In early larval stages, breathing is infrequent and irregular (Holeton, 1971; McDonald and McMahon, 1977; Jonz and Nurse, 2005). For example, breathing movements in zebrafish larvae post hatch are rare under normoxic conditions; the HVR is first observed at 3 dpf but even during hypoxia, the buccal movements are irregular until about 8 dpf (Jonz and Nurse, 2005). Qualitatively similar results were observed for rainbow trout and Arctic char, although the transition from irregular to regular breathing movements occurred over a longer time course, corresponding with their slower rate of development (Holeton, 1971; McDonald and McMahon, 1977). During these early developmental stages, the buccal/opercular movements are synchronized with rapid pectoral fin movements, a behaviour presumed to aid gas transfer across the skin by increasing water flow over the body (Rombough, 1988). The synchrony between gill ventilation and pectoral fin movements during hypoxia in zebrafish larvae was studied by Zimmer et al. (Zimmer et al., 2020). Their data demonstrated that the synchronous movements, which are obvious at 4 dpf, are absent by 21 dpf. It was suggested that the loss of synchrony reflects the transition from reliance on cutaneous gas transfer to reliance on branchial gas transfer. To our knowledge, the sensory mechanisms underlying the fin movements during hypoxia have not been investigated (see Zimmer et al., 2020).

The ventilatory responses of adult fish to hypoxia can be modified by the status of the environment during early development. For example, zebrafish embryos treated with hyperoxic water for 7 days (0–7 dpf) exhibited a blunted HVR when exposed as adults to acute hypoxia (Vulesevic and Perry, 2006). However, there was no effect of subjecting larvae to hypoxia or hypercapnia on the responses of adult fish to hypoxia. The plasticity exhibited by zebrafish exposed to hyperoxia during early development (Vulesevic and Perry, 2006) agrees with the results of studies on mammals showing that exposure to hyperoxia during a critical prenatal period may permanently impair the phrenic response to hypoxia (Ling et al., 1997; Donnelly, 2000; Carroll, 2003). The mechanisms responsible for blunting of the HVR in adult zebrafish subjected to hyperoxia as larvae are unknown but as in mammals, may reflect changes to the O₂ chemoreceptors or afferent neurons (Donnelly, 2000).

Not only can the HVR be shaped by developmental plasticity (Vulesevic and Perry, 2006), but it is possible that the HVR is modified epigenetically in offspring of parents that were exposed to hypoxia. Notably, zebrafish larvae of one (male) or both parents exposed to hypoxia exhibited significant differences in their hypoxia tolerance compared to offspring of parents that had never experienced hypoxia (Ho and Burggren, 2012; Ragsdale et al., 2022). Although no breathing data were obtained (Ho and Burggren, 2012; Ragsdale et al., 2022), it is tempting to speculate that the transgenerational changes in hypoxia tolerance are related to changes in the HVR. Assessing possible epigenetic effects on the HVR is an area that warrants further attention.

In naturally hypoxic aquatic habitats, a thin layer of water at the surface tends to be enriched with O₂ relative to the bulk water (Burggren, 1982). Fish exploit this microenvironment with a behavioural adaptation termed aquatic surface respiration (ASR) (Kramer and Mehegan, 1981). During ASR, fish rise to the surface and skim the uppermost layer to ventilate their gills with O₂-enriched water (Gee et al., 1978; Kramer and McClure, 1982). Aquatic surface respiration has been studied almost exclusively in adult fishes (Chapman and McKenzie, 2009) where it has been shown to increase blood oxygenation during severe hypoxia (Burggren, 1982) as well as survival (Kramer and McClure, 1982). To our knowledge, ASR has been examined rarely in juveniles (Weber and Kramer, 1983; Sloman et al., 2008) and in larvae of a single species, zebrafish (Abdallah et al., 2015b; Mandic et al., 2022), in which ASR increased as the severity of hypoxia was increased. In zebrafish, the age of onset of ASR during acute hypoxia is 5 dpf (Abdallah et al., 2015b). At earlier developmental stages, ASR may be constrained by the lack of swimbladder inflation that is needed to control buoyancy (Abdallah et al., 2015b). Similar to adult zebrafish, ASR improves survival during severe hypoxia in larvae older than 5 dpf (Mandic et al., 2022). Indeed, when exposed to a level of hypoxia of 16 mmHg, the survival of larvae (>5 dpf) performing ASR was similar to that of normoxic larvae at close to 100%. Larvae (>5 dpf) spent approximately 30% of their time engaged in ASR, which was similar to the situation in adults experiencing the same level of hypoxia (16 mmHg) (Mandic et al., 2022). The energetic costs associated with extended periods of ASR in zebrafish larvae are unknown but are likely to be low given their state of neutral buoyancy (Lindsey et al., 2010). Conversely, longer time spent at the surface is likely to increase the risk of predation.

The O₂ chemoreceptors in fishes are believed to be NECs that were first described by Dunel-Erb et al. (1982). In adult fish, they are found on the tips of gill filaments and in larvae, they are located on the skin prior to gill maturation (Jonz and Nurse, 2006; Coccimiglio and Jonz, 2012). NECs also may be present on cutaneous surfaces in adults of some species including the mangrove rivulus (Kryptolebias marmoratus) (Regan et al., 2011; Rossi et al., 2020) and giant mudskipper (Periophthalmodon schlosseri) (Zaccone et al., 2017).

Gill NECs are characterized by the presence of dense-cored vesicles containing serotonin (5-HT) and their innervation is derived from a plexus of nerve fibres (Dunel-Erb et al., 1982; Bailly et al., 1989; Bailly et al., 1992; Jonz and Nurse, 2003). Initial suggestions of gill NECs being O₂ chemoreceptors were based on their morphology, location and innervation patterns (Dunel-Erb et al., 1982; Bailly et al., 1992; Jonz and Nurse, 2003; Bailly, 2009). Substantial indirect evidence supports a role for the NECs as peripheral O₂ chemoreceptors, largely based on their morphological changes (hypertrophy, hyperplasia and increased numbers of neuron-like processes) during exposure to prolonged hypoxia (Jonz et al., 2004; Burleson et al., 2006; Regan et al., 2011; Shakarchi et al., 2013; Rossi et al., 2020; Pan et al., 2021a). Direct evidence that gill NECs are O₂ sensitive was provided by whole-cell patch-clamp recordings in vitro from NECs isolated from adult gill filaments. Specifically, patch clamp experiments demonstrated hypoxia-induced membrane depolarization in isolated gill NECs under current-clamp conditions in zebrafish (Jonz et al., 2004) and channel catfish (Ictalurus punctatus) (Burleson et al., 2006). Additional direct evidence was provided by measuring intracellular [Ca²⁺] ([Ca²⁺]_i) and synaptic vesicle activity in isolated goldfish (Carassius auratus) NECs (Zachar et al., 2017).

There is less evidence supporting the notion that skin NECs are O₂ chemoreceptors. In zebrafish, skin NECs exhibiting innervation are evident in embryos at 1 dpf, before the onset of behavioral responses to hypoxia (Coccimiglio and Jonz, 2012). A population of skin NECs also expresses synaptic vesicle protein, suggesting a capacity for the secretion of neurotransmitters (Coccimiglio and Jonz, 2012). However, all evidence of hypoxia sensitivity in skin NECs is indirect, based on their increasing density and/or size under hypoxic conditions (Regan et al., 2011; Coccimiglio and Jonz, 2012; Rossi et al., 2020), or elimination of the HVR following their partial chemical denervation (Coccimiglio and Jonz, 2012). In contrast, however, NEC development in larval mangrove rivulus is largely unaffected by environmental O₂ levels (Cochrane et al., 2021). Thus, the current evidence supports hypoxia sensitivity of isolated gill NECs in vitro, whereas direct evidence for hypoxia sensitivity in skin NECs is lacking. Additional research, especially in vivo, is required to establish both gill and skin NECs as piscine O₂ chemoreceptors.

The sensing of O₂ begins with the detection of PO₂ changes by an O₂ sensor within the chemoreceptor, which leads to the inhibition of K⁺ channels and subsequent membrane depolarization. Cytosolic Ca²⁺ levels increase, through either extracellular Ca²⁺ entry via voltage-dependant Ca²⁺ channels or the release of Ca²⁺ from intracellular stores, thereby facilitating vesicle fusion and the release of neurotransmitters. This neurosecretion activates the afferent nerve fibers innervating the chemoreceptors, resulting in the signal being transmitted to the central nervous system (CNS) where it is processed to elicit downstream responses (Gonzalez et al., 2010; Zachar and Jonz, 2012; Prabhakar, 2013; Lopez-Barneo et al., 2016; Ortega-Sáenz et al., 2020).

Regardless of vertebrate class, the molecular mechanisms underlying acute O₂ sensing remain elusive (Rakoczy and Wyatt, 2018). The earliest ideas for molecular O₂ sensing were based on mitochondrial inhibition. In this scheme, inhibition of the mitochondrial electron transport chain under hypoxic conditions leads to decreased ATP production that is sensed by ATP-sensitive K⁺ channels, resulting in depolarization and neurotransmitter release (Duchen and Biscoe, 1992a; Duchen and Biscoe, 1992b; Wyatt and Buckler, 2004; Varas et al., 2007). In addition, the decrease in ATP may cause AMP to rise, which could activate AMP-activated protein kinase (AMPK), further phosphorylating membrane ion channels and leading to depolarization; thus, AMPK is a candidate O₂ sensor (Evans, 2004). Furthermore, disruption of mitochondrial complex I on the electron transport chain could increase production of reactive oxygen species (ROS), changing the redox status of membrane ion channels and thus initiating excitation (Fernandez-Aguera et al., 2015; Gao et al., 2017). The gasotransmitters CO and H₂S also have been proposed as components of the molecular mechanism for O₂ sensing (Li et al., 2010; Buckler, 2012; Yuan et al., 2015). Lactate is one of the newest candidates to be proposed as a member of the O₂ sensing pathway. Chang et al. (2015) demonstrated that the olfactory receptor encoded by Olfr78 is sensitive to lactate, which may accumulate during hypoxia. Olfr78 knockout mice did not exhibit a HVR but responded normally to hypercapnia. However, the idea of lactate being involved in O₂ sensing is not universally accepted because the original results obtained using Olfr78 knockout mice were not fully reproducible (Chang et al., 2018; Torres-Torrelo et al., 2018). It is likely that additional molecular mechanisms for O₂ chemoreception will be proposed. Indeed, it is likely that ultimately multiple mechanisms will be identified for O₂ sensing within the chemoreceptors (Kumar and Bin-Jaliah, 2007).

In fishes, the molecular mechanisms of O₂ chemoreception have received less attention, and only H₂S and lactate have been examined in the context of cellular (NEC) O₂ sensing. In trout branchial tissues where gill NECs are located, genes for the H₂S-synthesizing enzymes, cystathionine ß-synthase and cystathionine γ-lyase, are expressed (Olson et al., 2008). Further, gill homogenates produce H₂S enzymatically, a process which is inhibited by high levels of O₂ (Olson et al., 2008). Behaviourally, intrabuccal injection of H₂S in anaesthetized trout produced increases in f _V and V_AMP similar to those of the HVR (Olson et al., 2008). Similar results were obtained in zebrafish larvae in which H₂S elicited hyperventilation that was blocked by preventing endogenous H₂S synthesis throughout the body, including in the cystathionine γ-lyase-containing skin NECs (Porteus et al., 2014). Additionally, H₂S exposure increased [Ca ²⁺]_i in NECs isolated from zebrafish gill and resulted in membrane depolarization similar to that observed under hypoxia (Olson et al., 2008; Perry et al., 2016). Lactate can also elevate [Ca ²⁺]_i in isolated killifish (Fundulus heteroclitus) NECs (Leonard et al., 2022). In addition, lactate also elevates gill ventilation in a dose-dependent manner independent of pH changes in both the striped catfish (Pangasianodon hypophthalmus) and trout, with this response being attenuated following denervation of the first gill arch (Thomsen et al., 2017; Thomsen et al., 2019). Thus the existing evidence suggests that both H₂S and lactate have the potential to participate in O₂ sensing within NECs in fish. Two other gasotransmitters, CO and NO, also can modify ventilatory responses in fishes, with CO inhibiting ventilation in goldfish (Tzaneva and Perry, 2014) and zebrafish (Tzaneva and Perry, 2016). Interestingly, the effects of NO on ventilation are not only dependent on developmental age–NO stimulates breathing in zebrafish larvae while inhibiting breathing in adults (Porteus et al., 2015)–but also on HIF-1α, as NO is unable to contribute to the HVR in mutant larvae lacking HIF-1 α. However, there is no direct evidence that CO or NO participates in O₂ sensing specifically at the level of NECs (Olson et al., 2012; Mandic et al., 2019). A recent single-cell transcriptomic analysis of the zebrafish gill (Pan et al., 2022) revealed high expression of genes encoding for NADH dehydrogenase and cytochrome c oxidase in NECs, genes related to mitochondrial function that have been implicated in O₂ sensing in mammals.

Under voltage-clamp conditions, isolated gill NECs from adult zebrafish (Jonz et al., 2004) and channel catfish (Burleson et al., 2006) respond with a decrease in K⁺ current that is insensitive to voltage-dependent K⁺ channel blockers but sensitive to the background K⁺ channel blocker, quinidine (Jonz et al., 2004). In addition, hypoxia caused a reversible depolarization that was associated with a conductance decrease in isolated zebrafish gill NECs under current-clamp conditions; this response was reduced in the presence of quinidine (Jonz et al., 2004). In isolated gill NECs from goldfish, hypoxia exposure resulted in increases in [Ca²⁺]_i and vesicular activity that were blocked by the Ca²⁺ channel blocker, cadmium, and the L-type Ca²⁺ channel blocker, nifedipine (Zachar et al., 2017). These data suggest that hypoxia sensing in gill NECs is mediated by inhibition of a background K⁺ channel leading to depolarization, an increase in [Ca²⁺]_i, and possibly neurosecretion, similar to that observed in glomus cells of the mammalian carotid body (Lopez-Barneo et al., 2016).

Although there is evidence that trout NECs appear “degranulated” after hypoxia exposure (Dunel-Erb et al., 1982), and that NECs possess the synaptic vesicle protein SV2 indicative of neurosecretion potential (Jonz et al., 2004), there is no direct evidence of any neurotransmitter being released by NECs under hypoxic conditions. Serotonin (5-HT) is the only neurotransmitter directly and routinely identified in fish NECs, but there is an array of neurotransmitters that may enable signal transduction between NECs and the afferent neurons, and which have been shown to modulate breathing. The neuroendocrine control of breathing in fish was recently reviewed (Reed and Jonz, 2022) and thus, only a brief summary will be provided here. Using isolated gill arches from rainbow trout, application of acetylcholine resulted in potent stimulation of afferent neural activity, while 5-HT and dopamine caused brief, small bursts in neural activity followed by mild inhibition (Burleson and Milsom, 1995). Adrenaline and noradrenaline did not elicit any neural responses (Burleson and Milsom, 1995). However, when neurotransmitters were administered through intra-vascular injection or applied externally, 5-HT, adrenaline, noradrenaline, acetylcholine and purines generally stimulated ventilation, whereas dopamine inhibited ventilation (Burleson and Milsom, 1995). A similar inhibitory effect of dopamine on ventilation was reported in larval zebrafish (Shakarchi et al., 2013). In addition, single-cell transcriptomic analysis of the zebrafish gill showed that NECs and neurons express genes encoding transmembrane receptors for serotonergic, cholinergic and dopaminergic neurotransmission (Pan et al., 2022). These results are broadly consistent with the mammalian situation in which acetylcholine and purines are excitatory carotid body neurotransmitters whereas dopamine is inhibitory (Leonard et al., 2018). However, unlike in mammals in which neurotransmitter secretion from the carotid body was demonstrated, studies in fish, while establishing a direct link between neurotransmitters and ventilation, have not yet revealed the specific site of action. Future studies are needed to directly examine neurosecretion from NECs under hypoxic conditions.

Upon activation of the chemoreceptors and the presumed release of neurotransmitters, the signal is transmitted to the CNS leading to downstream ventilatory responses. Existing data suggest that cranial nerves IX (glossopharyngeal) and X (vagus) are the afferent neurons innervating the gill chemoreceptors. Thus, total gill denervation of cranial nerves IX and X decreased the magnitude of the HVR in nine species over a wide range of taxonomic diversity (Milsom, 2012) whereas rhythmic stimulation of cranial nerve X entrains the respiratory rhythm in carp (Cyprinus carpio) (De Graaf and Roberts, 1991). Using genetically encoded calcium sensors GCaAMP6s under control of the pan-neuronal promoter elavl3, Rosales et al. (2019) showed that in zebrafish, the average magnitude of Ca²⁺ transients increased within the sensory ganglia of cranial nerve X upon exposure to hypoxia, providing arguably the first direct in vivo evidence for any of the O₂ sensing components in a fish. Central projections from the ganglia of cranial nerves IX and X enter the hindbrain at presumptive rhombomere 6 and via a series of nerve roots to form a ‘plexus’ (Kucenas et al., 2006). Given that the preBötzinger complex, which is the origin of respiratory motor output in mammals, arises from rhombomeres 6 and 7 (Mellen and Thoby-Brisson, 2012), and also that rhombomere 7 is essential for gill and buccal bursts in bullfrog tadpoles (Duchcherer et al., 2013), it is reasonable to hypothesize that O₂ signals originating in the chemoreceptors are conveyed to the respiratory motor output centres in the hindbrain via cranial nerves IX and X.

In water-breathing adult fishes, evidence for a primarily branchial location of CO₂ chemosensors has come from studies in which CO₂-stimulated ventilatory responses were eliminated by denervation or extirpation of one or more gill arches (Burleson and Smatresk, 2000; Reid et al., 2000; Sundin et al., 2000; McKendry et al., 2001; Perry and Reid, 2002; Florindo et al., 2004; Bojink et al., 2010). These studies also revealed a predominant role for the first gill arch in eliciting CO₂-stimulated ventilatory responses in some species [rainbow trout (Perry and Reid, 2002) and jeju Hoplerythrinus unitaeniatus (Bojink et al., 2010), but see also (Sundin et al., 2000)], and they identified cranial nerves IX and X as those responsible for transmitting branchial CO₂ chemoreceptor activity to the brain. In the tambaqui Colossoma macropomum, total gill denervation did not completely eliminate ventilatory responses to CO₂, suggesting that extra-branchial CO₂ chemoreceptors may be present in some species (Milsom et al., 2002; Florindo et al., 2004). These extra-branchial receptors are likely to be associated with the orobranchial cavity or other peripheral sites because the balance of evidence does not support the presence of central CO₂/pH chemoreceptors in strictly water-breathing fishes (for discussion, see Tresguerres et al., 2019; Milsom et al., 2022).

The branchial chemoreceptors appear to respond primarily to water CO₂ rather than to water pH or to changes in blood CO₂/pH (reviewed by Gilmour and Perry, 2007; Milsom, 2012). Evidence to support this consensus has come from studies that have attempted to independently manipulate water versus blood CO₂ versus pH. For example, injection of CO₂-equilibrated water into the buccal cavity elicits hyperventilation, whereas injection of isocapnic water adjusted to the pH corresponding to that of the CO₂-equilibrated water has little or no effect (Reid et al., 2000; Sundin et al., 2000; Perry and McKendry, 2001; Gilmour et al., 2005; Bojink et al., 2010). These data argue for the importance of water CO₂ rather than water pH as the proximate factor controlling ventilation during ambient hypercapnia. A comparison between injection of CO₂-enriched water into the buccal cavity and injection of CO₂-enriched saline into the vasculature similarly argues for the importance of water CO₂, with injection of CO₂-enriched saline being without effect on ventilation (Perry and McKendry, 2001; Perry and Reid, 2002; Gilmour et al., 2005; Bojink et al., 2010). In addition, treatments that raise internal (but not water) CO₂ levels do not elicit hyperventilatory responses. For example, arterial CO₂ tension more than doubled in tambaqui treated with the carbonic anhydrase inhibitor acetazolamide (to inhibit CO₂ excretion), yet ventilation did not change until the fish were exposed to aquatic hypercapnia (Gilmour et al., 2005). Where questions remain is with the possibility of internally-oriented receptors that detect acid-base status. Although injections of acidified saline into the vasculature (Reid et al., 2000; Sundin et al., 2000; Bojink et al., 2010) or acetazolamide treatment (Gilmour et al., 2005) generally have not evoked ventilatory responses, several studies have reported results that are consistent with ventilation being adjusted by blood acid-base status (for detailed discussion, see Gilmour and Perry, 2007; Tresguerres et al., 2019). In particular, a close correspondence between ventilation and blood acid-base status during recovery from exhaustive exercise was observed in rainbow trout. Treatment of trout with carbonic anhydrase to enhance CO₂ excretion alleviated not only the acid-base disturbance, but also the accompanying hyperventilation (Wood and Munger, 1994). Thus, the door remains open to the possibility that ventilation may be adjusted according to internal acid-base status, with further research needed to refute or confirm this possibility.

Taken as a whole, the studies of ventilatory responses to hypercapnia suggest that CO₂ chemoreceptors are located in the gill of (adult) fishes and respond primarily to changes in water CO₂ levels. As with hypoxia, NECs located in the gill of adult fishes are thought to serve as the CO₂ chemoreceptors. In comparison with hypoxia, however, very few studies have investigated ventilatory responses to CO₂ in larvae, and there are no data that directly link cutaneous NECs to CO₂-stimulated ventilatory responses. Changes in NEC abundance, size and transcriptome in response to changes in environmental O₂ levels have provided indirect evidence of NEC involvement in O₂ chemoreception (see above). For CO₂ chemoreception, however, the data are both sparse and mixed. Acclimation to hypercapnic conditions for 28 d had no impact on the density of branchial (serotonin-positive) NECs in adult zebrafish (Vulesevic et al., 2006). On the other hand, increases in branchial NEC density were reported in mangrove rivulus acclimated to hypercapnia or to acidic water, although whether there were functional consequences of the changes in NEC density was not explored (Robertson et al., 2015). Overall, the paucity of data makes it difficult to draw any firm conclusions. However, there is evidence that branchial NECs detect and respond to hypercapnia in a manner that is consistent with their involvement in CO₂-evoked reflexes.

Direct evidence that branchial NECs can serve as CO₂ chemoreceptors was obtained from two studies of cellular responses to CO₂/pH in NECs isolated from the gills of adult zebrafish. Qin et al. (2010) reported that exposure to hypercapnia caused depolarization in a subset of the NECs that were tested, with the magnitude of the response increasing with increasing partial pressure of CO₂. The NECs that responded to increases in CO₂ also depolarized in response to lowering of O₂ levels, indicating that at least some NECs are dual sensors of O₂ and CO₂. Subsequently, an increase in [Ca²⁺]_i in response to hypercapnia was reported by Abdallah et al. (2015a). Although both responses were elicited by hypercapnia, the sensory mechanisms involved appear to differ, with a CO₂-stimulated change in intracellular pH being required for depolarization of the NEC (Qin et al., 2010), but a fall in extracellular pH that is independent of CO₂ being required for the increase in [Ca²⁺]_i (Abdallah et al., 2015a). Further research is needed to resolve this difference, and to reconcile the apparent discrepancy between sensory mechanisms at the cellular level (i.e. where extracellular pH is an adequate stimulus for a change in [Ca²⁺]_i) and the stimuli that elicit ventilatory responses at the whole-animal level (i.e. changes in water CO₂ but not pH; see above). In addition, there is a need to link activation of CO₂-sensing NECs directly to the occurrence of CO₂-stimulated ventilatory reflexes.

Current understanding of the signalling associated with NEC CO₂ sensing is based on the two studies mentioned above on NECs isolated from adult zebrafish gills (Qin et al., 2010; Abdallah et al., 2015a), and from studies in which cardiorespiratory responses to CO₂ in larval zebrafish were investigated (Miller et al., 2014; Koudrina et al., 2020; Kunert et al., 2022). As with O₂ sensing (Jonz et al., 2004), the initial cellular response to CO₂ is depolarization, which is caused by the inhibition of K⁺ flux through background K⁺ channels (Qin et al., 2010). It is likely that multiple K⁺ channels are present in NECs, with evidence supporting the involvement of TASK-2 (Koudrina et al., 2020), a member of the tandem-pore domains in a weak inward rectifying K⁺ channel (TWIK)-related acid-sensitive K⁺ (TASK) channel family. Functional studies carried out on zebrafish TASK-2 expressed in HEK-293 cells revealed that it is inhibited by intracellular or extracellular acidification as well as by increases in CO₂ (Peña-Münzenmayer et al., 2013). Notably, the inhibition of TASK-2 by CO₂ included both a contribution of intracellular acidification and a direct effect of CO₂ that was not dependent on changes in intracellular pH. Therefore, the properties of TASK-2 channels are consistent with the requirements for CO₂ sensing in NECs. Immunohistochemistry was used to demonstrate that TASK-2 is expressed by serotonin-positive NECs in the skin of larvae and the gills of adult zebrafish (Koudrina et al., 2020). A functional role for TASK-2 was identified using antisense oligonucleotide morpholinos to knock down TASK-2 and/or its paralog TASK-2b, which attenuated the hyperventilatory responses to hypercapnia in 4 dpf zebrafish larvae (Koudrina et al., 2020). Interestingly, the genes that encode for TASK-2 (kcnk5a) and TASK-2b (kcnk5b) were not detected in a single-cell transcriptomic analysis of zebrafish NECs (Pan et al., 2022) (nor was ca17a, see below). However, the focus of Pan et al. (2022) was on NECs that respond to hypoxia. If only a subset of NECs serve as bimodal sensors of O₂ and CO₂ (Qin et al., 2010), then it could be difficult to detect differential expression of transcripts specific to CO₂ sensing.

CO₂-induced depolarization of NECs is expected to cause an increase in [Ca²⁺]_i. Consistent with this pathway, Abdallah et al. (2015a) documented increases in [Ca²⁺]_i in isolated zebrafish NECs exposed to hypercapnia, with Ca²⁺ being derived primarily from intracellular stores. The increase in [Ca²⁺]_i, in turn, is expected to lead to neurosecretion, transmission of the sensory signal to the brain and activation of the appropriate ventilatory response. However, currently there are no data to support this pathway, nor has any specific neurotransmitter been identified as being secreted. Serotonin is viewed as likely to be involved owing to its status as the predominant neurochemical in NECs (Porteus et al., 2012) and the ability of serotonin receptor agonists to elicit hyperventilatory responses in zebrafish larvae (Jonz et al., 2015b; for reviews see Pan and Perry, 2020; Reed and Jonz, 2022). Although, as alluded to above (section 3.1.2), the fact that certain neurochemicals evoke hyperventilation does not necessarily indicate that they are secreted by NECs.

The CO₂ sensing mechanism likely involves carbonic anhydrase (CA), specifically the cytosolic isoform Ca17a (Ferreira-Martins et al., 2016). Immunohistochemistry supports the presence of CA or Ca17a specifically in larval cutaneous and adult branchial NECs of zebrafish (Qin et al., 2010; Miller et al., 2014; Kunert et al., 2022). Pharmacological inhibition of CA, as well as specific knockdown or knockout of Ca17a, blunts the cardiorespiratory responses to hypercapnia in zebrafish larvae (Miller et al., 2014; Kunert et al., 2022). These data argue that Ca17a contributes to CO₂ sensing, yet its specific role at the cellular level remains unclear. Inhibition of CA both slowed the rate and reduced the magnitude of CO₂-stimulated depolarization in isolated branchial NECs (Qin et al., 2010). In conjunction with the observation that extracellular acidification was not needed to elicit depolarization, these data suggest that depolarization occurs in response to CO₂ entry into the cell and its CA-catalyzed hydration to H⁺, resulting in intracellular acidification. However, intracellular acidification was not sufficient to elicit a [Ca²⁺]_i response, nor did inhibition of CA alter the [Ca²⁺]_i response to hypercapnia even though intracellular acidification was slowed (Abdallah et al., 2015a). Indeed, extracellular acidification alone was sufficient to elicit a [Ca²⁺]_i response (Abdallah et al., 2015a). Resolving these discrepancies likely will provide insight into the role of Ca17a in CO₂ sensing.