Characterization of two novel ammonia transporters, Hiat1a and Hiat1b, in the teleost model system Danio rerio

Ammonia is a toxic waste product generated by the catabolism of proteins and nucleic acids, but also via ureolytic and uricolytic pathways (Larsen et al., 2014) (in the current study, NH₄⁺ refers to the ammonium ion, NH₃ to gaseous ammonia, and ammonia to the sum of both). Ammonia is a weak base with a pK_a of 9.0–9.5 and occurs accordingly in physiological solutions predominantly (>90%) in its ionic form, NH₄⁺, while only a small portion is present as the membrane-permeable NH₃ (Cameron and Heisler, 1983). In fishes, elevated plasma ammonia levels caused by exercise or feeding (Bucking, 2017; Fehsenfeld and Wood, 2018; Sinha et al., 2012a,b) need to be tightly regulated to avoid negative physiological implications. Furthermore, when exposed to non-lethal high environmental ammonia (HEA), the resulting internal plasma ammonia load can cause a wide range of deleterious effects in fish, including disruption of general ion homeostasis (Diricx et al., 2013; Wilkie, 1997), reduced swimming performance (McKenzie et al., 2009, 2003; Shingles et al., 2001), changes in energy metabolism (Arillo et al., 1981; Sinha et al., 2012a) and a reduced growth rate (Dosdat et al., 2003). To avoid its negative side effects, ammonia needs to be rapidly excreted from the body fluids. With a few exceptions, teleost fishes are ammoniotelic and excrete up to 90% of their metabolically produced ammonia directly through the gills (Sayer and Davenport, 1987; Smith et al., 2012; Smith, 1929; Zimmer et al., 2014).

In 2009, Wright and Wood (2009) published a working model for the branchial ammonia excretion mechanisms in a freshwater fish with the contribution of Rhesus-glycoproteins. It is assumed that plasma ammonia enters the epithelial cell in its gaseous form NH₃ via a basolateral localized Rhesus-glycoprotein, Rhbg, along its partial pressure gradient. Further, Mallery (1983) and Nawata et al. (2010) showed that similar to invertebrates (Adlimoghaddam et al., 2015; Hans et al., 2018; Masui et al., 2002) and the mammalian kidney (Wall and Koger, 1994), teleost Na⁺/K⁺-ATPase (NKA) is capable of accepting NH₄⁺ instead of K⁺ as a substrate. Consequently, in addition to NH₃ uptake via Rhbg, the basolateral NKA may mediate an active uptake of extracellular NH₄⁺ into the branchial epithelial cell. The capacity of NKA to accept NH₄⁺ as a substrate, however, seems to be dependent on external salinity and varies substantially among fishes (Wood and Nawata, 2011).

According to the model (Wright and Wood, 2009), it is further assumed that cytoplasmatic ammonia is excreted apically in a sodium-dependent manner via an apical Na⁺/NH₄⁺ exchange complex (Ito et al., 2013; Wright and Wood, 2009). This complex is thought to act as a metabolon that promotes the acidification of the unstirred boundary layer by means of a V-type H⁺-ATPase and Na⁺/H⁺-exchanger 3 (NHE3), with both transporters creating a partial pressure gradient for gaseous NH₃. NH₃ follows this gradient utilizing another Rhesus-glycoprotein, Rhcgb (formerly Rhcg1; Nakada et al., 2007), a process commonly referred as ammonia trapping. A membrane-bound carbonic anhydrase (CA2a and/or CA15a; Ito et al., 2013) supports this metabolon by catalysing the hydration of CO₂, providing the required H⁺ for the process (Wright and Wood, 2009). A similar ammonia excretion mechanism has also been proposed for freshwater planarians (Weihrauch et al., 2012). An equivalent mechanism, i.e. a sodium-dependent apical Na⁺/NH₄⁺ exchange metabolon, however, could not be verified for freshwater leeches, where the cutaneous ammonia excretion process was shown to be independent of functional apical Na⁺ channels and NHEs (Quijada-Rodriguez et al., 2017). Furthermore, in the nematode Caenorhabditis elegans, knock-out of the apically localized Rh-protein RHR-2 caused a reduction in ammonia excretion rates; the remaining flux was, however, independent of environmental pH, in contrast to the wild-type (Adlimoghaddam et al., 2016). However, in contrast to predictions stemming from the metabolon model (Wright and Wood, 2009), knocking out the apical Rhcgb in zebrafish via CRISPR/Cas9 did not reduce ammonia excretion and/or Na⁺ uptake. Rather, an increase in ammonia excretion was observed (Zimmer and Perry, 2020). The diversity of ammonia excretory mechanisms across fishes and other aquatic animals underlies their importance for nitrogen homeostasis, and at the same time warrants the consideration of additional routes and mechanisms.

In invertebrates, for example, NH₄⁺-transporting ammonium transporters (AMTs, also members of the AMT/MEP/Rh-protein ammonia transporter family), driven by the cytoplasm-to-environment electrochemical gradient of the cation, were suggested to be involved in the epithelial ammonia excretion process (clams: Boo et al., 2018; insects: Chasiotis et al., 2016; Durant and Donini, 2018; Pitts et al., 2014; polychaeta: Thiel et al., 2017). In vertebrates, however, AMTs have not been shown to be expressed to date.

An additional potential contributor to the branchial ammonia excretory process is the hippocampus-abundant transcript 1 protein (Hiat1). This presumptive ammonia transporter (CmHiat1) was associated with acid–base regulation in the green crab, Carcinus maenas, because its mRNA levels decreased when the animals were acclimated to elevated P_CO2 levels (Fehsenfeld et al., 2011). Recently, it was demonstrated that CmHiat1 acts as an ammonia (likely NH₄⁺) transporter in this species (S.F., A. R. Quijada-Rodriguez, H.Z., A. C. Durant, A. Donini, M. Sachs, M. F. Romero, P.E. and D.W., unpublished) with its mRNA abundance found to be up-regulated in response to high environmental ammonia.

In light of the results of these studies, the current study was designed to test the hypothesis that Hiat1 is involved in ammonia excretion in zebrafish larvae. To address these hypotheses, we investigated Hiat1 function using multiple physiological and molecular techniques. After identifying and characterizing Hiat1 paralogues in a phylogenetic analysis, we expressed DrHiat1a and DrHiat1b proteins in frog oocytes to verify their capability to transport methylamine (ammonia analogue) competitively with NH₄Cl. We then determined mRNA expression patterns of both paralogues throughout early development (4-cell stage to 4 days post-fertilization) by in situ hybridization. Lastly, we knocked down DrHiat1a and DrHiat1b using morpholino injections and applied the scanning ion-selective electrode technique (SIET) to identify regions of interest of Hiat1-related ammonia flux in D. rerio larvae in early development.

DrHiat1-like (GenBank accession no. XP_002663499.2; 500 amino acids), DrHiat1b (GenBank accession no. AAH97075.1, 485 amino acids) and the translated DrHiat1a protein based on the open reading frame (ORF) of zebrafish DrHiat1a (GenBank accession no. BC056817, 493 amino acids) were aligned with Clustal Omega (https://www.ebi.ac.uk/Tools/msa/clustalo/). Protter (Omasits et al., 2014) was used to predict transmembrane (TM) domains and the potential phosphorylation sites for all three Hiat1 proteins, and Scansite 4.0 (Obenauer et al., 2003) was used to identify the binding motifs. All transmembrane models and proteoform predictions were performed using the system's default settings.

DrHiat1b was then used to identify Hiat1 isoforms in other species via GenBank-BLASTp search (Table S1) (Altschul et al., 1997). Proteins were aligned with the MUSCLE algorithm as provided by the Molecular Evolutionary Genetics Analysis across computing platforms (MEGA X; Kumar et al., 2018) using default settings. The evolutionary history was inferred by using the Maximum Likelihood method and JTT matrix-based model (Jones et al., 1992) as identified by the testing function provided by MEGA X. Initial trees for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log-likelihood value. A discrete Gamma distribution was used to model evolutionary rate differences among sites [5 categories (+G, parameter=0.3614)]. All positions with less than 95% site coverage were eliminated, i.e. fewer than 5% alignment gaps, missing data and ambiguous bases were allowed at any position (partial deletion option).

cDNA clones containing the full ORFs of DrHiat1a (clone ID: 2640969; GenBank accession number BC056817.1) and DrHiat1b (clone ID: 2600163, GenBank accession number BC097075) were purchased from Dharmacon (Lafayette, CO, USA). DNA miniprep kits (Qiagen, Hilden, Germany) were used to extract the plasmids following the manufacturer's guidance. Regular PCR employing Q5 High-Fidelity DNA Polymerase was used to obtain the full ORFs of DrHiat1a and DrHiat1b (see Table S2, ORF). The annealing temperature used for the amplification of DrHiat1a and DrHiat1b ORFs was 60°C and 62°C, respectively, with a total of 35 cycles. Subsequently, regular PCR was used to add restriction enzyme sites to the 5′ and 3′ ends of the ORFs of DrHiat1a and DrHiat1b (see Table S2, oocytes). The annealing temperature used for the amplification of DrHiat1a and DrHiat1b ORFs was 65°C for both, with a total of 35 cycles. The quality of the DNA was assessed by gel electrophoresis and the concentration of DNA was determined spectrophotometrically after purification (GeneJET PCR purification kit, ThermoFisher, Waltham, MA, USA). The ORFs were then cloned into the pGEM^®-HE plasmid, a modified pGEM^®-3Z vector containing the Xenopus beta globin 5′- and 3′-UTR sequences, using T4 ligase following the manufacturer's guidance [New England Biolabs (NEB), Ipswich, MA, USA]. The in vitro transcription of capped mRNA (cRNA) was performed with HiScribe™ T7 ARCA mRNA kit (NEB) followed by column purification (RNeasy MinElute Cleanup Kit, Qiagen). The cRNA was quantified spectrophotometrically (Nanodrop, ND-1000, ThermoFisher) and integrity was assessed on a Mops agarose gel containing formaldehyde.

Stage VI–V oocytes were collected from mature female Xenopus laevis (VWR, Radnor, PA, USA) as previously described (Soreq and Seidman, 1992) (Protocol Reference F20-021). Briefly, the frogs were euthanized via decapitation prior to ovary collection. The ovary was placed in Ca²⁺-free OR2 solution (in mmol l⁻¹: 82.5 NaCl, 2.5 KCl, 1 MgCl₂, 1 Na₂HPO₄, 5 Hepes; pH 7.2) with collagenase type VI (1 mg ml⁻¹) (Gibco, Waltham, MA, USA) and gently agitated for 90 min at room temperature; 1 mmol l⁻¹ CaCl₂ was added to terminate the collagenase activity. Oocytes were then manually sorted and rinsed three additional times with standard OR2 solution. For overnight storage of isolated oocytes, OR2 was supplemented with 2.5 mmol l⁻¹ sodium pyruvate, 1 mg ml⁻¹ penicillin–streptomycin (Gibco, Long Island, NY, USA) and 50 μg ml⁻¹ gentamicin at 16°C. All procedures used were approved by the University of Manitoba Animal Research Ethics Board and are in accordance with the Guidelines of the Canadian Council on Animal Care.

Isolated oocytes after the overnight recovery were injected with 18.4 ng of cRNA (36.8 nl with 0.5 ng nl⁻¹) or nuclease-free water as control using a Nanoject II auto-nanolitre injector (Drummond Scientific, Broomall, PA, USA). Experiments were conducted 48 h post-injection.

Experiments were conducted at room temperature in 200 μl of standard OR2 solution, containing 3 μmol l⁻¹ (9.3 μCi) [H³]methylamine (Moravek Inc., Brea, CA, USA) with 60 min of incubation time. For each experiment 20 oocytes injected with either with cRNA or water (sham) were randomly selected. At the end of each experiment, ice-cold standard OR2 solution was used to terminate [H³]methylamine uptake by washing oocytes 3 times to remove the external radioactivity. The washed oocytes were immediately solubilized individually in 200 μl of 10% SDS (Sigma-Aldrich, St Louis, MO, USA) to assess the internal radioactivity, employing 5 ml of Ultima-Gold scintillation cocktail (PerkinElmer) and liquid scintillation counting (Tri-Carb 2900 TR; PerkinElmer). The effect of washing was determined by assessing the radioactivity in 200 μl standard OR2 solution used for the third wash post-radioactivity exposure compared with fresh, radioactivity-free standard OR2 solution of the same volume. For NH₄Cl competitive uptake experiments, the same procedures described above were conducted, but with 1 mmol l⁻¹ NH₄Cl added to the incubation solution.

To determine whether Hiat proteins mediate bi-directional methylamine transport, the efflux of methylamine from sham oocytes, and from DrHiat1a- and DrHiat1b-expressing oocytes was measured. Experiments were performed at room temperature in 200 μl of standard OR2 solution containing 3 μmol l⁻¹ (9.3 μCi) [H³]methylamine adjusted to pH 7.2, with 60 min incubation time. Two groups of cRNA-injected oocytes (n=20) and two groups of water-injected oocytes (sham; n=20) were assessed for each experimental point. Each group was considered as one replicate and, overall, two replicates were done. After termination of [H³]methylamine uptake as described above, one group of cRNA-injected oocytes and one group of water-injected oocytes were immediately solubilized individually to assess the uptake of [H³]methylamine as described above, to determine the starting point for the following release experiment. The remaining groups of cRNA-injected oocytes and water-injected oocytes were placed in 200 μl of room temperature, radioactivity-free standard OR2 solution for 60 min for release and afterwards washed 3 times with ice-cold standard OR2 solution, solubilized individually in 200 μl of 10% SDS and assessed for remaining internal radioactivity. The efficiency of the washing steps was determined by analysing 200 μl of the radioactivity of the third wash post-radioactivity exposure compared with that of fresh, radioactivity-free standard OR2 solution of the same volume.

DrHiat1a and DrHiat1b ORFs were amplified and cloned as described above with Q5 High Fidelity DNA Polymerase (NEB) and primers located at both ends of the ORFs with restriction enzyme sites (Table S2, In situ).

The deletion of the Xenopus beta-globin 5′-UTR was done by removing a 63 bp fragment containing the 5′-UTR with KpnI and XmaI followed by blunting with Quick Blunting Kit (NEB) and re-ligation. The Xenopus beta-globin 3′-UTR was removed in the course of linearization of the vector for the preparation of an antisense RNA probe. The antisense RNA probe was synthesized as previously described (Thisse and Thisse, 2008) and experiments were conducted at the University of Bayreuth, Germany, using the Casper strain (White et al., 2008). Adult fish were maintained under standard conditions (14 h:10 h light:dark cycle; 0.1 g l⁻¹ salinity; 300 µS conductivity; pH 7.5) at 27.5–28.5°C. Embryos were obtained from pairing events of 3 male and 2 female fish and were reared in E3-medium (in mmol l⁻¹: 5 NaCl, 0.17 KCl, 0.33 CaCl₂, 0.33 MgSO₄, 10.5% Methylene Blue; pH 7.2) at 28.5°C.

Zebrafish embryos and larvae were euthanized on ice, fixed in 4% PFA/PBS, washed in PBS/0.1% Tween 20 (PBTw), transferred to 100% methanol and stored at least overnight at −20°C. After rehydration using a methanol/PBTw series and post-fixation in 4% PFA/PBS, older developmental stages were treated with 10 µg ml⁻¹ proteinase K (Fermentas, Waltham, MA, USA) for 10 min (2 days post-fertilization, dpf), 15 min (3 dpf) and 20 min (4 dpf), post-fixed in 4% PFA/PBS and washed several times in PBTw. After hybridization with the corresponding antisense probes overnight at 68°C, the samples were washed in a formamide solution/PBTw series, blocked in 0.5% blocking reagent, incubated in a 1:2000 dilution of anti-DIG-AP antibody in 0.5% blocking reagent for 4 h at room temperature and washed several times in PBTw. For detection, samples were equilibrated in BCL buffer and stained with NBT and BCIP/X-phos until staining was clearly visible.

Danio rerio (F. Hamilton 1822) originally purchased from the pet trade (Big Al's Aquarium, Ottawa, ON, Canada) were maintained in flow-through aquaria receiving dechlorinated city of Ottawa tap water (‘system water’; in mmol l⁻¹: 0.8 Na⁺, 0.4 Cl⁻, 0.25 Ca²⁺; pH 7.6) and held at 28.5°C with a 14 h:10 h light:dark photoperiod. Embryos were obtained from breeding events between 1 male and 2 females held in static tanks, and reared in 60 ml Petri dishes containing system water with 0.05% Methylene Blue at a density of approximately 50 embryos per dish held at 28°C. Handling and experimentation of zebrafish was conducted in compliance with the University of Ottawa Animal Care and Veterinary Service (ACVS) guidelines under protocol BL-1700 and followed the recommendations from the Canadian Council for Animal Care (CCAC).

Splice-blocking morpholino oligonucleotides (Gene Tools, LLC, Philomath, OR, USA) targeting the intron–exon junction between intron and exon 4 of DrHiat1a (NCBI accession no.: NM_199584.1; 5′-CATGGTGAGAGCCTCTGAAATCAAG-3′) and the intron–exon junction between intron and exon 4 of DrHiat1b (NCBI accession no.: NM_213527.2; 5′-AAAGTGAGGAGCGTAACGAACCATG-3′), and a standard control morpholino with no biological target in zebrafish (5′-CCTCTTACCTCAGTTACAATTTATA-3′) were suspended to a final concentration of 2–4 ng morpholino nl⁻¹ in Danieau buffer [in mmol l⁻¹: 58 NaCl, 0.7 KCl, 0.4 MgSO₄, 0.6 Ca(NO₃)₂, and 5.0 Hepes buffer; pH 7.6] containing 0.05% Phenol Red for injection visualization. One-cell stage embryos were injected with 1 nl of this injection solution [4 ng per embryo for DrHiat1a morpholino (MO), 4 ng per embryo for DrHiat1b MO, and 4 ng per embryo for control (sham) MO] using an IM 300 microinjection system (Narishige, Amityville, NY, USA). However, we observed a large proportion (>50%) of morphological abnormalities (curved tail/spine) with the 4 ng DrHiat1b dose and therefore used a dose of 2 ng for DrHiat1b and observed virtually no morphological abnormalities (<5–10%). Injections were verified at 1 dpf by examining the distribution of carboxyfluorescein (tagged to each morpholino) in individual embryos under a fluorescence dissecting microscope (Nikon SMZ1500, Nikon Instruments, Melville, NY, USA). Embryos that were successfully injected, containing a homogeneous distribution of fluorescein signal, were reared as described above.

Knockdown was confirmed by reverse transcription PCR (RT-PCR). At 4 dpf, approximately 20 pooled larvae (n=1) from each treatment group (DrHiat1a MO, DrHiat1b MO and sham MO) were euthanized by an overdose of neutralized tricaine methanesulfonate (MS-222; Syndel Canada, Nanaimo, BC, Canada), flash frozen in liquid nitrogen, and stored at −80°C. RNA was extracted from pooled samples using Trizol (ThermoFisher) following the manufacturer’s protocol. Extracted RNA was treated with DNAse (ThermoFisher) and cDNA was then synthesized from 700 ng total RNA using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA). PCR was performed using DreamTaq DNA polymerase (ThermoFisher) following the manufacturer's guidance, with primers specific to DrHiat1a and DrHiat1b (see Table S2). The reaction volume for each PCR reaction was 20 µl with 55°C annealing temperature and the reaction was terminated after 40 cycles.

SIET was performed to assess NH₄⁺ flux in sham and DrHiat1b MO-treated 4 dpf larvae. Measurements were made in a K⁺-free medium which was prepared by adding stock solutions of Na₂SO₄, CaCl₂, MgSO₄·7H₂O, Na₂HPO₄ and NaH₂PO₄ to doubly distilled water (in mmol l⁻¹: 0.84 Na⁺, 0.5 Cl⁻, 0.25 Ca²⁺, 0.15 Mg²⁺; pH 7.6). The omission of K⁺ in the medium is necessary because this ion interferes with NH₄⁺ signals from the ion-selective probe. In addition, the medium contained MS-222 anaesthetic (0.2 g l⁻¹) and 0.05 mmol l⁻¹ (NH₄)₂SO₄, the latter being necessary because the Nernstian slope of NH₄⁺ tends to fall at concentrations below 0.1 mmol l⁻¹.

NH₄⁺-selective probes were constructed as described previously (Donini and O'Donnell, 2005). Glass capillary tubes (World Precision Instruments, Sarasota, FL, USA) were pulled to microelectrodes with a tip diameter of approximately 5 µm using a P-2000 micropipette puller (Sutter Instrument, Novato, CA, USA) and silanized with N,N-dimethyltrimethylsilylamine (Sigma-Aldrich) on a hot plate covered with a glass Petri dish. Silanized microelectrodes were then back-loaded with 100 mmol l⁻¹ NH₄Cl and front-loaded with an approximately 250 μl ionophore column of NH₄⁺ Ionophore I Cocktail A (Sigma-Aldrich). Probes were calibrated in the same K⁺-free medium containing MS-222 described above to which NH₄Cl was added to a final concentration of 0.1, 1 or 10 mmol l⁻¹. pH was maintained at 7.6 across all calibration solutions using NaOH or H₂SO₄. The slope (in mV) for a tenfold change in NH₄⁺ concentration was 59.2±1.1 (n=3).

All data were tested for normal distribution by Shapiro–Wilk's test. In cases where data were not normally distributed, data were log transformed. Homogeneity of variances was assessed by Levene's test. If data did not meet the criteria for parametric testing including normal distribution and homogeneity of variances, non-parametric test were applied. Statistical analyses were performed with PAST3 (Hammer et al., 2001). Graphs were generated with GraphPad Prism 9.0 for Windows (GraphPad Software, San Diego, CA, USA; www.graphpad.com) and Inkscape (https://inkscape.org).

The three isoforms of DrHiat1 are referred to as DrHiat1a, DrHiat1b and Hiat1-like according to the current nomenclature used by GenBank. DrHiat1a (493 amino acids), DrHiat1b (485 amino acids) and DrHiat1-like (500 amino acids) exhibited a conservation of 323 identical amino acids (ca. 66%), with 80 residues as strongly conserved groups (two out of three amino acids identical; ca. 16%) and 19 residues as weakly conserved groups (three different amino acids but with similar properties; ca. 4%). The structure analysis of DrHiat1a, DrHiat1b and DrHiat1-like by Protter (Omasits et al., 2014) identified 12 predicted TM domains. Furthermore, a sugar binding motif (amino acids 91–108 for DrHiat1a, amino acids 89–106 for DrHiat1b, amino acids 103–120 for Hiat1-like) seems to be associated with TM3 in all three cases. While DrHiat1a and DrHiat1b contained two equivalent phosphoserine/threonine binding groups (amino acids 5–21/amino acids 5–19, N-terminal; amino acids 299–312/amino acids 297–310, between TM8 and TM9) (Larner et al., 1993), DrHiat1b contained an additional one between TM1 and TM2 (amino acids 55–70), but DrHiat1-like did not contain any at all. DrHiat1b was the only isoform to contain a proline-dependent serine/threonine kinase group (Lew et al., 1992; Meyerson et al., 1992) (Cdk5, amino acids 456–469) (Fig. S1).

Our preliminary phylogenetic analysis showed Hiat1 to be conserved across many taxa and clustered in three distinct clades according to the three different paralogues (Fig. 1).

Both DrHiat1a and DrHiat1b-expressing oocytes showed significantly higher (2.7-fold) methylamine (MA) uptake compared with the sham (oocytes injected with water, negative control). There was no difference in the uptake of MA between oocytes expressing DrHiat1a or DrHiat1b (Fig. 2A; ANOVA with Tukey's pairwise comparisons with P<0.001, N=20).

When exposed to 1 mmol l⁻¹ non-labelled NH₄Cl in addition to MA in the medium, the uptake of MA mediated by DrHiat1a and DrHiat1b was reduced by 45% and 40%, respectively. The degree of inhibition of MA by NH₄Cl was not significantly different between oocytes expressing DrHiat1a and DrHiat1b (Fig. 2B).

Pre-loaded DrHiat1a- and DrHiat1b-expressing oocytes released significantly more ammonia compared with sham-injected oocytes (27% versus 15%; Fig. 3). There was no difference between the amount of MA released in oocytes expressing Hiat1a and Hiat1b.

To determine Hiat1 expression in zebrafish, DrHiat1a and DrHiat1b antisense RNA probes were generated and whole-mount in situ hybridization (WISH) was performed on zebrafish embryos and larvae of different developmental stages, beginning from the 4-cell stage until 4 dpf (Fig. 4). This revealed maternal expression of both DrHiat1a (Fig. 4Ai) and DrHiat1b (Fig. 4Bi), early onset of transcription during the maternal-to-zygotic transition (Fig. 4Ai′ and Bi′) and ubiquitous expression in all tissues during early segmentation stages (Fig. 4Aii,ii′,Bii,ii′). At 24 h post-fertilization (hpf), towards the end of the segmentation period, expression of both genes was mainly localized to the head region, especially to the forebrain, midbrain and hindbrain, including the midbrain–hindbrain boundary, the lens and the trigeminal ganglion (Fig. 4Aiii,iii′,Biii,iii′). In addition, DrHiat1a showed distinct expression in the olfactory placode and the hatching gland in both examined segmentation stages (Fig. 4Aii,ii′,iii,iii′), while DrHiat1b showed additional expression in the optic vesicle, posterior lateral line ganglion and pectoral fin bud at 24 hpf (see Fig. S1). From 2 to 4 dpf, DrHiat1a and DrHiat1b expression could be detected in the brain, heart, gills, digestive system, pectoral fins and parts of the sensory nervous system including the cranial and lateral line ganglia (DrHiat1a and DrHiat1b) and the retinal ganglion cell layer and lateral line neuromasts (DrHiat1b) (Fig. 4Aiv,iv′,v,v′,Biv,iv′,v,v′; Fig. S1). For the digestive system, DrHiat1a appeared to be more strongly expressed in the pancreas, whereas DrHiat1b was more apparent in the intestine (Fig. S1E–H). Because of the proteinase K treatment, which was performed from 2 dpf onwards, staining of some of the more external structures, such as the pectoral fins and lateral line neuromasts, was reduced or even completely abolished. For images without proteinase K treatment, see Fig. S2.

Successful knockdown of Hiat1a and Hiat1b using splice-blocking MOs was verified by PCR, where the reduced size of amplicons (from 450 to 300 bp) indicates successful splice blocking of the 4th exon in pre-mRNA DrHiat1a and DrHiat1b such that the respective exon was not included in mature mRNA (Fig. 5A).

A significant reduction of ammonia excretion was found in larvae with DrHiat1b knockdown, but not with DrHiat1a knockdown (Fig. 5B). SIET was then used to measure the regional flux of NH₄⁺ on larvae with DrHiat1b knockdown (Fig. 6A). NH₄⁺-selective electrodes were placed at the yolk sac and pharyngeal region/immature gill region (Fig. 6A). A significant reduction of NH₄⁺ flux was detected at both regions in DrHiat1b knockdown larvae compared with the sham larvae (Fig. 6B).

Despite their high level of sequence conservation (86%), Hiat1a, Hiat1b, Hiat1-like transporters (renamed mfsd14a/b in mammals; Doran et al., 2016; Lekholm et al., 2017) could be clearly distinguished in the phylogenetic analysis. Unfortunately, likely as a result of the high level of amino acid similarity and the lack of research on the phylogeny and function of these transporters, there is currently no consistent nomenclature available in the literature and databases. Consequently, the reader needs to be aware that many proteins annotated as one or the other in the NCBI database did not hold true in our analysis (i.e. many transcripts annotated as Hiat1-like instead cluster as Hiat1b or Hiat1a). It should also be noted that some sequences that were annotated as different isoforms turned out to be simply truncated/not properly assembled versions of the same gene. We therefore present here a potential reference for future annotations of this transporter based on the zebrafish names for each of the Hiat1 paralogues.

To date, only one isoform of Hiat1 has been identified in invertebrates (S.F., A. R. Quijada-Rodriguez, H.Z., A. C. Durant, A. Donini, M. Sachs, M. F. Romero, P.E. and D.W., unpublished). Interestingly, while all three isoforms can be identified in teleosts, other vertebrates seem to lack one or more: for lamprey, only Hiat1b could be identified, Hiat1-like was not found for elasmobranchs, and no Hiat1a was identified for terrestrial reptilians as well as terrestrial and aquatic mammals. At this point, further phylogenetic analysis would be needed to clarify the exact evolutionary history of the different isoforms.

The present study revealed that both DrHiat1a and DrHiat1b could promote MA uptake when expressed in frog oocytes. Furthermore, MA release was enhanced in pre-loaded oocytes when expressing either of the Hiat1 paralogues. Adding the fact that external ammonia was able to inhibit MA uptake in transgenic oocytes (as seen by the immediate inhibition of MA uptake in the presence of NH₄Cl in the medium) allows us to conclude that DrHiat1a and DrHiat1b also function as ammonia transporters. An equivalent observation was made in our recent study (S.F., A. R. Quijada-Rodriguez, H.Z., A. C. Durant, A. Donini, M. Sachs, M. F. Romero, P.E. and D.W., unpublished) in green crabs, where CmHiat1 promoted ammonium transport (likely NH₄⁺), shown either directly with scanning ion-selective microelectrodes, or indirectly with radiolabelled MA as a proxy. Despite possessing a sugar transporter-specific motif D-R/K-X-G-R-R/K between TM domain 2 and TM domain 3 (Matsuo et al., 1997), recent results clearly showed a lack of glucose transport in X. laevis oocytes when expressing crustacean Hiat1 (S.F., A. R. Quijada-Rodriguez, H.Z., A. C. Durant, A. Donini, M. Sachs, M. F. Romero, P.E. and D.W., unpublished). Similarly, mammalian Mfsd14a was not capable of mediating glucose uptake into the oocyte (Zhouyao et al., 2022).

Interestingly, competition of NH₄⁺ transport with MA uptake was only observed in transgenic oocytes, but barely in sham-injected oocytes. This might simply be based on the respective concentration of NH₄⁺ (1 mmol l⁻¹): competition might occur at higher concentrations of NH₄⁺ as the kinetics of ammonia/MA transport by endogenous pathways in Xenopus oocytes may just have a much higher inhibition constant. Or, in other words, MA uptake by DrHiat1 may have a lower inhibition constant than the endogenous pathways for ammonia uptake of the Xenopus oocytes (i.e. via Rhesus-glycoproteins and/or NHE3), such that we saw this large difference at 1 mmol l⁻¹ NH₄Cl.

Ammonia handling and excretion seem to be especially important in early life stages of fish for a variety of reasons. First, because of an amino acid-focused metabolism, the ammonia load in fish embryos is substantial (Zimmer et al., 2017). Second, embryos experience physiological constraints; these include the presence of a chorion capsule as a barrier that impedes diffusion between the embryo and its external environment, and the lack of fully functional gills which function as the main ammonia excretory organ in more mature fish (Zimmer et al., 2017).

Generally, ammonia excretion in larval zebrafish occurs in part via H⁺-ATPase-rich (HR) cells in the yolk sac and gills that also contain apical Rhesus-glycoprotein 1 (Rhcg1 or Rhcgb) (Nakada et al., 2007), NHE3 (Ito et al., 2013) and H⁺-ATPase (Lin et al., 2006). Furthermore, with their unique expression patterns along the skin, gills and yolk sac, Rhag, Rhbg, Rhcg1 (Rhcgb) and Rhcg2 (Rhcgl1) have all been shown to contribute to whole zebrafish larvae ammonia excretion to varying extents (Braun et al., 2009; Zimmer and Perry, 2020). Additionally, Rhbg and/or Rhcg1 (Rhcgb) also seem to more specifically participate in the excretion of ammonia via skin keratinocytes (Shih et al., 2008, 2013). In contrast, a Na⁺/NH₄⁺ exchange metabolon as mentioned in the Introduction, seems instead to contribute to ammonia excretion only under challenging conditions (i.e. low pH and/or low Na⁺ environments) and less so when fish were kept in unaltered laboratory conditions (i.e. circumneutral water with adequate salt concentrations) (Kumai and Perry, 2011; Shih et al., 2012).

Here, we provide the first evidence that ammonia excretion in this early life stage appears to be additionally mediated by DrHiat1b, hence giving an alternative route for basic ammonia transport. Interestingly, Hiat1 is expressed in frog oocytes and hence maternally provided (S.F., A. R. Quijada-Rodriguez, H.Z., A. C. Durant, A. Donini, M. Sachs, M. F. Romero, P.E. and D.W., unpublished), ensuring it is passed on with the maternal lineage, implying a crucial role for this transporter as early as development begins. The fact that both DrHiat1a and DrHita1b are ubiquitously expressed in the zebrafish embryo and early larvae emphasizes their importance as potential ammonia transporters in this crucial phase of life. They might provide a very general route for baseline ammonia excretion in every tissue, complemented by more differentially expressed and tissue-specific Rhesus-glycoproteins. It is not surprising that both Hiat1 transcripts can initially be found in the nervous system, resembling the first discovery of Hiat in the marsupial hippocampus (Matsuo et al., 1997). DrHiat1b being more abundantly expressed might indicate its more important role for basic ammonia maintenance and might also explain the subtle differences in tissue expression. For instance, while mRNA for DrHiat1b was present in the intestine, mRNA for DrHiat1a was not detectable.

In DrHiat1b MO knockdowns, whole-larvae ammonia excretion was reduced by ca. 30%. The fact that we observed an even further reduction of ammonia excretion when we specifically screened the yolk sac (−75%) and pharyngeal region (immature gill; −85%) underlines the complexity of excretory mechanisms in this stage. Similar discrepancies have been observed in earlier studies on zebrafish larvae, when for example, Zimmer and Perry (2020) found that a knockdown of Rhcgb resulted in an increase of whole-larvae ammonia excretion, whereas NH₄⁺ flux at the yolk sac assessed by SIET was reduced in response to knockdown using the same MO (Shih et al., 2008; 2012). Further thorough investigations are needed to fully understand these respective mechanisms, including potential off-target effects and/or compensatory responses caused by knockdown, i.e. the possibility that other ammonia transporters such as Rhesus-glycoproteins take over after the loss of Hiat1 (Kok et al., 2015; Rossi et al., 2015).

The findings of the present study warrant a reconsideration of ammonia-related transport processes in the gills and potentially other excretory organs of fish. We show here that especially early on in development, the involvement of Hiat1b and/or Hiat1a provides an alternative route for ammonia transport besides the involvement of Rhesus-glycoproteins and/or Na⁺/NH₄⁺ exchange metabolon. Alternatively, Hiat1a/1b could be part of this metabolon. It would be highly desirable for future studies to identify the transporters' exact cellular localization in the branchial membrane to allow for more detailed modelling of branchial ammonia transport. Clearly, Hiat1 transporters have to be taken into account in future studies on ammonia transport in fish and other animals, including the mammalian kidney.

We like to thank the Animal Care and Veterinary Services at the University of Ottawa, and in particular Christine Archer and Vishal Saxena for taking care of the zebrafish.