Inhibition of glutamine synthetase during ammonia exposure in rainbow trout indicates a high reserve capacity to prevent brain ammonia toxicity

Ammonia, a breakdown product of amino acid catabolism, can be a potent neurotoxin in vertebrates if allowed to accumulate (Felipo and Butterworth, 2002; Randall and Tsui, 2002). This potential for neurotoxicity makes it crucial for organisms to effectively excrete ammonia. In teleosts, when exogenous ammonia concentrations are low, ammonia is passively excreted across the gills down the blood-to-water diffusion gradient (Cameron and Heisler, 1983; Wright and Wood, 1985). In contrast, when fish encounter high environmental ammonia levels, excretion of nitrogenous waste can become impaired as the concentration gradient is reduced or even reversed (e.g. Cameron and Heisler, 1983; Claiborne and Evans, 1988). In post-larval teleosts such as rainbow trout (Oncorhynchus mykiss) where the ornithine urea cycle is not functional, evidence suggests that the major ammonia detoxifying mechanism involves the conversion of ammonia and glutamate to glutamine via the enzyme glutamine synthetase (GSase; EC6.3.1.2) (reviewed by Ip et al., 2001; Randall and Tsui, 2002). In both fish and mammals, the activity of GSase is much higher in the brain than in other tissues (Cooper and Plum, 1987; Wang and Walsh, 2000; Wright et al., 2007) and it is suspected that the primary functions of GSase in the brain are to provide protection against ammonia toxicity and maintain homeostasis of the excitatory neurotransmitter glutamate (Suárez et al., 2002).

In some fish species, but not in others, brain GSase activity (Wang and Walsh, 2000; Ip et al., 2001; Wicks and Randall, 2002) and mRNA levels (Wright et al., 2007) have been demonstrated to increase in response to ammonia exposure. However, regardless of whether GSase activity is stimulated or not, brain glutamine concentrations in ammonia-exposed teleosts generally increase (Levi et al., 1974; Wang and Walsh, 2000; Wicks and Randall, 2002; Ip et al., 2004; Veauvy et al., 2005; Wee et al., 2007). In some species, the increase in brain glutamine is accompanied by a decrease in glutamate (Arillo et al., 1981; Vedel et al., 1998; Wicks and Randall, 2002; Veauvy et al., 2005). Interestingly, despite the overall reduction in brain glutamate, there is evidence suggesting that ammonia toxicity in some fish species is at least partly mediated by an overactivation of the glutamate NMDA receptors (Walsh et al., 2007).

Glutamate can also stimulate the hypothalamic–pituitary–adrenal (HPA) stress axis in mammals and cause the release of corticotropin-releasing factor (CRF) from hypophysiotropic neurons (Joanny et al., 1997). CRF stimulates the secretion of adrenocorticotropin hormone from the pituitary, which in turn stimulates the synthesis and release of glucocorticoids from the adrenal cortex of mammals and the interrenal tissue of fish (Wenderlaar Bonga, 1997; Bernier et al., 2009). In rainbow trout, while chronic ammonia exposure leads to an increase in plasma cortisol concentrations and an increase in brain preoptic area CRF mRNA expression (Wicks and Randall, 2002; Ortega et al., 2005; Bernier et al., 2008), it is not known whether glutamate also plays a role in the regulation of the hypothalamic–pituitary–interrenal (HPI) stress axis. The increase in plasma cortisol during ammonia exposure is transient, with levels returning to baseline by 96 h (Ortega et al., 2005; Bernier et al., 2008). It has been suggested that the decrease in cortisol over time is due to the conversion of ammonia and glutamate to glutamine, catalysed by GSase (Ortega et al., 2005). Indeed, several teleost studies have shown an increase in tissue-specific GSase activity or an upregulation in GSase mRNA expression with elevations in plasma cortisol (Walsh et al., 1994; Hopkins et al., 1995; Aluru and Vijayan, 2007).

In this study, we hypothesized that high brain GSase activity is critical in preventing the accumulation of brain ammonia and glutamate during episodes of hyperammonia in the ammonia-intolerant rainbow trout. If so, then inhibition of brain GSase will result in decreased survival and increased brain ammonia and glutamate concentrations and plasma cortisol when fish are exposed to elevated external ammonia levels. To test this hypothesis, we used the GSase inhibitor methionine sulfoximine (MSOX), which has been employed effectively in previous studies on fish (Ip et al., 2005; Veauvy et al., 2005; Wee et al., 2007). Four treatment groups were established in order to investigate the role of brain GSase during ammonia exposure and the influence of MSOX: saline-control, MSOX-control, saline-ammonia and MSOX-ammonia. Finally, experiments were performed over 96 h in order to determine the response over time.

Juvenile rainbow trout (Oncorhynchus mykiss Walbaum 1792) were obtained from Rainbow Springs Trout Farm (Thamesford, ON, Canada). Fish were housed in 758 l tanks in the Hagan Aqualab (Guelph, ON, Canada) under a 12 h:12 h light:dark photoperiod (16°C), and were fed 1% body mass of commercial trout pellets three times a week. At least 2 weeks before experimentation, fish were transferred into 125 l tanks in groups of 8–10 and allowed to acclimate. Fish were fasted for 48 h prior to commencement of the experiments to reduce the effects of feeding history on the experimental results. The University of Guelph's Animal Care Committee approved care and use of the animals, as per the principles of the Canadian Council for Animal Care.

A series of trial dosages of the GSase inhibitor MSOX (Sigma-Aldrich Chemical Co., St Louis, MO, USA) were tested on rainbow trout to determine the quantity required to chronically reduce brain GSase activity by 75% for a 96 h period. Fish (N=2–6 per dosage) were anaesthetized with buffered (NaHCO, 0.2 g l^–1 3) tricaine methanesulphonate (MS-222; 0.1 g l^–1), weighed, and injected intraperitoneally (i.p.) with either saline (0.9% NaCl) or MSOX dosages ranging from 3 to 150 mg kg^–1 body mass (2.6 μl g^–1 body mass). Fish mass ranged between 56.2 and 74.6 g. After each trial, surviving fish were terminally anaesthetized with 2-phenoxyethanol (2 ml l^–1), and their brains were surgically removed, wrapped in foil and flash frozen in liquid nitrogen for later analysis of brain GSase activity. In a final trial, a dosage of 6 mg kg^–1 was tested in fish exposed to elevated environmental ammonia (740±10 μmol l^–1 NH₄Cl) for 96 h.

To determine the effects of NH₄Cl exposure with or without prior treatment with MSOX, rainbow trout were divided into 16 groups of 8–10 fish. The experiment consisted of four treatment groups: (1) fish in control water and injected with saline, (2) fish in control water and injected with MSOX, (3) fish exposed to NH₄Cl and injected with saline and (4) fish exposed to NH₄Cl and injected with MSOX. An experiment was conducted on fish exposed to 670 μmol l^–1 NH₄Cl (fish mass 80.7±1.0 g, N=160) and then the experiment was repeated on a separate group of fish exposed to 1000 μmol l^–1 NH₄Cl (fish mass 103.8±3.1 g; N=128). The lower ammonia concentration (670 μmol l^–1 NH₄Cl) was chosen to match that in our previous study on GSase gene expression in trout brain (Wright et al., 2007). The second experiment at a higher ammonia concentration (1000 μmol l^–1) was performed to maximize the ammonia load to the fish while not exceeding the toxic level (Ortega et al., 2005). It should be noted that this higher exposure level was close to the lethal limit of these fish under the conditions tested, as slightly higher levels (1200 μmol l^–1 NH₄Cl) resulted in 100% mortality.

Each treatment was replicated in duplicate tanks and the experiment lasted either 24 or 96 h. Fish were anaesthetized with buffered (NaHCO₃, 0.2 g l^–1) MS-222 (0.1 g l^–1), weighed, and injected i.p. with either saline (0.9% NaCl) or MSOX at a dosage of 6 mg kg^–1 body mass (2.6 μl g^–1 body mass) as determined from the dosage trials. Four hours later, eight groups of fish were exposed to chronic flow-through ammonia conditions and the other eight groups were maintained in control water and served as time controls. Using a prepared stock of NH₄Cl, ammonia levels in the tanks were increased instantaneously to 670 or 1000 μmol l^–1 and maintained at these levels using a peristaltic pump (Minipuls 3, Gilson, Villiers Le Bel, France). Water samples were periodically taken throughout the exposures and frozen at –20°C for later analysis of ammonia content. Average water temperature throughout the 670 μmol l^–1 NH₄Cl exposure was 16.4±0.2°C and pH was 8.02±0.01, whereas in the 1000 μmol l^–1 NH₄Cl exposure the values were 15.9±0.1°C and pH was 8.42±0.01. One mortality was recorded in a control MSOX 96 h tank of the 670 μmol l^–1 NH₄Cl exposure. In the 1000 μmol l^–1 NH₄Cl exposure, there were three mortalities recorded in one MSOX-control tank, two mortalities in one saline-ammonia tank, and one mortality in each of the MSOX-ammonia tanks. A few fish demonstrated a loss of equilibrium in response to 1000 μmol l^–1 NH₄Cl – they either recovered or it occurred towards the end of the treatment and they were thus kept within the experiment.

At the end of the 24 and 96 h exposures, fish were quickly netted and terminally anaesthetized with an overdose of 2-phenoxyethanol (2 ml l^–1). Blood was drawn via caudal puncture using a Na₂EDTA-treated syringe, centrifuged for 2 min at 12,000 g, and the separated plasma was stored at –20°C for later analysis of plasma ammonia and cortisol concentrations. To quantify CRF mRNA levels, the brain from half of the fish in each tank was regionally dissected to isolate the pre-optic area as described previously (Bernier et al., 2008) and immediately frozen in liquid nitrogen. Whole brains were collected from the remaining fish, flash-frozen in liquid nitrogen, and stored at –80°C until analysis of ammonia levels, GSase activity and amino acid concentrations. Similarly, livers were collected and stored as above until analysis of ammonia levels and GSase activity.

In order to determine whether MSOX has an effect on the cortisol response of rainbow trout subjected to stressors other than ammonia, we performed an additional experiment, a standardized 60 s handling stress as previously described by Barton and Iwama (Barton and Iwama, 1991). Three groups of 10 fish each (107.7±1.7 g, N=30) were acclimated for at least 2 weeks in separate 125 l tanks. One group was left undisturbed and served as a control. The two other groups were anaesthetized with buffered MS-222 as above and while one group was injected i.p. with 0.9% saline the other received MSOX at a dosage of 6 mg kg^–1 body mass (2.6 μl g^–1 body mass). Twenty-eight hours later, in accordance with the timeline established above for a 24 h ammonia exposure, the saline- and MSOX-injected groups were each netted out of the water for 60 s and returned to their respective tanks. One hour after the handling stressor, all three groups of fish were terminally anaesthetized with 2-phenoxyethanol (2 ml l^–1). Blood was drawn via caudal puncture, spun and the separated plasma was stored at –20°C for later analysis of plasma cortisol concentration. Whole brains were also collected, flash frozen in liquid nitrogen, and stored at –80°C until analysis of GSase activity.

Water ammonia concentration was determined using a colorimetric assay (Verdouw et al., 1978). To measure plasma, liver and brain ammonia concentrations, tissues were prepared as described previously (Wright et al., 2007). Plasma and liver ammonia levels were analysed using the glutamate dehydrogenase (GDH) method described by Kun and Kearney (Kun and Kearney, 1974) with micro-modifications for use in a 96-well microplate (SpectraMAX 190; Molecular Devices, Sunnyvale, CA, USA). Brain ammonia concentrations were analysed using a commercial kit (Raichem Catalog No. 85446; Cliniqa, San Marcos, CA, USA) and a spectrophotometer (Cary 300 Bio UV/Visible; Varian, Palo Alta, CA, USA).

Brain and liver GSase activity was assayed based on the production of γ-glutamyl hydroxamate as previously described (Shankar and Anderson, 1985), with the following modifications. Ground brain tissue was homogenized in 700 volumes of ice-cold homogenization buffer and the final reaction was performed over 5 min at 15°C. Ground liver tissue was homogenized in 9 volumes of ice-cold homogenization buffer and the final reaction was performed over 15 min at 15°C. Samples were read at 500 nm in a spectrophotometer (Ultrospec 3100 pro UV/Visible; Biochrom Ltd, Cambridge, UK).

Brain amino acid concentrations were quantified using the EZ:faast amino acid analysis kit (Phenomenex, Torrance, CA, USA) run on a Waters Micromass Quattro Micro API tandem mass spectrometer (Waters, Milford, MA, USA). Samples (80 mg) were ground to a fine powder under liquid nitrogen and homogenized in 4 volumes of 0.5% trifluoroacetic acid in methanol. A buffer (final concentrations, 0.11 mmol l^–1 sodium acetate, 0.056 mmol l^–1 NaOH) was added to bring the final volume to 1150 μl prior to centrifugation at 10,000 g (10 min, 4°C). Further details are provided in a previous publication (Speers-Roesch et al., 2006). Plasma cortisol was assayed using a validated RIA as described elsewhere (Bernier et al., 2008).

Total RNA was extracted using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and cDNA synthesis was carried out as previously described (Wright et al., 2007). mRNA expression of CRF was quantified from the above cDNA products using SYBR Green and an ABI Prism 7000 sequence detection system (Applied Biosystems, Foster City, CA, USA) as outlined before (Ortega et al., 2005). To correct for variations in template input and transcriptional efficiency, each sample was normalized to the expression of the reference gene, elongation factor 1α (EF1α). The primers used were as follows: EF1α forward 5′-GGG CAA GGG CTC TTT CAA GT-3′ and reverse 5′-CGC AAT CAG CCT GAG AGG T-3′ (GenBank accession no. AF498320), CRF forward 5′-ACA ACG ACT CAA CTG AAG ATC TCG-3′ and reverse 5′-AGG AAA TTG AGC TTC ATG TCA GG-3′ (GenBank accession no. AF296672).

Statistical analyses were performed using SigmaStat 3.0 (SPSS Inc., Chicago, IL, USA). Differences among treatments at a given time were analysed using one-way analysis of variance (ANOVA) followed by a Tukey post-hoc test if differences were found. If the data for a given variable were not normally distributed, the data were log transformed and retested. If data were not normalized after transformation, a Kruskal–Wallis one-way ANOVA on ranks was used. Differences among times for a given treatment were analysed using a t-test or a Mann–Whitney rank sum tests if log transformation failed to normalize the data. The significance level for all statistical tests was set at P<0.05. All data are presented as means ± s.e.m.

Concentrations of MSOX over 30 mg kg^–1 body mass were lethal within 24 h post-injection without the addition of ammonia. MSOX dosages of 3 and 6 mg kg^–1 resulted in 71% (24 h post-injection) and 88% (96 h post-injection) decreases in brain GSase activity relative to saline-injected controls, respectively, and caused no mortalities even with exposure to 740 μmol l^–1 NH₄Cl for 96 h.

The average water ammonia concentration in the control tanks was higher at 96 h (13±3 and 12±3 μmol l^–1 in the 670 and 1000 μmol l^–1 NH₄Cl experiments, respectively) than that at 24 h (2±0 and 2±1 μmol l^–1 in the 670 and 1000 μmol l^–1 NH₄Cl experiments, respectively), although overall water ammonia levels in the control tanks were very low relative to the ammonia-treated tanks. Water ammonia concentrations remained elevated throughout the experiment (24 h: 668±29 and 1040±31 μmol l^–1 in the 670 and 1000 μmol l^–1 NH₄Cl experiments, respectively; and 96 h: 678±9 and 934±10 μmol l^–1 in the 670 and 1000 μmol l^–1 NH₄Cl experiments, respectively) and there were no significant differences between duplicate treatment tanks or between time periods.

The physiological responses of fish to ammonia exposure and/or MSOX injection between 670 and 1000 μmol l^–1 NH₄Cl exposures were very similar. The magnitude of the increases in plasma, brain and liver ammonia concentrations were more pronounced during the 1000 μmol l^–1 NH₄Cl exposure relative to the 670 μmol l^–1 NH₄Cl exposure. We chose to present data from both exposures to demonstrate the overall consistency of the changes in brain chemistry and highlight a few dissimilar results with different concentrations of external ammonia.

Ammonia concentrations in the plasma, brain and liver were all significantly elevated in fish exposed to ammonia relative to control fish and there were no differences between the saline- and MSOX-injected fish after either 24 or 96 h of ammonia exposure (Fig. 1). On average, plasma ammonia concentrations increased 16-fold, brain ammonia concentrations increased 4.6-fold and liver ammonia concentrations increased 4.1-fold in response to ammonia exposure. Within a given treatment, there was no change in brain ammonia concentrations between exposure durations, increases in liver ammonia concentrations between the 24 h and 96 h exposures in the 670 μmol l^–1 NH₄Cl treatments, and no change in liver ammonia concentrations in the 1000 μmol l^–1 NH₄Cl exposures. Finally, while plasma ammonia levels increased between 24 h and 96 h of exposure to 670 μmol l^–1 NH₄Cl in the saline–ammonia treatment, the levels decreased between exposure durations in the fish exposed to 1000 μmol l^–1 NH₄Cl and pre-injected with MSOX.

Exposure to NH₄Cl had little effect on either brain or liver GSase activity in the saline-injected fish (Fig. 2) but a small increase in brain GSase after 96 h of exposure to 670 μmol l^–1 NH₄Cl was detected (Fig. 2A). In contrast, brain GSase activity was inhibited by 71–78% in fish injected with MSOX relative to saline-injected fish (Fig. 2A,B). In the liver, the effects of MSOX were less dramatic. MSOX inhibited liver GSase activity in the control and ammonia-exposed fish by 40–60% relative to the saline-injected fish at 24 h, but in most cases MSOX no longer had an inhibitory affect after 96 h (Fig. 2C,D).

In saline-injected fish, brain glutamate concentrations (Fig. 3A,B) were generally lower and glutamine concentrations (Fig. 3C,D) higher after the NH₄Cl treatments compared with control conditions. Brain glutamine levels during ammonia exposure were consistently higher after 96 h relative to 24 h (Fig. 3C,D). Pre-injection of MSOX had little impact on the changes in brain glutamate concentrations associated with hyperammonia (Fig. 3A,B). In contrast, by 96 h, the MSOX treatment was characterized by significant reductions in brain glutamine levels in both the control and ammonia-exposed fish (Fig. 3C,D).

In addition to glutamate and glutamine levels, we also quantified the impact of ammonia exposure and MSOX treatment on the brain concentration of another 20 amino acids. Elevated brain ammonia was associated with a decrease in the concentration of four amino acids (alanine, aspartate, GABA and serine; Fig. 4A,B; Tables 1 and 2) and an increase in the concentration of six amino acids (arginine, phenylalanine, threonine, tyrosine, tryptophan and valine; Fig. 4C,D; Fig. 5A,B; Tables 1 and 2). In fish pre-injected with MSOX prior to ammonia exposure, the above changes in brain amino acid concentrations were reversed or eliminated (Figs 4 and 5; Tables 1 and 2).

In saline-injected fish, ammonia exposure elicited a transient increase in plasma cortisol at 670 μmol l^–1 NH₄Cl (Fig. 5C), but a more prolonged increase at 1000 μmol l^–1 NH₄Cl (Fig. 5D). In fish pre-treated with MSOX, this increase in plasma cortisol with NH₄Cl exposure was eliminated at 670 μmol l^–1 NH₄Cl (Fig. 5C) but not at 1000 μmol l^–1 NH₄Cl (Fig. 5D). Accompanying these changes, brain CRF mRNA levels were significantly elevated in fish exposed to 1000 μmol l^–1 NH₄Cl, especially in the MSOX-treated fish (Table 3).

Pre-treatment of fish with MSOX resulted in significantly lower brain GSase activity (5-fold) relative to both unhandled fish and fish handled for 60 s and injected with saline (Fig. 6A). Independent of whether the fish were injected with saline or MSOX, the handling stressor resulted in significantly higher plasma cortisol concentrations relative to the unhandled fish 1 h after the disturbance (Fig. 6B).

The findings of this study demonstrate that rainbow trout have a considerable reserve capacity to prevent brain ammonia toxicity. Trout survived an inhibition of brain GSase by ∼75% coupled to 96 h of exposure to high external ammonia concentrations (1000 μmol l^–1), very close to their lethal limit. We predicted that if fish survived under these conditions, then brain ammonia and glutamate concentrations would increase over the exposure time. Remarkably, brain ammonia and glutamate levels were unchanged. This suggests that alternative mechanisms exist in the brain to maintain low levels of the neurotoxin ammonia and appropriate levels of the neurotransmitter glutamate under extreme conditions of ammonia loading. In other words, although GSase appears to be the key enzyme that catalyses glutamine synthesis, it is not the only effective mechanism to regulate brain ammonia and glutamate concentrations. Our results also confirm that exposure to elevated external ammonia is stressful to fish but, counter to our prediction, inhibition of GSase with MSOX does not result in a more dramatic stress response. Although our biochemical analysis suggests that the fish treated with MSOX were able to maintain glutamate homeostasis and prevent hyperammonia toxicity, we cannot rule out the possibility that GSase inhibition affected the overall health and the neural/behavioural repertoire of the fish.

To understand more fully the biochemical changes in the brain during ammonia loading, we measured the full suite of amino acids under all experimental conditions. Although the overall changes in brain amino acid concentrations in the 670 and 1000 μmol l^–1 NH₄Cl exposures were consistent, the concentrations of some of the individual amino acids were variable between the two exposures. Since the fish used in the two ammonia exposures were derived from the same stock, fed the same diet and reared under a common feeding regime, we suggest that seasonal differences may account for these differences. Overall, the interpretation of the changes in amino acids has been performed with caution, as changes in concentration provide only a ‘snapshot’ at a specific time point rather than a comprehensive view of changes in flux through a pathway. Nevertheless, the changes in brain amino acid concentrations with ammonia exposure with or without MSOX provide some valuable insights.

During exposure to elevated water ammonia concentrations, our results confirm that brain glutamine levels are increased similar to previous reports (Levi et al., 1974; Arillo et al., 1981; Peng et al., 1998; Ip et al., 2004; Wee et al., 2007), presumably because of increased flux through the GSase reaction. At the same time the lower concentration of brain glutamate, also shown by others (Arillo et al., 1981; Vedel et al., 1998; Wicks and Randall, 2002; Veauvy et al., 2005), implies that under high ammonia conditions the supply of glutamate is not perfectly matched to the demand. The pathways outlined in Fig. 7 provide a possible explanation of the data and explain how glutamate levels in the trout brain may be regulated. During exposure to elevated external ammonia, brain alanine, aspartate and GABA concentrations were significantly lower. These three amino acids can provide glutamate via transaminase reactions, along with the GDH reaction (Fig. 7) (Nelson and Cox, 2000). The transaminase and GDH reactions deplete α-ketoglutarate, a tricarboxylic acid cycle (TCA) intermediate, and reduce the provision of ATP needed for glutamine synthesis. The depletion of α-ketoglutarate must be addressed by supplying TCA cycle intermediates, such as acetyl-CoA. Acetyl-CoA can be supplied via the transaminases that have pyruvate as a product, for example alanine and serine (Fig. 7). Indeed, brain alanine and serine were significantly reduced in response to elevated ammonia levels. The significantly higher levels of brain arginine, phenylalanine, tyrosine and valine during ammonia exposure were also predicted as these amino acids are produced as by-products of pyruvate synthesis (Fig. 7). The proposed pathways above and in Fig. 7are consistent with the observed changes in 10 of the amino acids measured. Four amino acids showed no changes in concentration and changes in the remaining amino acids do not fit a recognizable pattern, except for tryptophan, discussed below.

Given the proposed scenario for the regulation of brain glutamate, how do fish respond to an inhibition of GSase and high external ammonia? The elevation of brain glutamine concentrations after 24 h of MSOX and ammonia treatment implies that glutamine continues to be synthesized despite the fact that GSase activities were only ∼25% of normal levels. By 96 h, however, glutamine concentrations in the MSOX-treated fish were only 5–16% of levels found in saline-injected fish exposed to elevated ammonia. This apparent delayed effect of MSOX in reducing brain glutamine stores is likely to be independent of the mechanism by which MSOX inhibits GSase. MSOX is a competitive inhibitor of GSase that binds irreversibly to the active site of the enzyme, permanently inactivating it (Cooper et al., 1976). Also, the predicted increase in brain glutamate concentrations in the MSOX-ammonia treatment did not occur, consistent with a study on the marine toadfish (Veauvy et al., 2005). It is possible that excess glutamate was used in the reversible transaminase reactions and the significant increase in alanine, aspartate and GABA concentrations in fish pre-injected with MSOX and exposed to ammonia supports this notion. It is also possible that brain GDH was downregulated; however, in the African sharptooth catfish Clarias gariepinus, injection with MSOX (100 mg kg^–1) did not alter brain GDH or alanine aminotransferase in vivo (Wee et al., 2007). In terms of the other amino acids presented in the model (Fig. 7), GSase inhibition by MSOX treatment during ammonia exposure for the most part reversed or eliminated the changes in brain amino acid concentrations discussed above. These results are consistent with a reduced depletion of glutamate and α-ketoglutarate during hyperammonaemic conditions and the recruitment of transaminase pathways to prevent glutamate excitotoxicity and brain ammonia toxicity during MSOX treatment (Fig. 7). In fact, the reversal of transaminase reactions with MSOX treatment was more pronounced after 96 h of ammonia exposure than after 24 h. Therefore, we suggest that the reduction in brain glutamine in the 96 h MSOX-ammonia treatment results from a gradual shift away from the condensation of glutamate to glutamine towards the transamination of glutamate to alternative amino acid products.

Trout exposed to elevated external ammonia in this study had significantly higher plasma and tissue ammonia concentrations that were constant between 24 and 96 h. The concentration of brain ammonia depends on the plasma-to-brain tissue ammonia gradient, the permeability of the blood–brain barrier and neural cells to ammonia and the rate of ammonia incorporation into glutamate/glutamine. Under these conditions of continuous ammonia exposure, GSase does not act as a ‘barrier’ against ammonia influx (Cooper and Plum, 1987) and this is reinforced by the fact that inhibition of brain GSase by MSOX did not affect brain ammonia concentrations in the rainbow trout (present study) or the gulf toadfish (Veauvy et al., 2005). The classic diffusion trapping model (Pitts, 1973) predicts that the distribution of ammonia between extracellular and intracellular compartments will be entirely influenced by the H⁺ gradient, if cells are impermeable. However, in the lemon sole (Parophrys vetulus) (Wright et al., 1988) brain ammonia concentrations under various physiological conditions far exceed those predicted by transmembrane pH gradients and were consistent with the membrane potential, indicating that NH ⁺₄ must be permeable across the blood–brain barrier and/or neural cell membranes. Brain ammonia permeability may also depend on the density of Rhesus (Rh) glycoproteins, the putative ammonia transporter identified in the gills of rainbow trout (Nawata et al., 2007) and other species (Nakada et al., 2007a; Nakada et al., 2007b; Hung et al., 2007) (reviewed by Wright and Wood, 2009). It is interesting to note that mRNA expression of Rhcg1 levels is downregulated in the brain of rainbow trout 48 h after exposure to 1500 μmol l^–1 NH₄HCO₃ (Nawata et al., 2007). Although brain Rh glycoprotein expression was not quantified in this study, the results of Nawata and colleagues (Nawata et al., 2007) suggest that brain ammonia concentrations in rainbow trout may plateau after 24 h of hyperammonia exposure due in part to a decrease in the permeability of neural cells to ammonia. Similarly, the gradual recovery and increase in gill ammonia excretion rates during chronic exposure to high external ammonia may contribute to the stabilization of brain and plasma ammonia (Cameron and Heisler, 1983; Clairborne and Evans, 1988; Wilson and Taylor, 1992; Nawata et al., 2007). Here also Rh glycoproteins may be playing an important role, as the recovery of ammonia excretion in rainbow trout exposed to high external ammonia is accompanied by an upregulation in Rhcg2 mRNA levels (Nawata et al., 2007).

Brain GSase activity was not altered by exposure to elevated external ammonia, in contrast to previous reports in rainbow trout (Wicks and Randall, 2002; Wright et al., 2007). It is possible that the higher resting brain GSase activity in the present study provided more than enough glutamine synthetic capacity. As well, the assay conditions for GSase between studies were not the same. In the present study, we chose to perform the GSase assay at the same temperature as the fish (16°C), in contrast to our previous study where the assay temperature was 12°C higher than the fish (26°C) (Wright et al., 2007), to enable comparison of the results with those of numerous other studies (e.g. Shankar and Anderson, 1985; Korte et al., 1997; Anderson et al., 2002; Felskie et al., 1998). Induction of brain GSase activity in response to ammonia loading is not a consistent response in fish (see Introduction) and the present data also indicate that excess GSase capacity may eliminate the need for additional enzyme activity under these conditions.

MSOX in this study was significantly more potent at inactivating GSase in the brain than in the liver. Although we do not know the proximal cause for the tissue-specific effects of MSOX in rainbow trout, the difference may result from the isoform-specific inhibitory actions of MSOX on GSase. In both mammals (Shin and Park, 2004) and bacteria (Shatters et al., 1993), the concentration of MSOX needed to inhibit different GSase isoforms varies greatly. Trout have four different GSase isoforms which are differentially expressed across tissues (Murray et al., 2003; Wright et al., 2007). The brain expresses Onmy-GS01 over the other forms, but the liver primarily expresses Onmy-GS02 (Murray et al., 2003; Essex-Fraser et al., 2005).

We predicted that ammonia exposure would result in an increase in plasma cortisol concentrations, consistent with previous rainbow trout studies (Wicks and Randall, 2002; Ortega et al., 2005; Bernier et al., 2008). Indeed, plasma cortisol concentrations were 7- to 15-fold higher after 24 h of NH₄Cl exposure, and at the higher ammonia level (1000 μmol l^–1 NH₄Cl) the surge in cortisol was also present at 96 h. The prolonged elevation of plasma cortisol at 96 h was correlated with an increase in brain CRF mRNA levels, underlining the severity of the stress under these conditions. With MSOX treatment, we predicted that plasma cortisol levels would be higher relative to the saline-injected fish exposed to ammonia based on the assumption that inhibition of GSase would result in elevated brain glutamate levels. In contrast, brain glutamate levels were not substantially elevated and overall plasma cortisol levels were not enhanced in MSOX-treated fish exposed to ammonia. In fact, we were surprised to observe that MSOX treatment significantly reduced the 24 h cortisol peak at 670 μmol l^–1 NH₄Cl. Therefore we performed a standard handling stress test (Barton and Iwama, 1991) to determine whether MSOX had a direct effect on cortisol release. Saline- and MSOX-injected trout had similar marked increases in plasma cortisol levels relative to the unhandled control fish, leading to the conclusion that MSOX does not directly influence the release of cortisol and is therefore not singularly responsible for the reduction in the cortisol response seen in the MSOX-ammonia fish. The reduction instead required the combination of MSOX and ammonia and coincided with the reduction in brain GSase.

In mammals, it has been shown that the increase in brain glutamine concentrations during an ammonia challenge results in the uptake of the large neutral amino acids (LNAAs) tryptophan, phenylalanine and tyrosine across the blood–brain barrier (Cangiano et al., 1983; Cardelli-Cangiano et al., 1984; Rigotti et al., 1985; Gorgievski-Hrisoho et al., 1986; Hilgier et al., 1992) (for a review, see Bachmann, 2002). The driving force for the increase in brain LNAAs are amino acid exchangers with 1:1 stoichiometries that transfer glutamine out of the brain at the same time as LNAAs are transported in. Tryptophan, phenylalanine and tyrosine can be synthesized into neurotransmitters implicated in HPI and HPA stress axis activation: tryptophan into serotonin, and phenylalanine and tyrosine into noradrenaline (Dinan, 1996; Winberg et al., 1997; Chaouloff, 2000; Herman et al., 2003; Pepels et al., 2004). This has not been studied previously in fish, but results from this study suggest that LNAA movement into the brain may occur in fish exposed to ammonia. Fish without MSOX treatment had higher overall concentrations of tryptophan, phenylalanine and tyrosine relative to MSOX-treated fish during both ammonia exposures. Changes in phenylalanine and tyrosine concentrations may also be the result of alterations in brain amino acid concentrations (see above and Fig. 7). We suggest that the higher concentrations of brain glutamine may have been the catalyst for the transport of LNAAs into the brain.

By promoting the uptake of LNAAs such as tryptophan during hyperammonaemia, GSase may be indirectly involved in the regulation of the HPI stress axis. In fish, as in other vertebrates, the rate-limiting step in brain serotonin synthesis is the availability of tryptophan (Aldegunde et al., 1998; Aldegunde et al., 2000). Ammonia exposure in teleosts is associated with an increase in brain serotonergic activity (Atwood et al., 2000; Ortega et al., 2005; Ronan et al., 2007), and after 24 h of hyperammonaemic conditions in rainbow trout there is a strong correlation between hypothalamic serotonergic activity and plasma cortisol levels (Ortega et al., 2005). The end product of HPI axis activation, cortisol, has also been shown to upregulate GSase expression in the gulf toadfish (Walsh et al., 1994; Hopkins et al., 1995), the sea raven (Hemitripterus americanus) (Vijyan et al., 1996) and the rainbow trout (Aluru and Vijayan, 2007) in a tissue-specific manner (Esbaugh and Walsh, 2009). Although the specific roles of LNAAs and cortisol in preventing brain ammonia toxicity in fish remain to be defined, the above observations suggest complex interactions between the ammonia detoxifying pathways regulated by GSase and the components of the HPI stress axis.

The authors wish to thank Sarah Alderman and Meghan Fuzzen for their assistance with tissue collection, Bob Frank and Matt Cornish for their technical assistance in the Hagen Aqualab, Lori Ferguson for typographical help with the manuscript, and two anonymous reviewers for constructive comments.