The effects of experimental alterations of aerial O2 partial pressure (PO2,air) on bimodal gas exchange and air-breathing behaviour were investigated in the aquatic air-breathing fish Trichogaster leeri in normoxic water. Fish responded to increasing PO2,air by decreasing air-breathing frequency,increasing aerial O2 consumption rate(O2), increasing mean O2 uptake per breath(O2/breath) and decreasing aquatic O2 to maintain a constant total O2. The rate of oxygen uptake from the air-breathing organ (ABO) during apnoea(O2,ap) was derived on a breath-by-breath basis from VO2/breath and apnoea duration. O2,ap and estimates of ABO volume were used to calculate the PO2 in the ABO at the end of apnoea. This increased with increasing PO2,air, suggesting that ABO-PO2 is not regulated at a constant level by internal chemoreceptors. Furthermore, mean O2,ap increased with increasing PO2,air, indicating that the observed increase in O2/breath with increasing PO2,air was facilitated not only by an increase in apnoea duration but also by an increase in the air–blood PO2 gradient.

Ventilation in bimodal fish (i.e. those that respire simultaneously in air and water) is characterised by variable branchial ventilation patterns and intermittent and arrhythmic air breathing(Shelton et al., 1986). It has been suggested therefore that air breathing in these animals is an on-demand phenomenon that is stimulated by afferent feedback from one or more peripheral receptors (Shelton et al.,1986). Studies on the air-breathing lungfish Protopterusand gar Lepisosteus show that O2 chemoreceptors are located diffusely throughout the gills, as is the case for most fish, and evidence points to the existence of both internally and externally oriented O2 chemoreceptors that respond to changes in blood O2partial pressure (PO2) and aquatic PO2, respectively(Johansen and Lenfant, 1968; Lahiri et al., 1970; Smatresk, 1986).

Since there is evidence to suggest that chemoreceptors are located at the site of aquatic respiration in air-breathing fish, it is reasonable to contemplate the existence of chemoreceptors at the site of aerial respiration,i.e. within the air-breathing organ (ABO). A common method used to identify sites of chemoreceptors in conscious air-breathing fish is to observe behavioural responses to independent changes in aquatic and aerial gas contents (e.g. Burggren, 1979; Graham et al., 1995; Hedrick and Jones, 1993; Hughes and Singh, 1970; Johansen and Lenfant, 1968; Sanchez et al., 2001). Studies involving manipulations of aquatic O2 content by far exceed studies involving manipulations of aerial O2 content. However, experimental manipulations of aerial O2 content have the potential to offer insight into the presence and role of ABO-O2 chemoreceptors. Air-breathing fish generally respond to aerial hypoxia by reducing apnoea(i.e. breath-hold) duration, whereas hyperoxia often lengthens it(Shelton et al., 1986). However, conclusions arising from experiments involving aerial O2manipulations are limited in that the activity of chemoreceptors within the ABO cannot be discriminated from those located remotely, in the efferent vasculature (Graham, 1997). Simulated breathing by injection of either hypoxic or hyperoxic gas into the ABO produces the same results as changes in aerial gas composition (e.g. Johansen and Lenfant, 1968). However, gas injection also increases the volume of the ABO, and such experiments may therefore be confounded by activation of ABO mechanoreceptors that transmit information about the rate or extent of organ wall deformation(Milsom, 1990; Pack et al., 1990).

Evidence for ABO-O2 chemoreceptors was found for the swamp eel, Monopterus albus, by Graham et al.(Graham et al., 1995). M. albus was observed to expel severely hypoxic or anoxic breaths within a few seconds of inspiration. The rapidity of the gas-voiding reflex suggested the presence of an ABO chemoreceptor, because it occurred about two to four times faster than would be expected if O2 levels in the ABO had to be conveyed by blood flow to remote vascular receptors located somewhere in the systemic circulation. By contrast, the response of the blue gourami, Trichogaster trichopterus, to changes in aerial PO2 was not immediate, suggesting that chemoreceptors were more centrally located(Burggren, 1979).

The present study investigates the effect of aerial O2 partial pressure (PO2,air) on bimodal gas exchange and air-breathing behaviour in an air-breathing fish, the pearl gourami, Trichogaster leeri Bleeker 1852 (sub-order Anabantoidei, family Belontiidae). T. leeri is a freshwater pelagic fish that has a pair of suprabranchial chambers serving as its ABO(Peters, 1978). Changes in air-breathing frequency (fab), aerial and aquatic O2 consumption rate(O2) and mean O2 uptake per breath (VO2/breath)are analysed. Using a novel approach, apnoeic O2(O2,ap=rate of O2 uptake from the air within the ABO while the fish is submerged)is derived by a breath-by-breath assessment of O2 uptake and apnoea duration. O2,apis used to address two questions: (1) is the response of T. leeri to changes in PO2,air regulated by ABO-O2 chemoreceptors and (2) is a change in mean VO2/breath with changing PO2,air facilitated by a change in the air–blood PO2 gradient as well as a change in apnoea duration?

To address the first question, O2,ap and estimated measurements of ABO volume (VABO) are used to calculate the PO2 in the ABO at the end of apnoea. It is hypothesised that if the PO2 in the ABO is regulated by ABO-O2 chemoreceptors, then the PO2 in the ABO at the end of apnoea should not be significantly different across treatments of varying PO2,air. This would suggest the existence of an ABO-PO2 threshold, which, once reached,triggers fish to renew the gas in their ABO.

The second question arises because O2 uptake during apnoea may change simply as a result of changes in apnoea duration, i.e. VO2/breath may increase during aerial hyperoxia because T. leeri holds its breath for longer. However, a change in the air–blood PO2 gradient may also facilitate O2 uptake during apnoea. Thus, it is hypothesised that if a change in apnoea duration is solely responsible for a change in mean VO2/breath then PO2,air will not have an effect on mean O2,ap. This is because mean O2,ap accounts for variations in apnoea duration between treatments, and thus if mean O2,ap changes with changing PO2,air then a change in the air–blood PO2 gradient must also be responsible. To our knowledge, this is the first study to examine how changes in the air–blood PO2 gradient associated with aerial hypoxia and hyperoxia influence O2 consumption in an air-breathing fish.

Experimental animals

Experiments were performed on seven adult pearl gourami Trichogaster leeri Bleeker 1852 (wet mass 5.15±0.97 g; length 59.1±1.2 mm; means ± s.d.) that were obtained from a local aquarium supplier. Fish were maintained in aquaria filled with dechlorinated Adelaide tap water at 25°C. All fish were maintained under these conditions for at least one year prior to experimentation.

Aerial respirometry

Prior to experimentation, food was withheld for 24 h and fish were individually placed in 1 litre bottles (Schott Duran, Germany) that were covered in black plastic to minimise stress. Bottles containing fish were held in a constant-temperature water bath maintained at 25±1°C by a heater (Thermomix 1419; B. Braun, Melsungen, Germany). Fish were allowed 12 h to acclimate to the chambers and to recover from handling prior to each experiment. An air stone gently aerated the water during the acclimation period.

Respiratory studies were conducted with an open-flow respirometer system. O2 and N2 were supplied from compressed cylinders (BOC Gases, Adelaide, Australia) and delivered to the respirometry chamber at a desired O2 content (FiO2)and flow rate (i; 100, 150 or 200 ml min-1). The flow rate for each trial was selected to minimise the washout time while providing reliably detectable changes in the excurrent O2 content(FeO2). O2 content and flow rate were controlled by mass flow controllers (model GFC171; Aalborg Instruments and Controls, New York, NY, USA; 0–1 l min-1,rated accuracy ±15 ml min-1) using a PC running digital–analogue control software and hardware (PowerDAQ™ PD2-AO and ProfessorDAQ™; United Electronic Industries, Canton, MA, USA). Flow controllers were calibrated to each gas with a 3.5 litre calibrator (model 1057A Vol-U-Meter Calibrator; Brooks Instruments, Hatfield, PA, USA; accuracy of calibrated controllers was better than 5% of reading, usually 1–2% of reading). To ensure uniform mixing of O2 and N2, these gases were passed through a 1 litre, rigid mixing chamber before entering the respirometer. Nevertheless, the O2 level was not as constant as expected in normal open-flow respirometry with atmospheric air, so the variability had to be accounted for in the analysis (see below).

Upon exiting the respirometry chamber, gases passed through a U-tube containing Drierite™ (Hammond Drierite Co. Ltd, Xenia, OH, USA) to remove water vapour and into a differential O2 analyser (FC-2 Oxzilla I; Sable Systems International, Las Vegas, NV, USA). The analyser was calibrated using dried (Drierite™), CO2-free(Ascarite™; Arthur H. Thomas Company, Swedesboro, NJ, USA) atmospheric air (0.2095 O2). The analyser measured FeO2 approximately every 1.25 s and used a running five-sample average for each data point. The data output from the O2 analyser was received simultaneously in an analogue and digital format. The analogue output was recorded at 2 s intervals with a digital multimeter (model TX3 True RMS and WaveStar™ Version 2.2;Tektronix, Beaverton, OR, USA) interfaced with a PC via the RS232 port. Depending on the analogue output scaling chosen, these voltage values were converted to FeO2 using relationships provided by Sable Systems International. The Oxzilla digital output was received by a DOS terminal program (SERIN; Sable Systems International) and recorded at intervals of approximately 1.1 s on a second PC. For analyses, WaveStar data were preferred due to SERIN's unstable timing interval, but SERIN data were used in two cases where high-frequency noise in the analogue data signal prevented reliable detection of breaths.

Experimental protocol

Aerial PO2(PO2,air) in the respirometry chamber was manipulated to nominal levels of 5, 10, 21 (control), 40 or 60 kPa (actual PO2,air range: 5.4–5.5, 9.6–10.1,19.9–20.7, 38.4–39.0 and 57.5–57.9 kPa, respectively). Treatments were executed in random order and an arbitrary exposure time of 1 h was set so as to reduce the effect of declining aquatic PO2. However, due to the instability of FeO2, some fish were exposed for longer than 1 h (maximum exposure time was 3 h), in order to accumulate 1 h worth of interpretable data. During trials, aquatic PO2 dropped 2.3, 1.8, 1.5, 1.2 and 0.6 kPa h-1 in the 5, 10, 21, 40 and 60 kPa treatments, respectively. To determine whether declining aquatic PO2 had an effect on aerial respiration, fab, aerial O2 and O2/breath in the first and final 15 min of each trial were compared. In each case, there was no significant difference between the first and final 15 min (P=0.22,0.94 and 0.56, respectively). Following exposure to each treatment, fish were given at least 24 h rest before the next treatment. During this time, either the water in the chamber was aerated with an air stone or the fish were returned to their aquarium.

Data processing

Calculation of aerial O2 consumption rate (aerial O2; ml h-1) was by integration of FeO2 inverted spikes. Each spike represented one exhalation, confirmed by visual observation. The duration of the spike depended on the washout characteristics of the respirometry system. Because these occurred on an unstable baseline(Fig. 1), a spreadsheet method was devised to isolate the spikes from the baseline. To select data representing periods of apnoea, a coefficient of variation (CV=s.d./mean) was calculated for 15 consecutive measurements, and a CV threshold value (e.g. 3.3×10–5) was manually set at a level that separated high-frequency, low-amplitude FeO2baseline noise from air-breathing events that were low frequency, high amplitude (Fig. 1B). Values below this threshold were considered to be baseline values, and, for every value above the threshold, a new baseline was linearly interpolated between previous and subsequent sub-threshold values. Where breaths were too frequent for the program to isolate values representing baseline FeO2, a polynomial regression was fitted to the baseline trace and entered in place of the CV threshold criteria. The final baseline was then produced by completing two rounds of 9-point nearest-neighbour averaging to remove high-frequency noise(Keller et al., 1994). The resulting baseline was considered to represent incurrent oxygen content(FiO2) and accounts for any exchange between the air and water in the system(Fig. 1A). Aerial O2 was then calculated for each data point (Fig. 1C) from the air flow rate through the chamber(i; ml min-1)and the respiratory quotient (RQ) according to Depocas and Hart(Depocas and Hart, 1957):
\[\ {\dot{V}}_{\mathrm{O}_{2}}=[{\dot{V}}\mathrm{I}(F\mathrm{I}_{\mathrm{O}_{2}}-F\mathrm{E}_{\mathrm{O}_{2}})]{/}[1-F\mathrm{E}_{\mathrm{O}_{2}}(1-\mathrm{RQ})].\]
We assumed a respiratory quotient of 0.25, as measured in the closely related blue gourami, Trichogaster trichopterus, under normoxic conditions(Burggren, 1979).

Each breath was integrated individually to arrive at a breath-by-breath estimate of O2 uptake from the ABO during the apnoeic period(VO2/breath; ml). This was summed across the trial and divided by trial duration to calculate aerial O2. Additionally, this approach measured fab (breaths h-1) by tallying the number of breaths during a trial and dividing by trial duration.

Aquatic respirometry

To evaluate the partitioning of aerial and aquatic respiration, aquatic O2 consumption (aquatic O2; ml h-1) was measured. For technical reasons, this was measured separately from aerial respirometry studies. Prior to experimentation, fish were treated in the same manner as they were for aerial respirometry. The aerial gas mix was produced as previously described but was vented to the atmosphere rather than fed through the O2 analyser. A fibre-optic O2 sensor (Implantable Oxygen Microoptode; Presens, Regensburg,Germany) encased within a Pasteur pipette was mounted through the respirometer lid to measure aquatic O2 content (% air-saturation). The O2 sensor was connected to a single-channel,temperature-compensated O2 meter and software (Microx TX3, OxyView TX3-V5.20; Presens) that recorded at 1 min intervals. A trial without a fish(control) was conducted for each PO2,airtreatment to account for aquatic O2 depletion not related to fish respiration. A linear regression was fitted to the data of O2content (% air-saturation) on time, and aquatic O2 was calculated using the equation:
\[\ {\dot{V}}_{\mathrm{O}_{2}}=-1{\times}[(m_{\mathrm{f}}-m_{\mathrm{c}}){/}100]{\times}V{\times}{\beta}_{\mathrm{O}2},\]
where mf is the slope derived from the trial with a fish(% air-saturation h-1), mc is the slope derived from the control (% air-saturation h-1), V is the water volume in the respirometry chamber (=1.088 litres) and βO2 is the O2 capacitance of air-saturated freshwater at 25°C (=5.77 ml l-1) (Riley and Chester,1971). Note that the difference between mf and mc is divided by 100 to convert the percentage of O2 in the water to a fraction.

Apnoeic V̇O2calculation

To calculate oxygen uptake from the ABO during apnoea(O2,ap;ml h-1), VO2/breath (ml) was plotted against the apnoea duration (h) for each breath of each fish under each treatment (e.g. Fig. 2). A linear regression was fitted to the data, with the derived slope of the regression representing O2,ap (ml h-1) (e.g. Fig. 2).

End-apnoea ABO-PO2 calculation

It would be possible to estimate end-apnoea PO2 in the ABO from the data if tidal volume or ABO volume were known. Because there are no measurements of either for T. leeri, VABO was calculated for each fish according to relationships derived for the dwarf gourami, Colisa lalia(Schuster, 1989). He analysed the time-course of air volume changes in the ABO during apnoea at different temperatures, finding that the relationship between VABO(μl) and fish length (l; mm) was well described by the function:
\[\ V_{\mathrm{ABO}}=(2.65{\times}10^{-4}){\times}l^{3.1}.\]
This function is a rough estimate of a single suprabranchial chamber volume at∼26°C after an apnoea duration of approximately 120 s[VABO(120)].
Schuster stated that if one measurement of VABO is known, the VABO at any instant time during apnoea(t; s) at 25°C is given by:
\[\ V_{\mathrm{ABO}}(t)=(0.897+0.166\mathrm{e}^{-0.0046t}){\times}[V_{\mathrm{ABO}}(t_{1}){/}(0.897+0.166\mathrm{e}^{-0.0046t_{1}})],\]
where VABO(t1) is the known VABO (μl) after an apnoea duration of t1 (s). Since Eqn 3 can be used to calculate VABO(120), this can be substituted as VABO(t1) in Eqn 4, and therefore VABO at the beginning of the apnoea period[VABO(t0)] can be calculated.
VABO(t0) and the PO2 in the ABO at the beginning of apnoea (i.e. initial PO2) are needed to calculate the initial volume of O2[VO2(t0); ml] in the ABO(Eqn 5). The initial PO2 was assumed to be equivalent to PO2,air. This is reasonable because during expiration practically all of the gas in the ABO is displaced out of the mouth with water from the opercular cavity(Peters, 1978):
\[\ V_{\mathrm{O}_{2}}(t_{0})=P_{\mathrm{O}_{2,\mathrm{air}}}{\times}{\beta}_{\mathrm{O}_{2}}{\times}V_{\mathrm{ABO}_{2}}(t_{0}),\]
where βO2 is the O2 capacitance of air at 25°C(=9.09 ml l-1 kPa-1)(Dejours, 1981), and VABO(t0) is the initial volume of one suprabranchial chamber (litres).
The volume of O2 consumed during apnoea is calculated using O2,ap (ml h-1) and mean apnoea duration (tap, h) for each fish under each treatment, and this is subtracted from O2(t0)to arrive at the volume of O2 in the ABO at the end of apnoea(VO2,end; ml):
\[\ V_{\mathrm{O}_{2,\mathrm{end}}}=V_{\mathrm{O}_{2}}(t_{0})-[({\dot{V}}_{\mathrm{O}_{2,\mathrm{ap}}}{/}2)t_{\mathrm{ap}}].\]
In Eqn 6, O2,ap is halved because it represents the rate of O2 uptake across the surface area of the entire ABO (i.e. both suprabranchial chambers) and VO2,end is calculated for a single suprabranchial chamber.
Although Schuster found that the ABO of C. lalia decreased in volume as O2 was consumed(Schuster, 1989), the ABO of T. leeri is a bony structure(Graham, 1997; Peters, 1978) (L.A.A.,personal observation) and the rate of change in VABO with declining ABO-O2 may be different from that in C. lalia,or may not occur at all. Therefore, end-apnoea ABO-PO2 was calculated with two assumptions that bracket reality (note that the following calculations make the simplifying assumption that CO2 is not present within the ABO): (1) VABO remained constant as O2 was consumed(Eqn 7) and (2) VABO decreased as if the ABO was totally compliant [i.e. VO2,end (ml) is subtracted from VABO(t0) (litres); Eqn 8]:
\[\ P_{\mathrm{O}_{2}}=V_{\mathrm{O}_{2,\mathrm{end}}}{/}[V_{\mathrm{ABO}}(t_{0}){\times}{\beta}_{\mathrm{O}_{2}}],\]
\[\ P_{\mathrm{O}_{2}}=V_{\mathrm{O}_{2,\mathrm{end}}}{/}\{[V_{\mathrm{ABO}}(t_{0})-({\dot{V}}_{\mathrm{O}_{2,\mathrm{end}}}{/}1000)]{\beta}_{\mathrm{O}_{2}}\}.\]
Although a totally compliant ABO is unlikely, these assumptions represent opposite, extreme situations and thus allow consideration of all possibilities and a broader interpretation of results.

End-apnoea ABO-PO2 assuming both a constant and totally compliant ABO volume was determined for each fish and compared between treatments.

Air–blood PO2 gradient

To ascertain whether a change in apnoea duration is solely responsible for a change in mean VO2/breath with changing PO2,air, or whether a change in the air–blood PO2 gradient is also a contributing factor, mean O2,ap was calculated for each treatment and plotted against PO2,air.

Statistical analysis

Where appropriate, mass-specific values were used to account for the variation attributed to mass differences between fish(McNab, 1999). Although this procedure does not completely remove mass effects(Packard and Boardman, 1999)because O2scales allometrically in fish (White et al., 2006), the body size range was too small to determine the intraspecific allometric exponent for T. leeri accurately and arrive at mass-independent values.

A repeated-measures analysis of variance (ANOVA) was performed in JMP Version 5.1 (SAS Institute, Cory, NC, USA) to determine the effect of PO2,air on all variables. To fulfil assumptions of normality and homogeneity of variance, PO2,air was log transformed for the analysis of fab, which was not transformed; aerial O2, aquatic O2, O2/breath,end-apnoea ABO-PO2 (assuming constant ABO volume) and end-apnoea ABO-PO2 (assuming a totally compliant ABO) were log transformed together with PO2,air; and O2,ap and total O2 were log transformed without PO2,air being transformed. A Tukey's HSD test was used in post-hoc analyses where repeated-measures ANOVA revealed significant treatment effects. Statistical significance for all tests was determined with α=0.05.

Aerial respiration

There was a significant negative effect of PO2,air on fab(F1,23=120.4, P<0.0001), with fab decreasing from 46.3 breaths h-1 at 5 kPa to 14.3 breaths h-1 at 60 kPa(Fig. 3A). There was a significant positive effect of PO2,air on aerial O2(F1,23=37.2, P<0.0001), with aerial O2 increasing from 20.1 ml kg-1 h-1 at 5 kPa to 60.8 ml kg-1 h-1 at 60 kPa(Fig. 3B). Complementary to this, there was a significant positive effect of PO2,air on VO2/breath (F1,23=221.5, P<0.0001), with VO2/breath increasing from 0.5 ml kg-1 at 5 kPa to 5.5 ml kg-1 at 60 kPa (Fig. 3C).

Aquatic and total respiration

There was a significant negative effect of PO2,air on aquatic O2(F1,27=17.8, P=0.0003), with a significant increase in aquatic O2 under the 5 kPa treatment (Fig. 3B). There was no significant effect of PO2,air on total O2(F1,23=2.49, P=0.13)(Fig. 3B).

End-apnoea ABO-PO2

Regardless of whether it was assumed that the ABO volume was constant or that the ABO was totally compliant, there was a significant positive effect of PO2,air on end-apnoea ABO-PO2 (F1,21=544, P<0.0001, and F1,21=1091, P<0.0001, respectively) (Fig. 4). When ABO volume was assumed to be constant, end-apnoea ABO-PO2 increased from 4.3 kPa at PO2,air=5 kPa to 40.6 kPa at PO2,air=60 kPa, and when ABO volume was assumed to be totally compliant, end-apnoea ABO-PO2increased from 4.4 to 47.3 kPa with the same change in PO2,air.

Air–blood PO2 gradient

A repeated-measures ANOVA revealed a significant positive correlation between mean O2,ap and PO2,air (F1,21=34.7, P<0.0001), with mean O2,ap increasing from 13.2 ml kg-1 h-1 at 5 kPa to 59.8 ml kg-1 h-1 at 60 kPa. The relationship between mean O2,ap and PO2,air was described by the logarithmic curve: O2,ap=45.8log(PO2,air)–19.2(r2 =0.63) (Fig. 5).

Aerial normoxia

Under normoxic conditions at 25°C, the mass-specific total (air +water) O2 of T. leeri was 108.2 ml kg-1 h-1. This value is lower than that found for the closely related T. trichopterus (156.2 ml kg-1 h-1)(Burggren and Haswell, 1979)and T. pectoralis (126.9 ml kg-1 h-1)(Natarajan and Rajulu, 1982),all normalised to a common body temperature of 25°C with a Q10of 1.65 (White et al., 2006). When viewed in comparison to an allometric relationship between O2 and body mass, the data obtained here for T. leeri are well within the 95%prediction intervals (31.8–290 ml kg-1 h-1) for new data (White et al., 2006),indicating that the data obtained in this study are reliable.

Aerial hypoxia

The breathing frequency (fab) of T. leeriunder normoxic conditions was 21.8±5.9 breaths h-1 (mean± s.d.), which is greater than that observed for T. trichopterus (12.8± 4.1 breaths h-1)(Burggren, 1979). However, like T. trichopterus, fab of T. leeri increased as the gas phase became increasingly hypoxic (Fig. 3A). At a PO2,air of 5 kPa, fab was twofold higher than control (normoxic) levels,whereas T. trichopterus increased fab almost threefold. However, fab calculated for T. leeriat 5 kPa may be an underestimate, because only three fish yielded results where some individual breaths could be isolated. For the remaining fish, no or few breathing events were apparent. This suggests that the air–blood PO2 gradient was insufficient to promote O2 uptake. The air–blood PO2gradient may have in fact been reversed, resulting in fish losing O2 gained from the water to the air. This futile behaviour would be expected, however, because severe aerial hypoxia is not encountered in the natural environment, and fish are unlikely to have evolved an appropriate response to such conditions.

An increase in fab in response to aerial hypoxia has been observed in species that have a reduced gill surface area and are therefore heavily reliant on aerial respiration. For example, M. albus showed a significant reduction in apnoea duration when inspiring gas mixtures containing 16 kPa O2 or less, and breaths containing 1.5 kPa O2 were exhaled almost immediately(Graham et al., 1995). Monopterus cuchia increased tidal volume as well as fab during hypoxic breathing(Lomholt and Johansen, 1974),but no such change in ABO tidal volume occurred in T. trichopterus(Burggren, 1979). Protopterus aethiopicus increased fab when N2 gas was injected into the ABO, and, interestingly, air breathing was stimulated by N2 injection even when arterial PO2 was higher than that towards the end of a normal apnoea period (Johansen and Lenfant, 1968). Amia calva, a fish less reliant on aerial respiration, increased fab twofold when air containing only 8 kPa O2 was inspired(Hedrick and Jones, 1993). In contrast to Amia, other fish with efficient aquatic gas exchange do not show a change in fab when N2 gas is injected into the lung (Neoceratodus forsteri)(Johansen et al., 1967) or air bladder (Lepisosteus oculatus)(Smatresk and Cameron, 1982b). The first and second gill arches of T. leeri are large and fully developed (Munshi, 1968), and T. leeri is sensitive to aquatic hypoxia(Miller, 2003), unlike Protopterus (Johansen and Lenfant, 1968) and Monopterus(Lomholt and Johansen, 1974). It therefore seems likely that, although T. leeri is not as reliant on aerial respiration as these species, aquatic respiration may not be sufficient to meet its metabolic demands, making it an obligate air breather(Graham, 1997). T. trichopterus is considered an obligate air breather at temperatures above 20–25°C, as it shows signs of distress if denied access to air(Burggren, 1979).

Despite an increase in fab during aerial hypoxic exposure, T. leeri was unable to sustain aerial O2 equal to that under normoxic conditions (Fig. 3B). Complementary to the observed decrease in aerial O2 with decreasing PO2,air, VO2/breath also showed a decline(Fig. 3C).

The contribution of aerial O2 to total O2 decreased from 33% in normoxia to 25% when PO2,air was 10 kPa, and to 16% at 5 kPa. Similarly, under normoxic conditions at 27°C,the ABO of T. trichopterus accounted for 42% of the total O2 and less than 15% when PO2,air was reduced to 7.2 kPa(Burggren, 1979). Both species showed an increase in aquatic O2 to compensate for their reduced ability to extract O2 from the air(Fig. 3B). It is reasonable to assume that this increase in aquatic O2 arises almost entirely from gas exchange via the gills, because cutaneous gas exchange in air-exposed T. trichopterus accounts for only ∼10% of the total gas exchange (Burggren and Haswell, 1979). Aquatic O2 may be increased via an increase in branchial ventilation frequency,branchial tidal volume or both. Branchial ventilation is known to increase initially in most air-breathing fish as aquatic PO2 falls(Hughes and Singh, 1971; Johansen et al., 1970; Pettit and Beitinger, 1985; Smatresk and Cameron, 1982a),but the effect of aerial hypoxia on branchial ventilation appears not to have been investigated.

Aerial hyperoxia

At a PO2,air of 60 kPa, fab significantly decreased in T. leeri(Fig. 3A). Similarly, T. trichopterus decreased fab when PO2,air was increased to 80 kPa(Burggren, 1979). A decrease in fab has been observed in almost all species exposed to aerial hyperoxia: P. ethiopicus(Lahiri et al., 1970), L. oculatus (Smatresk and Cameron,1982b), Electrophorus electricus(Johansen et al., 1968b) and M. albus (Graham et al.,1995).

Despite a decrease in fab in hyperoxic air, O2,ap increased by almost twofold when PO2,air was 40 kPa;however, no significant increase occurred when PO2,air was increased to 60 kPa(Fig. 3B).

Although aerial O2 increased significantly in aerial hyperoxia compared with normoxia, there was no corresponding decrease in aquatic O2, and total O2 was unchanged(Fig. 3B). However, because aerial and aquatic respiration were measured separately, inherent variability may have obscured the expected correlations.

ABO-O2 chemoreceptors

The responsiveness of T. leeri to changes in aerial O2content lends insight to the question of the existence of ABO-O2chemoreceptors. Graham et al. give support for the existence of ABO-O2 chemoreceptors in some air-breathing species(Graham et al., 1995). They found that the rapidity of the gas-voiding reflex of M. albus to changed aerial O2 content was indicative of ABO-O2chemoreceptors, not chemoreceptors in the systemic circulation. They suggested that an ABO-O2 chemoreceptor would be advantageous in the regulation of cardiac responses to an air-breathing event. A common pattern of cardiac arrhythmia found in fishes during an air-breathing event is inspiration-induced tachycardia followed by the gradual onset of bradycardia as the ABO-O2 content falls, leading to pronounced bradycardia with exhalation (Farrell, 1978; Johansen et al., 1968a; Singh and Hughes, 1973; Smatresk, 1988; Smatresk, 1990). Therefore, an ABO-O2 chemoreceptor may be important in the modulation of mechanoreceptor and other stimuli affecting air-breathing tachycardia, in attenuating tachycardia as ABO-O2 declines and in terminating the breath when ABO-PO2 becomes too low to promote O2 uptake (Graham and Baird,1984). This would result in the effective matching of ventilation and perfusion (Johansen, 1966; Johansen, 1970).

However, the findings of Graham et al.(Graham et al., 1995) contrast with Burggren's indication of more centrally located chemoreceptors (i.e. in the brain) in T. trichopterus, which was based on the lag in ventilation response time (several seconds) to stepwise changes in aerial O2 content (Burggren,1979). Burggren argued that because aerial hypoxia is rarely, if ever, encountered in the natural environment, selection pressures should not be strong for the evolution of a peripheral chemoreceptor control system able to differentiate between reduced systemic blood O2 resulting from gill ventilation with hypoxic water and that resulting from ABO ventilation with hypoxic gas.

More substantial evidence for ABO-O2 chemoreceptors in T. leeri would have been an end-apnoea ABO-PO2 that was independent of PO2,air. This would have suggested the existence of an ABO-PO2 threshold that, once reached, triggered T. leeri to renew the gas in its ABO. However,this was not the case in this study; end-apnoea PO2 increased with increasing PO2,air(Fig. 4). This lack of correlation between ABO-PO2 and renewal of ABO gas has also been recognised in lungfish(Johansen and Lenfant, 1968)and in Pacific tarpon that uses the swimbladder as an ABO(Seymour et al., 2007). It has also been shown that apnoea termination occurs at high ABO-PO2 levels when the rate of decline in ABO-PO2 is rapid(Shelton et al., 1986). This corresponds with the findings in this study; mean apnoeic O2(O2,ap)increased under hyperoxic conditions and end-apnoea ABO-PO2 was higher than that under normoxic conditions. These findings indicate that if ABO-O2 chemoreceptors are present, then their regulation of bimodal control in air-breathing fish is partial, and that respiration is mainly affected by chemoreceptors elsewhere in the central and peripheral nervous system and possibly by mechanoreceptors in the ABO.

Air–blood PO2 gradient

Mean O2,apincreased with increasing PO2,air(Fig. 5), supporting the hypothesis that the observed increase in VO2/breath with increasing PO2,air(Fig. 3C) was facilitated not only by an increase in apnoea duration but also by an increase in the air–blood PO2 gradient. The logarithmic function between O2,ap and PO2,air suggests that when there is no O2 uptake occurring in the ABO (i.e. O2,ap is equal to zero), PO2,air is equal to 2.62 kPa. Therefore, the air and blood are in equilibrium and the efferent blood from the gills can be assumed to have a PO2approximating this value. The relationship also indicates that in hyperoxia O2,ap reaches a maximum where it becomes independent of PO2,air(Fig. 5). Because the haemoglobin would be expected to be completely saturated in the hyperoxic ABO(Herbert and Wells, 2001),aerial O2plateaus because the blood reaches a point where it can take up no more than can be dissolved in the plasma.

List of symbols and abbreviations

  • ABO

    air-breathing organ

  • βO2

    oxygen capacitance

  • fab

    air-breathing frequency

  • FeO2

    excurrent oxygen content

  • FiO2

    incurrent oxygen content

  • PO2

    oxygen partial pressure

  • PO2,air

    aerial oxygen partial pressure

  • tap

    apnoea duration

  • VABO

    volume of the air-breathing organ

  • VABO(t0)

    ABO volume at the beginning of the apnoea period

  • i

    oxygen flow rate

  • O2

    oxygen consumption rate

  • O2,ap

    apnoeic O2

  • VO2,end

    volume of oxygen in the ABO at the end of apnoea

  • VO2/breath

    oxygen uptake per breath

This work was supported by the Australian Research Council. It was carried out under permit from the Animal Ethics Committee of the University of Adelaide. We would like to thank Philip Matthews for his technical assistance.

Burggren, W. W. (
). Bimodal gas exchange during variation in environmental oxygen and carbon dioxide in the air breathing fish Trichogaster trichopterus.
J. Exp. Biol.
Burggren, W. W. and Haswell, S. (
). Aerial CO2 excretion in the obligate air breathing fish, Trichogaster trichopterus: a role for carbonic anhydrase.
J. Exp. Biol.
Dejours, P. (
Principles of Comparative Respiratory Physiology
. Amsterdam:Elsevier/North-Holland Biomedical Press.
Depocas, F. and Hart, S. J. (
). Use of the Pauling oxygen analyzer for measurement of oxygen consumption of animals in open-circuit systems, and in a short-lag, closed-circuit apparatus.
J. Appl. Physiol.
Farrell, A. P. (
). Cardiovascular events associated with air-breathing in two teleosts, Hoplerythrinus unitaeniatus and Arapaima gigas.
Can. J. Zool.
Graham, J. B. (
Air-Breathing Fishes: Evolution, Diversity and Adaptation
. San Diego: Academic Press.
Graham, J. B. and Baird, T. A. (
). 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.
J. Exp. Biol.
Graham, J. B., Lai, N. C., Chiller, D. and Roberts, J. L.(
). The transition to air breathing in fishes. V. Comparative aspects of cardiorespiratory regulation in Synbranchus marmoratus and Monopterus albus (Synbranchidae).
J. Exp. Biol.
Hedrick, M. S. and Jones, D. R. (
). The effects of altered aquatic and aerial respiratory gas concentrations on air-breathing patterns in a primitive fish (Amia calva).
J. Exp. Biol.
Herbert, N. A. and Wells, R. M. G. (
). The aerobic physiology of the air-breathing blue gourami, Trichogaster trichopterus, necessitates behavioural regulation of breath-hold limits during hypoxic stress and predatory challenge.
J. Comp. Physiol. B
Hughes, G. M. and Singh, B. N. (
). Respiration in an air-breathing fish, the climbing perch Anabas testudineus. II. Respiratory patterns and the control of breathing.
J. Exp. Biol.
Hughes, G. M. and Singh, B. N. (
). Gas exchange with air and water in an air-breathing catfish, Saccobranchus (heteropneustes) fossilis.
J. Exp. Biol.
Johansen, K. (
). Air breathing in the teleost Symbranchus marmoratus.
Comp. Biochem. Physiol.
Johansen, K. (
). Air breathing in fishes. In
Fish Physiology
. Vol.
(ed. W. S. Hoar and D. J. Randall), pp.
-411. New York: Academic Press.
Johansen, K. and Lenfant, C. (
). Respiration in the African lungfish, Protopterus aethiopicus. II. Control of breathing.
J. Exp. Biol.
Johansen, K., Lenfant, C. and Grigg, G. C.(
). Respiratory control in the Lungfish, Neoceratodus forsteri (Krefft).
Comp. Biochem. Physiol.
Johansen, K., Lenfant, C. and Hanson, D.(
). Cardiovascular dynamics in lungfishes.
Z. Vergl. Physiol.
Johansen, K., Lenfant, C., Schmidt-Nielsen, K. and Petersen,J. (
). Gas exchange and control of breathing in the electric eel, Electrophorus electricus.
Z. Vergl. Physiol.
Johansen, K., Hanson, D. and Lenfant, C.(
). Respiration in a primitive air breather, Amia calva.
Respir. Physiol.
Keller, G., Warrack, B. and Bartel, H. (
Statistics for Management and Economics
. Belmont:Duxbury Press.
Lahiri, S., Szidon, J. P. and Fishman, A. P.(
). Potential respiratory and circulatory adjustments to hypoxia in the African lungfish.
Fed. Proc.
Lomholt, J. P. and Johansen, K. (
). Control of breathing in Amphipnous cuchia, an amphibious fish.
Respir. Physiol.
McNab, B. K. (
). On the comparative ecological and evolutionary significance of total and mass-specific rates of metabolism.
Physiol. Biochem. Zool.
Miller, B. C. (
The effect of activity, temperature and aquatic oxygen on bimodal respiration in the pearl gourami, Trichogaster leeri
. Honours thesis, Environmental Biology, University of Adelaide, Adelaide, Australia.
Milsom, W. K. (
). Mechanoreceptor modulation of endogenous respiratory rhythms in vertebrates.
Am. J. Physiol.
Munshi, J. S. D. (
). The accessory respiratory organs of Anabas testudineus (Bloch) (Anabantidae,Pisces).
Proc. Linn. Soc. Lond.
Natarajan, G. M. and Rajulu, G. S. R. (
). Bimodal oxygen uptake and some blood parameters in the bubble next builder, Trichogaster pectoralis (Regan).
Curr. Sci.
Pack, A. I., Galante, R. and Fishman, A. P.(
). Control of inter-breath interval in the African lungfish.
Am. J. Physiol.
Packard, G. C. and Boardman, T. J. (
). The use of percentages and size-specific indices to normalize physiological data for variation in body size: wasted time, wasted effort?
Comp. Biochem. Physiol.
Peters, H. M. (
). On the mechanisms of air ventilation in Anabantoids (Pisces: Teleostei).
Pettit, M. J. and Beitinger, T. (
). Oxygen acquisition of the reedfish, Erpetoichthys calabaracus.
J. Exp. Biol.
Riley, J. P. and Chester, R. (
Introduction to Marine Chemistry
. London: Academic Press.
Sanchez, A., Soncini, R., Wang, T., Koldkjaer, P., Taylor, E. W. and Glass, M. L. (
). The differential cardio-respiratory responses to ambient hypoxia and systemic hypoxaemia in the South American lungfish, Lepidosiren paradoxa.
Comp. Biochem. Physiol.
Schuster, S. (
). The volume of air within the swimbladder and breathing cavities of the Anabantoid fish Colisa lalia (Perciformes, Belontiidae).
J. Exp. Biol.
Seymour, R. S., Farrell, A. P., Christian, K., Clark, T. D.,Bennett, M. B., Wells, R. M. G. and Baldwin, J. (
). Continuous measurement of oxygen tensions in the air-breathing organ of Pacific tarpan (Megalops cyprinoides) in relation to aquatic hypoxia and exercise.
J. Comp. Physiol. B
Shelton, G., Jones, D. R. and Milsom, W. K.(
). Control of breathing in ectothermic vertebrates. In
Handbook of Physiology
. Vol.
,Sect. 3 (ed. N. S. Cherniack and J. G. Widdicombe), pp.
-909. Bethesda: American Physiological Society.
Singh, B. N. and Hughes, G. M. (
). Cardiac and respiratory responses in the climbing perch Anabas testudineus.
J. Comp. Physiol.
Smatresk, N. J. (
). Ventilatory and cardiac reflex responses to hypoxia and NaCN in Lepisosteus osseus, an air-breathing fish.
Physiol. Zool.
Smatresk, N. J. (
). Control of the respiratory mode in air-breathing fishes.
Can. J. Zool.
Smatresk, N. J. (
). Chemoreceptor modulation of endogenous respiratory rhythms in vertebrates.
Am. J. Physiol.
Smatresk, N. J. and Cameron, J. N. (
). Respiration and acid-base physiology of the spotted gar (Lepisosteus oculatus), a bimodal breather: I. Normal values and the response to severe hypoxia.
J. Exp. Biol.
Smatresk, N. J. and Cameron, J. N. (
). Respiration and acid–base physiology of the spotted gar, a bimodal breather. III. Response to a transfer from freshwater to 50% sea water, and control of ventilation.
J. Exp. Biol.
White, C. R., Phillips, N. F. and Seymour, R. S.(
). The scaling and temperature dependence of vertebrate metabolism.
Biol. Lett.