The effects of CO₂ and external buffering on ammonia excretion and Rhesus glycoprotein mRNA expression in rainbow trout

Rhesus (Rh) proteins are most commonly known for their role in erythrocyte antigenicity (Avent et al.,2006). Several non-erythrocytic Rh proteins have now been identified in a variety of species ranging from bacteria to mammals (e.g. Chain et al., 2003; Huang and Peng, 2005; Weihrauch, 2006). Structural and functional studies have revealed that Rh proteins are transporters of ammonia, a property shared with AmtB, their Escherichia colihomologue (for reviews, see Planelles,2007; Van Kim et al.,2006; Weiner and Hamm,2007). There is some conflict as to whether the primary substrate is NH₃ or NH₄⁺, but most evidence points to NH₃ (Javelle et al.,2007).

A number of different Rh orthologues have been characterized in fish(Huang and Peng, 2005; Hung et al., 2007; Nakada et al., 2007a; Nakada et al., 2007b; Nawata et al., 2007). While Rhag appears to be limited to erythrocytes in mammals(Huang and Peng, 2005), it occurs in both erythrocytes and pillar cells in the pufferfish Takifugu rubripes (Nakada et al.,2007b). Additionally, Nakada and colleagues showed in pufferfish that while Rhbg and Rhcg2 are situated basolaterally and apically,respectively, in the pavement cells, Rhcg1 is located apically in mitochondria-rich cells (Nakada et al.,2007b). Exactly how these Rh proteins function in ammonia transport in fish is not yet clear; however, Rh mRNA expression levels do change in response to high external ammonia in some fish species(Hung et al., 2007; Nawata et al., 2007), but not all (Nakada et al.,2007a).

The bulk of studies implicate Rh proteins in NH₃ transport,although there remains the possibility that they also facilitate transfer of CO₂, another readily hydrated gas of similar size(Endeward et al., 2006; Kustu and Inwood, 2006). Soupene and colleagues reported that the Rh1 protein of the green alga Chlamydomonas reinhardtii was upregulated by high CO₂ and that the absence of Rh1 impaired the growth of this alga under high CO₂, a condition that normally promotes rapid growth(Soupene et al., 2002; Soupene et al., 2004). Recently it was demonstrated, using ¹⁸O and mass spectrometry, that both NH₃ and CO₂ pass through the RhAG protein of human erythrocytes (Endeward et al.,2006). These two substrates compete for entrance and passage through the RhAG channel, although the affinity appears to be greater for NH₃ than CO₂. Furthermore, structural studies performed on the Rh1 protein of the ammonia-oxidizing bacterium Nitrosomonas europaea revealed that it may have a much lower ammonium affinity than the Amt proteins (Lupo et al.,2007) and a possible CO₂ binding site at the COOH-terminal end has been described (Li et al., 2007).

A few studies suggest, however, that CO₂ is not transported through the Rh channel. Ripoche and colleagues, for example, found no difference in the permeability to CO₂ in human erythrocyte membranes lacking RhAG, using a stopped-flow method(Ripoche et al., 2006). More recently it was shown that both wild-type(Weidinger et al., 2007) and Rh1 knockout mutants(Cherif-Zahar et al., 2007) of N. europaea were unresponsive to CO₂.

Like ammonia, CO₂ in fish is believed to pass through membranes mainly in its gaseous form (Henry and Heming, 1998), while the transport of CO₂ occurs in the plasma in its hydrated form, HCO ^–₃(Perry and Gilmour, 2002). A small amount of CO₂ exits the apical gill epithelium as HCO ^–₃ in exchange for Cl^– after being hydrated in the gill by carbonic anhydrase and thus serves a role in acid–base regulation (Claiborne et al., 2002; Perry,1986; Perry and Gilmour,2006).

In the present study, our goal was to determine whether exposure to high external CO₂ levels could elicit changes in Rh mRNA expression in rainbow trout gill, skin and erythrocytes, which would indicate a possible dual role for the Rh glycoproteins as NH₃ and CO₂channels. However, studying the effects of high CO₂ on Rh mRNA expression in fish is problematic in that hypercapnia causes an increase in endogenous ammonia production (Claiborne and Heisler, 1986). Increased ammonia excretion rates have also been associated with exposure to external hypercapnia(Claiborne and Heisler, 1986)and this may be due to the favourable plasma to water ammonia gradient created by the lowered water pH as a result of the hydration of CO₂ in the water (Claiborne and Heisler,1986; Larsen and Jensen,1997; Lloyd and Herbert,1960; Wright et al.,1989). Buffering minimizes acidification of the water during hypercapnia; however, this introduces the confounding effect of abolishing the acidified boundary layer at the gill that normally facilitates ammonia excretion (Salama et al.,1999; Wilson et al.,1994; Wright et al.,1986; Wright et al.,1989).

We therefore examined the effect of three experimental conditions on the mRNA expression of Rh glycoproteins in trout: hypercapnia alone, hypercapnia with Hepes buffering, and normocapnia with Hepes buffering. The hypercapnia treatment chosen (1% CO₂ for 12 h) directly duplicated a previous study in our lab in identical water quality where internal acid–base and ion status (both extracellular and intracellular) were monitored(Wood and LeMoigne, 1991). Additionally we analysed the ammonia excretion rates, plasma cortisol levels and the mRNA expression of other gill transporters possibly involved in ammonia excretion in each experimental condition. Finally, we isolated three full-length Rhbg2 cDNA variants and assessed the mRNA expression of these in the gill and skin.

Rainbow trout (Oncorhynchus mykiss, Walbaum), weighing 170–240 g, were obtained from Humber Springs Trout Hatchery, Ontario,Canada, and held in dechlorinated Hamilton tap water (moderately hard:[Na⁺]=0.6 mequiv. l^–1, [Cl^–]=0.8 mequiv. l^–1, [Ca²⁺]=0.8 mequiv. l^–1, [Mg²⁺]=0.3 mequiv. l^–1,[K⁺]=0.05 mequiv. l^–1; titration alkalinity 2.1 mequiv. l^–1; pH ∼8.0; hardness ∼140 mg l^–1 as CaCO₃ equivalents; temperature 12–16°C) and fed ad libitum until 1 week before experimentation, during which time food was withheld. Fish were transferred to individual, opaque boxes supplied with aerated, flowing dechlorinated tap water (15±0.5°C) and allowed to recover overnight. All procedures used were approved by the McMaster University Animal Research Ethics Board and are in accordance with the Guidelines of the Canadian Council on Animal Care.

Water flow was closed off to each box and the box water volume was set to 4 l. Fish were exposed for 12 h to a 1% CO₂ in air mixture(hypercapnia), provided by a Wösthoff gas-mixing pump (Calibrated Instruments, Ardsley, NY, USA), bubbled into the water. Control fish were treated identically with a 12 h exposure to 100% air (normocapnia).

Bubbling of 1% CO₂ in air into the external water reduced the pH by approximately 1.22 pH units, a similar reduction to that reported earlier by other investigators (Larsen and Jensen,1997; Wright et al.,1988a). In order to minimize acidification of the water while fish were exposed to hypercapnia, Hepes (Sigma, St Louis, MO, USA) was added to the external water to a nominal concentration of 10 mmol l^–1using a 1 mol l^–1 stock solution adjusted to pH 8.0 with KOH. This maintained the pH within the range of 7.76–7.39 during the course of 12 h. To examine the effects of Hepes buffering alone on ammonia excretion and mRNA expression, another set of fish were exposed to normocapnia for 12 h in the presence of 10 mmol l^–1 Hepes, as above.

In order to check some particular parameters highlighted in the results of the hypercapnia and Hepes-buffering exposures, a high environmental ammonia(HEA) exposure protocol identical to that used by Nawata and colleagues(Nawata et al., 2007) was repeated using an ammonia concentration similar to that used by previous investigators (Cameron, 1986; Cameron and Heisler, 1983; Claiborne and Evans, 1988; Wilson and Taylor, 1992; Wilson et al., 1994). Trout were exposed to a nominal total waterborne ammonia (T_Amm)concentration of 1.5 mmol l^–1 NH₄HCO₃(pH 7.95±0.05) for 12 h. During this period the water flow was stopped and the box volume was set to 4 l. Water samples were removed every 3 h and total water ammonia concentration remained virtually constant at 1.41±0.03 mmol l^–1 over the 12 h period.

Water samples (10 ml) were removed at 3 h intervals, stored at–20°C and later analysed in triplicate for total ammonia(T_Amm) using a modified salicylate-hypochlorite method(Verdouw et al., 1978). Because Hepes altered colour generation in this assay, additional ammonia standards containing 10 mmol l^–1 Hepes were prepared and analysed in conjunction with the samples from the Hepes experiments. At the end of each experiment, fish were anaesthetized with 0.1 g l^–1 MS222 (Sigma) and caudal blood samples were collected into a heparinized syringe. Blood cells were separated from plasma by centrifugation and both blood fractions were snap frozen in liquid nitrogen and stored at –70°C.

Net flux rates of total ammonia (J_Amm; μmol kg^–1 h^–1) into the water were calculated as: J_Amm=(T_Amm,i–T_Amm,f)×V/(t×M)where i and f refer to the initial and final concentration (μmol l^–1), V is the box water volume (l), t is the time elapsed (h) and M is the fish mass (kg). A negative J_Amm indicates net excretion into the water. Plasma total ammonia was measured enzymatically (Raichem, Hemagen Diagnostics, San Diego,CA, USA) and reported in μmol l^–1.

External water pH was monitored with a Radiometer GK2401C (Copenhagen,Denmark) low ionic strength combination electrode thermostatically set to the experimental water temperature.

Prior to tissue extraction, fish were perfused free of blood using Cortland saline (Wolf, 1963) as previously described (Nawata et al.,2007). Samples of gill, skin and blood were removed, snap frozen in liquid nitrogen and stored at –70°C until later processing. Gill samples from control fish were used for cloning of Rhbg2 and additional samples of brain, gill, intestine, kidney, liver, muscle and skin were removed for screening of Rhbg2 mRNA expression.

Total RNA was extracted from blood and tissues using Trizol (Invitrogen,Burlington, ON, Canada) and concentrations were quantified and checked for quality with a Nanodrop ND-1000 spectrophotometer (Nanodrop Technologies,Wilmington, DE, USA) and electrophoresed on 1% agarose gels stained with ethidium bromide, to verify integrity. First strand cDNA synthesis was performed using 1μg of DNaseI-treated (Invitrogen) total RNA with oligo(dT₁₇) primer and Superscript II reverse transcriptase(Invitrogen).

Full-length cDNA of Rhbg2 was obtained by 5′- and 3′-rapid amplification of cDNA ends (Smart RACE cDNA amplification kit,BD Bioscience Clontech, Mississauga, ON, Canada) using the RACE primers listed in Table 1 and the protocol described previously (Nawata et al.,2007). Sequence analyses were performed with BioEdit(Hall, 1999) and CLUSTAL W(Thompson et al., 1994). Hydropathy profile and N-linked glycosylation site predictions were made using SPLIT 4.0 (Juretic et al.,2002), and ScanProsite (de Castro et al., 2006), respectively.

Tissue distribution of Rhbg2 was determined by performing reverse transcriptase PCR (PCR) on the above-mentioned cDNA with the primer set listed in Table 1 at 35 cycles for Rhbg2 and 25 cycles for elongation factor EF-1α. Products were electrophoresed on 1.5% agarose gels stained with ethidium bromide and sequenced to confirm identity.

Quantitative real-time PCR (qPCR) was performed on the cDNA described above using the primers listed in Table 1. Rh mRNA expression was assessed in the gill, skin and erythrocytes and the expression of carbonic anhydrase (CA2;cytoplasmic), H⁺-ATPase (v-type, B-subunit),Na⁺/H⁺ exchanger (NHE) and Na⁺/K⁺-ATPase (NKA; α1a-subunit) was analysed in the gill. Each 20 μl reaction contained 4 μl of cDNA, 8 pmol of each primer and 10 μl of RT² Real-Time SYBR Green/ROX PCR Master Mix (SuperArray, Bioscience Corp., Frederick, MD, USA). Analyses were performed at 95°C for 2 min, followed by 40 cycles of 95°C for 30 s and 60°C for 1 min. Melt-curve analysis and gel electrophoresis verified the presence of a single product. No-template controls and non-reverse-transcribed controls were run in parallel. Data were extrapolated from standard curves generated by serial dilution of one randomly selected control sample. Three housekeeping genes encoding β-actin, EF-1αand 18S rRNA proved to be unstable across treatments and data were therefore normalized to ng total RNA, another acceptable method of normalization(Bustin, 2000; Bustin, 2002; Nolan et al., 2006), as further elaborated in the Discussion.

Plasma cortisol levels were measured in duplicate on 25 μl samples by radioimmunoassay (Cortisol ¹²⁵I RIA Kit, MP Biomedicals,Orangeburg, NY, USA) and values are reported in ng ml^–1.

Data are presented as means ± s.e.m. (N, number of fish). One-way analysis of variance (ANOVA) followed by Fisher's least significant difference post-hoc test was used to analyse both the effects of the experimental treatments and the relative abundance of Rhbg2 mRNA in the skin. One-way repeated measures ANOVA was used to analyse ammonia excretion rates. Significance was set at α=0.05.

Three Rhbg2 cDNA sequences were identified in this study (GenBank accession nos: EU660221, EU660222 and EU660223). Two of these, designated Rhbg2a and Rhbg2b, have identical open reading frames (ORFs)encoding a protein 461 amino acid residues long. Analysis showed a 98% amino acid sequence identity with the ORF of Rhbg1(Fig. 1A). Although these Rhbg2 variants are identical, Rhbg2b has a 414 base-pair deletion in the 3′-untranslated region (UTR). The ORF of the third cDNA,designated Rhbg2c, is a truncated variant of the full-length ORF,with 166 amino acids deleted from the COOH-terminal end. The last 18 amino acids form an alternative end (Fig. 1A). Rhbg2b and Rhbg2c additionally have four different 5′-UTR variants (herein termed variants 1, 2, 3 and 4, in order of decreasing length) with an upstream ORF (uORF) that would encode a polypeptide of six amino acids. The longest 5′-UTR variant (variant 1)has an additional uORF upstream to the common uORF that would encode a polypeptide of 26 amino acids (Fig. 1B). Rhbg2a has three of the four 5′-UTR variants described; the longest one was not detected by RT-PCR analysis(Fig. 1C). Hydrophobicity analysis revealed 12 predicted transmembrane domains and two putative N-linked glycosylation sites for Rhbg2a and Rhbg2b (Asn48 and Asn212). The truncated variant, Rhbg2c retains the same predicted N-linked glycosylation sites as the full-length variants, but has only eight predicted transmembrane domains(Fig. 1A).

PCR analysis of the various trout tissues from control fish revealed that Rhbg2 is expressed in the brain, gill, intestine, skin and possibly in the liver and muscle (Fig. 2). The truncated form, Rhbg2c was only detected in the gill (data not shown). qPCR analysis performed on control samples of gill and skin showed that the 5′-UTR variants were present in equal abundance in the gill. In the skin, however, variants 1 and 3 were significantly more abundant than the other two. To examine whether or not these variants respond differentially, we analysed the expression in the skin of fish exposed to hypercapnia in Hepes-buffered water. Expression of all four variants decreased such that they were no longer differentially expressed(Fig. 3).

Exposure to external hypercapnia resulted in ammonia excretion rates that were elevated significantly over both the 3 and 12 h control rates during the course of the exposure. The rates at 3 and 12 h were 69% and 54% higher than the control rates, respectively (Fig. 4). Plasma ammonia levels in these fish at 12 h were not significantly different from the control values(Fig. 5). There were no significant changes in Rh glycoprotein mRNA expression in the gill or skin,although Rhcg2 mRNA was decreased by almost 50% (Figs 6 and 7). Similarly, no significant differences in Rhag and Rh30-like2 mRNA expression were noted in the erythrocytes; however, Rhag was increased by about 1.7-fold (Fig. 8). Expression of CA2, H⁺-ATPase, NHE2 and NKA mRNA in the gill was unchanged (Fig. 9). The plasma cortisol concentration was not significantly different from that of the control fish(Fig. 10).

Throughout the 12 h exposure to hypercapnia in Hepes-buffered water, trout exhibited ammonia excretion rates that were significantly lower than those in fish exposed to hypercapnia alone, with a 57% reduction at 3 h and a 42%reduction at 12 h. At 9 h, however, the rates were significantly increased above the 3 h control rates and by 12 h they were not different from the 12 h control rates (Fig. 4). Plasma ammonia levels were significantly increased 2.3-fold above the control value after 12 h of exposure (Fig. 5). In the gill, CA2 and NHE mRNA levels were significantly increased (Fig. 9) and Rhcg2 mRNA was a significantly decreased by 60%(Fig. 6). No significant changes were noted in the mRNA expression of Rhag or Rh30-like2 in the erythrocytes, although expression of Rhagwas 2-fold higher than the control (Fig. 8). In the skin, there was a 67% reduction in Rhbg2 but a 30-fold increase in Rhcg2. Rhcg1 was decreased, but not significantly(Fig. 7). Plasma cortisol was not significantly different from the control value at 12 h(Fig. 10).

The ammonia excretion rate in fish exposed to normocapnia in Hepes-buffered water was not significantly different from the 3 h control rate, throughout the 12 h exposure. However, there was a trend towards decreased excretion and by 12 h the rate was reduced to 71% of the 12 h control value(Fig. 4). Plasma ammonia was significantly elevated over the control value at 12 h(Fig. 5). Rhcg2 mRNA levels were significantly reduced in the gill to about half of the control level (Fig. 6), but no changes were noted in the other gill mRNAs (CA2, H⁺-ATPase, NHE2,NKA; Fig. 9). In the skin,while there was a 2-fold reduction in Rhbg1 and a 4-fold reduction in Rhbg2 mRNA, Rhcg1 and Rhcg2 mRNA expression was increased by 4- and 58-fold, respectively(Fig. 7). In the erythrocytes,there was a 3-fold reduction in both Rhag and Rh30-like2mRNA expression (Fig. 8) and plasma cortisol levels were low, but not significantly different from control values (Fig. 10).

Because of the high nucleotide sequence identity between the Rhbg1and the Rhbg2 cDNAs (the highest being 91.5%, between the Rhbg1a and Rhbg2a variants), we designed a primer set more specific for Rhbg1 than that used previously by Nawata and colleagues(Nawata et al., 2007), and assessed the expression of Rhbg1 and Rhbg2 in the gill and skin of HEA-exposed fish. Both Rhbg1 and Rhbg2 mRNA expression levels remained unchanged in the gill during HEA(Fig. 6). In the skin, Rhbg1 mRNA levels were not significantly different from that of the control, but Rhbg2 mRNA expression was significantly downregulated by about 4-fold (Fig. 7). Expression of Rhcg1 and Rhcg2 mRNA in the gill and skin, Rhag and Rh30-like2 mRNA in the erythrocytes, and CA2,H⁺-ATPase, NHE2 and NKA mRNA in the gill was also measured and confirmed the results reported earlier by Nawata and colleagues (Nawata et al.,2007) and are therefore not reported here. Plasma ammonia and cortisol levels were both significantly elevated above the control values by 10- and 4-fold, respectively (Figs 5 and 10).

In the present study, quantitative expression data for mRNA by real-time PCR were normalized to total RNA rather than to reference control genes. Expression levels of endogenous reference genes such as β-actin or elongation factor (EF-1α) are frequently used as internal controls to normalize expression levels of target genes in real-time PCR studies(Guilietti et al., 2001; Nolan et al.,2006), as in our previous studies on Rh proteins(Hung et al., 2007; Nawata et al., 2007). Several recent reports, however, have shown that reference genes themselves may be modulated by experimental conditions, which in turn may result in inaccurate and misleading interpretations of data(Gibbs et al., 2003; Huggett et al., 2005; Ings and Van der Kraak, 2006; Kubista et al., 2006; Tricarico et al., 2002). Indeed, our initial experiments revealed that the expression levels of three reference genes (encoding β-actin, EF-1α and 18S rRNA) were elevated in tissues after exposure to Hepes. Currently, no universal method of normalization exists that accounts for all possible variables; however, an accepted method commonly used (e.g. Luqueti et al., 2005; Serrano et al., 2007; Tipsmark et al., 2008; Vera Cruz et al., 2006) is normalization against total RNA (Bustin,2000; Bustin, 2002; Huggett et al., 2005; Nolan et al., 2006). Therefore, to eliminate any artificial bias in the data due to reference gene variability, we chose to normalize quantitative expression data to total RNA. In this regard it is important to note that our approach is validated by the target gene responses to HEA normalized against total RNA reported in the present study, which were qualitatively and quantitatively similar to those reported earlier by Nawata and colleagues(Nawata et al., 2007) in parallel but separate HEA experiments on trout where β-actin or EF-1α was used for normalization.

Recently, we established a link between Rh proteins and ammonia by showing that rainbow trout exposed to high environmental ammonia (HEA) responded with upregulated Rhcg2 mRNA expression in the gill(Nawata et al., 2007). This response coincided with the re-establishment and enhancement of ammonia excretion that was initially inhibited by the reversed plasma to water ammonia gradient. In the present study, we used hypercapnia as a tool to assess the response of Rh mRNA transcription levels to high CO₂ in the rainbow trout. This proved problematic in that, like the study by Claiborne and Heisler (Claiborne and Heisler,1986), hypercapnia caused an increase in internal ammonia production as evidenced by the elevated rate of ammonia excretion(Fig. 4) without alteration of plasma total ammonia concentration (Fig. 5). Bubbling of CO₂ into the water lowered the pH,creating a sink for NH₃, thus enhancing the rate of excretion. This increased excretion rate was sufficient to maintain plasma ammonia at control levels (Fig. 5) and may explain why changes in Rh mRNA levels were not observed. Any effects that elevated CO₂ may have had were not reflected as changes in mRNA transcription levels in the gill or skin.

Buffering of the external water effectively reduced the acidification caused by the CO₂, but it also probably abolished the acidified gill boundary layer. It has been well documented that the boundary layer,whether acidified by hydration of CO₂ by carbonic anhydrase in the mucus (Wright et al., 1986; Wright et al., 1989) or by release of protons from an apical H⁺-ATPase(Lin et al., 1994; Lin and Randall, 1990),normally facilitates ammonia excretion. Indeed, the present study confirms previous reports (Salama et al.,1999; Wright et al.,1989) showing that there was a decline in the ammonia excretion rate when fish were exposed to normocapnia in Hepes-buffered water, with the rate reduced to a third of the control rate by 12 h(Fig. 4). As a consequence of the reduced excretion, plasma ammonia levels in these fish were elevated after 12 h of treatment (Fig. 5). Despite this elevation, and unlike HEA-exposed fish, which also have high plasma ammonia (Nawata et al.,2007), the Rhcg2 mRNA transcription level was decreased in the gill (Fig. 6). Upregulated or enhanced ammonia transport capacity at the gill would probably prove futile in the presence of Hepes since any protons formed or released at the apical gill surface would be rapidly removed by the buffer, effectively eliminating a favourable ammonia gradient.

However, in the skin of fish exposed to Hepes in the presence of normocapnia, both Rhcg1 and Rhcg2 mRNA levels were highly elevated (Fig. 7). Rhcg2 exhibited the greatest fold increase in mRNA expression, suggesting that this protein may have a dominant role in the skin, but this requires further investigation because only mRNA level and not protein function was measured. Upregulation of Rhcg2 mRNA expression in the skin also occurred when excretion was blocked at the gill after 12 h of exposure to HEA(Nawata et al., 2007). Fish skin is generally thought to have low permeability(Fromm, 1968) and therefore its role in gas and ion exchange has been largely ignored. A few studies have measured ammonia excretion from fish skin (for a review, see Wood, 1993), but it appears that the skin has a secondary role to the gill, if any, in freshwater fish. This may be true especially under normal circumstances, but under conditions in which excretion is blocked at the gill, the skin may become an important alternative site of ammonia release and this possibility needs to be explored further in freshwater species. Indeed the skin becomes a site of ammonia excretion via volatilization during aerial exposure in the mangrove killifish Kryptolebias marmoratus, and skin Rh mRNA levels are also upregulated during this time (Hung et al.,2007).

The acid–base disturbance in response to hypercapnia has been well described in catheterized fish (for reviews, see Claiborne, 1998; Heisler, 1993). Catheterization was avoided in the present study, as our initial trials indicated that catheterization alone may alter Rh expression (C.M.N. and C.M.W., unpublished data). However, we used an exposure regime in which the acid–base effects had been documented previously. In an identical 1%CO₂ exposure protocol in Hamilton tap water, chronically catheterized rainbow trout exhibited a 0.3 unit depression in arterial pH at 3 h with partial recovery by 12 h [see figure 7 of Wood and LeMoigne (Wood and LeMoigne, 1991)]. Although gill cell intracellular ion levels(Na⁺ and Cl^–) declined, gill intracellular pH did not change.

The associated increase in acid excretion which helps correct internal pH by building up internal HCO ^–₃ levels during hypercapnia involves the hydration of CO₂ by carbonic anhydrase within the gill (Henry and Heming, 1988; Perry, 1986; Perry and Gilmour, 2006). Apically located H⁺-ATPase and/or NHE may then be involved in the release of protons into the water (Edwards et al., 2005). Although buffering during hypercapnia would cause an internal elevation of both CO₂ and ammonia, the results from our study suggest that fish in this treatment group were responding to an acidosis induced by the high CO₂. The upregulation of NHE2 mRNA levels in the gill corresponds well with the findings of Edwards and colleagues, who reported an increase in NHE2-like protein expression in Fundulus heteroclitus exposed to external hypercapnia(Edwards et al., 2005). This suggests that, although H⁺-ATPase is believed to play the major role in acid secretion in freshwater teleosts(Edwards et al., 2005), NHE2 may also have a role during acidosis in rainbow trout, especially in Hamilton tap water where Na⁺ levels are relatively high (700 μmol l^–1). H⁺-ATPase mRNA expression levels were not upregulated at 12 h (Fig. 9), but this does not necessarily mean that the corresponding protein is non-functional during acidosis. The expression levels of H⁺-ATPase mRNA may in fact peak as early as 2 h after hypercapnia exposure (Perry et al.,2000), a response that could then result in a sufficient quantity of protein by 12 h. The increased CA2 mRNA levels that occurred are also in agreement with previous studies that reported increased carbonic anhydrase activity (Dimberg and Hoglund,1987) and increased CA2 mRNA and protein expression(Georgalis et al., 2006) in the gills of hypercapnia-exposed trout.

Fish exposed to hypercapnia in Hepes-buffered water also experienced inhibition of ammonia excretion at the gill as evidenced by the elevated plasma ammonia (Fig. 5) as well as excretion rates that were significantly lower than those of the hypercapnia-exposed fish in unbuffered water(Fig. 4). Although the excretion rates started to rise at 9 h(Fig. 4), the downregulation of Rhcg2 and H⁺-ATPase mRNA levels in the gill indicate that these fish were probably not using the same mechanism to enhance ammonia excretion as that used during HEA. Both treatments resulted in blockage of ammonia excretion from the gill, but it appears that in hypercapnic fish, the primary mRNA response in the gill was to CO₂ rather than high ammonia. Similar to the fish exposed to normocapnia in Hepes-buffered water,there was an upregulation of Rhcg2 mRNA in the skin(Fig. 7), again reinforcing the idea that this may be an alternative route for ammonia excretion.

Previously we reported that Rhag as well as the Rh30-like mRNAs were downregulated during HEA, when plasma ammonia levels were high(Nawata et al., 2007). In the present study, both groups of fish exposed to Hepes-buffered water had elevated plasma ammonia levels, albeit lower than HEA-exposed fish(Fig. 5), so we would have expected a similar downregulation of Rhag expression in these two groups. Fish exposed to normocapnia did indeed exhibit lowered RhagmRNA levels; however, the hypercapnia-exposed fish did not. Instead, the Rhag mRNA expression in the hypercapnia-exposed fish was increased 2-fold over that of the control fish, a response similar to that of the hypercapnia-exposed fish in unbuffered water(Fig. 8). This suggests that Rhag was responding to both high CO₂ and high ammonia, but in a differential fashion.

Ammonia accumulates in trout erythrocytes during hypercapnia(Wright et al., 1988b). If Rhag allows passage of NH₃, then downregulation when plasma ammonia levels are elevated may be a protective response. Transport of NH₃with CO₂ and H₂O into the erythrocyte would produce NH₄⁺ and HCO₃^– and once HCO₃^– is exchanged for Cl^–, the formation of NH₄Cl would lead to H₂O uptake and result in swelling (Bruce, 2008). Also, the buffering effect of NH₃ could interfere with the Bohr and Haldane effects, and therefore with O₂ and CO₂transport. In human erythrocyte membranes, RhAG forms a macrocomplex with a number of proteins including the band-3 anion exchanger (AE1) and carbonic anhydrase, and this complex is thought to function in CO₂/O₂ gas exchange(Bruce et al., 2003). Trout erythrocytes also have AE1 present (Michel and Rudloff, 1989), but whether or not it forms a similar complex with Rhag is unknown. However, if Rhag functions as a CO₂ channel,then upregulation along with AE1 could be beneficial during hypercapnia when plasma HCO ^–₃ is high.

It has also been speculated that the Rh30 proteins could allow CO₂ passage (Callebaut et al.,2006), although Endeward and colleagues found no difference in CO₂ uptake in erythrocytes lacking the RhD (Rh30) protein (Endeward et al., 2007). Data from our study also suggest that the Rh30-like proteins are not involved in CO₂ transport as there were no notable changes in mRNA expression in erythrocytes from either of the two groups exposed to hypercapnia (Fig. 8). Rh30-like2 mRNA expression, however, was downregulated in the fish exposed to normocapnia and Hepes, like HEA-exposed fish(Nawata et al., 2007),suggesting a response to high ammonia only.

So far, studies have indicated that fish have multiple Rh genes(Huang and Peng, 2005; Hung et al., 2007; Nakada et al., 2007a; Nakada et al., 2007b; Nawata et al., 2007) but the significance of this seeming redundancy is unclear. Nakada and colleagues(Nakada et al., 2007b)proposed that Rhag in the pillar cells of pufferfish may work in conjunction with the basolaterally located Rhbg and apically situated Rhcg2 in the lamellae to keep plasma ammonia levels low, with additional excretion aided by Rhcg1 in the mitochondria-rich cells. Others have shown that some Rh proteins may be dispensable like Rhbg in the mouse(Chambrey et al., 2005) and Rh1 in the slime mould (Benghezal et al.,2001). Likewise, the physiological significance of the multiple variants of Rhbg2 (Fig. 1) is puzzling, but the presence of uORFs and splicing in the 5′- and 3′-UTRs of the mRNAs suggests complex regulation of this protein. uORFs have the potential to impact gene expression and some may serve as cis-acting regulatory elements modulating translation of the main ORF (Meijer and Thomas, 2002; Morris and Geballe, 2000). Similarly, the UTRs are involved in many post-transcriptional pathways that control the localization, stability and translation efficiency of mRNAs(Pesole et al., 2001). The existence of a truncated version of the Rhbg2 protein suggests yet another level of complexity. Truncation alters function and/or intracellular location of some transporters while some alternative-splice products negatively regulate the wild-type protein (Kitayama et al., 1999; Mangravite et al., 2003). Whether the truncated version of Rhbg2 functionally interacts with the full form or whether it has a completely different function has yet to be analysed.

Interestingly, the same 5′-UTR Rhbg2 variants are present in both the skin and gill, although in the gill they are expressed equally while in the skin they are expressed differentially(Fig. 3). The downregulation of Rhbg2 mRNA expression that occurred in the skin with Hepes treatment during both hypercapnia and normocapnia appears to be in response to high plasma ammonia rather than high CO₂ as suggested by the unchanged expression in the hypercapnia-treated fish in unbuffered water. To support this idea, we found that HEA-exposed fish exhibited a similar downregulation(Fig. 7). Rhbg was reported to be non-essential during acidosis in knockout mice(Chambrey et al., 2005). An abundance of Rhbg2 in fish skin may similarly be unnecessary during hypercapnia and HEA, and instead Rhbg2 could have a role other than ammonia transport.

Glucocorticoids are known for their role in regulating gene expression in mammals and the same is likely to be true for fish(Mommsen et al., 1999). Data from this study, however, showed no correlation between Rh mRNA abundance and cortisol levels. Plasma cortisol and gill Rhcg2 mRNA levels were both elevated during HEA but in the two Hepes-exposed groups Rhcg2 mRNA was elevated in the skin while the cortisol remained at control levels(Fig. 10). A lack of correlation, however, does not exclude the possibility that cortisol could be involved either directly or indirectly in the control of Rh protein expression. Indeed, cortisol levels need not be elevated in order to have an influence on physiological processes (e.g. Wood et al., 2001) and other factors such as the clearance rate and non-genomic actions of cortisol need to be taken into consideration (Mommsen et al., 1999).

We have evaluated the expression of Rh protein mRNAs in rainbow trout exposed to hypercapnic and normocapnic conditions in the presence and absence of Hepes buffer. Previously, we showed that at the mRNA level, Rh proteins clearly respond to high ammonia conditions in rainbow trout tissues(Nawata et al., 2007). A similar response to high CO₂ would suggest that Rh proteins may have a dual role as CO₂ and NH₃ gas channels in trout tissues. Data from our present study revealed, however, that high CO₂ did not directly elicit changes in Rh mRNA transcription levels in the gill and skin. Instead, the changes that did occur probably reflected responses to high plasma ammonia, and thus reinforce the connection between the Rh proteins and ammonia. As such, a dual role for Rh proteins in the rainbow trout is not apparent from this study. There does remain, however, the possibility that Rhag interacts with both CO₂ and ammonia in the erythrocyte and further work is needed to clarify this.

We thank Drs P. A. Wright, J. M. Wilson, C. Y. C. Hung and K. N. T. Tsui for very useful advice and two anonymous referees for constructive comments. Supported by an NSERC (Canada) Discovery Grant to C.M.W., who is supported by the Canada Research Chair Program.