The amphibious fish Kryptolebias marmoratus uses different strategies to maintain oxygen delivery during aquatic hypoxia and air exposure

The transition from aquatic to terrestrial life represents a major step in vertebrate evolution because the physical conditions between these environments are dramatically different. Oxygen solubility is relatively low in water and even well-oxygenated aquatic environments contain only about 3% of the O₂ available in atmospheric air (Dejours, 1988). Low concentrations of aquatic O₂, particularly in hypoxic habitats, have often been hypothesized to be one of the driving forces behind the evolution of amphibious or terrestrial life histories because invasion of land would allow animals to exploit the O₂-rich aerial environment (Graham, 1997).

Taking advantage of aerial O₂ presents several challenges for fishes. Water-breathing fishes exchange respiratory gases across the gills but during emersion the gill lamellae typically collapse and coalesce, reducing the surface area available for respiration. Accumulation of CO₂ in the blood of emersed fishes, resulting from the low solubility of CO₂ in air versus water, can also impair the ability of hemoglobin (Hb) to bind and transport O₂ (Rahn, 1966; Ultsch, 1987). High blood partial pressure of CO₂ (P_CO2) may reduce intraerythrocytic pH and reduce the affinity of Hb for O₂ by the Bohr effect, which prevents O₂ loading at the gas-exchange surface (Bohr et al., 1904). Impaired O₂ transport can also occur via the Root effect: a phenomenon found only in fishes, which prevents Hb from becoming fully O₂ saturated at high P_CO2 (Root and Irving, 1941; Gillen and Riggs, 1971). It has been hypothesized that well-adapted air-breathing and amphibious fishes should possess relatively pH-insensitive Hb to reduce the consequences of Bohr and Root effects on O₂ transport (Graham, 1997; Shartau and Brauner, 2014); however, not all air-breathing fishes follow this pattern (e.g. Farmer et al., 1979; Wells et al., 1997). Thus, despite the fact that amphibious fishes are surrounded by O₂-rich air during emersion, high blood P_CO2 coupled with a reduced respiratory surface area may cause amphibious fishes to become hypoxemic until physiological and morphological adjustments to the O₂ transport system can be made.

Fishes can increase their capacity for O₂ uptake, binding and transport in response to hypoxia or hypoxemia. For example, lamellar perfusion can be increased during exposure to aquatic hypoxia to increase the surface area available for O₂ uptake (Booth, 1979; Soivio and Tuurala, 1981). Fish adapted or acclimated to hypoxia often have blood with a high O₂-carrying capacity and possess Hb with a relatively high O₂-binding affinity (Weber and Fago, 2004; Brauner and Val, 2006; Wells, 2009). Air-breathing and amphibious fishes similarly tend to possess blood with a high O₂-carrying capacity, which has been hypothesized to compensate for the impaired O₂ uptake and transport caused by elevated blood P_CO2 during emersion (Farmer, 1979; Farmer et al., 1979; Graham, 1997).

During the evolution of amphibious lifestyles by early tetrapods or extant amphibious fishes, the hypoxia response could have been used to maintain O₂ transport in the face of high P_CO2 during emersion. The ability to sense and respond to hypoxia likely predates the invasion of land by animals, because even the basal Caenorhabditis elegans is known to accumulate the transcription factor hypoxia inducible factor-1 during hypoxic exposure (Rytkönen et al., 2011). After 24h of emersion the amphibious fish Gillichthys mirabilis alters gene expression in a similar pattern to fish that experienced aquatic hypoxia, suggesting that similar strategies may be used to maintain metabolism (Gracey, 2008). However, it is unknown whether these transcriptional changes result in altered protein synthesis or have functional consequences. The response of O₂-transport systems in amphibious fishes that spend longer periods (days to weeks) out of water is similarly not well understood.

Given the reduced capacity for transport of blood O₂ as a result of high P_CO2 and the loss of respiratory surface area from the gills during emersion, we hypothesized that amphibious fishes maintain O₂ uptake and transport while out of water by co-opting their hypoxia response. This hypothesis predicts that amphibious fishes acclimated to either aquatic hypoxia or a terrestrial environment would use a conserved suite of phenotypic changes to increase the O₂-carrying capacity, the affinity of Hb for O₂ and the flow of blood to extrabranchial sites of gas exchange. Alternatively, amphibious fishes may use different responses to tolerate aquatic hypoxia and air exposure. The mangrove rivulus Kryptolebias marmoratus (Poey 1880) is an ideal model for studying the adaptations used by amphibious fishes to maintain O₂ uptake and transport. In the wild, mangrove rivulus typically inhabit extremely hypoxic (P_O2<2.5kPa) puddles and crab burrows filled with water, but these fish also frequently emerse and find refuge in terrestrial habitats such as damp leaf litter or inside rotting logs that may also be hypoxic (Taylor et al., 2008, Ellison et al., 2012; Taylor, 2012). Emersions can last upwards of 2months (Taylor, 1990), during which time mangrove rivulus remain active and are thought to obtain O₂ by cutaneous respiration (Grizzle and Thiyagarajah, 1987; Cooper et al., 2012; Wright, 2012). To test the hypothesis that air-exposed fish mount a hypoxia response, we exposed K. marmoratus to 7 days of aquatic normoxia, aquatic hypoxia (P_O2 ~3.6kPa), aerial normoxia or aerial hypoxia (P_O2 ~13.6kPa). We measured the O₂-carrying capacity of blood (number of red blood cells, hematocrit and [Hb]), Hb–O₂ affinity and extra-branchial angiogenesis, predicting that similar phenotypes should occur in fish exposed to aquatic hypoxia and air exposure.

Both hypoxia and air exposure significantly increased blood [Hb] (two-way ANOVA: oxygen, P<0.001; medium, P<0.05; interaction, P=0.92; Fig. 1A). Hematocrit was significantly increased in all three treatment groups relative to normoxic control fish (two-way ANOVA: oxygen, P<0.01; medium, P=0.23; interaction, P<0.05; Fig. 1B). Hypoxia significantly increased the number of red blood cells (RBCs), but air exposure had no effect (two-way ANOVA: oxygen, P<0.001; medium, P=0.46; interaction, P=0.52; Fig. 1C). As a result of the small blood volumes used in this study, repeated sampling was impossible and therefore mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH) and mean corpuscular hemoglobin concentration (MCHC) could only be calculated from the mean values for each treatment group, precluding statistical analysis. None of the values varied dramatically between treatment groups (Table 1).

Air exposure significantly decreased P₅₀, the blood P_O2 at which Hb is 50% saturated with O₂ and P₅₀ was increased by elevated in vitro CO₂, but there was no overall effect of oxygen availability (three-way ANOVA: medium, P<0.001; CO₂, P<0.001; O₂, P=0.06; Fig. 2). A significant interaction between respiratory medium and O₂ availability was found (P<0.001, all other interactions P>0.05) and post hoc tests revealed that air exposure significantly increased Hb–O₂ affinity (reduced P₅₀) to a greater extent in normoxic relative to aerial hypoxic conditions (Fig. 2). The Root effect was significantly larger between 0.3kPa and 4.0kPa P_CO2 than between 0.3kPa and 0.6kPa P_CO2, but was not influenced by any treatment condition (three-way ANOVA: medium, P=0.21; O₂, P=0.06; CO₂, P<0.05; all interactions, P>0.05; Table 1). Hb–O₂ binding cooperativity (Hill coefficient, n_H) was significantly increased by hypoxia and high P_CO2, but not air exposure (three-way ANOVA: medium, P=0.52; O₂, P<0.01; CO₂, P<0.01; all interactions, P>0.05; Table 1).

Under conditions of aquatic normoxia (control), the marker for angiogenesis CD31 was visible in the dorsal and ventral cutaneous epithelium, as well as within the buccal and opercular chambers (Fig. 3A). After 7 days of air exposure, there was a striking increase in fluorescence of the buccal and opercular chambers (Fig. 3B). Across all four body regions, CD31 fluorescent intensity was significantly increased after air exposure, independent of O₂ availability (three-way ANOVA: medium, P<0.001; O₂, P=0.94; all interactions, P>0.05; Fig. 3C). Significantly reduced fluorescence was observed in the ventral epithelium compared with the other body regions measured (three-way ANOVA: region, P<0.001; Fig. 3C).

We observed air-exposed mangrove rivulus opening their mouths and ‘gulping’ air when out of water. Rates of aerial ventilation after 1 h of air-breathing were more than double the rates observed after 1 day or 7 days of air exposure (P<0.05; Fig. 4A). The activity of air-exposed fish (body movementsh⁻¹) did not significantly change over the 7day air-exposure period, although there was a trend towards decreased movement after 1day (P=0.06; Fig. 4B). The majority of aerial ventilations occurred within 30s of a body movement, although some ventilations were separated from body movements by more than 10min (Fig. 4C).

Our results do not support the hypothesis that amphibious fishes maintain O₂ uptake and transport while out of water by co-opting the hypoxia response. Instead, we found that K. marmoratus have flexible O₂ uptake and delivery strategies that respond to changes in the respiratory environment. Behavioral and immunohistochemical data suggest that, during air exposure, the mouth and bucco-opercular chamber provide a supplementary gas exchange surface to the cutaneous epithelium. The hematological system also displays flexibility in response to changes in the respiratory environment. In contrast to our predictions, different responses were observed in fish exposed to air and hypoxia. Air exposure increased Hb–O₂ affinity and increased the concentration of Hb in the blood without increasing the number of circulating RBCs. In contrast, hypoxia in both aquatic and aerial environments substantially increased the O₂-carrying capacity of the blood via an increase in the density of RBCs, without any major effect on Hb–O₂ affinity. These results indicate that the changes in angiogenesis and binding affinity of Hb for O₂ observed during air exposure are not the result of altered O₂ availability, but instead may be controlled by a secondary pathway that can be activated concurrently with the hypoxia-activated pathway in the case of aerial hypoxia. Air-exposed mangrove rivulus might, for example, respond to changes in blood CO₂ or pH rather than O₂ (Putnam et al., 2004; Casey et al., 2010; Tresguerres et al., 2010).

If the binding affinity of Hb for O₂ reflects a balance between O₂ uptake and delivery, the abundance of O₂ in atmospheric air compared with water is predicted to result in relatively lower Hb–O₂affinity in air-breathing animals. Consistent with this hypothesis, the air-breathing fishes Protopterus annectens, Polypterus senegalus and Clarias lazera all reduce Hb–O₂ affinity during the ontogenetic shift from breathing in water to breathing in air (Graham, 1997). After 3weeks out of water, the amphibious Canterbury mudfish Neochanna burrowsius also decreases Hb–O₂ affinity (Wells et al., 1984). Conversely, hypoxia-tolerant organisms are generally characterized by high Hb–O₂ affinities to maximize O₂ uptake (Graham, 1997; Wells, 2009). Why do mangrove rivulus contradict this pattern and increase affinity of Hb for O₂ after 7 days of air exposure? The lungfish P. amphibius increases Hb–O₂ affinity while air-breathing during aestivation, probably to enable reduced lung tidal volumes while inhabiting hypoxic subterranean environments (Johansen et al., 1976). In the non-aestivating mangrove rivulus, however, there was no significant effect of O₂ availability on Hb–O₂ affinity, suggesting that these fish are using a different strategy to aestivating lungfish.

We propose that the elevated affinity of Hb for O₂ observed in air-exposed mangrove rivulus serves to maintain a relatively stable effective P₅₀ across respiratory media by counterbalancing the Bohr effect in emersed fish resulting from increased blood P_CO2 and/or decreased blood pH. The P₅₀ of control (aquatic normoxia) mangrove rivulus at a P_CO2 of 0.6kPa (P₅₀=24.7±1.3Torr; 1 Torr=133Pa) – a typical partial pressure in venous blood of aquatic tropical fishes – was approximately equal to that of blood from air-acclimated fish at a P_CO2 of 4.0kPa (P₅₀=24.3±1.5Torr), which is a typical level for emersed amphibious fishes. Thus, Hb–O₂ affinity may be maintained in K. marmoratus under a variety of environmental conditions, regardless of widely variable blood CO₂ levels. Like Hb in mangrove rivulus, protein function in other animals is often highly conserved during acclimation to changing environmental conditions (Somero, 1995). For example, expression of alternative troponin isoforms help to maintain cardiac function in Oncorhynchus mykiss and Danio rerio acclimated to varying temperatures (Alderman et al., 2012; Genge et al., 2013). Conservation of Hb–O₂ affinity might allow mangrove rivulus to maintain consistent O₂ delivery in both air and water without requiring modifications to O₂-accepting tissues such as skeletal muscle (Hoppeler and Vogt, 2001; Fraser et al., 2006).

Fishes are able to modify Hb–O₂ affinity primarily by modifying the RBC intracellular environment or by changing the relative expression of Hb isoforms with different O₂-binding affinities (Grigg, 1969; Marinsky et al., 1990; Jensen, 1991; Val, 2000; Weber, 2000; Rutjes et al., 2007). Most commonly, air-breathing fishes increase organic phosphate concentrations relative to [Hb] during behavioral or ontogenetic shifts to air-breathing in order to facilitate O₂ unloading at the tissues (Babiker, 1984; Graham, 1997; Val, 2000). Conversely, hypoxia-tolerant fish species often maximize O₂ uptake using low ratios of organic phosphate to Hb compared with fishes that inhabit well-oxygenated water (Wells, 2009). If air-exposed mangrove rivulus are able to decrease organic phosphate concentrations to increase Hb–O₂ affinity, it is puzzling why fish acclimated to hypoxia did not increase affinity to the same extent. At a P_CO2 of 0.3kPa, which is typical of arterial blood in a water-breathing fish, the P₅₀ of hypoxia-acclimated fish was 20.2±1.3Torr, compared with 21.0±0.7 Torr in control fish. When environmental O₂ is limited there is often a trade-off between optimizing O₂ uptake to RBCs and O₂ unloading at the tissues. Perhaps hypoxia-acclimated mangrove rivulus are constrained from increasing Hb–O₂ affinity by the need to facilitate O₂ delivery to the tissues and instead rely on other hypoxia-tolerance strategies such as increasing the density of red cells in the blood or enlarging gill surface area (Richards, 2011; Turko et al., 2012; Cook et al., 2013). At high partial pressures of CO₂ there was a trend towards increased Hb–O₂ affinity in hypoxia-acclimated fish; this may result from the significantly increased [Hb], which would provide extra pH-buffering capacity and thus lessen the magnitude of the Bohr effect at high P_CO2 (Weber, 2000; Wells, 2009).

Oxygen-carrying capacity was significantly increased, via increased RBC density, by hypoxic acclimation in either water or air but not during air exposure alone. Increasing the number of circulating RBCs in response to hypoxia to enhance O₂ transport, via erythropoiesis or splenic release, is a strategy used by many fishes (e.g. Tetens and Lykkeboe, 1981; Murad et al., 1990; Soldatov, 1996; Lai et al., 2006), but there are exceptions (e.g. McLeod et al., 1978; Routley et al., 2002). Increased O₂-carrying capacity allows fish to increase O₂ uptake from the environment without compromising unloading at the tissues and may therefore be a more efficient strategy for coping with hypoxia than adjusting Hb–O₂ affinity (Brauner and Wang, 1997). It is curious, then, why air exposure alone increased Hb–O₂ affinity but did not induce increased RBC density. Extra RBCs may increase blood viscosity and impede flow through the epidermal capillary beds used for aerial gas exchange (Baldwin and Wells, 1990).

Overall, we observed a relatively minor Root shift (12.2±2.1% mean reduction in maximum O₂ saturation in all treatment groups under normoxic conditions; P_O2=22kPa) across a large (3.7kPa) CO₂ gradient. Neither air-breathing nor hypoxia exposure affected the magnitude of this shift, which was generally small compared with other teleosts (Pelster and Randall, 1998). This suggests that mangrove rivulus Hb is well adapted to maintaining near-maximal O₂ saturation at high P_CO2 during emersion, even immediately after leaving water.

Angiogenesis was observed in the skin and bucco-opercular region of air-exposed mangrove rivulus, but not in fish exposed to aquatic hypoxia. Instead of recruiting additional sites of gas exchange to tolerate hypoxia, mangrove rivulus appear to increase their capacity for branchial O₂ uptake by increasing gill surface area by reduction of the inter-lamellar cell mass (Turko et al., 2012). During emersion, amphibious fishes often rely on air-breathing organs such as lungs, buccal pouches or capillary-rich cutaneous surfaces (Johansen, 1970; Satchell, 1976; Olson, 1994; Graham, 1997). For example, bucco-pharyngeal blood vessels in G. mirabilis are filled with blood after 10–20min of emersion (Todd and Ebeling, 1966). Phylogenetic analyses of mudskippers suggest that species with an increased reliance on air-breathing possess increased capillary density on the tongue and within the bucco-opercular chamber (Zhang et al., 2000; Gonzales et al., 2011). We also observed, for the first time, ‘gulping’ activity by emersed mangrove rivulus, strongly implicating this region as a site of gas exchange. These aerial ventilations were usually associated with body movements, suggesting that transient increases in O₂ demand might provide a stimulus to refresh the air in the bucco-opercular chamber. Aerial ventilation frequency significantly decreased between 1 h and 1 day of air exposure and was probably correlated with an increase in blood flow to extrabranchial respiratory surfaces. Augmented cutaneous blood flow increases O₂ uptake and probably allowed mangrove rivulus to reduce evaporative water loss by decreasing aerial ventilation frequency (Burggren and Mwalukoma, 1983; Burggren and Moalli, 1984). Air exposure also induces enlargement of the inter-lamellar cell mass, reducing the effective gill surface area, which may further limit water loss during bucco-opercular ventilation (Ong et al., 2007).

Observing bucco-opercular respiration was highly unexpected, considering gill surface area is reduced during air exposure and the branchial region was presumed to be non-functional (Ong et al., 2007; Cooper et al., 2012). Instead, aerial O₂ uptake in this species was thought to be achieved entirely across the cutaneous epithelium (Wright, 2012). The dorsal surface of mangrove rivulus has long been known to be covered with a dense network of epithelial capillaries (Grizzle and Thiyagarajah, 1987) and a recent paper from our laboratory demonstrated angiogenesis in the caudal fin of air-exposed fish (Cooper et al., 2012). However, given the small size (~100 mg) of mangrove rivulus and the infrequent nature of ‘gulping’ (1 breath every ~12min), it is not surprising that this behaviour has been overlooked, especially considering that disturbances typically cause fish in both air and water to cease ventilatory activity (A.T., unpublished observation).

Overall, these results provide the first evidence that mangrove rivulus have a phenotypically flexible respiratory system that appears to be independently regulated by both O₂ availability and the respiratory medium. Although it initially appears that air-exposed fish increase the affinity of Hb for O₂, in contrast to most other amphibious fishes, we propose that this change represents an effort to maintain a stable P₅₀ in the face of large changes in blood P_CO2 or pH when out of water. We also have shown for the first time that air exposure promotes angiogenesis in the bucco-opercular chamber, which mangrove rivulus ventilate by ‘gulping’ air. These results contradict the hypothesis that the hypoxia response was co-opted by fishes during the invasion of land and instead suggest that novel adaptations to the O₂ transport system have evolved.

Kryptolebias marmoratus hermaphrodites at least 6 months of age (Slc8E strain) (Tatarenkov et al., 2010) were obtained from the breeding colony maintained in the Hagen Aqualab, University of Guelph. Fish were reared individually in 120ml semi-transparent plastic containers (FisherBrand Collection Containers; Fisher Scientific) under constant conditions (25°C, 15–18‰, pH 8.1, 12h:12h light:dark cycle) (Frick and Wright, 2002). Water changes were performed weekly using artificial seawater (Crystal Sea Marinemix; Marine Enterprises International, Baltimore, MA, USA) made with reverse osmosis water. Fish were fed Artemia nauplii three to four times per week. The experiments in this study were approved by the University of Guelph Animal Care Committee (Animal Utilization Protocol 10G008).

Experimental fish were acclimated for 7 days (25°C) to one of four treatment groups: brackish water (control), air, aquatic hypoxia or aerial hypoxia. All fish were fasted for the duration of the acclimation period. Control fish were maintained under standard rearing conditions (~21kPa O₂). Air-acclimated fish were maintained on moist filter paper (15‰ brackish water, ~21kPa O₂) in plastic rearing containers (Ong et al., 2007). Hypoxia in both water and air was achieved by introducing a finely regulated flow of nitrogen into closed chambers containing experimental fish (Regan et al., 2011). Oxygen levels were measured daily (Hach LDO101 electrode connected to Hach HQ30d meter, Hach Company, Missisauga, ON, Canada) and averaged 3.59±0.06kPa (mean ± s.e.m.; here and throughout) in water and 13.57±0.06kPa in air over the course of the experiments. The P_O2 chosen for aquatic hypoxia has been previously found to be slightly above the critical O₂ tension of mangrove rivulus (Turko et al., 2012). The aerial O₂ concentration was selected after preliminary experiments indicated that more severe aerial hypoxia was lethal, whereas fish acclimated to less severe aerial hypoxia (~17kPa) did not show any hematological response that differed relative to fish acclimated to normoxic air.

To measure hematocrit, hemoglobin concentration and red blood cell counts, blood was collected by caudal severance using heparinised microhematocrit tubes (VWR, Missisauga, ON, Canada) (Bianchini and Wright, 2013). It should be noted that stress induced by this method of blood sampling may have caused the acute splenic release of red blood cells and thus our measurements may be consistent overestimates compared with the in vivo situation. The red blood cell counts at O₂-exchanging surfaces (e.g. the gills or epithelial capillaries) may also differ from the aortic blood we gathered (e.g. Klitzman and Duling, 1979; Hudetz et al., 1999). However, given the small size of K. marmoratus, this was the only practical method of blood collection. Hematocrit (Hct) was determined by centrifuging samples in sealed microhematocrit tubes at 7000 r.p.m. for 2min in an International Clinical Centrifuge (Model CL, International Equipment, Needham, MA, USA). To accurately measure hematocrit of small blood volumes, images were taken under a dissecting microscope and analyzed using ImageJ (Bianchini and Wright, 2013). Hemoglobin concentration was quantified using the cyanmethemoglobin method (Blaxhall and Daisley, 1973). RBCs were counted in a standard hemocytometer using a 1:400 dilution of whole blood: Cortland's isotonic saline (Wolf, 1963) further diluted 1:1 with 0.4% Trypan Blue solution. The RBC indices, mean RBC volume (MCV), mean cellular Hb content (MCH) and mean cellular [Hb] (MCHC), were calculated from Hct/RBC count, ([Hb]/RBC count)×10 and [Hb]/Hct, respectively.

Hb–O₂ equilibrium curves were determined (25°C) at three different CO₂ partial pressures (0.3, 0.6, 4.0kPa) using a spectrophotometer (P_wee50, La Trobe University, Melbourne, Australia) designed to utilize small blood volumes (Reeves, 1980; Henriksson et al., 2008; Bianchini and Wright, 2013). The lower CO₂ partial pressures approximated resting values for arterial (0.3kPa CO₂) and venous (0.6kPa CO₂) blood in tropical fishes (Heisler, 1982; Perry, 1986, Tufts and Perry, 1998), while the higher P_CO2 is typical of tropical amphibious fish during air exposure (Johansen, 1970; Graham, 1997). Blood samples were obtained via caudal severance and were immediately spread thinly between two gas-permeable membranes (Glad Go-Between Freezer Film, Hobart, Tasmania, Australia) held in place on a circular sample holder using an O-ring and inserted into the P_wee50. Oxygen saturation curves at each P_CO2 were created by measuring absorbance at 390nm and 435nm, respectively the isosbestic point and peak absorbance of deoxygenated Hb (Iuchi, 1973; Hoxter, 1979), at O₂ partial pressures between 0 and 22kPa in 1kPa increments. Each O₂ and CO₂ combination was mixed from compressed O₂, CO₂ and N₂ by the P_wee50, humidified and introduced to the sample within the temperature-controlled spectrophotometer. Hb-O₂ affinity (P₅₀), Root effect and cooperativity (Hill coefficient, n_H) were automatically calculated from the saturation curves by a manufacturer-supplied spreadsheet. The blood sampling method used may have overestimated Hb–O₂ affinity by causing the release of catecholamines, which activate NHE, increase red blood cell pH and ultimately reduce P₅₀ (Nikinmaa and Boutilier, 1995; Reid et al., 1998). Blood was collected as quickly as possible to minimize the effects of a catecholaminergic response, but because of the small size of K. marmoratus, this was the only practical method of collecting blood. If stress impacted blood parameters, then the effect would be more or less consistent across all treatment groups. We considered measuring the pH of whole blood at each P_CO2 used for the Hb–O₂ affinity measurements to calculate the magnitude of the Bohr effect, but the small blood volumes and tendency of the blood to clot made this unfeasible.

Angiogenesis was determined by immunohistochemical staining for the endothelial integral membrane protein cluster of differentiation 31 (CD31), also known as platelet endothelial cell adhesion molecule (PECAM-1) (Albelda et al., 1991; Baluk and McDonald, 2008), which has previously been used to measure angiogenesis in Kryptolebias marmoratus (Cooper et al., 2012). Euthanized fish were fixed in 10% neutral buffered formalin at 4°C for 24h, decalcified (Surgipath Decalcifier II, Winnipeg, Canada) for 1 h at 20°C and stored in 70% ethanol (4°C) until embedding. Gill arches were obtained from the left side. Dissected tissues were routinely embedded in paraffin, serially sectioned in 5 μm increments, deparaffinised in xylene and rehydrated with a graded ethanol series prior to staining. All samples were processed at the same time. Sections were incubated for 1h at room temperature in primary antibody (1:400 Rat anti-Mouse PECAM/CD31:PBS; BD Pharmingen), rinsed briefly with PBS and incubated overnight with Alexa-Fluor-488-labelled secondary antibody (1:100 goat anti-rat IgG:PBS; Invitrogen) in PBS. Samples were then washed five times with PBS, mounted with Fluoromount (Sigma-Aldrich) and sealed with nail polish.

All slides were viewed on the same afternoon using a Nikon epifluorescent microscope (Nikon Eclipse 90i microscope, Nikon, Melville, NY, USA) and images were captured (×40) using identical camera settings (NIS Elements, Nikon). The background of each image was subtracted using the ‘rolling ball’ subtraction tool in ImageJ (US National Institutes of Health, Bethesda, MD, USA) and then the same intensity threshold was applied (Cooper et al., 2012). A box was then drawn around one of four body regions (the mouth, branchial cavity, dorsal epithelium or ventral epithelium) and the integrated density was calculated using ImageJ.

Contrary to our predictions, the immunohistochemical data suggested that air exposure increased blood flow to the bucco-opercular cavity of mangrove rivulus. To confirm that mangrove rivulus are able to use these surfaces for gas exchange while emersed, we performed a study to measure rates of aerial ventilation. Mangrove rivulus were acclimated to air as described above. After 1 h, 1 day or 7days of air exposure, fish were video recorded (Logitech Quickcam Pro, Fremont, CA, USA) under a dissecting microscope for 30 min and ventilatory activity (defined as an opening of the mouth and/or opercula) and body movements were recorded. Each fish was used only once.

One-way ANOVA and post hoc Holm–Sidak tests were used to test whether the duration of air exposure influenced ventilation rate. Two-way ANOVA with post hoc Holm–Sidak tests were used to determine the overall effects of respiratory medium and O₂ availability on [Hb], Hct and RBC counts. Three-way ANOVAs with post hoc Holm–Sidak tests were used to determine the overall effects of respiratory medium, O₂ availability and CO₂ on P₅₀ and the magnitude of the Root effect. A three-way ANOVA was also used to determine the effect of respiratory medium, O₂ availability and body region on angiogenesis (CD31 fluorescence). Data were log transformed when necessary to meet the assumptions of normal distribution and equal variance. SigmaPlot 11 (Systat Software, San Jose, CA, USA) was used for all analyses (critical α=0.05). Throughout the text values are given as means ± s.e.m.

The authors with to thank Dr D. Fudge for use of the microscope and hemocytometer and Dr G. Tattersall for help optimizing the P_wee50 equipment. We also thank Drs G. Scott, K. Gamperl and K. Gilmour for thoughtful discussions regarding the data. Thank you to B. Frank, M. Cornish, M. Davies and several volunteers for help with fish maintenance.