Multiplicity of expression of Na⁺,K⁺–ATPaseα-subunit isoforms in the gill of Atlantic salmon (Salmo salar): cellular localisation and absolute quantification in response to salinity change

The branchial mechanisms and cellular targets responsible for ion-regulatory homeostasis in teleosts have been subject to extensive research since the pioneering investigations of Homer-Smith, Krogh and Keys in the early 20th century (Evans,2008). Nonetheless, important issues still need to be clarified regarding the molecular details of ion transport. Among the various cell-types in the gill, the mitochondrion-rich so-called chloride cells (MRCs) in conjunction with their neighbouring cells (accessory- and pavement cells) are of principal importance in ion-uptake and ion-excretion. Chloride-secreting cells were first discovered and named by Keys and Wilmer(Keys and Wilmer, 1932), and were assigned an important role in salt secretion in a number of marine teleosts. Subsequently, MRCs have been identified by different techniques based on their ultrastructural and biochemical properties (for reviews, see Evans et al., 2005; Hwang and Lee, 2007). One common way of visualising MRCs exploits the dense abundance of the Na⁺,K⁺–ATPase enzyme in the basolateral membrane networks of these cells. Thus, immunostaining with generic anti-Na⁺,K⁺–ATPase antibodies is a common procedure to identify these cells. MRCs have been classified into several subtypes, based on ultrastructure (α- and β-type), localisation within the gill (in lamellae, at base of lamellae, interlamellar in filament,etc.), dynamics in response to salinity change (recruitment to/from lamellae,hyperplasia in filament) or expression of additional ion-transport proteins(V-type H⁺–ATPase, Cl^–/HCO ^–₃ exchanger, CFTR-chloride channel,Na⁺,K⁺2Cl^– cotransporter) (for reviews,see Evans et al., 2005; Hwang and Lee, 2007). There is consensus that in euryhaline teleost species, ion transporting cells may appear in both filament and lamellar positions, in some cases depending on salinity and other stressors, and also with clear species specific differences(Tipsmark et al., 2007; Uchida et al., 1996; Varsamos et al., 2002). At least in salmonids, it has been convincingly demonstrated that ionocytes appear in both locations in freshwater (FW) whereas in seawater (SW) most lamellar MRCs regress and the filament MRCs increase in number and particularly in size (Seidelin et al.,2000; Uchida et al.,1996). This led these authors to propose that lamellar cells are involved in ion-uptake (FW–MRC) and filament cells may have a dual function in both ion-secretion (SW–MRC) and ion-uptake.

The active and passive steps in gill ion-translocation have been unravelled from studies involving both stenohaline and euryhaline teleosts of FW as well as SW origin. At both salinities, net ion-transfer is based on the activity of Na⁺,K⁺–ATPase as the primary energy-consuming step, even though in certain species this may be supplemented by the active transfer of protons by V-type H⁺–ATPase(Evans et al., 2005). The level of complexity has recently increased by the discovery of three different isoforms of the α₁-subunit in gill tissue of Oncorhynchus mykiss (Richards et al.,2003): α_1a, α_1b andα _1c. These are present in addition to an α₃isoform, which has also been described in a number of other species[Oreochromus mossambicus (Lee et al., 1998); Danio rerio(Rajarao et al., 2001); Trematomus spp (Guynn et al.,2002)]. The relative abundances of some of these isoforms in whole-gill RNA extracts have been investigated during salinity acclimation in a few studies [O. mykiss, Salvelinus alpinus, Salmo salar(Richards et al., 2003; Bystrianski et al., 2006;Bystrianski et al., 2007); O. nerka(Shrimpton et al., 2005); Fundulus heteroclitus (Scott et al., 2004)]. Based on analyses of relative mRNA levels by real-time quantitative polymerase chain reaction (RT-QPCR), the general picture emerging from these studies is that salinity induces a reciprocal regulation (switching) – primarily of the α_1a andα _1b isoforms, α_1a generally being upregulated in FW and α_1b being upregulated in SW. This led Richards et al. to suggest that the α_1a is a FW-isoform driving in ion-uptake whereas α_1b is a SW-isoform driving ion-secretion (Richards et al.,2003). Accordingly, the α_1b is upregulated during salmon smoltification in FW – a process preparing the juvenile fish for movement into the marine environment(Nilsen et al., 2007). In support of this, Jorgensen recently showed that a few particular amino acid differences between the two isoforms may critically influence cation binding properties, and reduce the Na⁺:ATP ratio from 3:1 inα _1b to 2:1 in the α_1a isoform(Jorgensen, 2008). In O. mykiss and S. salar, the presence of a lysine residue instead of asparagine at site 783 in transmembrane segment 5 of the α_1aisoform thus leads to this isoform being energetically better suited for Na⁺ transport against extreme electrochemical gradients, as is the case for ion uptake in FW.

Only relative measures of mRNA abundance have been reported so far, and nothing is known neither about the quantitative relationship between the four isoforms nor of their cellular localisation in the gill. Based on relative measures, attempts have been made to calculate changes in the total abundance of α-subunit mRNA during salinity changes(Bystrianski et al., 2006;Bystrianski et al., 2007); however, such an arithmetic exercise requires absolute quantification and is invalid when based on relative measures. The purpose of the present study was to investigate the cellular localisation of the α-subunit isoform mRNAs in the Atlantic salmon gill by in situ hybridisation (ISH) and to clarify the time course and the exact quantitative interrelationship of the four α-subunit isoforms in the gill when subject to salinity change. The β₁-subunit, which is needed in a 1:1 ratio with the α₁-subunit for functional maturation and membrane targeting (Scheiner-Bobis, 2002) was also quantified to give insight into the stoichiometric relationships between the two subunits.

One-year old Atlantic salmon (Salmo salar L., Ätran strain obtained from Danish Center for Wild Salmon, Randers, Denmark) were kept in outdoor tanks with flowing Odense city tap water (in mmol l^–1: 1.4 Cl^–, 1.5 SO ^2–₄, 1.5 Na⁺, 0.16 K⁺, 3 Ca²⁺, 0.6 Mg²⁺; 5.5 total CO₂ content; pH 8.3) at environmental temperatures and exposed to natural daylight through smoltification. After smoltification in August 2006, these fish were transferred to indoor conditions in 500-l fibre-glass tanks at 14°C at 12 h:12 h light:dark cycle and were considered to be post-smolts. In October 2006, subgroups were either transferred to 28 p.p.t. artificial, aerated,re-circulated and filtered SW (Red Sea Salt, Eliat, Israel) or sham transferred to FW. Eight fish from each salinity were sampled 12 h, 24 h, 3 days and 7 days after transfer together with eight fish at 0 h as a control group. Fish were stunned with a blow to the head and blood collected from the caudal vessel. After cutting the spinal chord and pithing of the brain, a second gill arch was immediately dry frozen in liquid nitrogen for RNA purification and the other second gill arch was frozen in sucrose–EDTA–imidazol buffer (SEI: 300 mmol l^–1sucrose, 5 mmol l^–1 Na₂EDTA, 50 mmol l^–1 imidazol, pH 7.3) for analysis of Na⁺,K⁺–ATPase enzymatic activity(McCormick, 1993). Experimental protocols were approved by the Danish Animal Experiments Inspectorate being in accordance with the European convention for the protection of vertebrate animals used for experiments and other scientific purposes (#86/609/EØF).

Total RNA was purified using GenElute™ Mammalian Total RNA kit (Sigma Chemical Co., St Louis, MO, USA) according to the manufacturer's recommendations. RNA concentration and purity was determined by measuring A260/A280 on undiluted samples on a Nanodrop spectrophotometer (Nanodrop Technologies, Wilmington, DE, USA). One μg RNA was DNase treated (Promega,Madison, WI, USA) and then reverse transcribed using 2 μg oligo dT primers(GE Healthcare Bio-Sciences, Little Chalfont, UK) and 200 units MMLV reverse transcriptase (Invitrogen, Carlsbad, CA, USA) for 1 h at 37°C in the presence of 40 units of RNAguard (GE Healthcare) in a total volume of 25μl. At the end the cDNA was diluted to 50 μl with milliQ H₂O. Real-time PCR analysis using SYBR Green detection was performed on a Mx3000-P PCR machine (Stratagene, La Jolla, CA, USA) using standard software settings including adaptive baseline for background detection, moving average and amplification based threshold settings with the built-in FAM(5-carboxyfluorescein) filter (excitation wavelength, 492 nm; emission wavelength, 516 nm). Reactions were carried out with 1 μl cDNA, 150 nmol l^–1 forward and reverse primer(Table 1) and 1×SYBR® Green JumpStart™ Taq ReadyMix™ (Sigma Chemical Co.) in a total volume of 25 μl. Cycling conditions were: 95°C for 120 s followed by 40 cycles of 95°C for 30 s and 60°C for 60 s. Melting curve analysis was performed following each reaction to confirm the presence of only a single product in the reaction. Negative control reactions using DNase treated total RNA from representative samples were used to analyse carry over of genomic DNA. In all samples and with all primers there was negligible genomic contamination.

The primers used to amplify the Na⁺,K⁺–ATPaseα _1b, and α_1c isoforms were the same as used by Richards et al. (Richards et al.,2003). These were designed for O. mykiss and have been successfully used for S. salar by Bystrianski et al.(Bystrianski et al., 2006). The sequences of the primers used for the α_1a isoform in O. mykiss (Richards et al.,2003) and S. salar (Bystrianski et al., 2007) in previous studies were identical to the α_1b sequence in the primer region (mismatch 1/18 and 2/18 for forward and reverse primers, respectively)and in our hands, were found not to be specific for the α_1aisoform (not shown). Therefore, we designed specific primers for theα _1a near the 3′terminus of the complete mRNA sequence(for O. mykiss) in order to obtain maximum isoform specificity. For the α₃ isoform we designed primers based on a S. salar EST sequence (acc. no. CX355357), which is 99% identical to the O. mykiss α₃ isoform. For theβ ₁-subunit, primers were designed based on a partial S. salar β₁-sequence (acc. no. AJ250810). Primers were designed using the NetPrimer software (Premier Biosoft International, CA, USA)with standard settings and double checked using the Primer3 software(Rozen and Skaletsky, 2000)and BLASTed. All primers were tested for non-specific product amplification and primer–dimer formation using both melting curve analysis and agarose gel verification. All primers were synthesised by DNA Technology A/S (Aarhus,Denmark).

In order to convert threshold cycle (Ct)-values from a given PCR reaction into absolute quantities of the corresponding transcript, it was necessary to generate a standard curve for each of the five transcripts of interest(α_1a, α_1b, α_1c,α ₃ and β₁). Therefore, a specific single stranded polynucleotide (approximately 100 bases long)(Table 1) based on the antisense chain of each α-isoform gene sequence was synthesised(Invitrogen), thus defining the amplicon of each primer pair. The amount of specific polynucleotides used per PCR reaction to generate the standard curves ranged from 0.01 amol to 1 fmol. This generated Ct-values in the range 10–28 depending on the amplicon. For each standard curve, the regression line was then used for calculations of cDNA quantity in the unknown samples. Ct-values of the unknown samples were always within the range of the actual standard curve. Amplification efficiencies of the five PCR reactions were in the range of 92 to 103% both when using polynucleotides or actual cDNAs amplified from tissue RNAs as template.

In a separate experiment using long term (>3 weeks) FW- and SW-acclimated S. salar, blocks of gill tissue (2 mm wide) were quickly removed from euthanised fish. After trimming away the cartilage rod, a 2–3 mm gill block was placed in a small alufoil mould, embedded in Optimal Cutting Temperature Compound (Miles, Eikhart, IN, USA) and quick frozen on dry ice. The specimens were stored desiccated at –80°C until further processing. Sections (8–10μm) of the frozen specimens were then cut on a cryostat at –14°C, transferred to Superfrost Plus glass slides (Erie Scientific Company, Portsmouth, NH, USA) and dried for 1 h at 50°C. The sections were then transferred to –80°C until further processing. Upon use, the sections were thawed at room temperature,soaked in 96% ethanol for 1 h and dried at 37°C for 30 min. They were incubated overnight under cover glass with 3 pmol ml^–1 of alkaline phosphatase (AP) labelled cDNA probes(Table 2) in hybridisation buffer at 37°C in a sealed chamber with 100% humidity. Depending on the base composition and theoretical melting point of the DNA–RNA duplexes(T_m) hybridisation conditions were varied (see Table 2): 30–50%formamide, 1.5–4× saline sodium citrate buffer (SSC: 165 mmol l^–1 NaCl, 3.3 mmol l^–1 sodium citrate),1× Denhardt's solution (0.2 mg ml^–1 Ficoll, 0.2 mg ml^–1 polyvinyl pyrrolidon, 0.2 mg ml^–1bovine serum albumin), 10% dextran sulphate, 10 μgml^–1single-stranded salmon sperm DNA. The hybridisation temperature was always 37°C, which is 20–25°C lower than the theoretical T_mof all probes in solution and thus near the optimal temperature for hybridisation in the tissue (Augood et al.,1993; Tecott et al.,1987). After incubation, the cover glass was carefully removed and the sections rinsed in three 30 min changes of 3× SSC (22.5 mmol l^–1 NaCl, 2.3 mmol l^–1 sodium citrate,pH8.0) at 37°C followed by two 10 min rinses in post-hybridisation buffer(0.01 mol Tris-HCl, 0.15 mol NaCl, pH9.5) at room temperature. For colour development, the specimens were incubated in freshly made AP developer buffer(0.32 mg ml^–1 nitro blue tetrazolium (NBT; Sigma Chemical Co.), 0.17 mg ml^–1 5-bromo-4-chloro-3-indolyl phosphate(BCIP; Sigma Chemical Co.), 0.1 mol Tris-HCl, 0.1 mol NaCl and 0.05 mol MgCl₂, pH9.5) for 1 to 3 days in the dark at room temperature until an appropriate colour deposition was observed. Finally, the specimens were rinsed in distilled water (30°C, 30 min) to stop the colorimetric reaction. Cover slips were mounted using Aquatex (Merck, Darmstadt, Germany). The above methodology is based on the protocol described by Lambertsen et al.(Lambertsen et al., 2001).

AP-labelled 28-mer cDNA probes were designed using the Oligo 6 software(Molecular Biology Insights, West Cascade, CO, USA) with standard settings(Table 2). Probes were designed in order to have minimal hairpin and dimer formation, and all sequences were BLASTed. Due to the high similarity of the α-isoform sequences, each probe sequence was aligned against the other isoforms. Identity generally ranged between 10/28 and 16/28 bases.

Two series' of control hybridisation experiments were done in order to validate the specificity of the 28-mer cDNA probes. In one series, the specificity of the α_1a and α_1b probes were checked by co-incubating gill sections with the AP-conjugated cDNA probe (3 pmol ml^–1) for a particular target (α_1avs α_1b) with 10-fold excess of the matching cDNA sequence of the alternate isoform in the probe region (see Table 3). In both cases, the identity between the AP-labelled probe and the cDNA sequence of the alternate isoform was 14/28. In another series, the ability of the unlabelled probe sequence to displace the AP-conjugated probe was checked by co-incubating tissue with the probe (3 pmol ml^–1) and 5-fold excess of the unlabelled cDNA sequence.

All data was analyzed using two-way analysis of variance (ANOVA) followed by Bonferroni adjusted Fisher's Least Significant Difference (LSD) test,taking into account the total number of paired comparisons. When necessary,data was log-transformed to obtain normality and homogeneity of variances. In all cases, a significance level of α<0.05 was used. All tests were performed using SAS (v. 9.1 for Windows, by SAS Institute, Cary, NC, USA).

Plasma sodium concentration (Fig. 1A) was elevated 12 h–3 days after SW transfer compared with corresponding FW groups. There was an overall effect of SW and a significant interaction between time and salinity. Gill Na⁺,K⁺–ATPase activity(Fig. 1B) varied insignificantly in the FW groups and was significantly elevated in the SW group at seven days post-transfer.

The α_1a transcript level was in the range 0.7–3.9 amol 20 ng^–1 total RNA(Fig. 2A). α_1aTranscript levels were affected by both time and salinity and there was a significant interaction between the two factors, with SW-values being lower than FW-values at 3 and 7 days after transfer. The α_1btranscript level was in the range 0.5–5.3 amol 20 ng^–1total RNA (Fig. 2B).α _1b Transcript levels were affected by both time and salinity and there was a significant interaction between the two factors, with SW-values being higher than FW-values at 1, 3 and 7 days after transfer. Theα _1c transcript level was in the range 0.3–0.4 amol 20 ng^–1 total RNA (Fig. 2C). Overall, it was affected by time but not salinity and with a significant interaction between the two factors. The α₃transcript level was in the range 0.02–0.05 amol 20 ng^–1 total RNA (Fig. 2D) and was affected by salinity and time. The level was generally lower in SW than in FW. The sum of α-transcripts was in the range 3–7 amol 20 ng^–1 total RNA(Fig. 2E) and was affected by both time and salinity with a significant interaction. The sum increased after SW-transfer being significantly higher at 3 and 7 days compared with the FW-level. The ratio of α_1a:α_1b was in the range 4–5 in FW and decreased to a lower value of 0.26 after SW-transfer(Fig. 2F). The ratio was significantly affected by both time and salinity and with a significant interaction.

The β₁ transcript level was in the range 0.07–0.3 amol 20 ng^–1 total RNA(Fig. 3). β₁Transcript levels were affected by time and salinity and there was a significant interaction between the two factors. The level was generally higher in SW than in FW, being significant at 12 h and 7 days.

The relative distributions of α-subunit isoforms in FW and after SW-transfer are shown in Fig. 4. In FW, the contribution of the four subunits wasα _1a, 68–74%; α_1b, 15–23%;α _1c, 7–13%; α₃, 0.7–1.2%. After SW-transfer, a major shift in α_1a and α_1bwas observed in favour of α_1b being the dominating isoform. Seven days post-transfer the relative proportions were: α_1a,19%; α_1b, 75%. No major changes occurred in the relative contribution of α_1c and α₃, their shares being 4% and 1.7%, respectively, on day 7 in SW.

The α_1a isoform was primarily located in cells in the interlamellar region at the base of the lamellae in both FW and SW gills(Fig. 5A,B). These cells were more numerous in FW than in SW gill sections(Fig. 5A–C,vsFig. 5D,E). Cells on the secondary lamellae also stained positive with the α_1a probe in FW (Fig. 5A-B) but this labelling disappeared in SW gills (Fig. 5D). The contour of the FW-cells was larger and they were often in contact with the surface of the epithelium(Fig. 5B) compared with SW-cells, which appeared smaller and deeper in the epithelium(Fig. 5D).

In FW, the α_1b staining occurred in cells almost exclusively in the interlamellar region(Fig. 6A). The cells appeared small and deep in the epithelium, without contact with the apical surface of the epithelium. Rarely, labelling also appeared in cells on the lamellae. In SW-acclimated fish, there was a much higher density of positively reacting cells – again primarily in the interlamellar region of the filament epithelium (Fig. 6B-D). These cells were larger than those observed in FW and generally were in contact with the apical surface. In SW, there was no labelling of cells on the lamellae(Fig. 6B,C). Application of theα _1c or the α₃ probe gave no clear staining of gill cells either in FW or SW (not shown). The use of the β-actin probe gave a strong and almost uniform staining of all cell types in the gill(Fig. 6E).

Simultaneous incubation of gill sections with a specific cDNA probe(α_1a or α_1b) and the corresponding unlabelled cDNA sequence of the alternate isoform (α_1b orα _1a) in 10-fold excess, gave no changes in staining pattern or staining intensity as shown in Fig. 7A–D. Thus, the probes were highly isoform specific. However, simultaneous incubation of gill sections with a specific cDNA probe(α_1a or α_1b) and the corresponding unlabelled cDNA sequence of the same isoform in 5-fold excess, completely displaced the hybridisation signal (Fig. 7E,F) (shown for α_1b only), showing that the hybridisation signal is target specific.

Based on the present quantitative estimates of α-subunit mRNA in whole gill RNA extracts and a few simple assumptions, we can make a rough calculation of the fraction of the mRNA pool in chloride cells made up ofα-subunit isoforms. Consider the following situation: the mRNA pool of gill tissue makes up 1–3% of the total RNA extract, thus in 20 ng of total RNA (1 QPCR reaction) the mRNA amounts to 0.2–0.6 ng. Assuming 100% efficiency in the first strand cDNA synthesis reaction this is all converted 1:1 into cDNA. The sum of α-subunit isoform transcript is presently found to be around 5 amol 20 ng^–1 total RNA. The mean transcript size (of α_1a and α_1bisoforms) is 3350 bases (based on full length sequences in O. mykiss). With an approximate nucleotide molar weight of 307 g mol^–1, the mean molar weight of α-subunit mRNA is 1×10⁶ g mol^–1. Thus, the mean weight of 5 amol α transcript is 5.2 pg. This amount equals 0.8–2.5% of the estimated mRNA pool in the gill extract. From the ISH analyses,α _1a and α_1b isoforms are predominantly located in presumed chloride cells in the gill. The relative fraction of mitochondrion-rich chloride cells have been estimated to be in the order of<10% of the total cell number in the gill (Goss et al., 2004). Assuming the mRNA content per cell is the same – irrespective of cell type – it can be estimated that the amount of α-subunit mRNA within the chloride cells makes up 8–25% of the total mRNA pool. Even though connected with some uncertainty, this gives a rough estimate of the total abundance ofα-subunit transcripts in the chloride cells.

This is the first attempt to localise the α-isoforms of Na⁺,K⁺–ATPase in the gill of any fish species and to quantify all α- and β-isoforms in absolute terms at the mRNA level in response to a salinity change. Two α-subunit isoforms(α_1a and α_1b) were predominant in the gill irrespective of salinity and were present in almost equimolar amounts. They both varied in response to salinity whereas the additionalα _1c and α₃ isoforms were expressed at lower levels – especially the α₃ isoform, which was present at negligible levels and showed little variation. Based on these measures, the sum of α-subunit transcripts is in the range of 5 amol α-subunit transcript per 20 ng total gill RNA, which we estimated to correspond to 8–25% of the mRNA pool within chloride cells. Even though some uncertainty is connected with this estimate, this is a very high fraction of total mRNA but matches well with the high abundance of Na⁺,K⁺–ATPase pumps estimated to reach 200 million pumps per chloride cell (Karnaky,1980).

The primers we used to quantify these two major transcripts were designed in order to obtain maximum isoform specificity. The α_1bprimer pair was the same as used by Richards et al.(Richards et al., 2003) in O. mykiss, by Bystrianski et al.(Bystrianski et al., 2006;Bystrianski et al., 2007) in S. alpinus and S. salar, and by Shrimpton et al. (Shrimpton et al.,2005) in O. nerka. These primers were designed near the 3′-terminus, where there is a region of maximum heterology between theα-isoforms. In all of the above studies, however, theα _1a-primers were designed in the middle of the transcript in a region with very high identity between the isoforms. These primers had only one (forward) and two (reverse) mismatches compared with theα _1b sequence, and when tested against specific cDNA sequences of the two isoforms, they annealed and amplified the two isoforms with almost equal efficiency (not shown). Thus, in the former studies there is a risk that the α_1a primers may have amplified both α_1aand α_1b transcript. In order to avoid this, we designed theα _1a-primers in the 3′-region in order to obtain maximum isoform specificity (zero and 10 mismatches in forward and reverse primers,respectively) and when tested against the synthetic polynucleotide sequences of the two isoforms, they amplified only the α_1a isoform, as expected.

Our present study confirms previous reports of a reciprocal change of the two major α-isoforms in response to salinity change. When the salinity was increased from FW to SW, α_1a decreased andα _1b increased within a 1–3 day time course, the sum of these two showing a doubling. This pattern is in accordance with Richards et al. [(Richards et al., 2003) O. mykiss], Mackie et al.[(Mackie et al., 2005) S. salar] and Bystrianski et al.[Bystrianski et al., 2006) S. alpinus], who speculated that this pattern of regulation may reflect major functional differences of these isoforms with respect to their putative roles in ion uptake (α_1a) and ion secretion(α_1b), respectively. In maturing O. nerka, Shrimpton et al. (Shrimpton et al.,2005) found upregulation of α_1a upon FW entry;however, α_1b was unchanged after FW-encounter and upregulated in fish on the spawning grounds, suggesting a non-specific response to stress in moribund fish.

There were already marked changes in the transcript levels on day 1 increasing until day 3 after SW-transfer. This is probably initiated by the osmotic stress apparent at 12 h and peaking on day 1, and mediated by osmoregulatory hormones, of which cortisol is a likely candidate(McCormick, 2001). Remarkably,Na⁺,K⁺–ATPase hydrolytic activity, which is a measure of total protein abundance, was not elevated until day 7. An overall increase in Na⁺,K⁺–ATPase activity is seen in response to hyperosmotic conditions in many euryhaline fish(Marshall, 2002), and a lag period of several days is not unusual(Bystrianski et al., 2006; Madsen et al., 1995). It has been argued that this reflects the time for protein synthesis in poikilothermic fish (Conte and Lin,1967), however, we propose that de novo synthesis of`secretory'-type Na<sup>+</sup>,K<sup>+</sup>–ATPase may be accomplished prior to that. Because the enzymatic assay does not separate between`FW-absorptive-type' and `SW-secretory-type' enzyme in the gill, the gross capacity is merely a static measure of a highly dynamic enzymatic pool reflecting the difference between protein being degraded and synthesised. Thus, new `SW-secretory-type' (α_1b)Na⁺,K⁺–ATPase may be synthesised well before day 7 without being observed in measures of catalytic capacity, which also explains why plasma [Na⁺] is stabilised and already regulated after <3 days. The presence and reciprocal regulation of two dominating isoforms may be specific to salmonids, as in killifish, the α_1a isoform increases in both hyper- and hyposaline environments compared with brackish water (Scott et al., 2004)whereas the α_1b isoform remains unchanged.

The α_1c variant of the α₁ isoform has only been described in salmonids [O. mykiss(Richards et al., 2003); S. salar (Nilsen et al.,2007)]. In the present study as well as the quoted studies, the transcript level of α_1c was unresponsive to salinity as well as during smoltification. The level was somewhat lower than theα _1a and α_1b isoforms. In O. mykiss,there is a universal tissue distribution of the transcript of this isoform(Richards et al., 2003) and if this also applies to cell types within the gill, this may lead to a very low transcript level per cell and thus explain why the isoform was not detected using AP-conjugated ISH probes. All together, the data conform with the`house-keeping' function suggested by Richards et al. and Nilsen et al.(Richards et al., 2003; Nilsen et al., 2007).

The levels of α₃ mRNA have only been reported in a single study in O. mykiss, where it was found to be unresponsive to SW (80%SW) (Richards et al., 2003)and is, thus, suggested to be a housekeeping isoform. In our present study,α ₃ mRNA levels were extremely low in whole-gill extracts,making up only approximately 1% of the total α-transcript level. Despite this low level, α₃ showed an overall significant decrease in response to SW. Even if the transcript may be concentrated in a few specific cell types, the level is still very low. At the protein level, three studies have investigated the presence of a α₃-like isoform in the gills of teleosts. Pressley reported lack of presence in the catfish gill by using a heterologous antibody (TED) raised against the rat α₃isoform (Pressley, 1992). D'Cotta et al. detected the α₃ in S. salar gills using the same antibody (D'Cotta et al.,2000). The isoform was found in both FW parr and in FW and SW smolt but no attempts were made to quantify the abundance. Theα ₃ isoform was also reported in a study of euryhaline tilapia(Lee et al., 1998) (use of a polyclonal antibody against rat α₃), where it was found not to change in the gills after transfer to SW. This was in contrast to theα ₁ in their study (use of polyclonal antibody against avianα ₁), which was higher in gills of SW than FW fish. Both theα ₁ and α₃ were localised by immunocytochemistry in MRC apparently belonging to different populations. Unfortunately, we were unable to localise the α₃ transcript in the salmon gill. We suspect that this is due to the extremely low abundance of the mRNA in this species.

Whereas the α-subunit is the catalytic component determining the transport kinetics of the Na⁺,K⁺-ATPase, theβ-subunit is essential for stabilisation of the α-subunit, membrane targeting, structural and functional maturation of the pump (Ackerman and Geering, 1990; Geering, 1990;Scheiner-Bobis, 2002). Compared with the α-subunit, very few studies have addressed the expression and regulation of the β-subunit in teleosts[Danio rerio (Appel et al.,1996); Anguilla anguilla(Cutler et al., 2000); Sparus sarba (Deane and Woo,2005); S. salar(Nilsen et al., 2007)]. In mammals, three isoforms (β₁–β₃) are expressed in different tissues and may assemble with differentα-isoforms, thus increasing the number of possible α,β-heterodimers considerably. In the eel, the β₁ and a duplicate β₁ isoform, named β₂₃₃ are the only isoforms expressed in the gill (Cutler et al., 1995; Cutler et al.,2000) whereas a β₃ isoform is found exclusively expressed in the brain (Cutler et al.,1997). Both the β₁ and the β₂₃₃isoform showed a strong response to salinity changes in the eel gill and the latter also in other ion-transporting epithelia. We found that theβ ₁ transcript overall increased (doubled) after transfer to hypersaline conditions, with an initial peak at 24 h followed by an increase on day 7 after transfer. Remarkably, the absolute level of β₁is considerably lower (approximately 1/50) than the sum of α-subunits,and assuming that the transcripts are localised in the same cell types, this suggests that the rate of β-subunit production may be a limiting step in generation of new pumps. β-Isoform multiplicity is unknown in salmonids at this moment but based on the present findings, it should be pursued in future studies. In mammalian and cellular model systems, there are examples of co-expression and regulation of the two subunits in both equimolar and different levels (Ewart and Klip,1995). In these situations, where different transcript levels are found, the β-subunit transcript is usually expressed at the lower level(Young and Lingrell, 1987). In addition to differentially regulated rates of transcription, there is evidence that the translational efficiency of the β-subunit may be several-fold higher than the α-subunit, which may account for equal accumulation of the two subunit proteins in those cells and tissues where α-mRNA is more abundant than β-mRNA. There are several examples of differential control of α- and β-subunit genes in cell systems(Corthésy-Theulaz et al.,1991; Farman et al.,1992; Geering et al.,1989) and one remarkable example among teleosts. Deane and Woo cloned both the α- and β-subunit in S. sarba gill and examined the response to various hormones(Deane and Woo, 2005). The two subunits were mostly regulated in parallel, even though absolute quantification was not performed. However, there was a remarkable divergence with regard to the effect of cortisol, which stimulated the α-subunit but had no effect on the β-subunit, neither at the mRNA nor the protein level. In agreement herewith, the authors found no effect of cortisol on Na⁺,K⁺–ATPase enzymatic activity. This clearly demonstrates that expression of the β-subunit in some instances may exert important control with overall pump abundance. Future studies should include analyses of α- as well as β-subunit expression to increase our understanding of their regulatory diversity.

The mRNAs of the two predominant α_1a andα _1b isoforms were localised by ISH in the gill of S. salar whereas we were unsuccessful in localising the α_1cand α₃ isoforms. Based on the quantitative expression data,the obvious reason for the lack of success with the latter isoforms is their low level of expression in the gill. This is particularly the case for theα ₃ isoform. Isoforms α_1a andα _1b were generally localised in discrete cells primarily in the filament epithelium but occasionally also in lamellar positions. Both isoforms were present in cells in FW and SW; however, their cellular expression pattern changed in a characteristic way, in accordance with the reciprocal regulation described above. Judged from their location, these cells correspond to mitochondrion-rich chloride cells of one type or another (see Introduction).

In FW, the α_1a isoform was localised in numerous cells in the interlamellar space and also frequently in cells on the lamellae. Most of these cells appeared elongated and large and were typically in contact with the apical (water) surface of the epithelium. The α_1b isoform was also expressed in cells in the interlamellar region but less so than theα _1a isoform. Occasionally, cells in the lamellae also stained positive for α_1b. However, in FW theα _1b-probe stained rounded cells, notably small and localised deep in the epithelium without contact with the apical surface. This suggests these cells to be non-functional, possibly immature differentiated chloride cells awaiting stimulus to become functionally mature. This stimulus for development appears to be under the influence of salinity, as the staining intensity and apparent many-fold increase in cell size in the SW-gill, the mRNA now being localised in large, elongated cells with apical contact exclusively in the interlamellar region. It seems probable that these cells represent the functional stage of the smaller cells described for the FW-gill– thus representing the secretory SW-type chloride cell. Not much is known about cellular differentiation mechanisms and morphogenesis of the various types of MRCs in the fish gill. However, a model of epidermal stem cell differentiation into various types of MRCs in zebrafish (Danio rerio) embryos has recently been proposed by Hsiao et al.(Hsiao et al., 2007).

In the SW gill, the α_1a mRNA is present in several cells– mostly in the interlamellar region and only rarely on the lamellae. These cells appear deeper in the epithelium than their FW-counterparts and may be non-functional. The general picture emerging from the localisation part of this study is in good agreement with the quantification data and shows a very morphoplastic gill epithelium with at least two cell types responsible for ion-transport being recruited to- and from- the epithelial surface in response to salinity. The mRNAs of α_1a and α_1b are present in the functional state as well as presumed precursor state of these cells, characterised by their deeper position below the epithelial surface. This picture supports the hypothesis by Richards et al. that theα _1a isoform is predominantly involved in ion-uptake andα _1b is the isoform driving ion-secretion, even though both isoforms are present in the basolateral membrane network and pump Na⁺ into the basolateral space(Richards et al., 2003). Whereas the localisation data suggest that expression of theα _1b isoform is exclusive to secretory-type chloride cells and their early differentiated stages, it cannot be excluded that theα _1a may also play a role in secretory-type chloride cells. This question shall await co-localisation studies with other proteins involved in ion secretion (e.g. CFTR, Na⁺K⁺2Cl^–cotransporter) or ion absorption (e.g. ENaC, V-type H⁺ATPase)respectively.

Successful acclimation to hypo- and hyperosmotic media requires a net reversal of ion transport across the gill epithelium. The recent, discovery of multiple Na⁺,K⁺–ATPase α-isoforms is an important step in understanding the molecular basis for this reversal(Richards et al., 2003). Our present study is the first to localise these isoforms in gill tissue, to investigate their dynamics during remodelling of the gill and to establish the real quantitative relationship between the transcript levels of theα-isoforms and the auxiliary β-subunit of Na⁺,K⁺–ATPase. Importantly, the histological evidence suggests that α_1b may be good marker for SW-type chloride cells and their putative precursor cells deeper in the filament whereas it cannot be excluded that α_1a may play a dual role in both FW- and SW-type chloride cells. The observedα _1a–α_1b isoform switch during FW–SW acclimation may form the molecular basis for the reversal of Na⁺ transport across the gill epithelium(Jorgensen, 2008). In the SW gill, Na⁺,K⁺–ATPase may be rooted in glycosphingolipid-enriched rafts whereas the dominant isoform in FW(α_1a) due to sequence differences probably is not associated with rafts and it has been suggested that this leads to uncoupled Na⁺ transport (Lingwood et al.,2005). A lysine substitution found in α_1a in the ion binding site may reduce the Na⁺:ATP ratio from 3 to 2 making ion extrusion in FW thermodynamically more favourable(Jorgensen, 2008). In addition, a substitution of glutamate953 with serine in theα _1a isoform may interfere with interaction with FXYD peptides and thereby alter the affinities for Na⁺ or K⁺. With the recent findings that several FXYD isoforms are expressed in the salmon gill(Tipsmark, 2008), this may have important implications for the tuning of ion transport during osmotic adjustments. However, two interesting questions are open for future research:are the two main types of MRCs interconvertible and do they differentiate from the same stem cells given the right stimuli? There is elegant evidence that,at least in some species, both transformations may occur(Hiroi et al., 1999; Hsiao et al., 2007; Hwang and Lee, 2007).

LIST OF ABBREVIATIONS

 
• AP

  alkaline phosphatase

 
• Ct

  threshold cycle

 
• FW

  freshwater

 
• ISH

  in situ hybridisation

 
• MRC

  mitochondrion-rich cell

 
• PCR

  polymerase chain reaction

 
• QPCR

  quantitative polymerase chain reaction

 
• SEI

  sucrose–EDTA–imidazol

 
• SW

  seawater