Dietary protein is a major source of amino acids in animals. The intestines of carnivorous fishes are adapted to process diets that are high in protein and low in carbohydrate (Buddington et al., 1997). Karlsson et al. (2006) determined changes in plasma concentrations of free amino acids and their metabolites in pre- and post-hepatic blood following a single meal in rainbow trout (Oncorhynchus mykiss), and confirmed that amino acids could be metabolized in the intestine before they reached the liver. The plasma ammonia level in the hepatic portal vein was higher than that in the dorsal aorta, and the difference between the two blood sampling sites increased during amino acid absorption after a meal. Thus, Karlsson et al. (2006) concluded that deamination of certain amino acids occurred in the intestine of the rainbow trout after feeding. In support of the conclusion of Karlsson et al. (2006), Tng et al. (2008) reported that postprandial amino acid metabolism indeed occurred in the intestine of juvenile Oxyeleotris marmorata. The major amino acid accumulated in the intestine and liver of juvenile O. marmorata after feeding was glutamate, and feeding led to a significant increase in glutamate dehydrogenase (GDH) activities in the intestine and liver of O. marmorata, which could lead to a high retention of the ingested nitrogen for somatic growth. Consequently, only 33% of the ingested nitrogen was excreted during the 24 h post-feeding period, and the brain was effectively prevented from exposure to postprandial ammonia toxicity (Tng et al., 2008).

Animals cannot store excess amino acids, unlike carbohydrates and lipids which can be stored as glycogen and triglycerides, respectively. Thus, dietary amino acids in excess of the amounts needed for growth and maintenance of protein turnover are preferentially degraded over carbohydrates and lipids in the liver (Campbell, 1991). For fishes with high-protein diets, their dietary carbon is extracted from the carbon chain of amino acids after the removal of the α-amino group. Several amino acids, including alanine, are converted to glucose by fish hepatocytes (French et al., 1981) and this process is regulated hormonally in much the same way as it is in mammals. Approximately 40–60% of the nitrogen intake from food is excreted within 24 h (Ip et al., 2004c; Lim et al., 2004b). In addition to diet, muscle proteins can act as a source of amino acids, which are catabolized for the production of ATP or carbohydrates, in fasting fishes (Houlihan et al., 1995). Under adverse environmental conditions where ammonia excretion is reduced, some fishes can reduce the rate of ammonia production from amino acid catabolism to slow down the build up of ammonia internally (Ip et al., 2001c, 2004a,b; Lim et al., 2001). During exercise or hypoxia, ammonia can also be produced through the deamination of AMP in the skeletal muscle.

Much of the ammonia produced in fish comes from the α-amino group of amino acids that are catabolized. The rate of alanine and glutamine degradation by catfish hepatocytes can account for 50 and 85%, respectively, of the total ammonia excreted by the fish (Campbell et al., 1983). In addition, the rate of glutamate deamination by intact catfish liver mitochondria can account for 160% of the rate of ammonia excretion (Campbell et al., 1983). For goldfish, the liver is responsible for 50–70% (van den Thillart and van Raaji, 1995) of ammonia production. Ammonia can be produced either directly in the cytosol of hepatocytes by specific deaminases (histidase, asparaginase, serine dehydratase, and threonine dehydratase; Youngson et al., 1982) or via the combined actions (transdeamination) of cytosolic aminotransferases and mitochondrial GDH (Walton and Cowey, 1977; French et al., 1981), but transdeamination is the primary mechanism for catabolism of amino acids in fish liver (Ballantyne, 2001). Since GDH is localized exclusively in the matrix of fish liver mitochondria, it is within this compartment that ammonia is released through the route of transdeamination which involves the deamination of glutamate (Figure 1). Glutaminase, which releases NH₃ from the amide-function of glutamine is also present in the mitochondrial matrix of some fish species. Thus, ammonia released in the matrix of liver mitochondria has to permeate the mitochondrial membranes before excretion.

In aqueous solution, ammonia can be present as gaseous NH₃ and cationic NH4+, the ratio of which is pH dependent. The equilibrium reaction between these two components of ammonia can be written as NH3+H3O+⇔NH4++H2O, and the pK of this NH₃/NH4+reaction is around 9.0–9.5. NH₃ reacts avidly with water and is moderately soluble in lipid. Since phospholipids of biological membranes are not very permeable to NH4+, ammonia crosses membranes as NH₃ in most cases. NH₃ has a high solubility in water and therefore contrasts strongly with lipophilic molecules such as CO₂; consequently, the permeability to NH₃ of nearly all cell membranes is much less than their permeability to CO₂ or O₂. Nevertheless, biomembranes are so thin that NH₃ can diffuse quite rapidly through nearly all of them, although the NH₃ permeability varies greatly and can be very low in some cases (Marcaggi and Coles, 2001). It has been reported that plasma membranes of several mammalian cell types facing the gastric and urinary tracts have relatively low permeability to NH₃ (see Marcaggi and Coles, 2001 for a review). On the other hand, ammonia permeation through plasma membranes can also be enhanced by protein channels in certain cell types.

It has been established that the aquaporin (AQP) family generally functions for the selective passage of water or glycerol, but particular aquaporin isoforms can conduct unconventional permeants, including ammonia (Wu and Beitz, 2007). Nakhoul et al. (2001) undertook a study to determine whether expressing AQP1 could affect transport of NH₃. Using ion-selective microelectrodes, they conducted experiments on frog oocytes (cells characterized by low NH₃ permeability) expressing AQP1, and concluded that the oocyte membrane, although highly permeable to NH4+, had a significant NH₃ permeability and that NH₃ permeability was enhanced by AQP1. Holm et al. (2005) expressed the mammalian aquaporins AQP8, AQP9, AQP3, and AQP1 in Xenopus oocytes to study the transport of NH₃ and NH4+ under open-circuit and voltage-clamped conditions, and concluded that apart from being water channels, they also supported significant fluxes of NH₃ and NH4+. By reconstituting purified APQ8 into planar bilayers, Saparov et al. (2007) demonstrated recently through stoichiometric study that it selectively transport NH₃, excluding NH4+and H⁺.

Additionally, members of the Rhesus (Rh) glycoprotein family may play an important role in transmembrane ammonia transport (Bakouh et al., 2006). Among the six clusters of the Rh family, RhAG, RhBG, and RhCG are present in most vertebrates and are known to be capable for transporting ammonia and methylammonium (Planelles, 2007). RhCG is expressed in distal nephron sites that are critically important for ammonia secretion, and exhibits increased expression in response to chronic metabolic acidosis. It has been suggested that RhCG contributes to both apical and basolateral membrane ammonia transport in the human kidney (Han et al., 2006). RhCG apparently functions as an NH₃ channel, and it can be fully active without the contribution of any protein partner (Mouro-Chanteloup et al., 2010). This, however, does not preclude the possibility that, in physiological conditions, RhCG activity could be modulated by other transporters such as the vesicular-type proton pump (H⁺-ATPase) and/or the Cl−/HCO3− exchanger (AE1) expressed in the same epithelial cells in the kidney.

Since NH4+has some ionic properties similar to those of K⁺, it can compete with K⁺ to be transported through the K⁺ channels (Thomas, 1984; Marcaggi and Coles, 2001). According to Choe et al. (2000), most of the values for PNH4/P_K through K⁺ channels range between 0.1 and 0.3. However, some K⁺ channels apparently has high specificity for K⁺, for example, those of the starfish egg, which have a PNH4/P_K value of 0.03, and the glial cells of bees’ retina (see Marcaggi and Coles, 2001 for a review). NH4+can also compete with K⁺ to be transported through Na⁺/K⁺-ATPase and Na⁺:K⁺:2Cl⁻ cotransporter. Additionally, NH4+ transport can also involve the Na⁺/H⁺ exchangers (NHE) with NH4+ substituting for H⁺. In some cases, influx of exogenous ammonia as NH4+ through various transporters would result in intracellular NH₃ cycling leading to a decrease in the intracellular pH (Marcaggi and Coles, 2001).

Many ammonotelic fishes, which excrete more than 50% of the nitrogenous waste as ammonia, are carnivorous and sustain high rates of hepatic amino acid catabolism (Bever et al., 1981). The liver mitochondria of ammonotelic fishes can sustain glutamate deamination for long periods because of the uncoupling between deamination and transamination. There is little aspartate formation by fish mitochondria from glutamate or glutamine in vitro, even after relative long incubation periods with these substrates (Walton and Cowey, 1977; Campbell et al., 1983). The resulting high rate of mitochondrial amino acid deamination should result in a high rate of ammonia production and therefore a high rate of ammonia efflux across the inner mitochondrial membrane. A relatively large portion of the ammonia would be present as NH₃ in the mitochondrial matrix which has a higher pH than the inter-membrane region of the mitochondrion and the cytosol. At present, no information is available on the presence of APQ, specifically APQ8 which facilitates the permeation of NH₃ in mammalian mitochondria (Soria et al., 2010), in the liver mitochondria of ammonotelic fishes. However, in theory, NH₃ efflux from the mitochondrial matrix would result in the uncoupling of oxidative phosphorylation (Figure 3), and hence there is a fundamental issue on the species of ammonia that is transported out of liver mitochondria of ammonotelic fishes. Three mechanisms, which avoid uncoupling oxidative phosphorylation, have been proposed for ammonia efflux from mammalian kidney mitochondria (LaNoue and Schoolwerth, 1979), but there is no evidence for any of these mechanisms to be operative in ammonotelic fishes.

The study by Campbell et al. (1983) is one of the very few that dealt with the enigma on the form of ammonia that exits the inner mitochondrial membrane of a freshwater ammonotelic fish. The basic assumption made by Campbell et al. (1983) was that any increase in ambient H⁺ concentration should cause an increase in the rate of ammonia efflux. However, when the pH of the medium in which catfish liver mitochondria respiring in the presence of a substrate (glutamine) and ADP was varied between 7.4 and 6.7, a decrease in ammonia efflux occurred with an increase in the proton concentration in the medium. If the ammonia formed in the catfish liver mitochondria via the removal of the amide group of the added glutamine exited the mitochondrial matrix as NH₃, an alkalinization would be expected based on the findings of Brierly and Stoner (1970). By contrast, if NH4+ was the exiting species, it should result in a slight acidification due to its dissociation into NH₃ and H⁺. Therefore, it would appear that NH4+ was the permeating species in catfish liver mitochondria. Furthermore, carbonyl cyanide p-(trifluoromethoxy)phenylhydrazone, an uncoupler of oxidative phosphorylation, caused a marked increase in the release of ammonia formed from glutamate in the catfish liver mitochondrial system (Campbell et al., 1983). Since carbonyl cyanide p-(trifluoromethoxy)phenylhydrazone is a protonophore and causes a dissipation of the H⁺ gradient across the inner mitochondrial membrane, it was concluded that diffusion-trapping of exiting NH₃ by the proton gradient formed by the electron transport system as proposed for kidney mitochondria could not be applied to the catfish liver mitochondria. Rotenone, which directly inhibits the electrogenic H⁺ pump, also had little effect on ammonia efflux from these mitochondria. While the co-transport of glutamate and H⁺ seems an appropriate mechanism for mammalian kidney mitochondria where NH4+ formation from glutamine degradation is presumably a means of excreting H⁺ (Flessner et al., 1991), such a mechanism would seem inappropriate for fish liver mitochondria whereby transdeamination, a process that involves the deamination of glutamate, is a major pathway of amino acid catabolism (Campbell et al., 1983; Campbell, 1997).

Overall, the available experimental data suggest that ammonia formed during amino acid catabolism in catfish liver mitochondria exits the mitochondrial matrix as NH4+, and not as NH₃ (Figure 1). This is different from carbon dioxide which is also produced in the matrix, but exits the mitochondria mainly as CO₂ gas. It is possible that a putative transport protein, which transports NH4+ out of the matrix thereby preventing dissipation of the proton gradient established by the electrogenic H⁺ pump, may be present in liver mitochondria of ammonotelic fish species (Campbell, 1997; Figure 1). Transport of NH4+ across the inner mitochondrial membrane may involve NHE, especially NHE6 in higher eukaryotes (Numata et al., 1998; Figure 1). Additionally, the presence of a H⁺ transporter independent of the electron transport chain that acts simultaneously with the translocation of NH₃ cannot be ruled out. More importantly, it has been established that ammonia transporters are commonly present in bacteria (Kleiner, 1992; Huang and Peng, 2005). Therefore, assuming a symbiotic origin of mitochondria in eukaryotes, it is logical to deduce their presence in mitochondria. More importantly, these putative mechanisms for NH4+ transport through the inner mitochondrial membrane of ammonotelic fishes have been reported to be present in the branchial epithelium of certain amphibious fish species (e.g., the giant mudskipper) and aquatic crustaceans which are capable of excretion of NH4+ against an electrochemical gradient (Peng et al., 1998; Randall et al., 1999, 2004; Wilson et al., 2000; Ip et al., 2004d; Weihrauch et al., 2009). Therefore, an examination of the presence of transporters that would facilitate the exit of NH4+ from the matrix of, but avoid uncoupling oxidative phosphorylation in, the liver mitochondria of ammonotelic fishes is warranted.

The gills are the primary site of ammonia excretion in fish (Wilkie, 1997, 2002; Weihrauch et al., 2009), because they have large surface area, perfusion by 100% of cardiac output, large ventilation rates, small diffusion distances, and contact with a voluminous mucosal medium (Evans et al., 2005). Although gill tissues exert an extremely high metabolic rate, accounting for almost 10% of the entire oxygen demand of teleosts for osmoregulatory purposes, the overall metabolic expenditures for the release of ammonia appear to be minimal (Evans et al., 2005). For most ammonotelic fishes, the majority of ammonia is excreted across the branchial epithelium as NH₃, down a favorable blood-to-water diffusion gradient (Wilkie, 1997, 2002; Evans et al., 2005). In freshwater fishes (Figure 4), excreted NH₃ can be trapped via H⁺ secretion or CO₂ excretion into the unstirred layer of water on the apical surface of the gills (Avella and Bornancin, 1989). H⁺ secretion can be achieved through an apical vesicular type of H⁺-ATPase (see Lin and Randall, 1995 for a review), and there is an apparent lack of Na⁺/H⁺ (NH4+) exchange via NHE in gills of freshwater fishes (Evans et al., 2005). For marine fishes (Figure 5), despite the presence of NHE, which facilitates Na⁺ absorption, little to no ammonia excretion occurs through Na⁺/ NH4+ exchange (Wilkie, 2002) because of the presence of favorable NH₃ and NH4+ diffusion gradients. Unlike freshwater fishes, marine fishes have leaky tight junctions between mitochondrion-rich cells, which increase cation permeability for Na⁺ secretion. Therefore, a significant portion of ammonia can be excreted through NH4+ diffusion through the paracellular route in seawater fishes (Goldstein et al., 1982).

Traditionally, it has been accepted that NH₃ diffusion occurs across the phospholipid bilayer of the branchial epithelium, but research in the past decade reveals that permeation of ammonia across fish gills can involve transporters (Huang and Peng, 2005; Weihrauch et al., 2009; Wright and Wood, 2009). Within the ammonoteles, Rh glycoproteins are beginning to emerge as vital ammonia conduits (Figures 4 and 5). Nawata et al. (2007) confirmed the expression of seven Rh genes, with the suggestive role in ammonia excretion, in the gills of the rainbow trout. In a study of Rh gene expression in early life history stages of rainbow trout, Hung et al. (2008) discovered substantial expression of Rhbg, Rhcg1, and Rhcg2 mRNAs in rainbow trout whole embryo extracts that exhibited temporal dynamics consistent with developmental changes in nitrogen excretion patterns. Nakada et al. (2007a) also confirmed the expression of Rh gene on the surface of the yolk sac of zebrafish. Using scanning ion-specific electrode technique and knockdown methods, Shih et al. (2008) obtained results which indicate that the proton-ATPase-rich cells of the zebrafish yolk sac skin were involved in ammonia and acid excretion. Furthermore, Nakada et al. (2007b) identified the Rh glycoprotein homologs, fRhag, fRhbg, fRhcg1, and fRhcg2, of the pufferfish, Takifugu rubripes. All four glycoproteins were localized in the gill; they lined the pillar cells, pavement cells, and the mitochondrion-rich cells. Heterologous expression of these glycoproteins in Xenopus oocytes showed that they mediated methylammonium (an analog of ammonium) transport. Results obtained by Nakada et al. (2007b) suggest that pufferfish Rh glycoproteins are involved in ammonia excretion from the gill, and challenge the classic view that ammonia excretion in the fish gill occurs by passive diffusion across the phospholipid bilayer. Nawata et al. (2010) studied the involvement of Rh glycoproteins in ammonia transport through the gills of the rainbow trout, using oocyte expression and scanning ion-specific electrode technique. Their results are consistent with the current model in which ammonia passes through the Rh channels as NH₃ (Mouro-Chanteloup et al., 2010).

Specifically, Rhcg appears to be localized in apical membranes and Rhbg in basolateral membranes in the branchial epithelium. Their gene expression is up-regulated during exposure to high environmental ammonia or internal ammonia infusion, and may be sensitive to synergistic stimulation by ammonia and cortisol. Rhcg in particular appears to be coupled to H⁺ excretion and Na⁺ uptake mechanisms. Wright and Wood (2009) proposed a new model for ammonia excretion in freshwater fish and its variable linkage to Na⁺ uptake and acid excretion. In their model, Rhcg facilitates NH₃ flux out of the erythrocyte, Rhbg moves it across the basolateral membrane of the branchial ionocyte, and an apical “Na⁺/ NH4+ exchange complex” consisting of several membrane transporters (Rhcg, vesicular-type H⁺-ATPase, Na⁺/H⁺ exchanger NHE-2 and/or NHE-3, Na⁺ channel) working together as a metabolon provides an acid trapping mechanism for apical excretion. Intracellular carbonic anhydrase and basolateral Na+/HCO3− cotransporter and Na⁺/K⁺-ATPase play indirect roles in ammonia transport. These mechanisms are normally superimposed on a substantial outward movement of NH₃ by simple diffusion, which is probably dependent on acid trapping in boundary layer water by H⁺ ions created by the catalyzed or non-catalyzed hydration of expired metabolic CO₂ (Wright and Wood, 2009).

Recently, Braun et al. (2009) used the injection of antisense oligonucleotide morpholinos to elicit selective gene knockdown of ammonia transporters (Rhag, Rhbg, and Rhcg1) to assess the relative importance of each transporter to nitrogen excretion in developing zebrafish. Knockdown of each of the Rh protein resulted in a 50% reduction in ammonia excretion. Contrary to what has been hypothesized previously for adult fish, each of the Rh proteins appeared to have a similar effect on total ammonia excretion, and thus all are required to facilitate normal ammonia excretion in the zebrafish larva. At the point of hatching, ammonia excretion rapidly increased and appeared to be triggered by a large increase in the mRNA expression of Rhag, Rhbg, and Rhcg1. Unlike the situation in the adult pufferfish, the various transporters are not specifically localized on the gills of the developing zebrafish (Nakada et al., 2007a), but each protein has a unique expression pattern along the skin, gills, and yolk sac. This disparate pattern of expression would appear to preclude interaction between the Rh proteins in zebrafish embryos. However, this may be a developmental feature of the delayed maturation of the gills, because as the embryos matured, expression of the transporters in and around the gills increased.

While excretion of ammonia is not a problem at low environmental pH, it is a major problem for fish exposed to high pH. This is because at high environmental pH, worst still with high buffering capacity (Weihrauch et al., 2009), the gradient for NH₃ diffusion is reduced and this may lead to a build up of ammonia inside the fish (Wilkie and Wood, 1996). Death can occur when the rise in plasma ammonia level is too rapid and/or toxic levels of ammonia are reached (Wilkie et al., 1993).

Air-breathing is one of several adaptive responses utilized by fishes dwelling in habitats where O₂ supplies may be severely depleted (Graham, 1997). While most air-breathing fishes remain aquatic, some evolved to emerge from water, make excursion onto land, or even burrow into mud when the external media dry up. When a fish is out of water, it is confronted with problems of endogenous ammonia excretion because there is a lack of water to flush the branchial and cutaneous surfaces. So, amphibious air-breathing fishes must have adaptation to ameliorate ammonia toxicity during long term emersion (see Ip et al., 2001c, 2004a; Chew et al., 2006b for reviews).

Some air-breathing fishes can be trapped in puddles of water occasionally, or in crevices for many days; continual excretion of endogenous ammonia into a small volume of external media can lead to high external ammonia concentrations. Furthermore, water evaporation at the high temperatures of the tropics can concentrate environmental ammonia to high levels. In addition, fishes can be exposed to high concentrations of environmental ammonia under several conditions. Some tropical fishes may have unique behaviors; for example, mudskippers build burrows in the mud in estuaries and stay therein during high tides. During the breeding season, the male fish stays inside the burrow for 1–2 months to take care of the developing embryos and fry. Since the burrow water is not well flushed, the ammonia concentrations can be high, and mudskippers have to deal with the toxicity of environmental ammonia in the burrow. Some fishes live in rice fields, where agricultural fertilization can lead to high concentrations of environmental ammonia. In the presence of high concentrations of environmental ammonia, fishes are confronted simultaneously with retention of endogenous ammonia and uptake of exogenous ammonia, and they have special adaptations to deal with ammonia toxicity (see Ip et al., 2004b; Chew et al., 2006b, for reviews).

Since the pH of fish blood and intracellular fluid is approximately 1–2 units below the pK of ammonia, more than 95% of the ammonia exists as NH4+ in the cytosol and plasma of fish. Once in the plasma, ammonia will be circulated to other part of the fish body; were it allowed to accumulate many deleterious effects will develop because ammonia is toxic and it affects various cellular processes (Ip et al., 2001b; Randall and Tsui, 2002). Ammonia stimulates glycolysis in fish by activating phosphofructokinase I in the cytosol. It also interferes with energy metabolism through impairment of the tricarboxylic acid cycle in the mitochondria (Arillo et al., 1981). Ammonia affects the ionic balance in fish at both cellular and at the organismal levels, reducing Na⁺ influx and K⁺ loss through substitution of NH4+ for K⁺ in Na⁺/K⁺-ATPase and/or Na⁺:K⁺:2Cl⁻ co-transport (see Wilkie, 2002 for a review). In addition, NH4+ can substitute for H⁺ in Na⁺/H⁺ exchange in marine fishes (probably NHE2 and/or NHE3, Randall et al., 1999).

The most acute effects of ammonia are probably related to the ability of NH4+ to substitute for K⁺ in ion transporters (Binstock and Lecar, 1969) and disrupt electrochemical gradients in central nervous systems (Cooper and Plum, 1987). Because fishes have less advanced central nervous systems than higher vertebrates, this may explain why fishes are more tolerant of ammonia than mammals (Evans et al., 2005). Several theories (glutamatergic dysfunction, glutamine accumulation leading to astrocyte swelling, activation of N-methyl-d-aspartate-type (NMDA) glutamate receptors) have been proposed to address the mechanisms of acute ammonia toxicity in mammalian brains (Brusilow, 2002; Felipo and Butterworth, 2002; Rose, 2002), but they have yet to be confirmed in fish.

The mechanisms of ammonia toxicity in the brains of some fishes with high ammonia tolerance apparently differ from those in mammalian brains (Opsanus beta, Veauvy et al., 2005; Periophthalmodon schlosseri and Boleophthalmus boddarti, Ip et al., 2005a; Clarias gariepinus, Wee et al., 2007; Monopterus albus, Tng et al., 2009). MSO, at a dosage (100 μg g⁻¹ fish) protective for rats, does not reduce the mortality of two species of mudskippers, P. schlosseri and B. boddarti, injected with a lethal dose of CH₃COONH₄, indicating that increased glutamine synthesis and accumulation in the brain is not the major cause of death (Ip et al., 2005a). The prior administration of MSO (100 μg g⁻¹ fish) reduces the mortality of the African catfish, C. gariepinus, injected with a lethal dose of CH₃COONH₄ from 100 to 80% and prolongs the time of death from 27 to 48 min (Wee et al., 2007). However, the protective effect of MSO may be unrelated to the inhibition of GS and prevention of glutamine accumulation. Wee et al. (2007) demonstrated that MSO also suppressed the rate of ammonia build up in the brain of the African catfish, C. gariepinus, probably through inhibiting activities of GDH and alanine aminotransferase therein, as it does in vitro when MSO was added to the enzyme assay mixture. The demonstration of unchanged maximal activities of these two enzymes extracted from the brain of C. gariepinus with or without MSO treatment, as reported by Wee et al. (2007), did not necessarily indicate the absence of changes in activities in vivo as suggested by Sanderson et al. (2010), since enzymes rarely operate at maximal activities and saturating substrate levels in situ. Sanderson et al. (2010) reported a lack of changes in ammonia and glutamate levels accompanied with increases in contents of several amino acids in the brain of rainbow trout pre-injected with MSO followed with exposure to environmental ammonia (0.6 or 1.0 mmol l⁻¹) as compared with the control without MSO treatment. They (Sanderson et al., 2010) concluded that rainbow trout possessed a relatively large reserve capacity for ammonia detoxification and for preventing glutamate accumulation during hyperammonemic conditions. However, these observations could also be explained by the possible suppression of GDH activities by MSO in vivo leading to decreases in catabolism of certain amino acids and consequently the accumulation of these amino acids and a reduction in ammonia production, although it is uncertain at present whether MSO would affect GDH activity from the rainbow trout, as in the case of the African catfish. As for the swamp eel, M. albus, it can survive a sublethal dose (10 μmol g⁻¹ fish) of intraperitoneal injection with CH₃COONH₄ by having high ammonia tolerance at the cellular level, detoxifying ammonia to glutamine in and outside the brain, and reducing the production of endogenous ammonia (Tng et al., 2009). Those fish that succumbed to a lethal dose (16 μmol g⁻¹ fish) of CH₃COONH₄ has an extraordinary high content of ammonia in the brain, and therefore, unlike mammals, increased glutamine synthesis and accumulation might not be the major cause of death. Although MSO exhibits a partial protective effect against acute ammonia toxicity in M. albus, it is unrelated to the prevention of glutamine accumulation (Tng et al., 2009). Instead, MSO apparently reduced the rate of ammonia build up in the brain of M. albus through its effect on GDH. Taken together, these results indirectly indicate that fish brains can suppress ammonia production.

In mammals, acute ammonia intoxication is mediated by activation of NMDA receptors (Hermenegildo et al., 1996; Kosenko et al., 1999); and, excessive activation of these receptors is neurotoxic, leading to oxidative stress, neuronal degeneration and death (Miñana et al., 1996). However, the brain of B. boddarti also experiences ammonia-induced oxidative stress (Ching et al., 2009), although ammonia apparently does not induce excessive activation of NMDA receptor in the brain of this fish (Ip et al., 2005a). Exposure of B. boddarti to 8 mmol l⁻¹ NH₄Cl for 12 or 24 h leads to the accumulation of carbonyl proteins, elevation in oxidized glutathione content and oxidized:reduced glutathione ratio, decreases in activities of glutathione reductase and catalase, and an increase in the activity of superoxide dismutase (Ching et al., 2009). The capacity to increase glutathione synthesis and reduced glutathione content could alleviate severe ammonia-induced oxidative and nitrosative stress in the brain. Furthermore, the ability to decrease the protein abundance of p38 and phosphorylated p53 might prevent cell swelling, contributing in part to the high ammonia tolerance in the brain of B. boddarti. More importantly, ammonia can also induce oxidative stress in the gills of B. boddarti, an organ that lacks NMDA receptors (Ching et al., 2009). Hence, there could be multiple routes through which ammonia induces oxidative stress in brain or non-brain tissues. It is possible that ammonia directly affected the intracellular NO and/or Ca²⁺ concentrations, the increase of which can lead to increased production of free radicals (Hernández-Fonseca et al., 2008), in the gills and brain of B. boddarti.

Unlike terrestrial tetrapods, fish respire mainly through gills, which would be directly exposed to exogenous ammonia during environmental ammonia exposure. Ammonia must permeate through the branchial and cutaneous epithelia before being transported through the blood to the brain and other organs. As an active respiratory and osmoregulatory organ, fish gills have a high capacity to produce reactive oxygen species, and Ching et al. (2009) indeed reported that exposure to environmental ammonia resulted in oxidative stress in the gills of B. boddarti. In addition, environmental ammonia has deleterious effects on branchial ion transport not associated with the accumulation of endogenous ammonia. These effects are not applicable to fish simply exposed to terrestrial conditions or to fish injected/infused with exogenous ammonia. Acute exposure to environmental ammonia results in the inhibition of Na⁺ influx in the temperate rainbow trout O. mykiss (Avella and Bornancin, 1989) and the goldfish Carassius auratus (Maetz, 1973). In C. auratus, the deleterious effect is specific to Na⁺ uptake and not general to the epithelium or all ion uptake mechanisms. In contrast, no deleterious effect of ammonia exposure (up to 28.2 μmol l⁻¹ NH₃-N or 5.2 mmol l⁻¹ total ammonia) is seen on Na⁺ uptake in juvenile rainbow trout, but Na⁺ efflux is stimulated by ammonia levels greater than 6.4 μmol l⁻¹ NH₃-N (1.2 mmol l⁻¹ total ammonia) (Twitchen and Eddy, 1994). This increase in efflux is likely through an increased Na⁺ permeability of the gills (Gonzalez and McDonald, 1994), mediated through a modulation of the paracellular pathway (Madara, 1998). In addition, exposure to environmental ammonia predisposes the gills to histopathological changes that may disrupt ion transport (Daoust and Ferguson, 1984). Disruption of epithelial integrity has adverse consequences for ion transport and other cellular processes, and the proliferation of branchial mucous cells induced by environmental ammonia increases diffusion distances across the gill (Ferguson et al., 1992).

In ureogenic and ureotelic animals, including humans, the main mechanism for keeping the internal concentration of ammonia low is by detoxifying ammonia to urea which is eliminated in urine. Urea synthesis occurs primarily in the liver through the ornithine-urea cycle (OUC). In other tissues, including the brain, that do not possess a functional OUC, ammonia is mainly detoxified to glutamine, which is then released to the blood stream and may serve as a non-toxic carrier of ammonia from different tissues to the liver. Glutamine formation plays a major role in detoxifying exogenous and endogenous ammonia in non-ureogenic fishes, especially in the brain, during exposure to exogenous/environmental ammonia (Arillo et al., 1981; Dabrowska and Wlasow, 1986; Mommsen and Walsh, 1992; Peng et al., 1998; Lim et al., 2004a; Ip et al., 2005a; Veauvy et al., 2005; Wee et al., 2007; Wright et al., 2007; Tng et al., 2009; Sanderson et al., 2010) or after feeding (Wicks and Randall, 2002; Lim et al., 2004b). In certain fish species, like the swamp eel, M. albus (Tok et al., 2009), and the Amazonian freshwater stingray, Potamotrygon motoro (Ip et al., 2009), glutamine synthesized from endogenous ammonia can act as an important osmolyte for cell volume regulation during acclimation to high salinity. As for ureogenic and ureosmotic fishes, glutamine can act as a substrate for the synthesis of urea which is essential for osmoregulation in a hyperosmotic environment.

Glutamine is produced from glutamate and NH4+, the reaction catalyzed by GS, and multiple GS genes have been reported for the rainbow trout (Murray et al., 2003; Walsh et al., 2003; Wright et al., 2007). Glutamate may in turn be produced from α-ketoglutarate (α-KG) and NH4+, catalyzed by the aminating activity of GDH, in the mitochondria of cells that are not geared toward the catabolism of amino acids through transdeamination. It can also be produced from α-KG and other amino acids catalyzed by various transaminases in the mitochondria and/or the cytosol. It has been reported that the brain of the rainbow trout exhibits higher GS activity than GDH activity (Wright et al., 2007), but it is important to take note that the GS activity was determined by the transferase assay, and the transferase activity is known to be ∼20-fold greater than the biosynthetic activity. Whether the synthesis of glutamine begins with glutamate or α-KG is determined by the compartmentalization of GS within the cell (Figures 1 and 2). Webb and Brown Jr. (1976, 1980) determined the distribution of GS in tissues of ureosmotic and non-ureosmotic fishes, and reported the presence of high levels of GS in the cytosolic fraction of brain tissues of both group of fishes (Figure 2). Interestingly, in mammalian liver, GS is also present in the cytosolic compartment of perivenule hepatocytes (Wu, 1963) and appears to function in these cells as “fail-safe” mechanism for ammonia detoxification when the capacity for urea synthesis is exceeded. However, for ureosmotic fishes, the liver tissues also contained high levels of GS activity in the mitochondria (Figure 1). Furthermore, results obtained by Webb and Brown Jr. (1976, 1980) showed that certain non-ureosmotic fishes possessed relatively low activities of cytosolic GS in the livers. GS is also detected from the mitochondria of elasmobranch kidney, which may function as part of a substrate cycle for ammonia excretion during acidosis (King and Goldstein, 1983).

A major portion of ammonia is produced in the liver, and since NH₃ is an uncoupler of oxidative phosphorylation and exerts its deleterious effect during exit from the matrix of the mitochondrion, it is logical for it to be detoxified to a less toxic (or non-toxic) product before crossing the inner membrane of liver mitochondria. For the detoxification of endogenous ammonia, which is produced through transdeamination, to glutamine (or finally to urea in ureogenic fishes), it is essential for GS to be located in the mitochondrial matrix (Figure 1). Under such a situation, glutamate serves the dual functions of providing NH₃ and acting as a direct substrate for glutamine formation inside the mitochondrion. Glutamine formed in the mitochondrial matrix can exit to the cytosol where it serves as a precursor for various biosynthetic pathways (Figure 1). Glutamine shows no protonation of its R chain over a wide range of pHs and is therefore proton-neutral (Campbell, 1991, 1997). The physiological consequence of this is that NH₃ in the mitochondrial matrix, whether arising by deamination of glutamate via GDH or the direct entry of NH₃ or NH4+, exits to the cytosol as a proton-neutral amide-function of glutamine. Through this, the problem of uncoupling oxidative phosphorylation by NH₃ is circumvented.

Ammonia circulating in the plasma asserts its toxic effects on the brain, and fish brains possess high levels of GS activities to protect them against ammonia toxicity (Peng et al., 1998; Ip et al., 2005a; Wright et al., 2007). If GS were to be located in the mitochondria, it would render the detoxification of exogenous ammonia inefficient as NH₃ would have to permeate through both the plasma and mitochondrial membranes to be processed by the mitochondrial GS. In fact, NH₃ would bind with H⁺ after entering the cytosol to form NH4+, and ammonia would begin to exert its toxic effects. Hence, GS in brains of fishes are located in the cytosol (Figure 2); this specific location facilitates the detoxification of ammonia circulated in the blood, and prevents the brain from ammonia intoxication. In the case of marine elasmobranchs which depend on urea synthesized through carbamoyl phosphate synthetase III (CPS III) in the liver for osmoregulation, two isoforms of GS exist separately in the brain and the liver, which are localized in the cytosol and the mitochondria, respectively (Smith Jr. et al., 1983). In many tropical fishes, high levels of cerebral GS activities correlate well with their high environmental ammonia tolerance. As a result, the brain is often the organ undergoing the largest increases in glutamine content in fish exposed to ammonia. In mammal, it has been established that glutamine accumulation would lead to astrocyte swelling and brain edema. Recent studies have shown that glutamine can directly exert toxic effects on cultured astrocytes by increasing reactive oxygen species production and by inducing the mitochondrial permeability transition (Rama Rao et al., 2003; Jayakumar et al., 2004; Figure 2). How fish brains, especially those with high brain ammonia tolerance and accumulate high levels of glutamine, avoid these supposedly deleterious effects is uncertain at present. Since inhibiting mitochondrial glutamine hydrolysis in astrocytes of mammals mitigates many of the toxic effects of ammonia, it has been proposed that mitochondrial glutamine metabolism is involved in the mechanism of ammonia neurotoxicity (Pichili et al., 2007). Treatment of cultured astrocytes with l-histidine, an inhibitor of mitochondrial glutamine transport, completely blocked or significantly attenuated ammonia-induced reactive oxygen species production, cell swelling, mitochondrial permeability transition, and loss of ATP, indicating the intricate relationships between mitochondrial glutamine transport and intra-mitochondrial glutaminase in the mechanism of ammonia neurotoxicity (Pichili et al., 2007; Figure 2). Hence, it would be essential to examine the regulation of glutamine translocation across the inner membrane and the degradation of glutamine by glutaminase in the matrix of the brain mitochondria of fish species that can accumulate high levels of glutamine in the brain without any apparent deleterious effect.

He et al. (2010) reported recently that muscle GS is involved in glutamine production in mouse during fasting or undergoing extra-hepatic ammonia detoxification, although it has been established for almost a decade that ammonia can be detoxified to glutamine in non-cerebral tissues, including muscle, of some fish species. Certain air-breathing fishes, for instance, sleepers (the marble goby, O. marmoratus, and four-eyed sleeper, B. sinensis) belonging to the family Eleotridae and the swamp eel, M. albus, belonging to family Synbranchidae, can detoxify endogenous ammonia to glutamine in their livers and muscles during aerial exposure (Jow et al., 1999; Ip et al., 2001a; Tay et al., 2003) and ammonia-loading (Anderson et al., 2002; Ip et al., 2004f; Chew et al., 2005a). In B. sinensis, virtually all the GS activity in the liver is located in the cytosol, and it can be up-regulated through increased expression of the gene during ammonia exposure (Anderson et al., 2002). It is unlikely that the increased glutamine synthesis in the cytosol under such conditions is supported by an increase in the production of glutamate through the GDH amination reaction in the mitochondrial matrix of liver cells since the GDH would be involved in the transdeamination reaction (Figure 1). Glutamine synthesized is stored within the body, and it can be used for other anabolic processes (e.g., syntheses of purine, pyrimidine and mucopolysaccharides) when the environmental conditions become more favorable. The aquatic-breathing Gulf toadfish, O. beta, is another unique example; it detoxifies endogenous ammonia to glutamine to suppress ammonia excretion during confinement stress (Walsh and Milligan, 1995).

The fish intestine is a complex multifunctional organ; besides digestion and absorption, it is crucial for water and electrolyte balance, endocrine regulation of digestion and metabolism, and immunity. For juvenile marble goby O. marmorata kept in freshwater, it is noteworthy that the activity of GS from the intestine is six-fold higher than that from the liver (Chew et al., 2009). Similarly, intestinal GS activity is ∼5.7-fold higher than hepatic GS activity in the four-eyed sleeper, B. sinensis (Anderson et al., 2002). Mommsen et al. (2003) also reported high levels of GS activity in the gastrointestinal tract of the tilapia, O. mossambicus, and hypothesized that GS was concentrated toward the posterior section of the intestine for the efficient trapping of ammonia generated upstream in the intestinal tract as a by-product of digestion or through microbial processes. However, upon feeding, the glutamine content in the intestine of juvenile O. marmorata does not increase throughout the 24-h postprandial period in spite of a significant increase in the glutamate content therein and an increase in the GS activity at hour 12 (Tng et al., 2008). Instead, intestinal GDH activities, in both amination and deamination directions, increase significantly in juvenile O. marmorata 12 h after feeding. Increased production of glutamate after feeding can be an important adaptation to avoid or reduce postprandial ammonia toxicity (Tng et al., 2008). Subsequently, Chew et al. (2010) reported that exposure to seawater for 14 days also led to a significant increase in GDH activities, in the amination directions, and in GDH protein abundance from the intestine of juvenile O. marmorata. Their results (Chew et al., 2010) suggest that excess glutamate formed in the intestine could be transported to the liver and muscle to facilitate increased amino acid synthesis for the purpose of cell volume regulation (Chew et al., 2009). Furthermore, Peh et al. (2009) reported recently that the intestinal GDH was involved in ammonia detoxification in B. sinensis exposed to environmental ammonia in seawater. It was essential for the intestine to supply glutamate through increased GDH amination activity to the liver (and other organs) to support increased glutamine synthesis because liver is the main site of glutamate catabolism (Campbell, 1991). Excess free amino acids that are not used for protein synthesis and other essential functions are catabolized to ammonia and the corresponding α-keto acids through transdeamination in the liver (Campbell, 1991). Since transdeamination involves the degradation of glutamate, the hepatic GDH would have to operate mainly in the deamination direction. Hence, it would be essential for the intestine and liver to function cooperatively to detoxify ammonia to glutamine in B. sinensis exposed to environmental ammonia in seawater (Figure 1).

Differences in water and ionic regulation in seawater as opposed to freshwater may have rendered ammonotely disadvantageous in the marine environment (Campbell, 1973). However, it would be advantageous to accumulate urea as an osmolyte, and the function of the OUC in synthesizing urea for osmoregulatory purposes is seen today in elasmobranchs, holocephalans and coelacanths. Marine elasmobranchs are ureogenic because they possess a functional OUC in mitochondria of their livers and muscles (Anderson, 2001; Steele et al., 2005). They are also ureotelic, and urea is the primary product (>50%) of nitrogen excretion. In order to be able to retain urea for osmoregulation, the effective urea permeability in these cartilaginous fishes is decreased. This is achieved as a result of the presence of specific secondarily active (Na⁺ coupled) urea transporters in gills and kidney and modification of lipid composition of gills to achieve higher cholesterol:phospholipid ratios (Fines et al., 2001; Walsh and Smith, 2001).

In ureogenic fishes, the OUC consists of the enzyme CPS III, ornithine transcarbamoylase (OTC), argininosuccinate synthetase, argininosuccinate lyase and arginase. The primary ammonia-fixing enzyme, GS, CPS III, and OTC are located in the matrix of liver mitochondria. CPS III utilizes glutamine as a substrate and therefore ureogenic fishes possess high levels of GS in liver mitochondria. Since NH₃ is already detoxified to glutamine which is proton-neutral, the formation of carbamoyl phosphate via CPS III from glutamine in the liver mitochondrial matrix of fish cannot be regarded as an ammonia-detoxifying system (Figure 1). This is different from higher vertebrates, whose OUC involves CPS I which utilizes NH₃ directly as a substrate. For marine elasmobranchs, the localization of GS and CPS III in the liver mitochondrial matrix (Anderson and Casey, 1984) probably evolved more as a mechanism for urea synthesis for osmoregulatory purposes (Ip et al., 2005b; Chew et al., 2006a). In this way, glutamine is utilized directly for carbamoyl phosphate, and ultimately urea synthesis, instead of exiting the mitochondria to act as a substrate for other anabolic purposes in the cytosol. The absence of aspartate transcarbamylase activity in the liver of spiny dogfish (Anderson, 1989) is consistent with this view as is the absence of glutamine from the blood serum of marine elasmobranchs (Leech et al., 1979). Working together, GS, CPS III, and OTC form citrulline in the matrix, which then exits the mitochondria (Figure 1). Citrulline is converted to arginine via argininosuccinate synthetase and argininosuccinate lyase in the cytosol. Unlike mammals, arginase is located in the mitochondrial matrix of fish (except lungfishes; Mommsen and Walsh, 1991); so, arginine has to re-enter the matrix where it is converted to urea, regenerating ornithine for citrulline synthesis (Mommsen and Walsh, 1991). Like glutamine, citrulline is proton-neutral (Campbell, 1991, 1997); so, uncoupling oxidative phosphorylation would not occur.

For those fishes which are ureogenic but non-ureosmotic, the OUC with the involvement of CPS III in the liver can be involved in converting glutamine into urea, a much smaller molecule, for excretion. The rates of urea synthesis and excretion increase in the giant mudskipper P. schlosseri (Ip et al., 2004c) and the slender lungfish Protopterus dolloi (Lim et al., 2004) within 24 h after feeding. Studies on marine sharks (Wood et al., 2005) and freshwater stingray (Chew et al., 2006a) confirm that ureogenic elasmobranchs also increase the rate of urea synthesis after feeding, although urea is retained for the purpose of osmoregulation instead of being excreted.

Only a few teleosts are ureotelic, excreting more than 50% of the nitrogenous wastes as urea (e.g., the gulf toadfish O. beta under confined or crowded conditions; Walsh et al., 1990), or able to detoxify a minor quantity of ammonia to urea during ammonia-loading (Mugilogobius abei; Iwata et al., 2000). Opsanus beta, in its natural habitat, appears to excrete roughly 50:50 mixture of ammonia and urea (Barimo et al., 2007) as part of an elaborate mechanism to cloak the scent of ammonia to predators by urea (Barimo and Walsh, 2006). The majority of tropical air-breathing teleosts studied so far do not use ureogenesis as a major strategy to deal with endogenous (during aerial exposure) or exogenous and endogenous ammonia (during ammonia-loading). These include the mudskippers P. schlosseri, B. boddarti, and Periophthalmus modestus (Iwata and Deguichi, 1995; Peng et al., 1998; Lim et al., 2001), the marble goby O. marmoratus (Jow et al., 1999), the four-eyed sleeper B. sinensis (Ip et al., 2001a; Anderson et al., 2002), the oriental weatherloach Misgurnus anguillicaudatus (Chew et al., 2001; Tsui et al., 2002), the mangrove killifish Rivulus marmoratus (Frick and Wright, 2002), the small snakehead Channa asiatica (Chew et al., 2003c), the swamp eel M. albus (Tay et al., 2003; Ip et al., 2004f) and the climbing perch Anabas testudineus (Tay et al., 2006) exposed to terrestrial conditions or ammonia-loading for various periods. Ureogenesis in fishes is energetically intensive. For teleosts and elasmobranchs, a total of 5 mol of ATP are hydrolyzed for each mole of urea synthesized, equivalent to 2.5 mol ATP used for each mole of nitrogen assimilated (Ip et al., 2001b). It is probably because of this that ureogenesis is not commonly adopted as a single major strategy to handle ammonia toxicity when air-breathing tropical teleosts (Chew et al., 2005b) are exposed to terrestrial conditions or environmental ammonia.

To date, the only adult teleost for which unequivocal evidence is available for the OUC to function primarily for the purpose of detoxifying ammonia to urea is the tilapia Alcolapia grahami in Lakes Magadi and Nakuru in Kenya. It thrives in a highly alkaline environment (pH 10), in which NH₃ excretion would be totally impeded. As a result, it develops a high capacity to detoxify endogenous ammonia to urea via the OUC (Walsh et al., 1993). In fact, this is the first known example of complete ureotely in an entirely aquatic teleost fish (Randall et al., 1989). Alcolapia grahami contains in its liver significant levels of OUC enzymes and considerable GS activity. In addition, CPS III and all other OUC enzyme activities are present in the muscle at levels more than sufficient to account for the rate of urea excretion (Lindley et al., 1999). It is highly unusual that the muscle CPS can use NH₃ as a substrate; but, because of this adaptation, there is no need for GS and OUC to be tightly coupled, and GS is not well expressed in muscle (Lindley et al., 1999). So, like the OUC in mammals, NH₃ is directly converted via CPS and OTC to citrulline which exits the muscle mitochondria of A. grahami without uncoupling oxidative phosphorylation (Campbell, 1997). This is an important development for OUC to function primarily for ammonia detoxification, so that ammonia can be excreted as urea without being affected by the ambient alkaline water. The majority of the urea synthesized in A. grahami is excreted through the gills (Wood et al., 1994).

In contrast, African lungfishes, which belong to class: Sarcopterygii (lobe-finned fishes), despite being ammonotelic in water synthesize and accumulate urea during emersion and aestivation on land. African lungfishes are ureogenic and they possess a full complement of OUC enzymes (Janssens and Cohen, 1968; Mommsen and Walsh, 1989), including CPS III, in their livers (Chew et al., 2003b; Loong et al., 2005). Chew et al. (2004) demonstrated that the slender lungfish, P. dolloi, detoxified endogenous ammonia to urea, and the rate of urea synthesis increased 2.4- and 3.8-fold during 6 and 40 days (equivalent to 12 and 46 days, respectively, inclusive of the induction period), respectively, of aestivation in air. Subsequently, Loong et al. (2005) reported that the respective rates of urea synthesis in Protopterus aethiopicus and Protopterus annectens exposed to terrestrial conditions (without aestivation) for 6 days increased 1.2- and 1.5-fold, respectively. Why would African lungfish detoxify ammonia to urea? When urea is injected intraperitoneally into P. dolloi, only a small percentage (34%) of it is excreted during the subsequent 24-h period (Ip et al., 2005c). At hour 24, significant quantities of urea are retained in various tissues of P. dolloi, which lead to an apparent reduction in endogenous ammonia production, a significant decrease in the hepatic arginine content, and a significantly lower level of brain tryptophan in this lungfish (Ip et al., 2005c). All these three phenomena have been observed previously in aestivating P. dolloi (Chew et al., 2004). So, increased urea synthesis and accumulation may have a physiological role in initiating and perpetuating aestivation in this lungfish (see Ip and Chew, 2010 for a review). Moreover, accumulation of urea in an aestivating lungfish may reduce the rate of evaporative water loss through vapor pressure depression.

Theoretically, the most effective way to defend against ammonia toxicity in fish in alkaline water, during emersion or when exposed to ammonia-loading conditions is active transport of NH4+ (as have been suggested for the inner mitochondrial membrane of ammonotelic fishes; Campbell, 1997), because it facilitates the maintenance of low internal ammonia levels and prevents the brain from ammonia toxicity. Indeed, some air-breathing tropical fishes with modified gill structures or accessory breathing organs are capable of doing so (the giant mudskipper, Randall et al., 1999; Chew et al., 2003a; Ip et al., 2004d; the African sharptooth catfish, Ip et al., 2004e; the climbing perch, Tay et al., 2006).

The gills of the giant mudskipper P. schlosseri are specially adapted for terrestrial survival (Ip et al., 1993; Kok et al., 1998) because of its specialized gill morphology and morphometry (Low et al., 1988, 1990; Wilson et al., 1999, 2000). There are intrafilamentous interlamellar fusions in its gills, which form numerous fenestrae on the surface of the gill filament. When 8 μmol g⁻¹ ammonium acetate was injected intraperitoneally into P. schlosseri followed with 24 h of aerial exposure, the cumulative ammonia excreted was significantly greater than the saline-injected control (Chew et al., 2007). At hour 2, the ammonia excretion rate in the experimental fish was transiently greater than that in fish kept in water. By hour 6, a large portion (33%) of the injected ammonia was excreted by the experimental fish through the head region, probably through the gills (Chew et al., 2007). Because ammonia could be excreted only into the small amount of water trapped in the fenestrae of the fused secondary lamellae in the gills of P. schlosseri during emersion, ammonia concentration build up quickly therein and reached high level (∼90 mmol l⁻¹). So, P. schlosseri can effectively excrete a high load of ammonia on land, which indirectly supports the proposition that active NH4+ excretion contributes in part to its high terrestrial affinity and high tolerance of aerial exposure.

In addition, P. schlosseri can maintain an unchanged ammonia excretion rate when exposed to alkaline water (Chew et al., 2003a). It can also tolerate very high concentrations of environmental ammonia (>100 mmol l⁻¹ NH₄Cl; Ip et al., 1993; Peng et al., 1998). When exposed to environmental ammonia, P. schlosseri is able to maintain low concentrations of ammonia in its plasma (Peng et al., 1998; Randall et al., 1999) as a result of its ability to actively excrete ammonia against large inward NH₃ and NH4+ gradients (up to at least 30 mmol l⁻¹ NH₄Cl, pH 7.2; Randall et al., 1999). The mechanisms apparently involve Na⁺/K⁺ (NH4+)-ATPase at the basalemma and Na⁺/H⁺ (NH4+) exchangers at the apical membrane of the branchial epithelia. NH4+ substitutes for K⁺ in a number of ion transport proteins owing to similarity in their hydration radii. Indeed, active ammonia excretion in P. schlosseri exposed to environmental ammonia is sensitive to ouabain, an inhibitor of Na⁺/K⁺-ATPase, and NH4+ can replace K⁺ as a substrate for ATPase activity in vitro (Randall et al., 1999). Moreover, the addition of amiloride, an inhibitor of NHE, to the external medium decreases the rate of ammonia excretion in P. schlosseri (Randall et al., 1999). NHE2- and NHE3-like proteins have been shown to be present in the apical crypts of branchial mitochondria-rich cells (Wilson et al., 2000). Since plasma Na⁺ level increases in specimens exposed to high concentrations of environmental ammonia (Randall et al., 1999), NH4+ must be actively transported across the branchial epithelium in exchange with Na⁺. The possibility of NH₃ trapping in the active NH4+ excretion of ammonia can be eliminated because the addition of buffer to the medium, which removes the acid boundary layer, has no effect on the rate of ammonia excretion (Wilson et al., 2000; Ip et al., 2004d). Also the addition of bafilomycin A1, a vesicular-type H⁺-ATPase inhibitor, significantly decreases the net acid flux but has no effect on the rate of ammonia excretion (Ip et al., 2004d). The possible role of Rh glycoproteins in active NH4+ excretion in the giant mudskipper is uncertain at present, but Weihrauch et al. (2009) has speculated that these proteins are expressed at high levels in the gills through which NH₃ is transported out of the fish along with CO₂.

For fishes living in stagnant water of a finite volume (puddle, tidepool, or water-filled burrow) it is possible for the acidifying effects of their excretions of H⁺ and CO₂ to have a significant impact on environmental pH. This lowering of environmental pH has advantages for dealing with elevations in environmental ammonia levels by reducing the concentration of NH₃, the more permeant species of ammonia, constituting “environmental ammonia detoxification” (Chew et al., 2003a).

Mudskippers build burrows on the mud flat, and lay eggs therein during the breeding season. In one study, the pH of the water in a canal that supplied water to a mud flat was 7.84, yet the pH of the water sampled from burrows of the giant mudskipper P. schlosseri was close to 7.0 (Ip et al., 2004d). This indicates that P. schlosseri modifies the pH of the burrow water, confirming laboratory observations made by Chew et al. (2003a). Lowering the pH of the burrow water from 8.3 (pH of full strength seawater) to 7.0 increases the ratio of NH4+ to NH₃ which lowers the toxicity of NH₃ to the fish and embryos in the burrows. When P. schlosseri is kept in 10 volumes (w/v) of 50% seawater with 2 mmol l⁻¹ of Tris–HCl at pH 8.5 or 9.0, the decrease in pH is large and rapid (Chew et al., 2003a; Ip et al., 2004d). Thus, large quantities of acid must be excreted to manipulate the external pH of alkaline waters. More importantly, P. schlosseri is capable of increasing the rate of net acid excretion in response to the presence of 20 or 30 mmol l⁻¹ NH₄Cl in the water at pH 7.0 or 8.0 (Ip et al., 2004d). This represents direct evidence linking net acid excretion with defense against environmental ammonia toxicity in P. schlosseri.

As a result of active excretion of NH4+ into a finite volume of water in the burrow, the ambient ammonia concentration increases. Therefore, it is imperative for the excreted NH4+ not to dissociate into NH₃ and H⁺, because NH₃ would diffuse back into the body down an inwardly directed ΔPNH3. Because acid excretion is responsive to environmental ammonia, there is a continuous excretion of acid even at neutral pH. This would maintain a low pH (high concentrations of H⁺) in the boundary water layer of the branchial epithelia, preventing the excreted NH4+ from dissociating into NH₃ and H⁺, and avoiding back flux of NH₃. In essence, this is the process of “ NH4+ trapping”. Indeed, both active NH4+ excretion and H⁺ excretion take place in the head region of P. schlosseri, where the gills and the opercular membranes are located (Ip et al., 2004d). It is essential for these two mechanisms to be located together, because the branchial and opercular surfaces have the important functions of allowing passage of gases and other ions. Excretion of acid to trap the actively excreted NH4+ is likely to be more effective than modifying the fluidity of these surfaces to change the permeability of NH₃, which would also affect the permeability of other gaseous molecules.

The branchial epithelial surface of aquatic teleosts has a higher permeability to NH₃ due to its major function in gaseous exchange. Since air-breathing fishes depend largely on the accessory breathing organ for respiration, their gills are usually degenerate (Graham, 1997). Moreover, they usually hold air in their buccal cavities during immersion, which means their gills would not be exposed to the external medium, and NH₃ excretion through the gills would be relatively ineffective. However, many tropical air-breathing fishes substitute branchial respiration with cutaneous respiration by having highly vascularized skins. Therefore, a substantial portion of ammonia excretion takes place through the cutaneous surface of air-breathing fishes (Graham, 1997), and fish skin is known to express Rh glycoproteins (see Weihrauch et al., 2009 for a review). For fishes which are exposed frequently to environmental ammonia, it would be essential for them to reduce the permeability of their skins to NH₃ despite the cell membranes being permeable to gaseous molecules like O₂ and CO₂, even though the permeability of NH₃ is less than those of O₂ and CO₂ (Marcaggi and Coles, 2001). Moreover, under ammonia-loading conditions, it would be essential to reduce or suppress the expression of transporters that would facilitate ammonia transport in the skin of these fishes. Although it has been reported that Rhcg1 gene expression increases in the skin of the mangrove killifish, K. marmoratus, exposed to environmental ammonia (Hung et al., 2007), it is important to take note that this fish, unlike many others, is capable of volatilizing ammonia through its skin under certain conditions (Frick and Wright, 2002).

A certain amount of NH₃ permeates the membrane by solvation and diffusion in the lipid bilayer. The lipid–water partition coefficient for ammonia is low (Evans and Cameron, 1986), suggesting that membrane permeability to ammonia is generally low. In certain biological situations, however, membrane permeability to ammonia is further reduced (Kikeri et al., 1989), and several mechanisms for reducing the permeability of membranes to ammonia are apparent from model studies. In particular, the cholesterol and phospholipid fatty acid contents of artificial membranes have been shown to affect the permeability or artificial membranes to ammonia (Lande et al., 1995). When a fish is confronted with ammonia-loading conditions, a reduction in NH₃ permeability of the skin would help reduce the influx of NH₃. In the giant mudskipper, this is an important adaptation which complements active NH4+ excretion through its gills (Randall et al., 1999; Ip et al., 2004d), because it would prevent a back diffusion of NH₃ through the cutaneous surfaces after the build up of high ammonia concentrations in the external medium (Ip et al., 2004d).

Indeed, the flux of NH₃ through the skin of P. schlosseri in an Ussing-like apparatus at 25°C down a 10-fold NH₄Cl gradient between two media with 1 unit pH difference (pH 8 → pH 7) is only 0.009 μmol min⁻¹ cm⁻² (Ip et al., 2004d). The low NH₃ permeability in the skin of P. schlosseri is likely due to its low membrane fluidity (Lande et al., 1995). The phosphatidylcholine:phosphatidylethanolamine ratio, an indication of membrane fluidity from the skin of P. schlosseri is 3.4, much higher than those of tissues of other fish species (Hazel and Landrey, 1988). Furthermore, the skin of P. schlosseri has very high cholesterol content (4.5 μmol g⁻¹), which lowers the fluidity of biomembranes. The cholesterol content in the skin of P. schlosseri increased significantly to 5.5 μmol g⁻¹ after 6 days of ammonia exposure (Ip et al., 2004d), suggesting a role for cholesterol as a defense mechanism against environmental ammonia toxicity.

African lungfishes which can tolerate at least 100 mmol l⁻¹ NH₄Cl may also have low permeability to NH₃ in the skin. Chew et al. (2005b) estimated the flux of NH₃ through the skin of P. dolloi to be only 0.003 μmol min⁻¹ cm⁻², which is even lower than that of the giant mudskipper P. schlosseri. Subsequently, Loong et al. (2007) reported that the skin of P. aethiopicus also had low permeability to NH₃ in an Ussing-like apparatus. Indeed, the influx of exogenous ammonia into fish exposed to 30 mmol l⁻¹ NH₄Cl was low (0.117 μmol min⁻¹ 100 g⁻¹ fish). As a result, P. aethiopicus could afford to maintain relatively low ammonia contents in its plasma, muscle, liver and brain even after 6 days of exposure to 100 mmol l⁻¹ NH₄Cl. Surprisingly, fish exposed to 30 or 100 mmol l⁻¹ NH₄Cl had comparable ammonia contents in the muscle and the brain in spite of the big difference (70 mmol l⁻¹) in environmental ammonia concentrations. Results of Loong et al. (2007) suggest that P. aethiopicus was capable of decreasing the NH₃ permeability of its body surface when exposed to high concentrations of environmental ammonia. Indeed, after 6 days of exposure to 100 mmol l⁻¹ NH₄Cl the NH₃ permeability constant of the skin decreased to half of that of the control. A decrease in the already low cutaneous NH₃ permeability and an increased urea synthesis, working in combination, allowed P. aethiopicus to effectively defend against environmental ammonia toxicity without elevating the plasma ammonia level. Therefore, unlike other fishes, glutamine and alanine contents did not increase in the muscle and liver, and there was no accumulation of glutamine in the brain, even when the fish was immersed in water containing 100 mmol l⁻¹ NH₄Cl.

Since ammonia can exist as NH₃ gas, it is logical that certain fish may be able to excrete NH₃ directly into the atmosphere. In teleosts, ammonia volatilization was first reported in the temperate intertidal blenny (Blennius pholis), but it only accounted for 8% of the total ammonia excreted during emersion (Davenport and Sayer, 1986). However, the ammonotelic tropical fishes Alticus kirki (the leaping blenny) (Rozemeijer and Plaut, 1993), K. marmoratus (the mangrove killifish) (Frick and Wright, 2002), and M. anguillicaudatus (the oriental weatherloach) (Tsui et al., 2002) are capable of volatilizing significant amounts of ammonia during aerial exposure. High temperatures and humidity increase the likelihood of the ammonia excreted into the film of water covering the body surface of tropical fishes being volatilized.

In the case of M. anguillicaudatus, ammonia built up to very high levels in the muscle (14.8 μmol g⁻¹), liver (15.2 μmol g⁻¹), and plasma (5.09 μmol ml⁻¹) after 48 h of emersion (Chew et al., 2001). Building up of internal ammonia may be an essential prerequisite for volatilization to occur. The blood pH simultaneously becomes more alkaline; this would lead to a high level of NH₃ in the blood. It is possible that a much higher ΔPNH3 gradient is required to facilitate the efflux of ammonia through non-branchial epithelial surfaces. In M. anguillicaudatus, there appears to be at least two sites of NH₃ volatilization, i.e., the skin and the digestive tract, under terrestrial conditions. During emersion, the surface of the skin becomes significantly more alkaline (Tsui et al., 2002). Moreover, the pH of the water underneath the fish is significantly higher than that of the submerged control. Since the fraction of NH₃ present is strongly influenced by the pH of the medium, the skin of M. anguillicaudatus can be a site of NH₃ volatilization. Simultaneously, the pH of the mucosal surface of the anterior portion of the digestive tract becomes significantly more alkaline in M. anguillicaudatus exposed to terrestrial conditions (Tsui et al., 2002). Although the skin and the digestive tract may both be involved in NH₃ volatilization when the fish is moving on land, excretion of ammonia through the skin would be ineffective when the fish burrows into the mud. In the mud, maintaining the skin surface or the surrounding film of water at pH 8 would aggravate the situation, because of the reversed ΔPNH3 gradient. When the fish is surrounded by mud, the digestive tract (either through the mouth or the cloaca) becomes the only avenue for NH₃ volatilization to occur.

Recently, Moreira-Silva et al. (2010) reported an increase in membrane fluidity, which would presumably enhance NH₃ permeation through the plasma membrane, of the gills of M. anguillicaudatus exposed to terrestrial conditions. Additionally, there was a significant increase in the mRNA expression of the ammonia transport protein, Rhcg1 (Slc42a3), in the gills during aerial exposure. The colocalization of H⁺-ATPase and Rhcg1 to the same cell type suggested a role for H⁺-ATPase in ammonia excretion via Rhcg1 by NH4+ trapping (Figure 4), confirming the importance of gill boundary layer acidification in net ammonia excretion in this fish. By contrast, there was no significant change in the mRNA expression of Rhcg1 in the gills of M. anguillicaudatus exposed to high concentrations of environmental ammonia.

As for K. marmoratus, the cutaneous surface is alkalinized by 0.4–0.5 pH units during emersion, promoting the accumulation of NH₃ on the skin surface (Litwiller et al., 2006). Hung et al. (2007) demonstrated Rh glycoprotein mRNA expression in the gills and skin and verified that its expression was sensitive to both increased environmental ammonia concentrations and air exposure. Since Rhcg1 and Rhcg2 expression increased many fold in the skin following air exposure, Hung et al. (2007) concluded that Rhc glycoproteins promoted NH₃ passage from the blood to skin surface during emersion.