Two closely related ureotelic fish species of the genus Alcolapia express different levels of ammonium transporters in gills

Ammonia is the major nitrogenous waste product in typical fish species (Wood, 1993) and is produced as a result of the catabolism of amino acids (Wright and Fyhn, 2001). At high concentrations ammonia is toxic (Handy and Poxton, 1993); it accumulates in the nervous system and can lead to neurotoxicity (Oja et al., 2017; Rangroo Thrane et al., 2013). Due to its toxic effects, ammonia must either be efficiently excreted or converted into less toxic compounds such as urea (Mommsen and Walsh, 1991). Most fish are ammonotelic and excrete their nitrogenous waste as ammonia, predominantly across gill tissue. This method of excretion avoids the metabolic costs of converting ammonia to less toxic compounds and, moreover, gill permeability to ammonia is at least double that of urea (Wright et al., 1995). However, under environmental conditions such as heightened pH, ammonia cannot efficiently diffuse across the gills and will accumulate in the body (Wright and Wood, 1985). Some fish species, including the Nile tilapia, Oreochromis niloticus (Wright, 1993) are known to excrete a proportion of their nitrogenous waste as urea under high pH conditions (Randall et al., 1989), (Walsh et al., 1990; Wright et al., 1993). The soda lake cichlid Oreochromis (Alcolapia) grahami is reported to be the only 100% ureotelic fish species (Walsh et al., 1990; Wright et al., 1993).

The cichlids of the subgenus Alcolapia, nested within the genus Oreochromis (Ford et al., 2019), are a unique radiation of extremophile fishes found in the East African soda lakes Magadi (A. grahami) and Natron (A. alcalica, A. ndalalani and A. latilabris) (Ford, 2015; Seegers et al., 1999). Here they have adapted to thrive in some of the most extreme environments supporting fish life, with water temperatures of 30-42.8°C, pH 9-11.5, fluctuating dissolved oxygen levels and high salt concentrations of >20 ppt (Ford, 2015). The geological evidence suggests that their adaptations, which include facultative airbreathing, a specialised gut morphology, and maintaining a heightened metabolic rate, have evolved rapidly over the past 10,000 years, during which these fishes diverged from more freshwater ancestors (Roberts et al., 1993), (Tichy and Seegers, 1999). Another key adaptation to the high pH conditions is their ability to convert almost all their ammonia to urea (Randall et al., 1989).

The discovery of the ammonium transporting function of Rhesus glycoproteins (Rh) has changed the way ammonia excretion is viewed in fish (reviewed in Wright and Wood, 2009; Zimmer et al., 2017). There are four teleost Rh proteins involved in ammonia transport and excretion. Rhag is present in the membrane of red blood cells, while Rhbg, Rhcg1 and Rhcg2 are predominantly detected in gill tissue (Nakada et al., 2007a,b; Wright and Wood, 2009). In addition, some expression of different Rh proteins has been reported in skin, kidney and brain tissue in different fish species (Hung et al., 2007; Braun et al., 2009; Nawata et al., 2007; Nawata and Wood, 2009). Their most important role is believed to be in the gills where most of the ammonia excretion takes place. Immunohistochemistry in pufferfish gills has shown that Rhbg is present on the basolateral membrane while Rhcg2 is on the apical membrane of pavement cells; this arrangement may allow the proteins to work together to transport ammonia across gill tissues into the surrounding water (Nakada et al., 2007a,b). Another Rh protein, Rhcg1, is found on the apical membrane of mitochondria rich cells (chloride cells) and is believed to act in conjunction with a basolaterally positioned Na⁺, K⁺-ATPase, with ammonia substituting for K⁺ allowing excretion across this cell type (Nakada et al., 2007b). Rhcg1 was also shown to be apically positioned in Danio rerio kidney tissue (Nakada et al., 2007a). While this organisation of Rh proteins has been described in the fish gill, the expression of Rhbg in larval skin during early development, in the absence of Rhcg2, suggests that the collaboration of these two Rh proteins is not always required for ammonia transport (Braun et al., 2009), although Rhcg1 could fill this role (Shih et al., 2013). Selective knockdown using morpholinos against Rhag, Rhbg and Rhcg1 in developing D. rerio resulted in around a 50% reduction in ammonia excretion, regardless of the Rh targeted. This suggests that all Rh proteins are required for maintaining full and efficient ammonia excretion. In contrast, a recent paper showed that knock-down of rhcg(b) (also known as rhcg1; Nakada et al., 2007a) in D. rerio did not affect ammonia excretion in larva, which suggests that there are compensatory mechanisms upregulating the expression of other rhesus protein genes such as Rhbg (Zimmer and Perry, 2020).

The presence of Rhbg and Rhcg2 transcripts in Alcolapia gill tissue is surprising since these fish are thought to excrete all nitrogenous waste as urea, as has been reported for A. grahami (Randall et al., 1989; Wood et al., 2013). Interestingly, the expression of Rhbg and Rhcg2 has been detected in liver, muscle and intestinal tissue (as well as the gills) of these fishes (Wood et al., 2013), suggesting that Rh proteins could be used to shuttle ammonia into tissues with known ornithine-urea cycle activity (Lindley et al., 1999; White et al., 2020). It has been suggested that Alcolapia Rhbg and Rhcg2 are structurally different from their orthologues in other species, with 10 rather than 12 transmembrane domains (Wood et al., 2013), raising the possibility that Alcolapia Rh proteins transport another compound or are acting to channel ammonia to tissues for conversion to urea rather than excreting ammonia directly. This suggestion has some support: a single point mutation in human RHAG converts it to a cation-exchanger due to a decrease in pore size (Bruce et al., 2009). Additionally, some Rh proteins have been found to transport other compounds, such as CO₂ or HCO₃⁻ (Huang and Ye, 2010), suggesting it is possible that the Rh proteins may transport something other than ammonia in Alcolapia. Because of these inconclusive reports around whether some amino acid changes could impact Rh protein structure/function in Alcolapia, we have directly tested the hypothesis that Alcolapia rhbg is structurally and functionally conserved as ammonium transporters.

A. grahami is reported to be fully ureotelic when assayed in its harsh native environment (Randall et al., 1989). We tested the hypothesis that this feature may be lost after several generations of lab breeding under temperate conditions and found that while lab bred A. alcalica living in the same aquarium excretes some of its waste as ammonia, A. grahami continues to be fully ureotelic. We use phylogenetics and gene expression analyses to show that the genes coding for Rh proteins in both species of Alcolapia are conserved and are expressed in the gills. Given the nearly identical amino acid sequence of the Rh proteins in the two recently diverged species, it would be expected that the observed difference in ammonium excretion would be a result of something other than distinct protein function. We tested the hypothesis that there could be a difference in the level of expression of the rhesus protein genes in these two very closely related species, and we used comparative expression analysis of gills dissected from fish living in the same environment and show a dramatic difference in the levels of Rh protein gene expression. We suggest that this differential expression of Rh protein genes provides a molecular mechanism whereby A. grahami does not excrete ammonia and A. alcalica does.

Phylogenetic analysis of the amino acid sequence of the Rh proteins confirm that Alcolapia Rh proteins are orthologous to those in other species included in the analysis, grouping with genes coding for Rhag, Rhbg, Rhcg1, and Rhcg2 (Fig. 1A). Analysis of gene synteny (Fig. 1B) in D. rerio and O. niloticus shows that the approximate position with respect to neighbouring genes and the orientation of rhesus protein genes Rhbg, Rhcg1 and Rhcg2 is conserved between O. niloticus and D. rerio. The data suggest that these genomic regions derive from the same ancestral genomic region and that the A. alcalica genes identified in this work are homologous to those previously studied in zebrafish (Nakada et al., 2007b; Shih et al., 2013), and points to Rhcg2a (rhcgl1) arising from a gene duplication in D. rerio. Sequence data for the A. alcalica rhesus protein genes has been submitted to NCBI [accession numbers: MW448158 (rhbg), MW448159 (rhcg1), MW448160 (rhcg2)].

Testing for variation in ratio of the rates of non-synonymous to synonymous substitutions (dN/dS) for the four Rh proteins among branches of the phylogeny demonstrates that while Rhag and Rhcg1 are under purifying selection, both Rhbg and Rhcg2 have evolved under positive selection in the lineage leading to Alcolapia (Fig. 1C and Table S1).​ This is shown by the comparison of Alcolapian Rhbg and Rhcg2 to the same genes from non-Alcolapian cichlids, as highlighted by the increased dN/dS ration for that comparison in Fig. 1C. The signal of positive selection is only present in the lineage leading to Alcolapia, and at no other points that were tested, such as the lineage leading to all Oreochromis, all African cichlids or all cichlids (Table S1). This suggests that the function of these proteins may be of adaptive significance in Alcolapia.

A computationally derived model of the A. alcalica Rhbg protein displays a 12 transmembrane helices structure with a long C-terminal tail; the tail is a feature known to be highly variable within the Amt/Mep/Rh family and has been posited to facilitate a protein–protein interaction in one member from Nitrosomonas europaea (Li et al., 2007; Lupo et al., 2007) (Fig. 2A). The model of A. alcalica Rhbg displays high confidence across most of its sequence, with the notable exception of the C-tail, which is to be expected given its divergent sequence across the Rh family. The 12 transmembrane helices and other key features of rhesus proteins are conserved, including the characteristic twin-His motif that is situated within the hydrophobic channel of Rhbg, and has been shown in AmtB to be critical for transport function and selectivity (Williamson et al., 2020) (Fig. 2B). Two conserved Phe residues reside just above the twin-His motif and act as a hydrophobic barrier which has been suggested to contribute to selectivity (Huang and Ye, 2010; Li et al., 2007). The extracellular side of A. alcalica Rhbg contains the highly negatively charged surface characteristic of the Rh family (Fig. 2C). In contrast, the intracellular side typically contains a higher prevalence of basic residues, although in A. alcalica the presence of acidic residues at the intracellular pore opening creates a mixed environment (Fig. 2D). The structural features predicted by this modelling are consistent with A. alcalica Rhbg acting as an ammonia transporter which we went on to test using two heterologous expression systems.

To determine whether A. alcalica rhesus proteins are able to move ammonium in vivo, synthetic mRNA coding for A. alcalica Rhbg was injected into zebrafish embryos and ammonia excretion was measured after 24 h. Injection of mRNA coding for Rhbg from the ammonotelic D. rerio was also analysed for comparison. Analysis of nitrogenous waste produced by injected and uninjected control embryos shows that overexpression of A. alcalica Rhbg significantly increases the amount ammonia detected in holding water, while the small increase resulting from overexpressing D. rerio Rhbg is not statistically significant (Fig. 3A). No significant difference was found in the amount of urea measured in holding water of any of the samples (Fig. 3B). These data suggest that the Alcolapia Rhbg protein is functional and, like Rhbg proteins in ammoniotelic species, has the capacity to transport ammonium.

To further examine the biological activity of Rhbg proteins from Alcolapia and D. rerio, expression plasmids were generated such that A. alcalica Rhbg and D. rerio Rhbg were constitutively expressed in a S. cerevisiae strain lacking its endogenous Mep ammonium transporters. In this assay, the Rhbg proteins are tested for their ability to complement the growth defect displayed by the mutant strain in a limited ammonium environment, which would indicate ammonium transport activity (Marini et al., 1997). Growth in minimal media supplemented with known concentrations of NH₄⁺ showed that at 5 mM ammonium all yeast grow (Fig. 3C inset), as expected due to passive diffusion of NH₃ at high concentrations. A further experiment using 20 mM NH₄⁺ showed all strains reaching the same final optical density by 40 h (data not shown), indicating each strain has the capacity for growth and is only restricted by its ability to take up ammonium. In a more challenging environment, where ammonium is reduced to 1 mM, only strains expressing an active ammonium transporter will be complemented and able to grow. Fig. 3C shows that at 1 mM NH₄⁺ the expression of Rhbg, from both A. alcalica and D. rerio, facilitated growth. These data further indicate that both of these proteins can transport ammonium.

To determine the percentage of nitrogen excreted as either ammonia or urea by A. alcalica, the amount of NH₄ and urea in the water holding either adult or embryonic A. alcalica was measured and compared to the same analysis using D. rerio (Fig. 4A and B). As expected, both adult and embryonic A. alcalica excreted much more of their nitrogenous waste as urea than D. rerio. Adult A. alcalica excreted 64% (n=6) of their nitrogenous waste as urea compared to 14% in adult D. rerio (n=14) (Fig. 4A). A. alcalica embryos excreted 79% of their nitrogenous waste as urea (n=7) compared to 18% in D. rerio embryos (n=5) (Fig. 4B).

Our finding that A. alcalica are not fully ureotelic as has been previously reported for A. grahami (Randall et al., 1989) is surprising given the very close genetic relatedness between the two species and that they both live in similar soda lakes with high pH. We speculated that this difference could be due to the A. alcalica being kept under more benign laboratory conditions for over two years, while the published urea excretion rates in A. grahami were derived from wild fish. To test this hypothesis, we assessed urea excretion in A. alcalica and A. grahami that had both been housed for multiple generations in the same system water (pH 8.1). We found that A. alcalica excrete 42-64% of their nitrogenous waste as urea, compared to 91-100% in A. grahami (Fig. 3C).

The almost identical amino acid sequences of the Rhesus glycoproteins from these two species that have only very recently diverged (Tichy and Seegers, 1999) indicates no differences in their structures that would point to any distinct biochemical activity. The interpretation is that A. grahami Rh glycoproteins could move ammonia but do not. Transcripts for the genes coding for Rhbg and Rhcg2 have been detected in A. grahami, as has protein expression in the gill filaments (Wood et al., 2013). However, the expression level of Rh glycoprotein gene transcription in A. grahami compared to fish known to move ammonia has not been investigated and the modulation of gene expression could underlie this difference in ammonia transport.

In situ hybridisation using DIG labelled cRNA probes for Rh protein genes was used to analyse gene expression in A. alcalica embryos at 5 dpf (Fig. 5). A. alcalica embryos show a restricted expression domain where Rhbg (Fig. 5A,B), rhcg1 (Fig. 5D,E), and rhcg2 (Fig. 5G,H) are detected only in gill tissue (arrows). Rhbg has the strongest expression in both developing embryos (Fig. 5A,B) and adult gill tissue (Fig. 5C). In adults, Rhbg is detected solely in the pavement cells of the gill filaments. Rhcg2 is found in the same region of the gill as Rhbg although expressed at a lower level (Fig. 5I). Rhcg1, also restricted to expression in gill tissue, is only detected at the base of gill filaments, presumably in the mitochondria rich cells (ionocytes) as described in zebrafish by Nakada et al. (2007a). We have also analysed the expression of the orthologous genes in zebrafish embryos and adult gills (Fig. S1) and conclude that the expression patterns of A. alcalica rhesus protein genes are well conserved with that of D. rerio embryos and gill tissue (see Table S2).

Gills were dissected from A. grahami and A. alcalica adults that had been housed in the same environmental conditions for several generations. In situ hybridisation was carried out using DIG-labelled antisense probes to detect mRNAs Rhbg, Rhcg1 and Rhcg2. The expression of these genes in A. alcalica was detected within 90 min after adding the substrate (BM Purple, Roche), while none were seen in A. grahami (Fig. S4). Fig. 6A-C shows robust expression of Rhbg (Fig. 6A), Rhcg1 (Fig. 6B), and Rhcg2 (Fig. 6C) in gills of A. alcalica detected after 16 h in substrate. Fig. 6D-F shows the corresponding low-level expression of Rhbg (Fig. 6D), Rhcg1 (Fig. 6E), and Rhcg2 (Fig. 6F) in gills of A. grahami analysed in parallel, alongside A. alcalica and photographed after the same period of time. For both species, expression is apparent in the same regions: Rhbg and Rhcg2 in the gill filaments, and Rhcg1 in the mitochondria rich cells; but the expression appears much weaker in A. grahami. To validate these observations using another method, we extracted RNA from the same samples and used RT-PCR to amplify each of the Rhesus glycoprotein genes. Fig. 6G shows the relative amount of expression of these genes in A. grahami as compared to A. alcalica. We conclude that A. grahami remains 100% ureotelic due to low expression of Rhbg, Rhcg1 and Rhcg2; while the high expression of these functional rhesus glycoprotein genes in A.alcalica allows it to excrete a portion of its waste as ammonia.

This work demonstrates that the extremophile fish A. alcalica can excrete a proportion of its nitrogenous waste as ammonia, while A. grahami does not. This observation is consistent with the robust expression of functional ammonium transporters, including Rhbg, in A. alcalica. We show that A. alcalica Rhbg (as well as Rhcg1 and Rhcg2) are expressed in the gill in both A. alcalica embryos and adults and that Alcolapia Rhbg protein is capable of moving ammonium in two heterologous expression systems, zebrafish embryos and a yeast growth assay. The Rhbg protein sequence is almost identical in these two species, suggesting that at the biochemical level A. grahami has the means/tools to move ammonia, but does not, even under temperate environmental conditions.

While we show that A. grahami expresses transcripts of Rhesus protein genes in the gill, and the expression of rhbg and rhcg2 proteins have previously been detected in A. grahami gill tissue (Wood et al., 2013), our comparison of gene expression levels reveals a large difference in expression levels in the two species. The much higher gene expression of rhbg, rhcg1 and rhcg2 in A. alcalica may also be reflected at the level of protein expression, but this analysis has yet to be undertaken. Rhesus protein genes in the gills of these two very closely related Alcolapia species that are both able to survive extreme environments of high temperature, salt and alkalinity.

Enhancer modularity is a known mechanism for selectable variation (Shapiro et al., 2004). While the protein coding regions of the Rhesus protein genes are almost identical between A. alcalica and A. grahami, their differential expression suggests divergence in a regulatory region that controls levels of transcription in the gills. The data we present here suggest that the comparatively low level of Rhesus protein gene expression in A. grahami explains their ureotelism. Even after generations in non-extreme conditions, there is no recovery of Rhbg/cg expression in A. grahami and they remain fully ureotelic. It is intriguing to speculate that there is a strong transcriptional silencer acting to repress these genes in A. grahami that is not present in the very closely related A. alcalica; more molecular genetic studies should be undertaken to inform further on this.

The evolution of obligatory ureotelism in A. grahami but not in A. alcalica may be an adaptive response to the more extreme conditions of Lake Magadi. Lake Natron (where A. alcalica live) is much larger than Lake Magadi (∼5× surface area) and has two inflowing rivers and several perennial streams that provide an influx of freshwater (Tebbs et al., 2013). In contrast, Lake Magadi (where A. grahami are found) lacks any inflowing streams. While the hydro-chemical conditions are similar between both lakes, Magadi is generally the more hostile of the two, with higher temperatures and salinity and lower flow rates recorded at most spring sites (Trewavas, 1982; Seegers and Tichy, 1999; Ford, 2015). This difference in native environment could provide an evolutionary explanation for the entrenched transcriptional downregulation of the rhesus protein genes in A. grahami.

Our in situ hybridisation experiments demonstrate that the expression of Rh genes in A. alcalica is largely restricted to adult and embryonic gills, with expression patterns similar to that in D. rerio (Nakada et al., 2007a; Zimmer and Perry, 2020) and other species such as rainbow trout (Tsui et al., 2009), mangrove killifish (Hung et al., 2007) and common carp (Sinha et al., 2016). Although the subcellular localisation would need to be assessed using antibodies specific for each of the Rh proteins. Nevertheless, these expression data point to Alcolapia Rhbg, rhcg1 and rhcg2 as orthologues to those of other fishes.

Like other members of the Rh family, and unlike the Amt family, no defined NH₄⁺ binding site is present in A. alcalica Rhbg. However, it does contain a conserved aspartate, which interacts with NH₄⁺ and likely facilitates translocation as it does in AmtB (Williamson et al., 2020); this residue is also known to be important for bidirectional ammonia transport in human RHAG and RHCG (Marini et al., 2000). The model of A. alcalica Rhbg presented in this report strongly supports a protein that functions as a typical Rh protein. It is surprising that when overexpressed in zebrafish embryos, Danio Rhbg did not significantly increase excreted ammonia while Alcolapia Rhbg did. It is possible that the positive selection of rhbg in Alcolapia species (Fig. 1) has resulted in a more active protein in some contexts, although this is not the case in our yeast assay.

Given the near identical sequence of Rhbg in A. alcalica and A. grahami (99.5%; see Fig. S4) it can be expected that A. grahami Rhbg also forms a functional ammonia transporter, aligning with previous work showing that Rh proteins facilitate ammonia excretion under high environmental ammonia (HEA) exposure based on the response of A. grahami Rhbg at both the mRNA and protein level in the gills under HEA stress (Wood et al., 2013).

An accepted model for Rhesus protein activity is that they facilitate ammonia transport by recruiting ammonium, NH₄⁺, for deprotonation before transporting it across the gill as ammonia, NH₃ (Khademi et al., 2004; Wright and Wood, 2012; Abdulnour-Nakhoul et al., 2016; Yeam et al., 2017). NH₃ is then protonated in the gill acid boundary layer and released into the surrounding water (Weihrauch et al., 2009). A recent study of AmtB from Escherichia coli, part of the large Amt/MEP/Rh family of ammonium transporters, has revealed further molecular detail to this mechanism in Amt proteins indicating that after ammonium deprotonation, the charged H⁺ and neutral NH₃ are transported separately across the membrane with the transporter providing the substrate selectivity by specifically recruiting ammonium (Williamson et al., 2020). Whether this precise mechanism also occurs in the Rh family is unknown, but a multiple sequence alignment of A. alcalica rhesus proteins and D. rerio Rhbg with E. coli AmtB shows conservation of key ammonium transport residues. The mechanism is particularly interesting to consider in Alcolapia fish given their native environment in Lakes Magadi and Natron in eastern Africa. The high pH and the large amounts of carbonates present in the water would severely limit the supply of protons proximal to the gill epithelium (Weihrauch et al., 2009). In most teleosts, the movement of ammonia relies on a concentration gradient, and as high external pH impedes this mechanism, the fish facing such extreme conditions require novel adaptations. Moreover, at external pH of ∼11, given the pK_a of ammonia/ammonia is 9.25, the uncharged ammonia molecule will dominate in solution, and combined with the high temperature of the lake, over 99% of the NH₃/NH₄⁺ will be found as NH₃, acting to significantly reduce the diffusion gradients across the membrane for spontaneous ammonia diffusion from the gills. Alcolapia have acquired the ability to efficiently convert ammonia to urea through adulthood and express a distinct carbamoyl-phosphate synthase (CPS), initiating the ornithine urea cycle (Lindley et al., 1999; White et al., 2020). This adaptation may be an example of convergent evolution of Alcolapia with terrestrial vertebrates where amino acids changed in the CPS substrate binding site allowing direct interaction with ammonia rather that glutamine, a feature unique in fish (White et al., 2020). Even with the constant movement of water past the gill, the extracellular environment is not permissive for ammonia diffusion alone to be sufficient for nitrogen excretion, and persistent urotelism is a unique adaptation in Alcolapia fish species.

The evidence presented here indicates that Alcolapia Rhbg protein is capable of transporting ammonia; this is consistent with our finding that A. alcalica excrete a proportion of their nitrogenous waste as ammonia. However, while A. grahami retain and express the genes coding for the rhesus protein genes, they do not excrete ammonia, even after generations in temperate aquarium conditions; this is likely underpinned by an A. grahami species-specific mechanism maintaining very low expression of the Rh protein genes.

In York, a stand-alone, recirculating aquarium (Aquatic Habitats) was adapted to house A. alcalica in 10 or 30 L tanks at a constant temperature of 30°C, pH between 9.0 and 9.5 (buffered with a 40X stock of 2 M sodium bicarbonate and 2 M sodium carbonate) and salt concentration of 3800 µS (using Instant Ocean and magnesium sulphate). Fish were bred and fully acclimated to laboratory conditions for more than 2 years. A separate zebrafish system was maintained at 27°C, pH 7.4, and conductivity 800 µS. In Hull, A. grahami and A. alcalica bred for 3-4 generations in aquaria were kept in 120 L species stock tanks on the same recirculatory system, maintained at 26-29°C, pH 8.1. This study was approved by the Animal Welfare and Ethical Review Body at the Universities of York and Hull, and the UK Home Office project licence to M.E.P. (POF245295).

To understand the evolutionary relationships among Rh proteins, a phylogeny was constructed in MEGA X (Kumar et al., 2018) (LG method with gamma distribution) using an amino acid alignment of published sequence of the four Rh proteins from Alcolapia and seven other cichlid species, with D. rerio and Takifugu rubripes as outgroup taxa (Fig. 1A). Gene synteny analysis was conducted using the NCBI genome viewer on the genomes of D. rerio and the close relative of Alcolapia, O. niloticus (Ford et al., 2019), to further support that the genes discussed here are true homologues. A fully assembled genome is not available for Alcolapia species hence the use of O. niloticus for this analysis.

To test for a signal of positive selection acting on Rh proteins in the Alcolapia lineage, the nucleotide alignments sequences were analysed with the codeml package within PAML (Yang, 2007). Where available, sequences from the following taxa were used: oreochromiine cichlids (Alcolapia grahami, Alcolapia alcalica, Oreochromis niloticus), non-oreochromiine African cichlids (Maylandia zebra, Astatotilapia calliptera, Pundamila nyererei, Astatotilapia burtoni, Simochromis diagramma, Neolamprologus brichardi), New world cichlids (Archocentrus centrarchus), outgroups (Seriola lalandi, Toxotes jaculatrix). The tree topology specified was the currently best accepted hypothesis of phylogenetic relatedness amongst the chosen taxa (Chen et al., 2004; Wagner et al., 2012; Cao and Xia, 2016). The branch model was used to test for variation in the dN/dS ratio (ω) among branches of the phylogeny. Several two ω models were fitted, including separate ω for Alcolapia, all oreochromiine, all African cichlids, and all cichlids (Table S1). Chi-square tests were used to assess all models with two ω compared to a model with a single ω across all taxa.

To generate plasmids for probes to use for in situ hybridisation analysis, and for synthesising full-length mRNAs, we used reverse transcriptase PCR to clone A. alcalica rhesus protein genes. RNA was extracted from 5-day-old A. alcalica embryos using TriReagent (Sigma-Aldrich) according to the manufacturer's guidelines. For cDNA synthesis, 1 µg of total RNA was reverse transcribed with random hexamers (Thermo Fisher Scientific) and superscript IV (Invitrogen). A. alcalica cDNA was amplified using gene specific primers (Table S2) and cloned into pGEM T-Easy (Promega) for in situ probes using run-off transcription was used to incorporate a DIG labelled UTP analogue into an antisense cRNA probe. Full-length cDNAs coding for Rhbg proteins were cloned into the expression plasmid pCS2+ and sequenced. mRNA for injections was synthesised using messageMACHINE® SP6 kit as per manufacturer's instructions.

Since A. alcalica Rhbg appears to have evolved under positive selection, we used AlphaFold to predict the extent to which whether the amino acid changes may have altered its structure and its ability to transport ammonia. The AlphaFold model of A. alcalica Rhbg was calculated using the protein sequence of Rhbg from A. alcalica (accession number MW448158) on the AlphaFold Google Colab platform (Jumper et al., 2021). The Alphafold program provides a confidence score prediction for each residue of the structure within a numerical range of 0-100, called predicted local-distance difference test (pLDDT). Values >90 indicate high confidence in the predicted structure, including side chain orientation; 70-80 indicates less confidence suggesting only the backbone is modelled well; and values <50 denotes possible regions of disorder in which no confidence in the structural prediction can be given.

Differences in expression patterns of Rhesus protein genes were visualised using in situ hybridisation. A. alcalica embryos were collected at 5 days post fertilisation (between 15 and 20 embryos) and gills were dissected from adult A. alcalica and A. grahami. All specimens were fixed for 1 h in MEMFA (0.1 M MOPS pH 7.4, 2 mM EGTA, 1 mM MgSO₄, 3.7% formaldehyde) at room temperature and stored at −20°C in 100% methanol. In situ hybridisation was carried out according to (Harland, 1991), and briefly summarised here: embryos were rehydrated and treated with 10 μg/mg proteinase K at room temperature. After post-fixation and pre-hybridisation, embryos were hybridised with DIG probes (for rhbg, rhcg1 and rhcg2) at 68°C in a 50% formamide hybridisation buffer in 6X SSC. After extensive washing, hybridisation was detected using an antibody against DIG conjugated to alkaline phosphatase and visualised with the substrate BM purple (Roche).

Differences in expression levels of Rhesus protein genes were assessed via semi-quantitative RT-PCR. Total RNA extracted from adult gills of A. alcalica and A. grahami using TriReagent (Sigma-Aldrich) according to the manufacturer's guidelines followed by purification on an RNA clean and concentrator column (Zymo Research). cDNA was synthesised from 1 µg of total RNA using random hexamers (Thermo Fisher Scientific) and superscript IV (Invitrogen). Gene specific primers (Table S2) were designed to amplify Alcolapia Rhbg, Rhcg1, Rhcg2, and eF1ɑ. The nucleotides sequences in these regions are identical in A. alcalica and A. grahami so the same primers and PCR conditions were used (22 cycles). PCR product resolved on an agarose gel were analysed using Image-J to normalise intensity to that of eF1ɑ.

mRNA coding for rhesus proteins were expressed in zebrafish embryos to test whether they were capable of moving ammonium in ammoniotelic fishes. For this, early cleavage stage D. rerio embryos (AB strain) were used, and 1 nl of 150 ng µl⁻¹ mRNA was injected into the yolk cell just under the early blastomeres of 1-4 cell embryos using a glass needle and a Harvard apparatus gas micro-injector. To test the activity of Alcolapia Rhbg in moving ammonia we included uninjected embryos and embryos where Rhbg had been overexpressed by injection of synthetic mRNA from either Danio or A. alcalica. For these overexpression analyses, 30 zebrafish embryos were cultured in 2 ml tank water for 24 h after injection, after which the chorions were opened with forceps before the surrounding water was collected for analysis. This was to ensure all excretory products were collected including those which had not diffused across the chorion. In all cases, two water samples were taken from each group and frozen at −20°C. The experiment was repeated thrice with embryos from different breedings. The concentration of nitrogen as ammonia and urea for the three treatments (uninjected, Danio Rhbg, Alcolapia Rhbg) were measured and compared using two-sample t-tests.

In order to test the ability of the A. alcalica rhesus protein genes to move ammonium in yeast, the Rhbg genes from A. alcalica and D. rerio were cloned into the vector pDR195 by homologous recombination and transformed into Saccharomyces cerevisiae 31019b (MATα ura3 mep1Δ mep2Δ::Leu2 mep3Δ::KanMX2), a yeast mutant with deletions in the three endogenous ammonium-transporter genes and defective in ammonium transport. The empty vector pRS316 was transformed into S. cerevisiae strains 31019b and 23344c (Matα ura3) as controls in subsequent growth assays. The pDR195 plasmid was kindly gifted by Arnaud Javelle.

Saccharomyces cerevisiae strains transformed with the indicated plasmids were grown overnight in synthetic complete media lacking uracil (Formedium). Cells were resuspended in a minimal buffered (pH 6.1) medium with 3% glucose as the carbon source (Jacobs et al., 1980) and used to inoculate fresh minimal buffered media containing a range of concentrations of (NH₄)₂SO₄ as the sole nitrogen source in a Nunc 96-well plate. Growth assays were performed at 30°C for 40 h in an Epoch 2 microplate reader, with growth quantified by optical density every 30 min (Abs 600 nm).

The concentrations of ammonia and urea excreted into surrounding water was measured for adult and embryonic A. alcalica and D. rerio and adult A. grahami. For adults: single adult A. grahami, A. alcalica and D. rerio were incubated in 600 ml of water from their respective tanks for 3 h, after which each fish was individually weighed before being returned to their home systems. Two water samples were taken from each experiment (n=4) and stored at −20°C until processing. For embryos, 5-day-old embryos were grouped (five embryos for A. alcalica and 30 embryos for D. rerio) in 30 mm petri dishes in 2 ml of tank water for 6 h on a rocker at 28°C, after which water was collected and stored at −20°C until processing.

To calculate the concentration of ammonia in the water samples, 400 µl of sample was combined with 100 µl of reaction buffer (40 gl⁻¹ sodium tetraborate, 40 mgl⁻¹ sodium sulphite, and 50 mll⁻¹ of phthalaldehyde dissolved in ethanol). Samples were incubated for 3 h in the dark alongside calibration samples (known ammonia concentrations of 0 to 100 µM). After incubation, 300 µl of each sample were transferred into a 96-well plate and the absorbance read at 350 nm on an Infinite m200 Pro (Tecan). This protocol is adapted from (Holmes et al., 1999).

To calculate the urea concentrations, 350 µl of water samples was combined with 25 µl of DAMO-TSC [8.5 g of diacetylmonoxomine to 240 ml of dH₂O and 10 ml thiosemicarbazide solution (0.95 g in 100 ml) and 80 µl of acid-ferric solution (30 ml of sulphuric acid in 23.5 ml of dH₂O and 50 µl of ferric chloride 0.15 g in 10 ml)]. Reagents were all made fresh on the day of analysis. Samples were incubated at 85°C for 20 min. Calibration samples of known concentration of urea (0-100 µM) were analysed in the same way, this was multiplied by two to represent the amount of nitrogen present (CH₄N₂O). 300 µl of each sample was moved into a 96-well plate and the absorbance read at 585 nm. This protocol is adapted from (Mulvenna and Savidge, 1992). Concentrations of urea and ammonia in samples were determined using the equation of the straight line produced from the respective calibration curves. To control for differences in the size of specimens or number of embryos, measurements of ammonia and urea were calculated as a percentage of total nitrogen. Data for adult waste was calculated as a rate per gram per hour (µM N g⁻¹ h⁻¹).

The fish were collected from Lakes Natron and Magadi by L.J.W. in collaboration with the Tanzania Fishery Research Institute (TAFIRI) (COSTECH permit no. 2017-259- NA-2011-182). Animal work was approved by the Animal Welfare and Ethical Review Body at the Universities of York and Hull, and the UK Home Office project licence to M.E.P. (POF245295).