Aquatic surface respiration and swimming behaviour in adult and developing zebrafish exposed to hypoxia

The regulation of O₂ uptake, and its matching with metabolic requirements, are critical factors in the survival of all vertebrates. For most aquatic organisms that are incapable of using air as a respiratory medium, there is a reliance on gills for gas exchange (Randall and Daxboeck, 1984; Evans et al., 2005). In fish, environmental hypoxia produces reflex physiological responses, such as hyperventilation, bradycardia and changes in vascular resistance (Randall and Shelton, 1963; Holeton and Randall, 1967; Perry et al., 2009), as well as behavioural changes to decrease the physiological and biochemical demands imposed by hypoxia (Dalla Via et al., 1998; Domenici et al., 2000; Sloman et al., 2006; Chapman and McKenzie, 2009). One such behaviour is aquatic surface respiration, or ASR (Sloman et al., 2008; Chapman and McKenzie, 2009; Perry et al., 2009; Taylor et al., 2010; Richards, 2011). ASR involves ventilation at the air–water interface, where equilibration with the atmosphere raises the partial pressure of O₂ (P_O2) above that of the rest of the water column (Kramer and Mehegan, 1981). ASR is typically initiated near the P_O2 of peak gill ventilation (reviewed by Chapman and McKenzie, 2009; Perry et al., 2009), and the cumulative time spent in ASR increases rapidly as P_O2 decreases (Kramer and Mehegan, 1981; Kramer and McClure, 1982). Once at the air–water interface, fish actively skim the surface to ventilate their gills whilst maintaining a posture in which the dorsal aspect of the head lies just at or below the surface, with the body at a slight angle (Kramer and McClure, 1982; Timmerman and Chapman, 2004; Perry et al., 2009).

Performance of ASR under conditions of hypoxia effectively maintains or decreases gill ventilation rate (McNeil and Closs, 2007), increases arterial blood O₂ content (Burggren, 1982), and increases survival (Kramer and Mehegan, 1981; Kramer and McClure, 1982; Perry et al., 2009). The behaviour is believed to have evolved as an adaptation for surviving environmental hypoxia, and is thought to be an evolutionary precursor to aerial respiration in vertebrates (Kramer and McClure, 1982; Gee and Gee, 1995; Shingles et al., 2005). Despite extensive documentation of ASR in the literature, very little attention has been placed on the control of this behaviour. Specifically, it is not adequately understood how ASR is initiated by hypoxia, and the underlying neurochemical pathways associated with this response have not been sufficiently addressed.

In the tambaqui (Colossoma macropomum), ASR was abolished by bilateral section of the mandibular branch of the trigeminal nerve (cranial nerve V, which innervates the orobranchial region), whilst section of the glossopharyngeal (IX) or vagus (X) nerves to the gills had no effect (Sundin et al., 2000; Florindo et al., 2006); and in the flathead grey mullet (Mugil cephalus), application of sodium cyanide, a potent metabolic inhibitor and chemoreceptor stimulant, induced ASR (Shingles et al., 2005). These studies suggest that, like hyperventilation and bradycardia, ASR may be a reflex response to hypoxia driven by peripheral O₂ chemoreceptors.

O₂-chemosensitive neuroepithelial cells (NECs) of the gill filaments have been characterized in three fish species, including the model vertebrate, the zebrafish Danio rerio (Jonz et al., 2004; Burleson et al., 2006; Zachar and Jonz, 2012). NECs receive innervation primarily from cranial nerves IX and X (Dunel-Erb et al., 1982; Nilsson, 1984; Jonz and Nurse, 2003, 2005). Hypoxia stimulates zebrafish NECs via inhibition of membrane-bound K⁺ channels and subsequent membrane depolarization, and this presumably leads to the generation of cardioventilatory reflexes (Jonz et al., 2004). Interestingly, adult zebrafish immersed in quinidine, a blocker of O₂-sensitive K⁺ channels in NECs (Jonz et al., 2004), induced hyperventilation and ASR (Jonz and Nurse, 2005).

Gill NECs retain serotonin (5-HT) as well as a number of other neurochemicals (reviewed in Jonz, 2014). These cells first appear in the gill filaments of zebrafish larvae between 5 and 7 days postfertilization (d.p.f.) and are preceded by a cutaneous population of NEC-like cells in embryos (Jonz and Nurse, 2005; Coccimiglio and Jonz, 2012). In addition, cells expressing an acetylcholine (ACh) transporter were reported in the gills of zebrafish (Shakarchi et al., 2013) and the amphibious fish Kryptolebias marmoratus (Regan et al., 2011). Both 5-HT and ACh have been implicated in the chemoreceptor control of ventilatory reflexes in developing and adult fish (Burleson and Milsom, 1995a,b; Shakarchi et al., 2013; Jonz et al., 2015).

Given the advantages of understanding the biological features of O₂ sensing and respiratory development in zebrafish, we designed experiments to characterize ASR in this animal model, and to begin studies aimed at defining the control of this behaviour. We hypothesized that at P_O2 levels near maximal gill ventilation, zebrafish would begin to perform ASR; and that this behavioural response would be subject to modulation by factors that affect O₂ chemoreception, such as progressive changes in water P_O2, acclimation to hypoxia, development, and administration of 5-HT or ACh.

Adult zebrafish exposed to water of progressively decreasing P_O2 first display an increase in ventilation frequency at ∼110 mmHg (Vulesevic and Perry, 2006; Vulesevic et al., 2006). We observed similar hyperventilatory responses to mild hypoxia in the test chamber in the present study (data not shown). However, when zebrafish were exposed to severe levels of hypoxia, aquatic surface respiration (ASR) was observed in addition to hyperventilation. Once at the surface, the zebrafish adopted a position in which the body was at an angle with the air–water interface, with the top of the head in contact with the water surface and the mouth just under the surface layer. In some instances, zebrafish would perform ASR while hyperventilating and remaining stationary at the surface. As P_O2 decreased further, fish became more active and would skim the surface layer.

Fig. 1 illustrates the results from tracking experiments in the test chamber at three levels of P_O2, where vertical position is plotted against time. These experiments were performed on 11 adults at each P_O2 level. When zebrafish were exposed to normoxic levels of O₂ (P_O2∼158 mmHg; Fig. 1A), movements to the surface were uncommon and zebrafish were more likely to explore the lower region of the chamber. By contrast, zebrafish exposed to a P_O2 of 23 mmHg or 15 mmHg displayed regular and sustained trips to the surface (i.e. within 10 mm of the air–water interface) associated with ASR behaviour, and swimming activity was reduced elsewhere in the chamber (Fig. 1B,C). Raw data from tracking experiments indicate the preferred regions of the test chamber for zebrafish during the trials at these levels of P_O2 (Fig. 1D–F).

During 5 min trials, in which adult zebrafish were subjected to acute progressive hypoxia (P_O2 from 158 to 15 mmHg), zebrafish generally spent progressively more time engaged in ASR as water P_O2 was decreased (Fig. 2). From the mean of 11 individuals (Fig. 2C), it was evident that over a range of P_O2 values zebrafish displayed no observable ASR activity until the P_O2 reached ∼50 mmHg, where 2 out of 11 fish participated in ASR and cumulative time began to increase (Fig. 2C, filled circles). A significant increase in cumulative time in ASR above that of normoxic zebrafish was apparent at P_O2 values of 30 mmHg and below (P<0.001, repeated-measures ANOVA, Bonferroni post hoc test, N=11). We therefore estimate a threshold P_O2 for ASR behaviour at 30 mmHg. Additional analysis by two-way ANOVA indicated significant effects of acclimation to hypoxia on the ASR response (Fig. 2C; P<0.001, N=11). Acclimation to hypoxia (P_O2=30–45 mmHg) for 7 days significantly reduced cumulative time in ASR during acute hypoxic exposure, compared with unacclimated zebrafish, at P_O2 values of 25 mmHg and below. In addition, the line of best fit for acclimated zebrafish was shifted to the left by ∼7 mmHg (at 70.5 s), indicating a potential decrease in threshold P_O2 compared with unacclimated controls.

A biphasic change in swimming behaviour was associated with an increased ASR response with decreasing P_O2. Adult zebrafish exposed to progressive, acute hypoxia displayed only a marginal decline in swimming behaviour from 110 to 32 mmHg, but minimum values for total distance travelled (4.8±1.1 m), velocity (16.1±5.6 mm s⁻¹) and acceleration (42.4±13.3 mm s⁻²) were recorded at a P_O2 of 24 mmHg (Fig. 3A,B), just below the ASR threshold of 30 mmHg. When the P_O2 was reduced below 24 mmHg, zebrafish then began to display an increase in all three parameters up to 8 mmHg, where maximum distance (10.9±1.1 m), velocity (37.4±3.7 mm s⁻¹) and acceleration (108.1±15.2 mm s⁻²) were significantly greater compared with minimum values at 24 mmHg (paired t-test; P<0.05; N=6).

Through development of our procedures for intraperitoneal drug injection and behavioural assay, we found that injections of phosphate-buffered solution (PBS, as a sham control) resulted in a decrease in the ability of adults to perform ASR during exposure to acute hypoxia. Adults that received intraperitoneal injections of PBS achieved a lower cumulative time spent in ASR of 10.6±2.8 s (N=38) during a 5 min trial exposed to a P_O2 of 15 mmHg, compared with 179.9±24.7 s at the same P_O2 in uninjected zebrafish (see Fig. 2C). Despite the attenuating effects of the injection regime on ASR behaviour, it was evident that injection of 10 μg g⁻¹ 5-HT or 20 μg g⁻¹ ACh significantly increased cumulative time engaged in ASR during hypoxia by 5.5 and 4.9 times, respectively, compared with PBS-injected controls; and by 9.8 and 3.3 times, compared with lower concentrations of the same drug (Fig. 4; Kruskal–Wallis, Dunn's post hoc test; P<0.05). Neither 5 μg g⁻¹ACh nor 10 μg g⁻¹ 5-HT had a significant effect compared with PBS-injected controls.

As illustrated in Fig. 5, an increase in the number of frames s⁻¹ in which younger larvae were first observed near the surface increased at higher O₂ tensions compared with larvae at 14–21 dpf. Larvae displayed ASR behaviour at 32 mmHg at a mean frequency of 1.9±0.6 frames s⁻¹ at 5 dpf and 1.3±0.4 frames s⁻¹ at 7 dpf. At 14 dpf, the frequency of ASR events at 32 mmHg had decreased to 0.14±0.1 frames s⁻¹. At all stages of development tested, the frequency of larvae observed at the air–water interface was significantly greater only at 8 and 16 mmHg compared with the number of frames s⁻¹ at higher levels of P_O2 (ANOVA; Bonferroni post hoc test; P<0.05). We therefore estimate an ASR threshold of 16 mmHg for zebrafish larvae at all stages tested, although this may be an underestimate in younger larvae.

Developing zebrafish generally displayed a progressive reduction in swimming behaviour as P_O2 was decreased (Fig. 6). Significant reductions in total distance travelled, velocity and acceleration were observed at 5, 18 and 21 dpf; and these first occurred at a P_O2 of 16, 32 and 47 mmHg, respectively (P<0.05; ANOVA; Bonferroni post hoc test). Though not as prominent as was observed in adults, parameters of swimming behaviour began to increase below the estimated ASR threshold of 16 mmHg at 5, 14 and 18 dpf.

Given the prominent effects of 5-HT and ACh on increasing ASR behaviour in adults upon exposure to hypoxia, the same drugs were tested in behavioural assays of larvae. We recorded frames having a range of only 0–2 ASR events in response to hypoxia (as a control) between 7 and 25 dpf. The number of frames containing these events is plotted in Fig. 7. Note that in groups where few ASR events are present, the data distribution is skewed to the left. ASR events in response to hypoxia in 40 dpf larvae were more frequent. When larvae from 14 to 25 dpf were first immersed in 10 μmol l⁻¹ 5-HT or ACh and then exposed to hypoxia, the number of frames having 3 ASR events recorded increased, and the number of ASR events increased to 4 in 40 dpf larvae previously immersed in 5-HT (Fig. 7). Mean ASR events frame⁻¹ were significantly greater in hypoxia for 18 dpf (1.29±0.15 frame⁻¹), 25 dpf (1.63±0.19 frame⁻¹) and 40 dpf (1.9±0.19 frame⁻¹) larvae pre-exposed to 5-HT, compared with hypoxia alone (0.63±0.11, 0.37±0.09 and 0.79±0.15 frame⁻¹, respectively); whilst pre-exposure to ACh had a significant effect only in 18 dpf larvae (1.24±0.14 frame⁻¹) compared with hypoxia alone (rank sum Kruskal–Wallis and Dunn's multiple comparison test; P<0.05).

The present study has demonstrated that zebrafish exhibit ASR behaviour in response to severe hypoxic stress as early as 5 days postfertilization, and that ASR is increased with decreasing water P_O2. We have further shown that the threshold of the ASR response is dependent upon development and previous exposure to hypoxia, and that ASR can be enhanced by exogenous application of neurochemicals known to activate O₂-chemosensory pathways.

Behavioural changes performed by fish during environmental hypoxia play an integral role in survival at times when O₂ is critically limited and changes in gill ventilation can no longer support needs for increased O₂ uptake. Among these behavioural changes are: transitions from water to air breathing, emersion from water and ASR (Graham, 1997; Ong et al., 2007; Chapman and McKenzie, 2009; Perry et al., 2009; Taylor et al., 2010; Richards, 2011; Milsom, 2012). In adult zebrafish, threshold P_O2 for the ASR response, i.e. the P_O2 at which a significant increase in ASR behaviour first occurred, was 30 mmHg. Zebrafish exposed to hypoxia begin to hyperventilate at a P_O2 of approximately 110 mmHg and reach maximal gill ventilation near 40 mmHg (Vulesevic and Perry, 2006; Vulesevic et al., 2006). Therefore, in agreement with previous reports of ASR behaviour in different species (see table 5.5 in Perry et al., 2009), zebrafish engage in ASR at or below the P_O2 of maximal gill ventilation. Moreover, Chapman and McKenzie (2009) summarized the ASR thresholds of 63 species and suggested a negative correlation with hypoxia tolerance. McNeil and Closs (2007) determined ASR thresholds of ∼14 mmHg for goldfish (Carassius auratus) and common carp (Cyprinus carpio) – two cyprinid species related to zebrafish. Both goldfish and carp exhibit high levels of tolerance to hypoxia (Stecyk and Farrell, 2002; Lutz and Nilsson, 2004; Bickler and Buck, 2007), and this may account for the relatively lower ASR thresholds observed in these species compared with zebrafish.

The effects of hypoxia on spontaneous swimming activity have been studied in a number of fish species and appear to be both variable and species dependent (reviewed by Chapman and McKenzie, 2009). Generally, if swimming activity is reduced during hypoxia at levels above the critical P_O2 (P_crit), this may be regarded as a strategy to conserve energy in an environment of potentially limited O₂ availability. By contrast, the intense agitation of fish in severe hypoxia below P_crit has been interpreted as an acute escape or avoidance response (Bejda et al., 1987; Randall et al., 1992; van Raaij et al., 1996). In adult zebrafish, we measured a gradual decline in total distance travelled, velocity and acceleration as water P_O2 fell. These parameters reached minimum values at a P_O2 at 24 mmHg, and then increased as P_O2 was further reduced. A similar biphasic change in activity associated with decreasing P_O2 was observed in sole (Solea solea; Dalla Via et al., 1998; McKenzie et al., 2008). Reports for P_crit in adult zebrafish are ∼20 mmHg (Barrionuevo and Burggren, 1999; Barrionuevo et al., 2010), although a more recent study presented a considerably higher value that varied between males and females (Robertson et al., 2014). We therefore interpret the biphasic trends in swimming behaviour in zebrafish exposed to progressive hypoxia as a strategy to reduce activity and conserve energy at water P_O2 levels above P_crit; but as P_O2 falls below P_crit, zebrafish may become agitated and initiate a response in an effort to escape hypoxia. An alternative interpretation of increased swimming activity under these conditions of severe hypoxia may be to increase gas exchange at the surface, as in the case of ram-assisted ASR (Chapman et al., 1994). These data may match our observations that zebrafish first remained stationary when at the air–water interface during the early stages of ASR, but increased swimming activity as P_O2 continued to decline. However, when performing ASR at any level of P_O2, zebrafish did not remain at the surface of the chamber but would alternate between states of ASR and brief bouts of rapid swimming behaviour throughout the chamber.

The development of ASR behaviour in fish is not well defined in the literature. Sloman et al. (2008) reported that juvenile sculpins (Oligocottus maculosus) performed ASR, and this occurred at lower O₂ tensions compared with adults; and in juvenile guppies (Poecilia reticulata), ASR behaviour increased growth rate during hypoxia (Weber and Kramer, 1983). But the progression in ASR throughout early life in fish has not previously been correlated with events of respiratory development within the same species.

Zebrafish embryos respond to hypoxia by first increasing the frequency of pectoral fin and whole body movements as early as 2 dpf in order to increase flow across respiratory surfaces and improve diffusion; they develop a maximal ventilatory response to hypoxia between 5 and 7 dpf that coincides with innervation of gill NECs (Jonz and Nurse, 2005; Coccimiglio and Jonz, 2012), and this response develops an increasing reliance on central N-methyl-d-aspartate (NMDA) receptors at 8 dpf (Turesson et al., 2006). In the present study, developing zebrafish initiated ASR behaviour at lower O₂ tensions than adults. We have estimated an ASR threshold of 16 mmHg in zebrafish larvae, although our observations (Fig. 5) indicate that this may be an underestimate in larvae younger than 14 dpf. P_crit for developing zebrafish is ∼75 mmHg in embryos and is later reduced to ∼70 mmHg at 20 dpf (Barrionuevo and Burggren, 1999; Barrionuevo et al., 2010). Thus, unlike in adults, the ASR threshold in larvae is markedly lower than P_crit. This suggests that larvae utilize a different strategy when confronted with severe hypoxia.

Larvae reduced swimming behaviour with hypoxia, as did adults, but statistically significant reductions in total distance travelled, velocity and acceleration were first observed at a P_O2 of 32 and 47 mmHg for 18 and 21 dpf larvae, respectively. This indicates that larvae reduce swimming behaviour at higher O₂ tensions compared with adults, and thus may rely primarily upon conserving energy and upon gas exchange via simple diffusion during severe hypoxia. During early developmental stages, the gills may not be fully developed and so zebrafish larvae use both cutaneous and gill respiration to satisfy metabolic needs for O₂ (Rombough, 2002). In addition, a model has predicted that zebrafish can survive by simple diffusion in air-saturated water without the need of a circulatory system until 42 dpf (Rombough and Drader, 2009). ASR may then become important as a survival strategy in larvae at O₂ tensions closer to anoxia, where reduced activity no longer provides any benefit and an escape response is elicited. Indeed, we found that at O₂ tensions below the ASR threshold, at some developmental stages, parameters of swimming behaviour began to increase. Consistent with these results, silverside larvae (Menidia beryllina) between approximately 6 and 9 dpf increased swimming activity and moved in an upward direction to avoid hypoxia (Weltzien et al., 1999).

We show that zebrafish displayed ASR as early as 5 dpf. Zebrafish larvae from the time of hatching at 2 dpf have an average density above that of the surrounding water, and so must actively swim to the surface using rapid tail movements and by adhering intermittently to vertical surfaces, in order to take in a small bubble of air and inflate the swim bladder by the end of 4 dpf (Lindsey et al., 2010). This ‘swim-up’ behaviour is one of the first coordinated behaviours in developing larvae, and must occur in order for larvae to achieve buoyancy control (Spence et al., 2008). We suggest that development of ASR in zebrafish at about 5 dpf is limited primarily by coordination of swim-up behaviour and swim bladder inflation. Moreover, swim-up behaviour may even be a rudimentary form of ASR, occurring over a longer period of time as a result of the inability of newly hatched larvae to rise rapidly to the surface.

An element of some controversy in defining ASR behaviour in fish is that of its control – whether ASR is a reflex initiated by peripheral sensory receptors, or if it is a more complex behaviour. Although ASR may be initiated or dependent upon sensory receptors, it appears at least to have a strong behavioural component that can be modulated by higher brain centres. Shingles et al. (2005) found that even the perceived threat of aerial predation delayed the onset of ASR in the flathead grey mullet (Mugil cephalus) and reduced the proportion of fish performing the behaviour. Yet another possibility is that severe hypoxia may be perceived by fish as a noxious or painful stimulus. Some authors have identified an aversive behaviour that may lead fish away from regions of severe hypoxia and toward the surface (Petersen and Petersen, 1990; Weltzien et al., 1999; Poulsen et al., 2011; Cook et al., 2013). In the present study, ASR behaviour in zebrafish was punctuated with brief trips to the bottom of the chamber before returning rapidly to the surface. This may have been a searching behaviour directed by a repeated aversive response away from hypoxia. Nociceptors have been identified on the head of rainbow trout (Oncorhynchus mykiss) and mediate responses to noxious stimuli, including high CO₂ and acidity (Sneddon, 2003; Mettam et al., 2012).

Evidence in favour of ASR as a reflex response to hypoxia comes primarily from four studies. Application of sodium cyanide, a potent metabolic inhibitor and chemoreceptor stimulant, induced ASR in the mullet (Shingles et al., 2005); and in adult zebrafish immersed in quinidine, a drug that targets O₂-sensitive K⁺ channels in gill NECs (Jonz et al., 2004), hyperventilation and ASR were induced (Jonz and Nurse, 2005). In the tambaqui (Colossoma macropomum), ASR was abolished by section of the mandibular branch of the trigeminal nerve (cranial nerve V), which innervates the orobranchial region (Sundin et al., 2000; Florindo et al., 2006). The authors therefore suggested the orobranchial cavity as a site for O₂ chemoreceptors that produced the response. However, the trigeminal nerve is also an important source of innervation to cutaneous receptors of the somatic sensory system (Nilsson, 1984), and fibres of the mandibular branch innervate nociceptors of the head and lower jaw (Sneddon, 2003; Mettam et al., 2012).

The present study has demonstrated that administration of 5-HT and ACh increased ASR responses to severe hypoxia in zebrafish adults and larvae. Both 5-HT and ACh have been implicated in chemoreceptor control of ventilatory reflexes in developing and adult fish (Burleson and Milsom, 1995a,b; Shakarchi et al., 2013; Jonz et al., 2015), as they have in mammals (Shirahata et al., 2007; Nurse, 2010). In addition, O₂-chemoreceptive gill NECs in zebrafish retain 5-HT (Jonz and Nurse, 2003), and cells expressing an ACh transporter were reported in the gills of zebrafish (Shakarchi et al., 2013) and the amphibious fish, Kryptolebias marmoratus (Regan et al., 2011). In zebrafish larvae, 5-HT₂, 5-HT₃, and nicotinic receptors of the gills appear to play a role in initiation of the hyperventilatory response to hypoxia (Shakarchi et al., 2013; Jonz et al., 2015) and may be involved in initiating the ASR response. Sensory Merkel-like cells containing 5-HT were described in the oropharyngeal epithelium of zebrafish (Zachar and Jonz, 2012), but these cells did not display any morphological changes following long-term exposure to hypoxia (M.G.J., unpublished observations), as did gill NECs (Jonz et al., 2004), suggesting that the former are not O₂ sensitive. Together, these data may suggest that ASR is susceptible to control via O₂ chemoreceptors of the gills. In tambaqui, however, evidence suggests the contrary. Sectioning of the glossopharyngeal and vagus nerves, which supply sensory innervation to gill NECs, did not abolish ASR (Sundin et al., 2000; Florindo et al., 2006); although Florindo et al. (2006) did observe an attenuation of ASR response frequency that was not significantly different from controls.

From the present experiments, adults acclimated to chronic hypoxia for 7 days displayed a lower ASR threshold and spent less time at the surface during severe hypoxia. Other studies have likewise reported that ASR thresholds can be depressed by prior acclimation to hypoxia (Kramer and Mehegan, 1981; Timmerman and Chapman, 2004; Chapman and McKenzie, 2009). Previous studies in zebrafish have demonstrated that acclimation to hypoxia increases the size of O₂-sensitive NECs, and increases the number of nonserotonergic NECs, in the gills (Jonz et al., 2004). Whilst changes in chemoreceptor morphology and proliferation may contribute to the enhanced ventilatory responses observed in fish following acclimation to hypoxia for 1 to 3 weeks (Kerstens et al., 1979; Burleson et al., 2002), as it does in ventilatory acclimatization to hypoxia in mammals (Bisgard, 2000; Powell et al., 1998; Porteus et al., 2011; Kumar and Prabhakar, 2012), it remains to be determined if the blunted ASR response is due directly to these chemoreceptor changes, or due to a putative increased hyperventilatory response resulting from the acclimation period that may have mitigated the need for ASR and effectively reduced the ASR threshold. Confirmation of this will require studies directed at determining the ventilatory response to acute hypoxia in zebrafish after a similar period of acclimation, especially since time-dependent changes in the ventilatory response to hypoxia appear to occur in fish (Porteus et al., 2011). One report has indicated that acclimation of zebrafish to hypoxia for 28 days reduced basal ventilation without a significant effect on the acute response (Vulesevic et al., 2006).

As noted above, ASR behaviour was first observed in zebrafish larvae at 5 dpf. Unlike in adults, at this early developmental stage O₂-sensitive gill chemoreceptors are not yet fully innervated (Jonz and Nurse, 2005) and so are not likely to participate in reflex responses. However, cutaneous NECs that receive innervation, retain 5-HT and exhibit O₂-sensitive properties appear in zebrafish as early as 1 dpf (Coccimiglio and Jonz, 2012). Although these cells may be candidate chemoreceptors for initiation of ASR in early larvae, we were not able to accordingly modify ASR responses in larvae by immersion in 5-HT or ACh before 18 dpf. The reason for this is unclear, since serotonergic receptors participate in hypoxic hyperventilation as early as 7–10 dpf, and cholinergic receptors as early as 12 dpf (Shakarchi et al., 2013). Marginal increases in the frequency of ASR events observed with 5-HT at 7 and 14 dpf suggest that a higher sample size in our experiments might have revealed a more prominent change in behaviour. In addition, the loss of response to ACh in 25 and 40 dpf larvae may be due to the increased size in larvae during these later juvenile stages associated with decreased drug diffusion to tissues during immersion experiments. Indeed, only intraperitoneal injection of 5-HT or ACh enhanced ASR responses in adults. Immersion of adults in these drugs had no effect on ASR.

It is recognized that the transition from water to air breathing in vertebrates was an important adaptive change that led away from the diffuse distribution of peripheral O₂ chemoreceptors in the gills of anamniotes and towards the evolution of the carotid body, the primary peripheral chemoreceptive organ that controls cardiorespiratory reflexes in mammals (Milsom and Burleson, 2007). In addition, it has been proposed by some authors that ASR behaviour in fish may have been an adaptive response that facilitated the evolution of air breathing in vertebrates and the transition to terrestrial life (Kramer and McClure, 1982; Gee and Gee, 1995; Shingles et al., 2005). Considering these two hypotheses, and if ASR is dependent upon input from peripheral chemoreceptors as suggested in this and previous studies (Sundin et al., 2000; Florindo et al., 2006; Jonz and Nurse, 2005; Shingles et al., 2005), it would follow that air breathing arose as a result of evolutionary change enhanced by positive feedback. Positive feedback can be generally defined as the evolution of phenotypes in response to selection that serve to further reinforce or strengthen selection for more extreme phenotypes (Robertson, 1991; Crespi, 2004). In this scenario, movement to the surface resulting from chemoreceptor input would have, in turn, modified chemoreceptor distribution so as to enhance or favour further transitions toward aerial respiration. Observations of chemoreceptor distribution in some aquatic vertebrates may support the positive-feedback hypothesis. In amphibians, NECs of the gills are present in aquatic larval and neotenic forms (Jonz and Nurse, 2006; Saltys et al., 2006), but a carotid labyrinth develops after metamorphosis in adults as the primary O₂-chemoreceptive organ, when the lungs become functional (Kusakabe, 2002). Although variable between species, in some air-breathing fish specific chemoreceptor loci may stimulate air breathing while others do not (Milsom, 2012). In addition, air-breathing organs in fish are endowed with chemoreceptor-like cells (Zaccone et al., 1997, 2006), although there is currently no physiological evidence that these cells participate in O₂ sensing. Direct confirmation of a link between chemoreceptors and ASR behaviour will strengthen the positive-feedback hypothesis. It will help us define and understand the diversity of O₂-sensing mechanisms and the evolution of air breathing in vertebrates.

Wild-type zebrafish, Danio rerio (Hamilton 1822), were obtained from a local commercial supplier (Mirdo Importations, Montreal, Canada) and held in a closed recirculated facility. Animals were maintained at 28.5°C on a 14 h:10 h light:dark cycle (Westerfield, 2000). All animal handling and care was carried out in accordance with institutional guidelines, which adhere to those of the Canadian Council on Animal Care (CCAC). Embryos were bred in accordance with methods described in Westerfield (2000). Embryos were stored in an embryo medium and incubated at 28.5°C. Embryo medium consisted of: 5 mmol l⁻¹ NaCl, 0.17 mmol l⁻¹ KCl, 0.33 mmol l⁻¹ CaCl₂, 0.33 mmol l⁻¹ MgSO₄ at pH 7.8. After hatching, larvae were transferred to 1 litre aquaria filled with dechlorinated water at 28.5°C. The water was replaced every other day and treated with 0.01% Methylene Blue to reduce fungal or bacterial growth.

The ASR response, and the relationship between water P_O2 and ASR behaviour, was first established in zebrafish by exposing adults to acute, progressive hypoxia. We identified ASR behaviour based on previous descriptions in other species (Kramer and Mehegan, 1981; Kramer and McClure, 1982). Zebrafish were engaged in ASR behaviour when the top of the head was near the water surface, with the mouth just below the air–water interface, and the body oriented at a slight upward angle. For each level of P_O2, 4 groups of 2–4 adult zebrafish (N=11) were placed in a 1 litre test chamber, with dimensions 80 mm high by 150 mm wide, marked with gradations on the outside that measured distance (in mm) from the bottom to the air–water interface, or surface. 100% N₂ was delivered to the chamber via polyethylene tubing and a porous stone until the desired water P_O2 was achieved. Water P_O2 ranged from normoxic values of ∼158 mmHg to as low as 8 mmHg and were verified before and after each trial with an O₂ meter (Model 550A, YSI, Yellow Springs, OH, USA). Once water P_O2 had stabilized, zebrafish were introduced into the test chamber and given 2 min to recover from handling. At normoxic levels of P_O2, this was sufficient time for all fish to adopt normal swimming and breathing patterns. A 5 min trial was then performed, after which zebrafish were transferred to a recovery tank for an additional 5 min for recovery. For trials characterizing behaviour over a range of P_O2 levels, a group of zebrafish was reintroduced to the test chamber after recovery and an additional trial was performed at a lower P_O2. In this manner, zebrafish were exposed to successive bouts of progressively more severe acute hypoxia.

Adult zebrafish were video-recorded using a digital camera (SD780 IS, Canon, Tokyo, Japan), and their swimming behaviour was analyzed using Videopoint video-tracking software (Lenox Softworks, Lenox, MA, USA) at a sampling rate of 1 frame every 200 ms (5 Hz). The distance between each position was calibrated to the gradations marked directly on the outside of the test chamber to obtain the vertical distance (in mm) from the chamber bottom versus time, and vertical distance versus horizontal position (Fig. 1). Video recordings were analyzed post hoc with a desktop computer and high-resolution LCD display to determine the effects on ASR behaviour. Cumulative time engaged in ASR was obtained by reviewing videos and recording the total time (in s) each adult spent near the air–water interface (Figs 2 and 4). Additional experiments were performed to study parameters of swimming behaviour. In these experiments, and in order to facilitate video tracking, zebrafish were placed individually into the test chamber. During 5 min trials adults (N=6) were tracked, as above, and the following behavioural parameters were extracted using MatLab software (MathWorks, MA, USA): total distance travelled (in m), velocity (in mm s⁻¹) and acceleration (in mm s⁻², e.g. Fig. 3). Acceleration is an indicator of the rate of change in velocity and was measured because our early observations indicated that zebrafish continuously modify their velocity while swimming.

The behaviour of larvae at different developmental stages and P_O2 was observed. A small test chamber was constructed in order to assay one larva at a time at pre-determined levels of hypoxia. The chamber (30×25×5 mm) was placed inside a transparent 100 ml water bath. The chamber was designed to produce a vertical water column of sufficient height and width for observation, while at the same time minimizing thickness to prevent zebrafish from moving out of focus during video recording. N₂ was bubbled into the bath until the desired P_O2 in the chamber was achieved. Water flow was continuous between the chamber and the bath, and thus allowed for accurate control of chamber P_O2 while minimizing turbulence. The P_O2 of the chamber was measured directly before and after each trial using an O₂ meter, as above.

A single larva was introduced into the test chamber using a pipette and given 2 min to recover from handling stress. Throughout the subsequent 3 min trial, movement of the larva was video recorded (Sony, Handycam, Model HDR-XR350V, Tokyo, Japan). These assays were performed at P_O2 levels of 8, 16, 32, 47, 63 and 158 mmHg. Larvae of ages 5, 7, 14, 18, and 21 days postfertilization (dpf) were tested. Younger larvae lack the ability to make rapid vertical changes in the water column (Lindsey et al., 2010) and so were not tested for ASR. Each larva was tested individually at a single P_O2 value. The videos were reviewed, and the position of larvae was noted in 200 ms intervals (i.e. for a 3 min trial, 900 frames in total). ASR was defined as the presence of a larva within 3 mm of the air–water interface in a given frame. The total number of frames in which the larva was performing ASR was noted, and the data are expressed as the number of frames s⁻¹ for each developmental stage across a range of water P_O2 (Fig. 5). These data were further analyzed as a measure of swimming behaviour using tracking software (Videopoint, Lenox, MA, USA). The x and y coordinates obtained from Videopoint were exported into MatLab (MathWorks, MA, USA) in order to obtain the total distance travelled (m), velocity (mm s⁻¹) and acceleration (mm s⁻²) of each individual larva (Fig. 6).

Adult zebrafish were placed in a 2.5 litre aquarium and acclimated for 7 days within a P_O2 range of 30 to 45 mmHg. Water P_O2 was lowered to the desired levels by introducing a mixture of O₂ and N₂ (Pegas 4000 MF gas mixer, Columbus Instruments, Columbus, OH, USA) delivered through a porous stone. Control zebrafish were simultaneously maintained in an identical aquarium at a normoxic level (∼158 mmHg) for the same period of time. Both aquaria were maintained at 28.5°C. Water P_O2 was measured as above. Both aquaria were covered with a lid, and a 50% water change with dechlorinated water was performed each day to minimize potential changes in water quality. Acclimated and control adults were tested for acute responses to severe hypoxia as above.

In order to assess the potential role of O₂-chemoreceptive neuroepithelial cells (NECs) of the gill in initiation of the ASR response, zebrafish were treated with serotonin (5-hydroxytryptamine or 5-HT, cat. no. H9523) and acetylcholine (ACh, cat. no. A6625), both of which stimulate gill O₂ chemoreceptors (Burleson and Milsom, 1995a) and induce hyperventilation in fish, including zebrafish (Burleson and Milsom, 1995b; Shakarchi et al., 2013). Both chemicals were obtained from Sigma-Aldrich (Oakville, ON, Canada).

Adult zebrafish ranging from a body weight of 0.7 to 1.0 g were lightly anaesthetized with 0.25 mg ml⁻¹ MS 222 (Syndel Laboratories, Vancouver, BC, Canada) dissolved in dechlorinated system water. Fish were injected intraperitoneally (i.p.) with 5-HT (5 or 10 µg g⁻¹ body weight; N=12 or 17, respectively), ACh (10 or 20 µg g⁻¹; N=29 or 19, respectively) in phosphate-buffered solution (PBS), or PBS as a sham control (N=34). Injections were administered at the ventral aspect below the swim bladder. PBS contained the following: 137 mmol l⁻¹ NaCl, 15.2 mmol l⁻¹ Na₂HPO₄, 2.7 mmol l⁻¹ KCl, 1.5 mmol l⁻¹ KH₂PO₄ at pH 7.8 (Jonz et al., 2004). Preliminary experiments indicated that these drug concentrations were optimal in producing a measurable response with minimal mortality. A 10 µl syringe with a 33 gauge needle (Hamilton Company, Reno, NV, USA) was used to deliver a 5 µl g⁻¹ body weight volume of these drugs dissolved in PBS. Each adult zebrafish was given 4 injections over a 10 day period and returned to aquaria specific to each treatment group. Preliminary experiments indicated that single injections of these drugs had no observable effect on behaviour. At 48 h after the final injection, all fish were tested for their response to acute hypoxia, as described above.

Developing zebrafish (7–40 dpf) were pre-exposed to 10 µmol l⁻¹ 5-HT or 10 µmol l⁻¹ ACh dissolved in 10 ml dechlorinated system water in a 20 ml vial for 20 min. Controls were treated in the same way but were not exposed to any drug. Following pre-exposures, 10 larvae from each developmental stage were placed in the test chamber and given 2 min to recover. The effects of hypoxia (P_O2=25 mmHg) on ASR behaviour were recorded for each group and developmental stage by collecting an image (frame) every 10 s for 3 min (for a total of 38 frames). The experiment was repeated with an additional 10 larvae and data were pooled. The number of frames having 0, 1, 2, 3 or 4 ASR events was quantified (Fig. 7). An ASR event was defined as the presence of a larva within 3 mm of the air–water interface in the test chamber during exposure to hypoxia. The total number of events and mean±s.e.m. events per frame were recorded.

For data sets containing normal data, the paired t-test was used to compare means, or an analysis of variance (ANOVA) was performed to test for differences among multiple groups. In the latter case, where a statistical difference was detected, a post hoc Bonferroni test was used to indicate which groups were significantly different. For data that was not normally distributed, the rank sum Kruskal–Wallis and Dunn's multiple comparison test were used to test for differences between groups. For the data presented in Fig. 2C, cumulative time in ASR versus P_O2 was analyzed using a mixed model approach in which the repeated measures ANOVA test was used combined with blocking for groups using SAS (Statistical Analysis System Institute Inc., Cary, NC, USA). All other statistical analyses were performed using Prism v5.0 (GraphPad Software Inc., La Jolla, CA, USA). For cumulative time in ASR versus P_O2 (Fig. 2C), lines of best fit were produced following the equation: y=100/(1+10^logEC50−x). Note that in some figures, bars indicating s.e.m. were smaller than data points and are not visible.

The authors would like to thank Ginette Hupe, Isabelle Shank, and especially Imad Abdallah for their assistance with MatLab software.