in this and virtually all previous studies on the Magadi tilapia, was chosen as the study site. We suspected that large diurnal fluctuations in dissolved 02 might occur because the lagoon supports an abundant growth of cyanobacteria (the principal food source for the fish). Cyanobacteria perform photosynthesis and respiration during the day but only the latter at night. The second objective was to determine the effect of environmental temperature and Po2 on the oxygen consumption rate (Mo2) of these fish and to assess their resistance to hypoxia. A third, and related, objective was to investigate the possible respiratory significance of apparent air and/or surface breathing that we had observed in fish in the lagoon during our first several days at Lake Magadi and the effects of water PO2 and temperature on this behavior. The final objective was to determine the blood 02 transport characteristics (P50, pH sensitivity) of intact, freshly collected whole blood of 0. alcalicus grahami for comparison with the earlier observations of Lykkeboe et al. (1975) on stripped hemoglobin. All experiments were performed in a makeshift outdoor laboratory on fish freshly collected from the wild.

Experimental Animals
Small adult Oreochromis alcalicus grahami (1-4 g) used in most experiments were collected with a seine in January and February, 1992, from the Fish Springs Lagoon site on Lake Magadi, Kenya, described by Coe (1966). Much larger adult fish (10-40 g) used as blood donors for the 02 dissociation curve determinations were seined from a cement holding tank supplied with a continuous flow of lagoon water. An incidental population of O. alcalicus grahami lives in the tank, the initial stock having been pumped in with the water. Because this population is protected from intense avian predation by a chain-link fence and by the water depth in the tank, average fish size is much larger than in the lagoon population. Fish were immediately transported to an outdoor laboratory set up on the porch of the chemistry building of the Magadi Soda Company and transferred to fresh aerated lagoon water in 20-L buckets kept in the shade. Temperature ranged from 300 to 36oC diurnally, similar to that at the fish collection site (see Results). These fish are extremely territorial and aggressive; we found higher survival rates when fish density was less than 10 fish per 20 L and when the bucket was covered. Fish were deprived of food after capture and used within 24 h for experiments; survivors were returned to the wild.

Diurnal Dissolved 02 Concentrations
Figure 1 is a map of the Fish Springs Lagoon field site, with individual data collection points depicted for Po2, pH, and temperature measurements. The site has been modified by a retaining wall and an overflow dam of many years' standing (Coe 1966); the barrier serves to collect water for the carbonate trona processing operations of Magadi Soda Company. Some of this water is pumped elsewhere, and overflow from the pump house collects in a large pond. Temperature, pH, and Po2 of the Fish Springs Lagoon sites (typical depth = 1 m) were determined by a mercury thermometer and a YSI Model 54 pH and 02 meter; the standard polarographic membrane 02 probe was calibrated with air, with appropriate corrections for ambient temperature and the 700-m elevation of the site. The dissolved 02 (DO), pH, and temperature readings were collected on most days in January, and systematic measurements over discrete intervals and depths were made January 22 and 29-30. Measurements were made at 14 sites at the surface (5-10 cm below surface) and bottom (2-3 cm from bottom). DO values were converted to Po2 with appropriate salinity-corrected solubility coefficients from Boutilier et al. (1984).
Effect of Temperature on M0o, Closed-system respirometry in 530-mL Tusker chambers (amber Tusker beer bottles fitted with aeration tubes and ports for drawing 1-mL syringe samples anaerobically) kept in the shade was used to determine the effect of temperature on routine Mo2 of 0. alcalicusgrahami. We use the term "routine" as defined by Fry (1957) to refer to Mo2 in unfed fish with spontaneous rather than directed movements. Small adult fish (1-4 g) were loaded into Tusker chambers filled with continuously aerated lagoon water at approximately 270C and then allowed to settle for 30 min. Aeration was then stopped, and the chambers were sealed, with care taken not to trap any air bubbles, thereby eliminating any attempts of the fish to obtain 02 from the air or the air-water interface. Subsequent runs were done on the same fish with fresh lagoon water at higher temperatures (up to 420C, water preheated by exposure to direct sunlight) as ambient temperature increased over the day in the outdoor laboratory. The decrease in water Po2 over 10-60-min intervals was monitored with a Radiometer E5046 02 electrode attached to a pHM 71 meter thermostated to the experimental temperature. The Po2 calibration at time zero was based on the current experimental temperature, with a correction factor of 1.03 for measuring the Po2 Of water when calibrating on air. A chamber containing water but no fish was included in each series to serve as a blank.

Effect of Temperature on Threshold Po for Surface Breathing
Small adult fish (1-4 g) were placed into aerated Tusker chambers at either 310C (N = 12) or 37.50C (N = 11) and allowed to settle for 1 h. Aeration was then stopped, and an air bubble introduced into the neck of the chamber. The fish were then observed continuously as they depleted the 02 from the water. Pco2 did not increase, because of the very high absorptive capacity of the water for CO2 at pH 10. As soon as the fish was observed to first breathe from the air or air-water interface, a water sample was drawn for measurement (as above) of the threshold Po2 for this behavior.

Effect of Water Po2 on io,2
Flow-through respirometry was used to measure changes in Mo2 in response to changes in Po2. Respirometers were darkened 50-mL centrifuge tubes cut off at one end and sealed with rubber stoppers; three-way valves at both ends facilitated anaerobic sampling of inflow and outflow water into 1-mL syringes. A 20-L bucket filled with aerated lagoon water at 310-320C served as a head tank. Flow rates were adjusted by volumetric measurement to 9-11 mL min-'. Small adult fish (2-4 g; N = 12) were allowed to settle for about 1.5 h as the respirometers were flushed with near-saturated water (Po2 > 130 Torr, where 1 Torr = 133.322 Pa) before control Mo2 measurements were made in duplicate. The Po2 was then lowered to a new constant level by gassing the head tank with N2; the fish was allowed to adjust to the new Po2 for about 30 min, and then Mo2 was measured again in duplicate. Each fish was exposed to seven or eight levels of Po2 between 135 and 25 Torr. In each case, the exposure Po2 was taken as the mean of the measured inflow and outflow PO2 values.

Critical Po2 for Survival
In the course of a separate study comparing urea excretion and Mo2 (Wood et al. 1994), Mlo2 measurements were made by closed-system respirometry in 530-mL Tusker chambers on several hundred small adult fish (1-5 g) at 300-33'C. Seven of these fish died because the water Po2 was accidentally permitted to fall too low. Water Po2 measurements were made within 5 min of death for these fish and recorded as the critical Po2 for survival. In addition, records were kept of unusually low Po2's at which fish survived and of Po2's associated with death of some individuals in the preceding flow-through respirometry experiments.

Behavioral Observations of Respiration
A 20-L glass aquarium was built for behavioral observations of 0. alcalicus grahami in aerated and nonaerated Fish Springs water, which was changed daily. The temperature was allowed to fluctuate diurnally with ambient air temperature. The sides were masked to reduce the effects of outside stimuli on behavior, and the top was left open for natural light cycles. One to three fish were held in the aquarium at any one time. We noted periods of air or surface breathing and measured the incidental water temperature and Po2. We also noted any other unusual behaviors by the fish.

ofrom Water and Air during Hypoxic Conditions
In the laboratory, air breathing was observed only at times of low water Po2. Therefore, the objective of this series of experiments was to partition Mo2 from air versus Mo2 from water during aquatic hypoxia and to compare the total Mo2 with that seen under normoxic control conditions. Flow-through respirometry was used for Mlo2 from water (to maintain near-constant water Po2) and closed-system respirometry for Mo102 from air. Smaller, 350-mL Tusker chambers were set up on a flow-through basis (9-11 mL min-), with water entering at the bottom and leaving by siphon close to the top. In addition to sampling ports for water inflow and outflow, an additional port allowed sampling of a 10-mL air bubble placed in the neck of the chamber. The placement of the bubble was designed to minimize contact between water and air yet still allow the fish to access it. Small adult fish (2-5 g; N = 7) were allowed to settle for at least 1 h as the respirometers were flushed with water of Po2 greater than 120 Torr at 310-320C from the head tank (as described above). This produced a mean water Po2 of about 110 Torr in the respirometers. Mo2 was measured in duplicate, and then the inflowing Po2 was lowered to 50 Torr, which produced a mean water Po2 of about 40 Torr in the respirometers. A 10-mL air bubble was introduced into the neck of the chamber. After a 1-h adjustment period, Mo2 from air was measured by monitoring the depletion of 02 from the air bubble over the next hour, and Mo2 from water was measured by flow-through respirometry twice during the period. Appropriate blank chambers were run to correct for the very slight transfer of 02 from the air phase to the water phase. Measurements of both air and water Po2's were made with the Radiometer 02 electrode, as described above.

Blood 02 Dissociation Curves
A blood pool of 5.5 mL was obtained from 16 larger adult fish (10-40 g). Fish were first anesthetized in the anesthetic Transmore as described by Wood et al. (1994), and blood was drawn from the hemal arch into syringes prerinsed with 1,000 i.u. mL-1 sodium heparin (Richter) in 180 mmol L-1 NaCl. The blood was diluted about 15% by this heparinization. The blood was then divided into two 2.75-mL pools and equilibrated with either air or N2 in a homemade tonometry setup consisting of two 10-mL volumetric flasks suspended in a water bath at 300-320C and mounted on a shaker table. Tonometry gas was bubbled first through 180 mmol L-1 NaCl at the experimental temperature so as to match vapor pressure with that of the blood. A preliminary experiment demonstrated that there was no loss or gain of volume over 4 h of equilibration. Blood 02 dissociation curves were constructed through the mixing technique of Scheid and Meyer (1978). Various aliquots of oxygenated and deoxygenated blood were mixed in a modified gastight 250-jiL Hamilton syringe by using a small bead of mercury; 10 mixtures in total were prepared in this way. The Po2 of each mixture was measured with a Radiometer E5046 02 electrode thermostated to the experimental temperature, and the percentage saturation was calculated. The procedure was then repeated after adding 6 L.tmol HCI (in 3 pL of 180 mmol L-' NaC1) per milliliter of whole blood to each tonometer flask, so as to assess the extent of the Bohr effect.
In both control and acidified runs, extracellular pH and red blood cell intracellular pH (freeze-thaw method of Zeidler and Kim [1977]) were determined on the 0%, 50%, and 100% saturation mixtures, and hematocrit was determined gravimetrically after spinning at 13,000 g for 2 min. The plasma from the hematocrit determinations was decanted and used to measure plasma total CO2 concentrations by the method of Cameron (1971), and plasma protein concentration by refractometry (Alexander and Ingram 1980). Radiometer pH (E5021) and CO2 electrodes (E5036) were used for these determinations and displayed on the same Radiometer pHM 71 meter as the 02 electrode.

Diurnal Temperature and Po2
The Po2 levels varied in approximate synchrony with temperature at all 14 sites in Fish Springs Lagoon (Fig. 1). Very probably, this was a result of net photosynthetic production of 02 during the day and consumption at night by cyanobacteria. The pH of this extremely well buffered water did not significantly vary spatially or temporally; mean pH for all 14 sites was 9.84 (range = 9.76-9.97). Oreochromis alcalicus grahami were present at all 14 sites, sometimes in such quantity that the water appeared dark in areas where the fish congregated in large masses. In Table 1 we present the absolute range of surface data for temperature and Po2 at each site, although we did obtain even lower values for Po2 in a few bottom samples. In general, there was little variation with depth, as the water was well mixed by the high inflow from the springs and by frequent high winds combined with the shallow depth of the lagoon (<1 m). Temperatures in 12 of 14 sites fluctuated diurnally, ranging from a low of 220C recorded just before dawn to a high of 380C recorded in midafternoon. The temperature in the two other sites (small warm springs feeding the main lagoon) remained fairly constant (350-370C), but Po2 did fluctuate diurnally at these sites from 0.30 mg L-1 (7.2 Torr) during the night to 3.16 mg L-' (77.1 Torr) at midmorning. Figure 2 shows surface temperature and Po2 data for one site (no. 11) where fish were particularly abundant and that is representative of all 12 sites in which temperature fluctuated over a 24-h period. Po2 fluctuated diurnally with temperature at this site; the lowest Po2 was 18.9 Torr at 250C at midnight, and the highest Po2 was 402.4 Torr at 35.50 C in midafternoon.

Behavioral Observations
Many different fish behaviors observed in both the field and the lab appeared to be related to respiration. Fish congregated in dense numbers in the more oxygenated water just below the dam overflow (sites 2 and 3; Fig. 1) during both day and night, even though piscivorous birds used this site for feeding during daylight hours. Other areas in and adjacent to the lagoon (e.g., pump house pond sites 11-14) were supersaturated (hyperoxic) during the day and almost anoxic at night (Table 1, Fig. 2) but still contained abundant fish. In general, fish were most numerous in the shallow, warmest areas during the hottest part of the day and retreated to deeper areas at night.
We collected fish by seining in the early morning (0700-0800) when temperatures were lower than those recorded at any diurnal peak (Fig.  2). While surface breathing was seen only occasionally before seining, many fish that had escaped the seine were observed breathing at the surface for up to 30 min after seining. At times, the mouths protruded above the water surface and the fish seemed to be gulping. This did not appear to be a feeding behavior. Oreochromis alcalicus grahami were normally observed grazing on periphytic cyanobacteria; we are not aware of filter feeding in these fish. It is interesting that fish in the pump house pond (sites 11-14) were observed during the day repeatedly jumping  entirely out of the water to bite off pieces of cyanobacteria growing in mats down the side of the damp stone retaining wall that separated the higher water level in the Fish Springs Lagoon from that in the pump house pond (see Fig. 1). A graze line on the cyanobacterial mat could be seen above the water level.
In the lab, we observed fish gulping at the surface in the 20-L holding buckets when densities were too high or aeration inadequate. This behavior became more common in the late afternoon as water warmed up over the day in the outdoor laboratory. The glass aquarium allowed closer observation of the surface breathing events. While at times the fish appeared to be skimming the better oxygenated surface water, we clearly observed many events where the entire mouth was out of the water and in which air bubbles were later released from the mouth and operculi underwater. We conclude that these fish are not just surface skimming but are actually breathing air as well (though perhaps mixed with water). When males in breeding coloration were placed in the aquarium, they were extremely territorial and spent much of their time in high-energy activities such as chasing, charging, and locking mouths. Under normoxic conditions, these behaviors were not followed by surface breathing. However, territorial behavior ceased when hypoxic conditions forced fish to surface breathe.

Temperature Effects on AMo2
MO2 was extremely sensitive to temperature over the naturally occurring range. In a single batch of fish observed during the day as ambient temperature increased, routine Mo2 under normoxic conditions increased progressively from about 6.6 jtmol g-' h-1 at 270C to about 34.5 .mol g-' h-' at 360C (Fig. 3). Above 360C, Mo2 leveled off abruptly; temperature had no significant effect between 360 and 41'C. We were able to measure routine Mo2 at a mean temperature of 41.20 + 0.30C, but at 42.50C, fish begin dying. The effective Qio for the temperature range of 270-360C was 6.2, a very high value.

Temperature Effects on Threshold Po2 for Surface Breathing
Temperature also had a pronounced effect on the Po02 threshold for surface breathing, which increased significantly from 29.6 + 1.9 (N = 12) Torr at 310C to 54.7 + 9.5 (N= 11) Torr at 37.50C.

Critical Po2 for Survival
In the closed-system respirometry experiments (no access to air) at 300-330C, seven fish died when Po2 was allowed to drop below the required 02 level for survival. The mean Po2 measured at the time of death was 12.6 + 0.9 (range = 8.0-15.5) Torr. The three lowest Po2'S recorded that did not cause death were 16.5, 17.5, and 17.5 Torr, so the threshold was well defined at about 16 Torr.

Effect of Water Po2 on Mo2
Mo2, measured at 310-320C by flow-through respirometry, increased slightly as the Po2 of the water was initially decreased from fully saturated down to about 110 Torr, but metabolic rate did not change significantly between 140 and 60 Torr (Fig. 4). Mlo2 then rapidly declined as Po2 was decreased in stages below this level. At a Po2 of less than 50 Torr, some fish began to show signs of respiratory stress, and five of the 12 fish died at a mean Po2 of 31.6 + 4.6 (range = 20-47.5) Torr, while the other seven remained alive at a Po2 of about 25 Torr, the lowest level tested. These results indicate a somewhat higher critical Po2 for survival than in the closed-system respi- rometry experiment, but the protocol was different, involving exposure to progressive hypoxia over 5-7 h rather than in 1 h or less.

Po2 from Water and Air during Hypoxic Conditions
When mean water Po2 was reduced from about 110 Torr to about 40 Torr at 31o-320C in the flow-through respirometry system, five of the seven fish tested used the air bubble for surface breathing. Total Mo2 fell significantly by about a third relative to the normoxic values; of the remaining Mo2, 12.5% was taken from air and 87.5% from water (Fig. 5).

Blood 02 Dissociation Curves
The blood 02 dissociation curve of 0. alcalicus grahami at 300-320C was hyperbolic, with very low cooperativity (Hill coefficient = 1.18) and a high 02 affinity (P50 = 6 Torr at mean extracellular pH 8.58) (Fig. 6) Cco2 in either control (mean = 8.20 mmol L-1) or acidified runs (1.91 mmol L-1). In contrast, the pH measurements (Table 2)   icant change when temperature was gradually increased over this range (Fig. 3). Both studies indicate that routine Mo2 in the Magadi tilapia at the normal peak daytime temperature in the Fish Springs Lagoon area (about 380C; Table 1, Fig. 2) is extremely high and relatively temperature insensitive. To our knowledge, these routine metabolic rates are the highest reported in poikilothermic teleosts. This is perhaps not surprising in view of the high activity level of Magadi tilapia (we almost never observed stationary fish under normoxia, either in the wild or in the respirometers) and the high environmental temperature that would be lethal to most teleosts. Several other tilapia do tolerate this high temperature range but exhibit 25%-60% lower Mo2's at comparable temperatures above 350C (Job 1969a(Job , 1969bCaulton 1977Caulton , 1978Fernandes and Rantin 1989). A more striking difference from other tilapia is the transition from temperature insensitivity of Mo2 above 360C to marked sensitivity below this temperature (Fig. 3). The Q10 from 270C to 360C was 6.2, in comparison to 1.0-3.0 over the same range in other tilapia. This may be suitable to the natural diurnal cycle of dissolved 02 and temperature in the Fish Springs Lagoon area. Thus, at night, when 02 availability in water is lowest (Fig. 2) due to cyanobacterial respiration, metabolic 02 demand is only a fraction of that during the day because of the simultaneous decline of water temperature to the 230-280C range (Fig.  3). Birds at Lake Magadi do not feed at night, so the need for energy expenditure for predator avoidance at this time is likely reduced. In contrast, during the heat of the day, the fish can sustain a high metabolic rate for feeding, social, and predator escape behaviors, and this rate is more or less independent of exact environmental temperature right up to the lethal limit (--42.50C; Fig. 3). This upper lethal temperature is almost identical to that reported in other studies on 0. alcalicus grahami (Coe 1966; Reite et al. 1974;Johnston et al. 1994).
In their study on Blenniuspholis, Bridges et al. (1984) found a diurnal fluctuation in Po2 in intertidal rock pool sites similar to the diurnal fluctuation in our study (Fig. 2) but with two noticeable differences-no corresponding fluctuation in temperature and a pH fluctuation similar in magnitude to Po2, In comparable studies on pupfish from saline waters in the southwest United States, Stuenkel and Hillyard (1981) found that the metabolic rates of fish acclimated to freshwater and half-strength seawater increased with increased acclimation temperatures. However, the plasma osmolality of fish acclimated to seawater decreased significantly, and the fish were able to osmoregulate efficiently at higher temperatures (250-300C).
At the intermediate temperature where Mo2 was measured (310-320C), 0. alcalicus grahami proved to be a good 02 regulator, maintaining Mo2 until a Po2 of about 60 Torr (Fig. 4) We observed movement of the fish into the shallowest, warmest water where food supply and 02 availability (due to cyanobacterial growth and photosynthesis) were greatest during the day, and a retreat to deeper, cooler water and/or sites of aeration such as the dam outflow at night. This pattern was originally noted in the Magadi tilapia by Coe (1966) and has been reported in other tilapia as well (Caulton 1982). Thus a combination of metabolic and behavioral characteristics match this fish to the severe natural regime.

Short-Term Tolerance of Severe Hypoxia and the Importance of Surface Breathing
When the above-mentioned metabolic and behavioral adaptations prove insufficient, there appear to be two more that the animal can recruit. The first fallback mechanism, which is essentially metabolic, is an ability to resist extremely low P02 for short periods. The second, essentially behavioral, is an ability to breathe at the water surface.
The survival threshold at 300-33oC for short-term hypoxic exposure (i.e., < 1 h) was well-defined at about 16 Torr, and somewhat higher (--32 Torr) when hypoxia was imposed in progressive steps over 5-7 h. In the lagoon, we observed fish living in conditions as severe as approximately 7 Torr (0.3 mg L-1 at 360C), although it is likely that they did not spend long periods in such hypoxic areas. This tolerance appears to be characteristic of tilapia in general; comparable figures for short-term tolerance are approximately 10 Torr at 230C in Oreochromis shirana chilwae (Morgan 1972), approximately 5 Torr at 24oC in Oreochromis macrochir (Dusart 1963), and almost 0 Torr at 25 OC in Oreochromis niloticus (Verheyen et al. 1985). This characteristic may result from an ability to suppress metabolic rate and/or a capacity for anaerobic metabolism, both of which may prove fruitful areas for future research on the Magadi tilapia. (1982). In some species, the behavior involves only the extraction of 02 from the upper several millimeters of water, which are better oxygenated, while in others additional benefits are achieved by mixing water with air in the mouth, by holding air bubbles next to a well-vascularized buccal epithelium, and even by taking air into the swim bladder. Based on our observations, both in the field and in the aquarium, and on our anatomical studies of the swim bladder and buccal epithelium (Maina et al. 1996; J. N. Maina, unpublished data), we suspect that 0. acalicus grahami exploits all of these behaviors. Both the buccal epithelium and the physostomous swim bladder are very well vascularized; the latter has surface infoldings and can be inflated by the fish when held in air, and blood draining from the swim bladder joins with the branchial arches (Maina et al. 1996). Tilapia are not usually considered to be air or surface breathers, but we have found two anecdotal reports of apparent surface-breathing behavior in other tilapia species subjected to acute hypoxia (Dusart 1963; Morgan 1972). Our observations on fish chased by seining in the lagoon suggest that surface breathing may also be used to help pay off excess postexercise 02 COnsumption (formerly "02 debt").

Surface breathing (aquatic surface respiration; reviewed by Kramer and McLure [1982]; Gee and Gee [1995]) is a common adaptation to aquatic hypoxia in tropical fish. The benefits of this strategy in improving arterial oxygenation under hypoxic conditions have been shown by Burggren
Experiments at 310-320C demonstrated that under moderate hypoxia, surface-breathing fish obtained a small but significant fraction (12.5%) of their Mo2 from the air phase, although this was not sufficient to sustain routine Mo2 at normoxic levels (Fig. 5). We predicted that this fraction would become more important at a higher temperature, but our intention to repeat the experiment at 37.50C was thwarted by logistic problems. Instead, we conducted the simple experiment of measuring the Po2 threshold at 31'C versus 37.50C. The threshold was significantly higher (1.8-fold) at the higher temperature, in accord with these ideas.

Blood 02 Transport Characteristics
The 02 dissociation characteristics of 0. alcalicus grahami (determined at 300-320C) exhibit an extremely high 02 affinity (P50 = 6 Torr), low cooperativity (Hill coefficient = 1.18), and an insensitivity to pH in the tested range (8.2-8.6; Fig. 6). It must be noted that these curves were determined only at a Pco2 close to 0 Torr (i.e., with air/N2 equilibration only), again a logistic consequence of research in an isolated location. However, inasmuch as this fish lives in a "Pco2 vacuum" caused by an external pH close to 10 (Johansen et al. 1975; Wood et al. 1989), it is likely that arterial Pco2 is extremely low in vivo anyway. On the basis of reviewing studies of other teleosts at high environmental pH, we have argued elsewhere (Wood et al. 1994) that true resting arterial extracellular pH in vivo may be as high as 8.7, and true arterial Pco2 is well below 1 Torr. Therefore, the conditions of the equilibrations appear realistic, and the pH range tested probably represents the normal variation between rest (8.6) and exercise (8.2).
The present results agree with those of the only previous study on the blood of this species, by Lykkeboe et al. (1975). These workers reported a P50 of 4 Torr, a Hill coefficient of 1.2, and the absence of a Bohr effect (and Root effect) over the pH range 7.5-8.5 at 350C. However, what is remarkable is that Lykkeboe et al. (1975) determined these properties using a dialysed hemolysate (stripped hemoglobin), whereas we found almost identical values using intact whole blood. Assuming that the two studies can be compared, this suggests that the intracellular milieu (red blood cell intracellular pH, ions, organic phosphates) has negligible effect on 02 transport in the Magadi tilapia, a conclusion that is supported by the intracellular and extracellular pH insensitivity (Fig. 6, Table  2). In most other species, the intracellular milieu alters the binding properties of the hemoglobin such that whole blood behaves differently from stripped hemoglobin ( The absence of a Bohr effect in the Magadi tilapia is not due to a failure of extracellular pH changes to alter red blood cell intracellular pH. The measurements of Table 2 indicate a fairly standard relationship between extracellular and intracellular pH, explicable by normal Donnan distribution, such that the extracellular [H+]/intracellular [H+] rises as extracellular pH falls; that is, changes in red blood cell intracellular pH reflect, but are smaller than, changes in extracellular pH. Surprisingly, these measurements also suggest that there is a definite Haldane effect manifested as a 0.2-0.4-unit decrease in red blood cell intracellular pH with oxygenation, despite the lack of Bohr effect. The Haldane effect is generally considered the reciprocal of the Bohr effect: oxygenation releases Bohr protons from the hemoglobin molecule so that oxygenated blood exhibits a lower pH and Cco2 (Jensen 1989). In the present study, the effect was seen clearly only in red blood cell intracellular pH and not in extracellular pH or plasma CO2 measurements. Unfortunately, erythrocytic Cco2 was not measured in these experiments. In the absence of more data, speculation on mechanisms is unwarranted, but the possible uncoupling of the Haldane effect from the Bohr effect in this species deserves further investigation.
In conclusion, our blood 02 results reinforce the arguments of Lykkeboe et al. (1975) that the high 02 affinity and pH insensitivity of blood 02 binding are present for an extremely active fish with a high metabolic rate living in a frequently hypoxic environment. 02 uptake will be maximized regardless of exercise-induced pH changes or environmental Po2 fluctuations over a wide range. In combination with the unusually high respiratory surface area and low diffusion distance of the gills (Maina 1990), the unusual temperature sensitivity of metabolic 02 demand, numerous behavioral adaptations, and the ability to withstand acute hypoxia and perform supplementary surface and/or air breathing, these blood characteristics allow 0. alcalicus grahami to thrive in one of the most hostile aquatic environments on earth. DHL Nairobi, who came through for us, unlike the competition. We thank Drs. C. Franklin and I. Johnston for access to unpublished data. And, we thank the PZ referees for their helpful comments.