Corticotropin-releasing factor protects against ammonia neurotoxicity in isolated larval zebrafish brains

Beyond their role as regulators of the endocrine stress response, corticotropin-releasing factor (CRF)-related peptides work as autocrine/paracrine pro-survival factors. For example, release of the CRF-related peptides urocortin 1 (UCN1), UCN2 and UCN3 from the mammalian heart during ischemia/reperfusion injury improves cardiac function, reduces infarct size and prevents cell death (Onorati et al., 2013; Walczewska et al., 2014). Similarly, in humans, UCN1 is cytoprotective in chondrocytes (Lawrence et al., 2017), CRF reduces apoptosis induced by cytotoxins in retinoblastoma cells (Radulovic et al., 2003), and both UCN1 and CRF counter cytokine-induced apoptosis in pancreatic β-cells (Blaabjerg et al., 2014). In rodents, CRF is neuroprotective against toxins in primary neuronal cultures (Bayatti et al., 2003), protects neurons from hypoxia in brain slice preparations (Fox et al., 1993), and counters the damaging effects of glucocorticoids on neural stem cells (Koutmani et al., 2013). Limited evidence suggests that CRF-related peptides are also cytoprotective in non-mammalian taxa. CRF reduces apoptosis in Xenopus laevis tail buds (Boorse et al., 2006), and both CRF and UCN3 protect the adult zebrafish (Danio rerio) heart from hypoxia-induced apoptosis (Williams et al., 2017a). To date, however, it remains to be determined whether the neuroprotective effects of CRF-related peptides are also evolutionary conserved.

In zebrafish, the ontogeny, expression pattern and cytoprotective properties of the CRF system suggest that CRF-related peptides have functional roles during embryogenesis that are independent of those that regulate the endocrine stress response. All members of the zebrafish CRF system, including the ligands CRF, urotensin 1 (UTS1) and UCN3, receptors CRF-R1 and -R2, and binding protein CRF-BP, are expressed early during embryogenesis prior to maturation of the hypothalamic–pituitary–interrenal (HPI) stress axis (Alsop and Vijayan, 2008; Alderman and Bernier, 2009; Bräutigam et al., 2010). Moreover, while the broad expression patterns of CRF-related peptides in the zebrafish larval brain are consistent with their role in the regulation of the stress response, they also support HPI axis-independent functions (Chandrasekar et al., 2007; Alderman and Bernier, 2009). Recently, we showed that CRF overexpression defends zebrafish embryos from heat shock-induced apoptosis and that CRF-R1 blockade has the opposite effect (Alderman et al., 2018). These findings, together with the observation that CRF is cytoprotective in the developing mouse brain (Koutmani et al., 2013), suggest that these peptides may have neuroprotective properties during zebrafish development.

A neurotoxic stressor faced by all vertebrates is hyperammonemia. Whether the product of liver failure or eutrophic conditions, chronically elevated plasma ammonia levels can lead to irreversible brain damage (Walsh et al., 2007; Wijdicks, 2016). In fish, while ammonia tolerance varies considerably among species, brain edema, hyperactivation of N-methyl-d-aspartate (NMDA) glutamate receptors, oxidative stress and cerebral energy depletion have been implicated as mechanisms of ammonia toxicity (Ip et al., 2005; Wilkie et al., 2011, 2015; Feldman et al., 2014; Lisser et al., 2017; Zielonka et al., 2018). Potentially exacerbating the neurotoxic effects of high environmental ammonia (HEA) in embryonic and larval freshwater fish is their limited dispersal potential during vulnerable periods of nervous system development. In zebrafish larvae, although acute hyperammonemia is associated with increased whole-body glutamate levels and morphological signs of brain damage (Zielonka et al., 2018), it remains to be determined whether hyperammonemia-induced brain cell death is caused by hyperactivation of NMDA glutamate receptors and whether HEA-induced neurotoxicity is cell type specific.

Therefore, the aims of our study were to identify the neural specificity of HEA neurotoxicity, to elucidate whether NMDA glutamate receptors are implicated in mediating the neurotoxic effects of acute hyperammonemia, and to establish whether CRF protects against the neurotoxic effects of ammonia in zebrafish larvae. To this end, we first determined the effects of NH₄Cl exposure on the expression of cell type-specific markers of neurogenesis and differentiation in 1 day post-fertilization (dpf) zebrafish embryos and 5 dpf larvae, two developmental stages that differ markedly in their tolerance to ammonia (Williams et al., 2017b; Zielonka et al., 2018). Next, to test the hypothesis that activation of NMDA glutamate receptors mediates the neurotoxic effects of HEA, we quantified the effects of NH₄Cl exposure on the expression of cell type-specific markers with or without the NMDA receptor antagonist MK801. Finally, to test the hypothesis that CRF confers neuroprotection against HEA-induced brain cell death during zebrafish development via CRF-R1, we used isolated larval brains to quantify the effects of NH₄Cl exposure on the expression of cell type-specific markers and cell death with or without CRF and the CRF-R1 antagonist antalarmin.

Zebrafish, Danio rerio (F. Hamilton 1822), were obtained from AQuality (Mississauga, ON, Canada) and maintained in accordance with standard methods established in Westerfield (2000). Adult zebrafish were maintained at 27°C on a 14 h:10 h light:dark cycle and fed artemia and fish flakes to satiation twice daily. Eight to 12 tanks of fish were spawned at the same time each day, using a 2:1 female to male ratio; eggs were harvested within 1 h of spawning. Larvae were maintained at 28.5°C in larvae medium (60 µg ml⁻¹ Instant Ocean, Spectrum Brands Inc., Atlanta, GA, USA; 0.0004% Methylene Blue). After experiments, zebrafish embryos (1 dpf) and larvae (5 dpf) were anesthetized in ice-cold tricaine methanesulfonate (MS-222, 0.6%; Syndel International, Qualicum Beach, BC, Canada) followed by immediate storage on dry ice or at −80°C for subsequent analysis. All experiments were performed in accordance with guidelines set by the Canadian Council for Animal Care and were approved by the University of Guelph's Animal Care Committee.

Embryos and larvae were exposed to 0, 500, 1000 or 2000 μmol l⁻¹ NH₄Cl for 16 h ending at 1 or 5 dpf, respectively, to assess the effects of ammonia on the expression of cell type-specific markers of neurogenesis and differentiation. Specifically, we quantified the gene expression of nestin (nes; Mahler and Driever, 2007) and SRY (sex determining region Y)-box 2 (sox2; Okuda et al., 2010) as markers of pluripotent neural progenitor cells (NPCs), neurogenin1 (neurog1; Blader et al., 2003) as a marker of pro-neuronal cells, neurogenic differentiation 1 (neurod1; Mueller and Wullimann, 2003) as an indicator of post-mitotic determined neurons, and finally glial fibrillary acidic protein (gfap; Nielsen and Jørgensen, 2003) and myelin basic protein A (mbpa; Brösamle and Halpern, 2002) as markers of radial glia (or astrocytes in mammals) and oligodendrocytes, respectively. Note that the mRNA levels of mbpa were assessed at 5 dpf only as it is not expressed at 1 dpf. The duration and levels of NH₄Cl exposure used in this study were selected to match those used in Williams et al. (2017b). The 16 h exposure duration was chosen to approximate the length of one cell cycle in 5 dpf zebrafish larvae (Mueller and Wullimann, 2005), and the 96 h LC₅₀ for ammonia in adult zebrafish is 677 μmol l⁻¹ (Alsop and Wood, 2013). All exposures, unless noted otherwise, were performed in 6-well plates (Sarstedt, Nümbrecht, Germany) using pools of 25 embryos or larvae (1 dpf n=7–9, 5 dpf n=12–14) at 28.5°C using freshly made NH₄Cl solutions adjusted to pH 8.0. All NH₄Cl solutions were prepared with well water (Ca²⁺, 2.7 mmol l⁻¹; Cl⁻, 2.4 mmol l⁻¹; Mg²⁺, 1.7 mmol l⁻¹; K⁺, 0.05 mmol l⁻¹; Na⁺, 1.9 mmol l⁻¹).

Two approaches were taken to assess whether ammonia had a general toxic effect at the levels used in this study, or whether the effects of ammonia were specifically deleterious to neurogenic cell types. First, we assessed the effects of ammonia exposure on the gene expression of the muscle developmental marker myogenic differentiation 1 (myod1) in the 5 dpf larvae exposed to 0, 500, 1000 and 2000 µmol l⁻¹ NH₄Cl for 16 h as described above (n=10–16). Second, we assessed the effects of ammonia exposure on whole-body caspase-3 activity to identify whether the NH₄Cl treatments induce non-specific cytotoxic effects. Given that the brain is roughly 8% of the body mass (Hill et al., 2003), we expected minimal change in overall caspase-3 activity due to neural apoptosis when compared with general toxicity effects on the whole body. Pools of 25 larvae were treated with 0, 500, 1000, 2000, 3000 and 4000 µmol l⁻¹ NH₄Cl for 16 h ending at 5 dpf (n=7–10). Following the NH₄Cl exposures, larvae were immediately homogenized for quantification of caspase-3 activity.

To determine whether the ammonia-induced changes in the expression of neurogenic markers are NMDA glutamate receptor dependent, larvae were exposed to 0 or 1000 µmol l⁻¹ NH₄Cl for 16 h ending at 5 dpf in either the absence (0 µmol l⁻¹) or the presence (1 µmol l⁻¹) of MK801 (n=8–12). Given the relatively long duration of ammonia exposure, continuous MK801 treatment was chosen instead of pre-treatment. Based on a pilot study, 1 μmol l⁻¹ MK801 was selected as 10 and 100 μmol l⁻¹ MK801 caused mortality when used for this duration at pH 8.0 (Wilkie et al., 2011).

As there is no apparent transport of CRF into the brain from blood across the blood–brain barrier (Martins et al., 1996), isolated larval brains were cultured in vitro to assess the neuroprotective effects of CRF. To validate the use of this approach to study the effects of ammonia on the gene expression of neurogenic markers, we first assessed the effects of NH₄Cl exposure (0, 500, 750, 1000 μmol l⁻¹) in vitro on the expression of the housekeeping gene elongation factor 1α (ef1α; n=5 pools of 2 brains). We then tested whether the 16 h in vitro culture altered ef1α expression by comparing in vivo with in vitro ammonia exposure. For this purpose, larval brains (n=5 pools of 2 brains) were dissected before (in vitro exposure) or after (in vivo exposure) 16 h of 0, 500 or 750 μmol l⁻¹ NH₄Cl exposure and expression of ef1α was compared. To assess the effects of ammonia on neurod1 expression in culture, 4 dpf larval brains (n=5 pools of 2 brains) were exposed to 0, 500 or 750 μmol l⁻¹ NH₄Cl for 16 h ending at 5 dpf. In order to test whether CRF mitigates the effects of ammonia on nes, sox2, neurog1, neurod1, gfap or mbpa expression, 4 dpf larval brains were exposed to 0 or 750 µmol l⁻¹ NH₄Cl for 16 h ending at 5 dpf in either the absence or the presence (1 nmol l⁻¹) of rat/human CRF (American Peptide Company, Sunnyvale, CA, USA; n=5–7 pools of 2 brains). To determine whether the effects of CRF on ammonia-induced changes in neurogenic marker expression are CRF-R1 dependent, 4 dpf larval brains were exposed to 0 or 750 µmol l⁻¹ NH₄Cl for 16 h ending at 5 dpf in the presence of the non-peptide CRF-R1-specific antagonist antalarmin (10 nmol l⁻¹; Tocris Bioscience, Bristol, UK) either alone or with CRF (1 nmol l⁻¹; n=5–7). All incubations were performed in 12-well plates (Sarstedt) at 28.5°C. Immediately after NH₄Cl exposure, larval brains were stored at −80°C for subsequent analysis. Two brains were pooled for each sample.

To assess the effects of CRF on ammonia-induced cell death in cultured zebrafish brains, 4 dpf larval brains were exposed as above to 0 or 750 µmol l⁻¹ NH₄Cl for 16 h ending at 5 dpf in either the absence or the presence (1 nmol l⁻¹) of CRF (n=5–8). Similarly, to determine whether the effects of CRF on ammonia-induced cell death are CRF-R1 dependent, 4 dpf larval brains were exposed to 0 or 750 µmol l⁻¹ NH₄Cl for 16 h ending at 5 dpf in the presence of antalarmin (10 nmol l⁻¹) either alone or with CRF (1 nmol l⁻¹; n=5–8). Immediately after the NH₄Cl exposure, single larval brains were fixed in 2% PFA overnight at 4°C and processed for TUNEL staining to quantify cell death.

Water ammonia was measured at the end of each exposure and was within 10% of the stated value for all exposures, and <40 μmol l⁻¹ in the control (no NH₄Cl) treatment. Ammonia was measured by colorimetric assay (Verdouw et al., 1978) adapted to 96-well plates on a SpectraMax 190 plate reader (Molecular Devices, Menlo Park, CA, USA) using SOFTmax version 4.6 (Molecular Devices). Reactions were performed in triplicate, with shaking, at 24°C until the reaction was complete. The results were quantified from a serially diluted standard curve using freshly made NH₄Cl.

Larvae at the 4 dpf stage were anesthetized in MS-222 (0.04%) dissolved in artificial cerebro-spinal fluid (aCSF: 100 mmol l⁻¹ NaCl, 25 mmol l⁻¹ NaHCO₃, 4 mmol l⁻¹ KCl, 2.4 mmol l⁻¹ CaCl₂, 1.4 mmol l⁻¹ MgCl₂, 10 mmol l⁻¹ glucose, 280 mOsm l⁻¹) with penicillin/streptomycin (100 U ml⁻¹). Under a stereomicroscope (SMZ1500, Nikon Instruments Inc., Melville, NY, USA), an electrolytically sharpened tungsten wire (0.25 mm) was used to remove both eyes then create multiple medial incisions through the overlying skin from the nares to the caudal end of the hindbrain. The spinal cord was severed and extracted brains were rinsed with fresh aCSF (with antibiotics) before transfer to 12-well plates (Sarstedt). All subsequent chemical, peptide or drug treatments were diluted in aCSF. Brains that had been subject to piercing or stretching were discarded.

Individual pools of frozen embryos or larvae were homogenized in 0.5 ml of RNA extraction reagent (Bio Basic, Markham, ON, Canada) using a 21 G needle followed by sonication at 50 W for 2×8 s on wet ice. Larval brains were homogenized using the same extraction reagent and a micropestle. RNA extraction followed the manufacturer's instructions except for the addition of 10 μg RNA grade glycogen (Roche Diagnostics, Laval, QC, Canada) and an overnight incubation at −80°C during the isopropanol precipitation step to maximize recovery. Total RNA was quantified using spectrophotometry (Nanodrop 8000, Nanodrop Products, Wilmington, DE, USA) and treated with Ambion DNase (Thermo Fisher, Waltham, MA, USA) prior to use in cDNA synthesis (1 μg for pooled embryos or larvae, and 140 ng for pooled larval brains). SuperScript II RNase H⁻ reverse transcriptase (Invitrogen, Carlsbad, CA, USA) was used as directed by the manufacturer to synthesize cDNA with oligo-dT primer. Separate samples were treated identically without the addition of reverse transcriptase or without the presence of RNA to verify the absence of genomic DNA or contaminated reagents.

Relative mRNA levels of target genes were determined by qPCR using a CFX96 system (BioRad, Hercules, CA, USA) with Quanta Perfecta Supermix (Quanta Biosciences, Gaithersburg, MD, USA). Each 20 µl reaction contained 5 µl of diluted cDNA (1/40 for pooled embryos and larvae, and 1/20 for pooled dissected larval brains) or no reverse transcriptase controls (see above), 10 µl 2× SYBR green mix and 2.5 µl of both forward and reverse primers (0.4 µmol l⁻¹; Table S1). Default cycling conditions were used and followed by a melting curve analysis to verify the specificity of each PCR product. Only samples with a unimodal melt curve at the appropriate temperature were included for data analysis. To account for differences in amplification efficiency, standard curves were constructed for each gene using known dilutions of cDNA. Input values for each gene were obtained by fitting the average threshold cycle (Ct) value to the antilog of the gene-specific standard curve, thereby correcting for differences in primer amplification efficiency. To control for variations in loading, expression of target genes was standardized to that of the housekeeping gene ef1α. Note that the expression of ef1a did not differ between any of the treatments (P>0.05), with the exception of the 1000 μmol l⁻¹ NH₄Cl treatment on the isolated larval brains, which caused a decrease in expression. Gene expression data are reported as fold-change relative to the control (no NH₄Cl) treatment mean value.

Whole-body caspase-3 activity was quantified based on the methods described by Stankiewicz et al. (2005) as a measure of apoptosis. Briefly, the maximum rate of product formation from the fluorogenic caspase-3 substrate Ac-DEVD-AMC was quantified by fluorescence spectrophotometry. Fresh pools of 5 dpf larvae were homogenized in lysis buffer (50 mmol l⁻¹ Hepes, 0.1% w/v Chaps, 0.1 mmol l⁻¹ EDTA, 1 mmol l⁻¹ DTT, pH 7.4, 10 mg ml⁻¹ aprotinin, leupeptin and pepstatin A; Sigma-Aldrich, Oakville, ON, Canada) by 21 G needle and sonication (2×8 s at 50 W) while on ice. The lysate was centrifuged at 12,000 g for 10 min and the supernatant was mixed with 3 volumes of assay buffer (50 mmol l⁻¹ Hepes, 100 mmol l⁻¹ NaCl, 0.1% w/v Chaps, 10 mmol l⁻¹ DTT, 1 mmol l⁻¹ EDTA, 10% glycerol) containing 40 mmol l⁻¹ Ac-DEVD-AMC (Amresco, Solon, OH, USA). Fluorescence was measured at 460 nm (with excitation at 380 nm) for 1 h at 28°C using a fluorescence plate reader (BioTek FLx800, BioTek, Winooski, VT, USA). The maximum reaction rate was normalized to total protein determined by the Bradford assay (BioRad Protein Assay Dye Reagent, Bio-Rad, Hercules, CA, USA) on the remaining lysate.

Fixed larval brains were stained using an In Situ Cell Death Detection Kit (TMR Red, Roche, Mississauga, ON, Canada) to quantify cell death. Briefly, larval brains were permeabilized by incubation with proteinase K (20 µg ml⁻¹) for 10 min at 70°C, rinsed thrice, incubated in 0.1% Triton-X for 30 min, rinsed thrice, then incubated in ice-cold citrate buffer for 30 min (all solutions made with 1× PBS+1 mmol l⁻¹ EDTA). Larval brains were rinsed in 1× PBS thrice and incubated in a reaction mixture containing TMR-UTP, TdT enzyme and buffer for 2 h in a 37°C humidified chamber. Following the reaction, brains were rinsed thrice in 1× PBS+1 mmol l⁻¹ EDTA. Finally, brains were incubated with the nuclear stain Hoechst 33342 (1 μg ml⁻¹; Sigma-Aldrich, Oakville, ON, Canada) for 20 min and then rinsed thrice in 1× PBS+1 mmol l⁻¹ EDTA. To confirm TUNEL, labeling solution without terminal transferase was used to replace the TUNEL reaction mixture as a negative control, and DNase I treatment was used as a positive control. Stained brains were imaged using a Nikon Ti microscope at 580 nm for TUNEL (TMR Red) and 461 nm for all nuclei (Hoechst 33342). Images were processed in the image processing program ImageJ/Fiji (Schindelin et al., 2012). TUNEL- and Hoechst-positive cells were manually counted in focused optical stacks to establish automatic cell counting parameters that gave equal results. Results presented are the percentage of TUNEL-positive cells. Sections through the whole brain were quantified and averaged with one brain being considered an N of one.

Results are presented as mean values±s.e.m. All comparisons were done by one- or two-way ANOVA followed by Holm–Šidák post hoc test (SigmaPlot 12.5, San Jose, CA, USA) where assumptions of normality and equal variance were met. If these assumptions were not met, the data were log₁₀-transformed prior to the above analysis. Data that still failed equal variance assumptions were tested using a one-way ANOVA followed by Dunnett's T3 post hoc test (IBM SPSS 24, New York, NY, USA). Statistical significance was defined as P<0.05.

Ammonia significantly affected nes gene expression and this effect was dependent on developmental stage (ammonia: F_3,73=6.12, P<0.001; stage: F_1,73=37.27, P<0.001; ammonia×stage: F_3,73=5.96, P=0.001) (Fig. 1A). While ammonia had no effect on nes expression at 1 dpf (P>0.05), exposure to 500, 1000 and 2000 μmol l⁻¹ NH₄Cl increased nes expression at 5 dpf by 70% (P=0.006), 185% (P=0.006) and 182% (P<0.001), respectively, compared with control (no NH₄Cl) (Fig. 1A). Developmental stage, but not ammonia, had a significant effect on sox2 expression (stage: F_1,76=52.29, P<0.001; ammonia: P>0.05; ammonia×stage: P>0.05) (Fig. 1B). The apparent ammonia-induced increases in sox2 expression at 5 dpf (of 47%, 46% and 46% in the 500, 1000 and 2000 μmol l⁻¹ NH₄Cl treatments, respectively) did not reach statistical significance. Exposure to ammonia significantly affected the expression of neurog1 (Fig. 1C) and this effect was dependent on developmental stage (ammonia: F_3,71=7.05, P<0001; stage: P>0.05; ammonia×stage: F_3,73=5.96, P=0.001). Although no change was observed at 1 dpf, ammonia at 5 dpf caused neurog1 expression to decrease 31% at 500 μmol l⁻¹ (P<0.001), 55% at 1000 μmol l⁻¹ (P<0.001) and 57% at 2000 μmol l⁻¹ NH₄Cl (P<0.001). The concentration of ammonia had a significant effect on neurod1 gene expression (Fig. 1D; ammonia: F_3,73=94.15, P<0001). Expression of neurod1 changed with developmental stage and the effect of ammonia on neurod1 expression depended on stage (stage: F_1,73=249.76, P<0.001; ammonia×stage: F_1,73=10.43, P<0.001). Exposure to ammonia resulted in a dose-dependent decline of neurod1 expression at all concentrations irrespective of stage (P<0.001) and the magnitude of this decrease was greater at 1 dpf for any given concentration (P<0.001). At 1 dpf, ammonia caused neurod1 expression to decrease by 29%, 53% and 70% at 500, 1000 and 2000 μmol l⁻¹ NH₄Cl, respectively. At 5 dpf, the 500 μmol l⁻¹ NH₄Cl treatment had no effect on neurod1 expression, and the 1000 and 2000 μmol l⁻¹ NH₄Cl treatments reduced the expression of this transcript by 35% (P<0.001) and 71% (P<0.001), respectively. Exposure to ammonia significantly decreased gfap expression depending on developmental stage (ammonia: F_3,73=2.97, P=0.037; stage: F_1,73=8.84, P=0.004; ammonia×stage: F_1,73=4.19, P=0.008) (Fig. 1E). While no change was observed at 1 dpf, at 5 dpf the 1000 and 2000 μmol l⁻¹ NH₄Cl treatments caused gfap expression to decrease by 34% (P=0.004) and 38% (P=0.001), respectively. Relative to the control treatment, ammonia exposure reduced mbpa expression at 5 dpf (ammonia: F_3,34=11.59, P<0.001) in the 1000 and 2000 μmol l⁻¹ NH₄Cl treatments by 32% (P=0.014) and 62% (P<0.001), respectively (Fig. 1F). It should be noted that mbpa was not detectable at 1 dpf, consistent with previous reports (Brösamle and Halpern, 2002; Buckley et al., 2010).

Ammonia had a significant effect on myod1 mRNA levels (F_3,49=5.62; P=0.002), but only the 2000 μmol l⁻¹ NH₄Cl treatment reduced gene expression (by 49%; P=0.002) (Fig. 2A). Doses of 500 and 1000 μmol l⁻¹ NH₄Cl caused no significant change in myod1 expression compared with control (P>0.05). Exposure to ammonia increased whole-body caspase-3 activity (F_5,56=16.47, P<0.001); however, this was only significant at concentrations of 3000 μmol l⁻¹ (P=0.004) and 4000 μmol l⁻¹ NH₄Cl (P=0.007) (Fig. 2B).

Exposure of 5 dpf larvae to 1000 μmol l⁻¹ NH₄Cl significantly increased nes expression by ∼175% irrespective of whether MK801 was added to the incubation medium (ammonia F_1,20=24.17, P<0.001; MK801: P>0.05; ammonia×MK801: P>0.05) (Table 1). Likewise, ammonia increased sox2 expression by ∼204% and this increase was unaffected by MK801 (ammonia: F_1,20=24.39, P<0.001; MK801: P>0.05; ammonia×MK801: P>0.05) (Table 1).

Ammonia exposure significantly decreased neurog1 expression by ∼40% while MK801 had no effect (ammonia: F_1,20=21.19, P<0.001; MK801: P>0.05; ammonia×MK801: P>0.05) (Table 1). In contrast, both ammonia and MK801 affected neurod1 expression (ammonia: F_1,41=14.39, P<0.001; MK801: F_1,41=22.20, P<0.001; ammonia×MK801: F_1,41=5.20, P=0.02; Fig. 3). Ammonia alone (1000 μmol l⁻¹) reduced neurod1 expression by 40% relative to the control treatment (P<0.001) and this effect was abolished in the presence of MK801 (P>0.05). As previously observed, ammonia caused gfap expression to decrease by 46% but MK801 did not have any effect alone or in combination with ammonia on this gene (ammonia: F_1,22=4.83, P=0.040; MK801: P>0.05; ammonia×MK801: P>0.05; Table 1). Similarly, while ammonia caused a significant 37% decrease in mbpa expression, MK801 had no effect alone or in combination with ammonia on the expression of this gene (ammonia: F_1,22=11.16, P=0.039; MK801: P>0.05; ammonia×MK801: P>0.05; Table 1).

The basal expression of the housekeeping gene ef1α in the cultured larval brains did not change with NH₄Cl concentration between 0, 500 and 750 μmol l⁻¹ (F_3,12=9.13, P<0.01; 0, 500, 750 μmol l⁻¹ NH₄Cl P>0.05; Table S2). In contrast, exposure to 1000 μmol l⁻¹ NH₄Cl for 16 h reduced ef1α expression by 60% (1000 μmol l⁻¹ NH₄Cl: P=0.034). The expression of ef1α did not differ between brains exposed to 0, 500 and 750 μmol l⁻¹ NH₄Cl for 16 h in vivo and in vitro (exposure conditions: F_1,12=12.68, P>0.05; ammonia: F_2,12=0.559, P>0.05; exposure conditions×ammonia: F_2,12=0.18, P>0.05; Table S3). Under control conditions (no NH₄Cl), CRF (1 nmol l⁻¹) and antalarmin (10 nmol l⁻¹) did not affect basal ef1α expression (CRF: F_1,21=0.09, P>0.05; antalarmin: F_1,21=13.96, P>0.05; CRF×antalarmin: F_2,12=0.20, P>0.05; Table S4). Ammonia alone had a significant effect on neurod1 expression (F_2,17=7.28, P=0.005; Fig. 4A) with a 13% decrease at 500 μmol l⁻¹ (P>0.05) and a 33% decrease at 750 μmol l⁻¹ NH₄Cl (P=0.006). The addition of CRF had a significant effect on the ammonia-induced changes in neurod1 expression (ammonia: P>0.05; CRF: F_1,20=5.34, P=0.03; ammonia×CRF: F_1,32=593, P=0.02; Fig. 4B). Ammonia (750 μmol l⁻¹ NH₄Cl) reduced neurod1 expression in the absence of CRF (P=0.015), and CRF (1 nmol l⁻¹) under control conditions (no NH₄Cl) had no effect (P>0.05); however, in the presence of 750 μmol l⁻¹ NH₄Cl, CRF prevented the ammonia-induced decrease in neurod1 mRNA levels (P>0.05). When the above experiment was repeated in the presence of antalarmin (10 nmol l⁻¹), ammonia exposure reduced neurod1 expression (F_1,20=18.38, P<0.001) but CRF no longer had an effect (P>0.05) and there was no interaction between ammonia and CRF (P>0.05; Fig. 4C). In contrast, the addition of CRF did not affect the ammonia-induced changes in the expression of the other neural markers assessed (Table 2). Overall, irrespective of whether or not CRF was added to the incubation medium, exposure of larval brains to 750 μmol l⁻¹ NH₄Cl for 16 h increased nes and sox2 expression by ∼720% (F_1,21=157.75, P<0.001) and ∼189% (F_1,21=16.88, P<0.001), respectively. On average, ammonia exposure also decreased the expression of neurog1 (F_1,21=13.89, P=0.001), gfap (F_1,21=5.24, P=0.032) and mbpa (F_1,21=12.10, P=0.002) ∼41%, ∼21% and ∼30%, respectively. Likewise, in the presence of antalarmin, CRF did not alter the ammonia-induced changes in the expression of nes (F_1,20=136.68, P<0.001), sox2 (F_1,20=6.48, P=0.019), neurog1 (F_1,20=18.03, P<0.001), gfap (F_1,21=157.75, P<0.001) and mbpa (F_1,20=5.20, P=0.034; Table 2). Overall, in the combined CRF and antalarmin treatment, ammonia treatment increased the expression of nes and sox2 by ∼542% and ∼150%, respectively, and decreased neurog1, gfap and mbpa expression by ∼48%, ∼28% and ∼33%, respectively.

Ammonia exposure elicited a 195% increase in TUNEL staining in cultured larval brains (ammonia: F_1,16=20.66, P<0.001), and although CRF alone had no effect (CRF: P>0.05), the effects of ammonia were abolished by the addition of CRF (ammonia×CRF: F_1,16=9.09, P=0.008; Fig. 5A). In the presence of antalarmin, ammonia exposure increased the percentage of TUNEL-positive cells (ammonia: F_1,16=10.71 P=0.004) independent of whether or not CRF was present (ammonia×CRF: P=0.05), and CRF alone had no effect (CRF: P>0.05) (Fig. 5B).

This is the first study to demonstrate that the neurotoxic effects of HEA in fish are neural cell type specific, that NMDA glutamate receptors specifically mediate the deleterious effects of HEA on determined post-mitotic neurons and that exogenous CRF has neuroprotective effects against HEA neurotoxicity. In general, the neural cell type-specific effects of HEA in zebrafish embryos and larvae highlight the complexity of HEA neurotoxicity and the importance of developmental stage for determining the potential impact of hyperammonemic conditions in fish. Similarly, the cell type-specific effects of MK801 in counteracting the effects of HEA on the expression of various neural markers further implicate multiple mechanisms as effectors of ammonia toxicity. Finally, using antalarmin, we demonstrate that CRF-R1 is responsible for mediating the pro-survival effects of CRF on neurons in zebrafish larval brains.

HEA-induced changes in the expression of key neuronal markers show that hyperammonemic conditions in zebrafish can have complex effects on neuronal proliferation and differentiation that are developmental stage and cell type specific. The limited effect of HEA on the expression of neuronal markers at 1 dpf relative to 5 dpf is consistent with the observation that ammonia toxicity tolerance in zebrafish decreases significantly between the embryonic and larval developmental stages (Williams et al., 2017b; Zielonka et al., 2018). Several factors likely contribute to this developmental stage-dependent tolerance to ammonia toxicity in zebrafish: (1) both metabolism and the rate of nitrogenous excretion increase rapidly post-hatching (Braun et al., 2009); (2) while embryos excrete the majority of their nitrogenous waste as urea, larvae are primarily ammonotelic (Braun et al., 2009; Bucking et al., 2013; LeMoine and Walsh, 2013); and (3) the loss of the yolk sac through absorption post-hatching reduces the storage capacity for ammonia and is associated with a deficit in ammonia excretory capacity (Steele et al., 2001; Bucking et al., 2013; Braun et al., 2009). While our results suggest that NPCs, pro-neuronal cells, determined neurons and glial cells are cellular targets of hyperammonemia in 5 dpf larvae, the only cell lineage marker affected by HEA in 1 dpf embryo was neurod1, a marker of determined neurons. By converting excess glutamate to glutamine and reducing oxidative stress, astrocytes play a key role in defending against ammonia neurotoxicity (Rao et al., 2005). However, neurogenesis precedes gliogenesis during the development of the central nervous system (Jacobson, 1985). For example, astrocytes mature around or following parturition in mice (Ge et al., 2012), and trout glutamine synthetase activity, a function of mature astrocytes, is first observed around hatching (Essex-Fraser et al., 2005), equivalent to 3 dpf in zebrafish. Therefore, in 1 dpf larvae, without radial glia protection against ammonia toxicity, determined neurons may be more sensitive than other neuronal cell types to HEA and undergo necrotic and apoptotic damage in response to hyperammonemic conditions.

Although resilient to hyperammonemic conditions at the embryonic stage, the increase in nes and sox2 expression at 5 dpf suggests that HEA exposure stimulates NPC proliferation in zebrafish larvae. The dose-dependent increase in nes expression in response to HEA exposure observed here is consistent with the effects of the developmental neurotoxicant ethanol on this NPC marker in zebrafish embryos (Fan et al., 2010), and with the increase in nestin immunoreactivity that characterizes mice with hepatic encephalopathy (Tallis et al., 2014). Nestin expression also increases in response to neuronal inflammation (Krishnasamy et al., 2017) and retinal injury (Ooto et al., 2004; Xue et al., 2010). Similarly, although the regulation of sox2 expression has not been previously investigated in the context of ammonia toxicity, sox2-positive cells proliferate and sox2 is upregulated in response to spinal cord injury and regeneration (Gaete et al., 2012; Ogai et al., 2014). Therefore, as previously observed in response to various neuronal damage-causing injuries, our results suggest that ammonia toxicity in larval zebrafish enhances NPC proliferation. Interestingly, as Sox2 can also activate repressors of neuronal differentiation (Schmidt et al., 2013), the HEA-induced increase in sox2 expression may also regulate the fate of other brain cell lineages.

Unlike its effects on markers of NPCs, HEA exposure in zebrafish larvae reduced the expression of markers for more differentiated cell types including neurons. For example, HEA reduced the expression of the proneuronal transcription factor neurog1. neurog1 is expressed in primary mitotic neurons (Blader et al., 2003) and represents the marker of the least differentiated state to be negatively regulated by ammonia. Similarly, neurog1 expression decreases in response to toxicants such as dioxins (Theunissen et al., 2011) and ethanol (Fan et al., 2010). The reduced neurog1 expression we observed may represent a decrease in the differentiation of NPCs or may indicate that these determined but mitotic neurons are susceptible to ammonia toxicity. As the effects of HEA on neurog1 expression could affect the development of neuronal circuitries (Dixit et al., 2014), a more detailed look at this neuronal population may be important in understanding the impact of ammonia on neural development. Similar to Neurog1, the transcription factor NeuroD1 plays a vital role in neuronal differentiation but is indicative of determined post-mitotic neurons (Mueller and Wullimann, 2009). The HEA-induced reduction in neurod1 mRNA levels is consistent with the effects of the toxicants cadmium (Chow et al., 2008) and ketamine (Kanungo et al., 2013) on the expression of this neuronal marker, and with the observation that ammonia is directly toxic to neurons in mammals (Klejman et al., 2005). Given that early life HEA exposure can alter the cortisol stress response to a novel stressor in larval and juvenile zebrafish (Williams et al., 2017b), it would be valuable to test in future work whether blockade of the HEA-induced reduction in neurod1 via MK801 can mitigate the early life effects of ammonia on the later life stress response.

HEA exposure in zebrafish larvae also reduced the expression of markers for radial glia and oligodendrocytes. In general, however, the radial glia marker gfap showed a relatively small decrease in expression compared with markers of neurons and oligodendrocytes. This response may be due to the confounding expression of gfap in both NPCs and radial glia (Schmidt et al., 2013), cell types which may have different sensitivities to HEA. Alternatively, radial glia may have higher ammonia tolerance than neurons and oligodendrocytes because of their functional role in detoxifying excess ammonia, as described in mammals (Suárez et al., 2002; Rao et al., 2005; Cagnon and Braissant, 2009) and fish (Grupp et al., 2010). In contrast, the HEA-induced reduction in mbpa expression was similar to that of neurod1, indicating a similar ammonia tolerance of oligodendrocytes and neurons in zebrafish larvae. Given that hyperammonemic conditions in neonates and infants can lead to myelination delays, hypomyelination and demyelination (Cagnon and Braissant, 2007), our results suggest that HEA exposure in zebrafish larvae may also have an impact on neuron function through defects of myelination.

To address the neural specificity of ammonia toxicity, gene expression of the muscle differentiation factor myod1 (Weinberg et al., 1996) was used as an indicator of non-neural effects. HEA exposure did not affect myod1 expression using NH₄Cl concentrations below 2000 μmol l⁻¹, indicating that neural populations are in fact much more sensitive to ammonia toxicity than muscle cells. When examining overall toxicity through whole-body caspase-3 activity, a key effector in the apoptotic cascade (Strasser et al., 2000), no significant increase in caspase was detectable below 3000 μmol l⁻¹ NH₄Cl. These findings demonstrate that while neural markers are strongly impacted by ammonia, a non-neural developmental marker is largely unaffected and generalized toxic effects leading to apoptosis require higher concentrations than used in other experiments of this study. These findings agree with the consensus within mammalian literature which has attributed the pathological effects of hyperammonemia to central nervous system damage (Braissant et al., 2013).

As NMDA glutamate receptor antagonists can reduce HEA-induced mortality in several fish species (Tsui et al., 2004; Ip et al., 2005; Wilkie et al., 2011; Feldman et al., 2014; Zielonka et al., 2018), we tested whether MK801 may also counter the HEA-induced changes in the expression of key neural markers. While nes, sox2, neurog1, gfap and mbpa expression were unaffected, MK801 effectively prevented the ammonia-induced reduction in neurod1 mRNA levels in 5 dpf larvae. These results suggest that hyperactivation of NMDA receptors contributes to the deleterious effects of ammonia in zebrafish post-mitotic neurons but not in NPCs, pro-neuronal cells, radial glia or oligodendrocytes. Consistent with these findings is the observation that differentiation of immature neuronal precursor cells into mature neurons renders susceptibility to glutamate excitotoxicity via newly expressed NMDA receptors (He et al., 2013).

Moreover, although it is unclear whether NMDA receptors are expressed in zebrafish glial cells, mammalian studies have shown that NMDA receptors are expressed in astrocytes (Verkhratsky and Kirchhoff, 2007) and oligodendrocytes (Káradóttir et al., 2005) but do not contribute to glutamate excitotoxicity. Instead, as a result of differences in the properties of the NMDA receptor subunits between glia and neurons (Dzamba et al., 2013), glutamate excitotoxicity in glial cells is exclusively mediated by AMPA and kainate glutamate receptor subtypes (Alberdi et al., 2002; Matute et al., 2002, 2007). Together, these results do not preclude other mechanisms of ammonia toxicity from affecting the zebrafish brain, given that other markers were not rescued by MK801 treatment as previously reported in mammals (Matute et al., 2002), but support the hypothesis that hyperactivation of NMDA receptors is an important mechanism of the deleterious effects of ammonia on differentiated post-mitotic neurons in zebrafish.

In isolated larval zebrafish brains, our finding that CRF can reverse both an HEA-induced reduction in neurod1 expression and an increase in HEA-induced cell death shows for the first time that this neuropeptide has neuroprotective properties in a non-mammalian species. These results are consistent with reported neuroprotective effects of CRF in mammalian studies using organotypic brain slices and primary neuronal cultures (Fox et al., 1993; Bayatti et al., 2003; Facci et al., 2003; Madtes et al., 2004; Koutmani et al., 2013). The fact that both CRF and UCN1 can prevent glutamate-induced cell death in rat neuronal cultures (Elliott-Hunt et al., 2002; Pedersen et al., 2001; Valadas et al., 2012) suggests that CRF protects against the neurotoxic effects of HEA in zebrafish larvae by counteracting the excitotoxic events associated with excessive activation of NMDA receptors in post-mitotic neurons. In contrast, our observation that CRF does not reverse the effects of HEA on the expression of gfap and mbpa suggests that CRF does not protect against the excitotoxic effects of glutamate in glial cells which are mediated by AMPA and kainate receptors (Alberdi et al., 2002; Matute et al., 2002, 2007). Although CRF has neuroprotective effects on NPCs in mice (Koutmani et al., 2013), it did not affect the HEA-induced transcriptional changes of the NPC markers nestin and SOX2 in this study. While this may represent an area of divergence between fish and mammals, previous studies have shown that the signal transduction mechanisms mediating the neuroprotective effects of CRF vary by cell type and brain region in rodents (Fox et al., 1993; Elliott-Hunt et al., 2002; Bayatti et al., 2003; Facci et al., 2003; Koutmani et al., 2013), an important consideration for future studies in zebrafish. Our observation that the CRF-R1-specific antagonist antalarmin prevents the neuroprotective actions of CRF against HEA in isolated zebrafish brains is consistent with mammalian studies which have attributed the neuroprotective effects of CRF to this receptor subtype (Fox et al., 1993; Elliott-Hunt et al., 2002; Pedersen et al., 2001; Bayatti et al., 2003; Koutmani et al., 2013). However, in rat cortical neurons, there is also evidence that the protective actions of CRF-related peptides against glutamate excitotoxicity depend on both CRF-R1 and CRF-R2 (Valadas et al., 2012). Therefore, additional studies are needed to assess whether CRF-R2 is involved in the neuroprotection against HEA and other stressors.

The stimulatory effects of HEA on CRF transcription in the brain of fish and the growing body of evidence implicating CRF-related peptides in the modulation of ion channel activity and oxidative stress may provide a mechanistic basis by which endogenous CRF confers neuroprotection against HEA-mediated overactivation of NMDA receptors in zebrafish larvae.

As evidence of the role of CRF and UTS1 in the activation of the HPI axis, HEA exposure in adult rainbow trout (Oncorhynchus mykiss) increases their hypothalamic mRNA levels (Ortega et al., 2005; Bernier et al., 2008). Similarly, we have recently shown that the hyperammonemic conditions used in this study increase crfa and crfb expression and whole-body cortisol levels in 5 dpf larvae (Williams et al., 2017b). Although further work is needed to determine whether extrahypothalamic CRF-expressing neurons in the zebrafish brain (Alderman and Bernier, 2007) are also HEA responsive, the above studies suggest that CRF-related peptides are released from neurons in response to HEA. Characteristic cellular events associated with overactivation of NMDA receptors include Ca²⁺ overload and an increase in reactive oxygen species (Choi, 1988).

In this light, it is interesting to note that the cytoprotective effects of CRF-related peptides in cardiomyocytes (Tao and Li, 2005) and chondrocytes (Lawrence et al., 2017) involve gating of Ca²⁺ channels and a reduction in Ca²⁺ overload, and that activation of CRF-R1 protects neurons from injury by reducing oxidative stress (Lezoualc'h et al., 2000; Bayatti and Behl, 2005; Hanstein et al., 2009), mechanisms which could provide a link between CRF-R1 activation and protection against HEA neurotoxicity.

Overall, our findings raise numerous questions on the potential impacts of HEA exposure on brain development in fish and the functional roles of the CRF system in mitigating HEA neurotoxicity. Hyperammonemic conditions caused by liver failure in mammals can lead to a wide range of cognitive impairments (Felipo, 2013). Although zebrafish embryos have much higher ammonia tolerance than larvae (Williams et al., 2017b; Zielonka et al., 2018), here we show that post-mitotic neurons are susceptible to HEA toxicity at both developmental stages, and that HEA exposure in larvae can affect various neural cell types and cause cell death in an isolated brain. Whether such HEA-mediated neural alterations can interfere with functional brain development in fish remains to be determined. Similarly, our discovery that exogenous CRF has pro-survival properties in an isolated larval zebrafish brain raises questions about the role of endogenous CRF in brain resilience, i.e. in the capacity of the brain to withstand environmental insults. Despite the evolution of four CRF-related peptides in fish and mammals, all have cytoprotective properties under various conditions (Facci et al., 2003; Brar et al., 2004; Williams et al., 2017a). The fact that this role has escaped subfunctionalization within the vertebrate lineage since the last common ancestor of fish and mammals suggests a strong and vital role for cytoprotection by CRF-related peptides. The role that CRF plays in neuroprotection may prove to be as significant as its role in stress signaling, given its apparent conservation, and these two roles maybe intertwined.

We would like to extend sincere thanks to Dr Terry Van Raay for technical advice and sharing his expertise. We are also thankful to Jonathan Shum for his assistance with the experiments on the effects of ammonia exposure on the expression of neural markers. Finally, we are very appreciative of Matt Cornish and Mike Davies for their expert support in the Aqualab.