Immunolocalization of Ion-Transport Proteins to Branchial Epithelium Mitochondria-Rich Cells in the Mudskipper (Periophthalmodon Schlosseri)

The mudskipper Periophthalmodon schlosseri is an amphibious, obligate air-breathing, teleost fish with the ability to eliminate ammonia actively (Randall et al., 1999). It is an active carnivore with a high terrestrial affinity, inhabiting the brackish-water mangrove swamps and mudflats of southeast Asia (Murdy, 1989). Related to its ability to excrete ammonia actively is the high tolerance of this species of mudskipper to environmental ammonia (Peng et al., 1998).

P. schlosseri has a 96 h LC₅₀ value of 514 μmol l⁻¹NH₃ (Peng et al., 1998) and has an ammonia tolerance similar to that of the equally impressive Lake Magadi tilapia (Oreochromis alcalicus grahami; Randall et al., 1989) and the air-breathing catfish Heteropneustes fossilis (Saha and Ratha, 1990). However, these species make use of the ornithine–urea cycle to produce urea as a means of detoxifying ammonia, while this mudskipper species does not (Peng et al., 1998).

Mudskippers make use of free amino acids and have a powerful glutamate dehydrogenase/glutamine synthetase system for ammonia detoxification in the brain (Iwata, 1988; Peng et al., 1998). Incidentally, the ammonia levels in the burrow water of the mudskipper have been measured at 2 mmol l⁻¹(Y.-K. Ip, unpublished observations). Also, when out of water, P. schlosseri remain ammoniotelic, increasing ammonia, urea and free amino acid stores (Ip et al., 1993) and excreting the remainder into the boundary layer of water covering the animal (Wilson et al., 1999). Terrestrial excursions by other amphibious teleosts are generally characterized by a loss of the capacity to eliminate ammonia (Morii et al., 1978; Iwata et al., 1981).

Randall et al. (1999) have demonstrated high levels of Na⁺/K⁺-ATPase activity associated with the gills that can be activated by physiological levels of NH₄⁺instead of K⁺. Net ammonia excretion rates were also sensitive to in vivo pharmacological inhibition of Na⁺/K⁺-ATPase by ouabain and of the Na⁺/H⁺exchanger (NHE) by amiloride. Measured transepithelial potentials could not account for ammonia effluxes in high environmental ammonia levels, indicating that ammonia efflux is active.

The branchial epithelium of this species is unique among species of mudskipper and teleost fishes in general (Schöttle, 1931; Low et al., 1988, 1990; Wilson et al., 1999). The lamellar epithelium is densely packed with mitochondria-rich (MR) cells, and the interlamellar water spaces are restricted by interlamellar fusions. The opercular epithelium, which is lined with intraepithelial capillaries, appears to be better suited to gas exchange (Schöttle 1931; Wilson et al., 1999). The branchial MR cells have a tubular system continuous with the basolateral membrane that is typical of teleost fishes (Pisam, 1981; Philpott, 1980). However, the MR cells of this species have deep apical crypts, and accessory-type cells were seldom observed (Wilson et al., 1999). Accessory cells are typically associated with active NaCl elimination in marine teleost fishes. It is this apparent lack of accessory cells and the over-abundance of MR cells that suggest a function in addition to NaCl elimination for the gill population of MR cells. It would appear that the ability of these animals to eliminate ammonia actively is probably associated with the branchial MR cells.

In the present study, we investigate the possibility that the branchial epithelium is the site of active NH₄⁺elimination. Ion-transport proteins involved in the active elimination of NH₄⁺are localized using an immunological approach employing non-homologous antibodies. Antibodies against the following proteins were used: NHE2 and NHE3, the cystic fibrosis transmembrane regulator-like anion channel (CFTR), vacuolar-type H⁺-ATPase (V-ATPase), carbonic anhydrase, Na⁺/K⁺-ATPase and the Na⁺/K⁺/2Cl⁻cotransporter (NKCC). We also investigated the role of carbonic anhydrase in ammonia elimination by using the carbonic anhydrase inhibitor acetazolamide and the role of boundary-layer pH effects by buffering the bath water with Hepes at pH 8.0 and 7.0.

Adult Periophthalmodon schlosseri (Pall.) of both sexes were collected from the Pasir Ris, a small estuary on the east coast of Singapore, and transported to City University of Hong Kong. In the laboratory, fish in individual plastic aquaria were kept partially submerged in 50 % sea water (salinity 15 ‰). Every second day, the water was changed and the animals were fed goldfish (Carassius auratus) ad libitum. Fish mass ranged from 55 to 120 g. The room air and water temperatures remained constant at approximately 26 °C. No attempts were made to control lighting conditions, and a natural photoperiod was followed.

Animals were fasted for 3 days prior to the start of experimentation. Fish kept in 500 or 1000 ml of 50 % sea water were exposed to a series of three flux periods. An initial 3 h control flux period was followed by a 3 h experimental flux period and finally a 3 h recovery flux period. The experimental exposures consisted of exposing the fish either to 0.1 mmol l⁻¹acetazolamide to inhibit carbonic anhydrase or to 5.0 mmol l⁻¹Hepes to maintain water pH at the gill surface at either pH 7 or pH 8. Control and recovery fluxes were conducted in 50 % sea water. For the first two flux periods, water samples were taken at 0, 1, 2 and 3 h, but water samples were taken only at 0 and 3 h for the recovery period.

Water samples were assayed for total ammonia concentration (Verdouw et al., 1978), from which ammonia flux rates (J_Amm) were calculated (μmol ammonia kg⁻¹fish h⁻¹). Appropriate standards were prepared with acetazolamide or Hepes (pH 7 and 8) in 50 % sea water. Titratable acidity (TA) flux (J_TA) was measured according to the method of McDonald and Wood (1981) from the difference between 0 and 3 h water samples from the control and experimental periods. In brief, 100 ml water samples were aerated overnight to remove respiratory CO₂, and 25 ml subsamples (weighed to 0.01 g) were titrated to pH 4.00 with 0.01 mol l⁻¹HCl using a Radiometer autotitrator and Orion Ross-type combination electrode. Aeration continued during titration to remove liberated CO₂. The net acid flux (J_Acid; μequiv H⁺kg⁻¹h⁻¹) was calculated as the sum of J_TA and J_Amm, signs considered.

Whole gill arches were immersion-fixed for immunolocalization studies at the light microscopic level using 3 % paraformaldehyde/phosphate-buffered saline (PFA/PBS), pH 7.4, and at the electron microscopic level using 2 % paraformaldehyde, 75 mmol l^{−1 L}-lysine, 10 mmol l⁻¹sodium m-periodate fixative (McLean and Nakane, 1974) overnight at 4 °C. Following fixation in 3 % PFA/PBS, tissues were rinsed in PBS and 10 % sucrose/PBS and either frozen in liquid nitrogen or processed for paraffin embedding (Paraplast, Fisher Scientific). Following fixation for immunoelectron microscopy, tissue was rinsed in PBS, and free aldehyde groups were quenched in a 50 mmol l⁻¹NH₄Cl solution for 20 min. The tissue was then dehydrated through a decreasing temperature (room temperature to −20 °C) ethanol series (1 h each in 30 %, 50 %, 70 %, 95 % and three times for 1 h in 100 % ethanol), gradually infiltrated with resin (ethanol:Unicryl; 2:1, 1:1, 1:2 for 30 min at −20 °C and 100 % Unicryl twice for 1 h and overnight at −20 °C) and transferred to Beem caps for embedding (BioCell International UK). The resin was polymerized using ultraviolet light at −10 °C over 3 days.

Gill Na⁺/K⁺-ATPase was immunolocalized using a monoclonal antibody specific for the α-subunit of chicken Na⁺/K⁺-ATPase (Takeyasu et al., 1988). The antibody (α5) developed by D. M. Fambrough (Johns Hopkins University, MD, USA) was obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa Department of Biological Sciences, Iowa City, IA 52242, USA, under contract NO1-HD-7-3263 from the National Institute of Child Health and Human Development (NICHD). The antibody was purchased as cell culture supernatant (0.9 mg ml⁻¹). This antibody is now in routine use for identifying gill MR cells by way of their high Na⁺/K⁺-ATPase levels (Witters et al., 1996; T. H. Lee et al., 1998).

The gill Na⁺/K⁺/2Cl⁻cotransporter was immunolocalized using a monoclonal antibody specific for human colonic NKCC1 (Lytle et al., 1995). The antibody (T4) developed by Christian Lytle (Division of Biomedical Sciences, University of California, Riverside, CA, USA) was obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa Department of Biological Sciences, Iowa City, IA 52242, USA, under contract NO1-HD-7-3263 from the NICHD. The antibody was purchased as cell culture supernatant (0.8 mg ml⁻¹). Fixed/frozen dogfish (Squalus acanthias) rectal gland tissue was used as a positive control. The T4 monoclonal antibody has greater species cross-reactivity than the J3 monoclonal antibody, the immunoreactivity of which is limited to shark tissues (Lytle et al., 1995).

Rabbit polyclonal antibodies generated against glutathione S-transferase fusion proteins incorporating the last 87 amino acid residues of NHE2 (antibody 597; Tse et al., 1994) and the last 85 amino acid residues of NHE3 (antibodies 1380 and 1381; Hoogerwerf et al., 1996) were used. Antibody specificity has been determined in NHE-expressing PS120 cell lines. These antibodies have been used in a number of different studies for immunolocalization and western analysis of NHE2 and NHE3 isoforms (He et al., 1997; M. G. Lee et al., 1998; Levine et al., 1993; Sun et al., 1997) including a recent study on two teleost fishes (Oncorhynchus mykiss and Pseudolabrus tetrious) by Edwards et al. (1999). Three monoclonal antibodies generated against a maltose-binding protein fusion protein that contained the carboxyl-terminal 131 amino acid residues of NHE3 were also tested (Biemesderfer et al., 1997). We were unable to demonstrate cross-reactivity of gill tissue by either western analysis or immunohistochemistry with these three commercially available monoclonal antibodies (Chemicon International, CA, USA).

A commercial monoclonal antibody specific to human CFTR (165–170 kDa) was purchased from NeoMarkers Inc., CA, USA. The antibody was raised against a full-length human CFTR recombinant protein, and the IgM was purified from ascites fluid by ammonium sulphate precipitation (0.2 mg ml⁻¹of 10 mmol l⁻¹PBS, pH 7.4, with 0.2 % bovine serum albumin, BSA, and 15 mmol l⁻¹sodium azide). This antibody did not cross-react with dogfish transmembrane regulator (DFTR) in cryosections and has been reported to cross-react only with human tissue (NeoMarkers Inc.).

The V-ATPase was immunolocalized using a rabbit polyclonal antibody raised against a synthetic peptide corresponding to a sequence from the catalytic 70 kDa A-subunit of the bovine vacuolar-type H⁺-ATPase complex (CSHITGGDIYGIVNEN; Südhof et al., 1989) (Protein Service Laboratory, University of British Columbia, Canada). The peptide was conjugated to keyhole limpet haemocyanin using a maleimide linker (Pierce) and mixed (300 μg) with Freund’s complete adjuvant (Sigma). New Zealand White rabbits were immunized by subcutaneous injection and boost injections (300 μg) with incomplete Freund’s adjuvant followed at biweekly intervals with test bleed. The rabbits were terminally bled and the serum collected. A pre-immunization serum was collected for use as a control. Sera were tested by peptide enzyme-linked immunosorbent assay (ELISA) and western blot analysis of trout gill homogenates.

A polyclonal antibody raised against the same peptide (Südhof et al., 1989) has been used to identify the distribution of the V-ATPase A-subunit in mammalian kidney (Madsen et al., 1991; Kim et al., 1992) and trout gill (Lin et al., 1994). Using rat kidney as a positive control tissue, we found identical results to those cited above.

Carbonic anhydrase was immunolocalized using a rabbit polyclonal antibody generated against chick retina carbonic anhydrase II (CAII). The antiserum was kindly provided by P. Linser (The Whitney Laboratory, University of Florida, USA). The antibody cross-reacts strongly with teleost and dogfish erythrocyte and branchial carbonic anhydrase isoforms (Wilson et al., 2000). Western analysis reveals a strong immunoreaction with bands in the 30 kDa molecular mass range.

Cryosections (5–10 μm) thick were cut on a cryostat at −20 °C, collected onto either 0.01 % poly-^L-lysine-coated (Sigma) slides or electrostatically charged slides (SuperFrost Plus, Fisher Scientific) and fixed in acetone at −20 °C for 5 min. Sections were then air-dried. Paraffin sections (5 μm thick) were collected onto charged slides, air-dried at 37 °C overnight, and dewaxed though a series of xylene baths and rehydrated through an ethanol series finishing in PBS. Sections were circled with a hydrophobic barrier (DakoPen, Dako DK) and then blocked with 5 % normal goat serum (NGS)/0.1 % BSA/0.05 % Tween-20 in PBS (TPBS), pH 7.4, for 20 min. Sections were then incubated with primary antibody diluted 1:50 to 1:200 in 1 %NGS/0.1 % BSA/TPBS, pH 7.3, for 1–2 h at 37 °C. They were then rinsed in 0.1 % BSA/TPBS or PBS followed by incubation with goat anti-mouse or anti-rabbit secondary antibody conjugated to fluorescein isothiocyanate (1:50 FITC, Chemicon Intl. Inc.), Texas Red (1:100, Molecular Probes or Jackson Labs) or Cy3 (1:200, Sigma) for 1 h at 37 °C. Following rinsing with 0.1 % BSA/TPBS, sections were mounted with VectaShield mounting medium and viewed on a Zeiss AxioPhot photomicroscope or Bio-Rad 600 confocal microscope with the appropriate filter sets.

To obtain a representative sample, gill tissue from a minimum of 4–6 animals was sectioned and immunolabelled using each of the above-mentioned antibodies. Immunolabelling was repeated on at least on two occasions on slides in duplicate or triplicate for each antibody and animal. Normal mouse or rabbit serum negative controls were also processed at the same time (see below).

Ultrathin sections were prepared on a Reichart ultramicrotome and collected onto either Formvar- or Formvar/carbon-coated nickel grids. Following air-drying, sections were rehydrated by floating the grids on drops of PBS. The grids were then transferred to drops of diluted primary antiserum (1:100) and incubated at room temperature for 1 h. They were then rinsed and transferred to drops of diluted secondary goat anti-mouse antiserum (1:100) conjugated to colloidal gold (10 or 20 nm, Sigma, Chemicon) and incubated for 1 h at room temperature. They were rinsed with PBS and fixed for 10 min in 1 % glutaraldehyde/PBS and rinsed again in double-distilled water before counter-staining with lead citrate and saturated uranyl acetate. Sections were viewed on a Philips 300 transmission electron microscope and photographed with Kodak EM plate film 4489.

Tissue was thawed in ice-cold SEI buffer (300 mmol l⁻¹sucrose, 20 mmol l⁻¹EDTA, 100 mmol l⁻¹imidazole, pH 7.3), and filaments were scraped from the arch using a blunt razor blade. Filaments were then homogenized with a Potter–Elvehjem tissue grinder on ice (1000 revs min⁻¹, 10 strokes). The homogenate was centrifuged at 2000 g, and the supernatant was discarded. The pellet was resuspended in 2.4 mmol l⁻¹deoxycholate in SEI buffer and rehomogenized. Following a second centrifugation at 2000 g, the supernatant was saved.

Total protein was measured using the Bradford method (Bradford, 1976) with a BSA standard and the homogenate diluted to 1 μg μl⁻¹in Laemmli’s buffer (Laemmli, 1970). Proteins were separated by polyacrylamide gel electrophoresis (PAGE) under denaturing conditions, as described by Laemmli (1970), using a vertical mini-slab apparatus (Bio-Rad, Richmond, CA, USA). Proteins were transferred to Immobilon-P membranes (Millipore) using a semi-dry transfer apparatus (Bio-Rad). Blots were then blocked in 3 % skim milk/TTBS (0.05 % Tween 20 in Tris-buffered saline: 20 mmol l⁻¹Tris-HCl; 500 mmol l⁻¹NaCl, 5 mmol l⁻¹KCl, pH 7.5).

Blots were incubated with primary antiserum for 1 h at room temperature or overnight at 4 °C. Following a series of washes with TTBS, blots were incubated with either goat anti-rabbit or anti-mouse horseradish-peroxidase-conjugated antibody (Sigma). Bands were visualized by enhanced chemiluminescence (Amersham). Tissue homogenates from three animals were electrophoretically separated and probed for each antibody.

Normal mouse serum and normal rabbit serum were substituted for primary antibodies in the immunohistochemical and blotting protocols for use as negative controls. Except for the V-ATPase, pre-immune sera were not available. Also, antigens were not available, so pre-absorption controls were also not conducted.

Data are presented as means ± S.E.M. (N). Ammonia excretion rates from the different treatments were compared using a one-way repeated-measures analysis of variance (ANOVA) and post-hoc Student–Newman–Keuls test. Net ammonia, titratable acidity and acid fluxes under control and experimental conditions were compared using paired t-tests. The fudicial limit of significance was set at 0.05.

The carbonic anhydrase inhibitor (0.1 mmol l⁻¹) acetazolamide reduces J_Amm by 48 % yet has no significant effect on J_Acid (Fig. 1). J_Amm is inhibited immediately upon exposure to acetazolamide and starts to recover slightly during the second and third hours of exposure. Removal of acetazolamide results in a complete recovery of J_Amm.

The addition of 5 mmol l⁻¹Hepes to 50 % sea water is sufficient to fix the water pH at either 7 and 8 during the 3 h experimental period (Fig. 2E). Buffering the water at pH 7.0 does not significantly affect either J_Amm or J_Acid (Fig. 2A,B). However, addition of 5 mmol l⁻¹Hepes, pH 8, significantly increases J_Acid while having no effect on J_Amm (Fig. 2C,D).

Figs 3A, 4 and 5A show an abundance of Na⁺/K⁺-ATPase immunoreactivity associated with the lamellar epithelium. Immunopositive cells are also found in the epithelium of the gill arch. Immunoreactivity for Na⁺/K⁺-ATPase is restricted to the large epithelial MR cells (Fig. 5A). Labelling occurs throughout the cell body with the exception of the nucleus and apical crypt area. Immunogold labelling for Na⁺/K⁺-ATPase reveals that the intracellular staining observed by immunofluorescence is restricted to elements of the tubular system and that no labelling is associated with the apical crypt (Fig. 4A,B). Control incubations with normal mouse serum result in negligible levels of immunoreactivity (Fig. 3B). In western blots, the α5 antibody cross-reacts strongly with a band with an apparent molecular mass of 116 kDa (see Fig. 9A).

The Na⁺/K⁺/2Cl⁻cotransporter has an essentially identical distribution to that of the Na⁺/K⁺-ATPase (Fig. 5E,F). Na⁺/K⁺/2Cl⁻immunofluorescence labelling shows a similar diffuse cytoplasmic distribution, and immunogold labelling indicates that the immunoreaction is associated with the MR cell tubular system. No labelling is associated with the apical crypt. We were, however, unable to demonstrate immunoreactivity on western blots using the T4 Na⁺/K⁺/2Cl⁻antibody.

The mouse monoclonal antibody against the human cystic fibrosis transmembrane regulator (CFTR) cross-reacts specifically with the MR cell apical crypt (Fig. 5C,D). In sections, labelled crypts appear as either a ring or a U shape indicative of cross or longitudinal sections through the crypt, respectively. This antibody was only useful with fixed/ frozen tissue and not fixed/paraffin-embedded tissue. No immunolabelling is associated with pavement cells, pillar cells, mucocytes, erythrocytes or the basal portion of MR cells. Western blots indicate an immunoreactive band with an apparent molecular mass of 150 kDa (see Fig. 9B).

The distributions of apical isoforms of the Na⁺/H⁺exchanger (NHE2 and NHE3) were determined in the mudskipper gill using antibodies Ab597 and Ab1380 specific for apical NHE2 and NHE3 isoforms, respectively. Both NHE isoforms are immunolocalized to the apical crypts of MR cells (Figs 6, 7). From low-power photomicrographs, most crypts can be seen to be labelled (Fig. 6A). The NHE3 antibody Ab1381 was not useful for tissue localization. A control incubation with normal rabbit serum reveals non-specific labelling associated with the contents of the afferent and efferent vessels (Fig. 6E). In western blots of crude membrane preparations, immunoreactivity is seen with bands in the predicted molecular mass ranges (approximately 85 kDa); however, additional major bands are also recognized at lower molecular masses (see Fig. 9D,E).

Immunoperoxidase staining for carbonic anhydrase in fixed/frozen sections is restricted to the apical crypt region of the MR cells (Fig. 8A,B). Staining is not associated with the rest of the cell. Erythrocytes also show immunoreactivity, while no immunoreactivity is associated with pavement cells or pillar cells. Incubation of control sections with normal rabbit serum at an equivalent dilution results in insignificant levels of labelling.

The V-ATPase was immunolocalized using a polyclonal antibody directed against the A-subunit of the vacuolar proton pump complex. Immunoperoxidase staining shows that the V-ATPase is restricted to the apical crypt region of MR cells (Fig. 8C). It should be noted that some difficulty was encountered in trying to reproduce these results. Western blots show an immunoreactive band with an apparent molecular mass of 78 kDa (Fig. 9C).

These results support the hypothesis that, in the amphibious air-breathing mudskipper Periophthalmodon schlosseri, ion-transport proteins in gill MR cells contribute to active ammonia (NH₄⁺) excretion. In vivo pharmacological studies have shown that a significant proportion of the total ammonia efflux (J_Amm) is sensitive to inhibition of the Na⁺/H⁺(NH₄⁺) exchanger by amiloride (Randall et al., 1999) and to inhibition of carbonic anhydrase by acetazolamide (Fig. 1). The elimination of boundary-layer acidification by the addition of 5 mmol l⁻¹Hepes buffer, pH 8.0, to the 50 % sea water was without effect on J_Amm, indicating that NH₃ trapping is not an important component of total ammonia excretion (Fig. 2). These animals are also capable of excreting ammonia against inward NH₄⁺and NH₃ gradients (2 mmol l⁻¹NH₄Cl); however, a portion of J_Amm is then sensitive to inhibition of Na⁺/K⁺(NH₄⁺)-ATPase by ouabain (Randall et al., 1999).

The branchial epithelium contains an abundance of MR cells with deep apical crypts and an extensive tubular system continuous with the basolateral membrane (Wilson et al., 1999). Immunological localization studies have demonstrated the presence of NHE2- and NHE3-like isoforms, a CFTR-like anion channel and carbonic anhydrase associated with the apical crypt and a Na⁺/K⁺-ATPase and a Na⁺/K⁺/2Cl⁻cotransporter (NKCC) associated with the tubular system of these MR cells. The demonstration of the presence of these transporters within the branchial MR cells strongly indicates that these cells are probably involved in the active elimination of ammonia, although their involvement in active Cl⁻elimination (typical of brackish-water teleosts) is also expected (see Marshall and Bryson, 1998).

In P. schlosseri, excretion of ammonia is not dependent on boundary-layer pH. This would indicate that P_NH gradients are not as important a mechanism of total ammonia elimination as in freshwater fishes (Cameron and Heisler, 1983; Wilson et al., 1994). The facilitation of NH₃ diffusion by trapping of NH₃ (NH₃+H⁺→NH₄⁺) through the acidification of the boundary layer by hydration of respiratory CO₂ or direct H⁺excretion (by the Na⁺/H⁺exchanger or V-ATPase) is important in freshwater fish (Wright et al., 1989; Wilson et al., 1994; Salama et al., 1999) and probably also in seawater fish (Wilson and Taylor, 1992). The mudskippers kept in unbuffered 50 % sea water (pH 8) were able to acidify their bath water to a final pH of approximately pH 7.4 (Fig. 2E). The acidification of the gill boundary layer is dependent on the water buffer capacity. Sea water tends to be poorly buffered, and the gill boundary layer will therefore be acidified. Experimentally, the acidification of the boundary layer can be eliminated by the addition of buffer to the water. In the mudskipper, the elimination of boundary-layer acidification by buffering with 5 mmol l⁻¹Hepes, pH 8.0, results in no change in J_Amm. This did, however, stimulate a large J_Acid. An increase in J_Acid has also been observed in freshwater trout Oncorhynchus mykiss (Salama et al., 1999), but not to the same extent.

The gill MR cells are associated with high levels of Na⁺/K⁺-ATPase immunoreactivity (Fig. 3), and in vitro assays show that (ouabain-sensitive) enzyme activity is also very high, being 3–4 times higher than in another mudskipper species (Boleophthalmus boddaeti; Randall et al., 1999). The levels of immunoreactivity in the other gill cell types (pavement cells, erythrocytes, pillar cells, filament-rich cells) are below the level of detection, although the absence of immunoreactivity is not a good criterion for concluding that the enzyme is absent. There are limits to the sensitivity of the technique in addition to the possibility that other isoforms of the α-subunit are present in these cell types. Indeed, Na⁺/K⁺- ATPase is probably present in all cell types performing cellular housekeeping roles.

K⁺and NH₄⁺share a similar hydration radius, and NH₄⁺has been shown in vitro to compete with K⁺for binding sites on Na⁺/K⁺-ATPase (Mallery, 1983; Kurtz and Balaban, 1986; Garvin et al., 1985; Towle and Hølleland, 1987; Wall and Koger, 1994) and also on the Na⁺/K⁺/2Cl⁻cotransporter (Kinne et al., 1986). Randall et al. (1999) were able to demonstrate in P. schlosseri that physiological levels of NH₄⁺could stimulate the ATPase in vitro as well. Only during exposure of these animals to high environmental ammonia levels, when NH₄⁺and NH₃ gradients (2 mmol l⁻¹) are directed inwards, is J_Amm sensitive to ouabain. This inhibition of J_Amm is associated with a significant increase in plasma ammonia levels (Randall et al., 1999). Thus, in the absence of favourable diffusion gradients NH₄⁺substitution becomes physiologically significant.

Although the Na⁺/K⁺/2Cl⁻cotransporter is found in association with the MR cell tubular system (Fig. 5E) and has been found to substitute NH₄⁺for K⁺in other vertebrates (Kinne et al., 1986), it is not known how significant a contribution this basolateral pathway makes to the net ammonia efflux in the mudskipper. The Na⁺/K⁺/2Cl⁻cotransporter is driven by the Na⁺gradient maintained by the Na⁺/K⁺-ATPase, and thus Na⁺/K⁺/2Cl⁻cotransport may also only be important when the animal is faced with inward ammonia gradients. Evans et al. (1989) could find no effect of the Na⁺/K⁺/2Cl⁻cotransporter inhibitor bumetanide on ammonia efflux in the Gulf toadfish Opsanus beta.

The inhibition of the Na⁺/H⁺exchanger by amiloride reduces J_Amm by 50 % (Randall et al., 1999), yet the elimination of boundary-layer acidification is without effect on J_Amm (Fig. 2). It seems likely, therefore, that an important component of ammonia elimination is achieved by direct NH₄⁺transport by the Na⁺/H⁺exchanger rather than a boundary-layer pH effect on Na⁺/H⁺exchange. In renal microvillus membrane preparations, NH₄⁺has been shown to interact with the Na⁺/H⁺exchanger (Kinsella and Aronson, 1981; Nagami, 1988; Blanchard et al., 1998).

There are at least five NHE isoforms (NHE1–4 and NHEβ; see Claiborne et al., 1999), although only NHE2 and NHE3 are found apically. Immunoreactivity for both these isoforms is present in the apical crypt of the mudskipper gill MR cells. This contrasts with the finding of NHE3-like cytoplasmic immunoreactivity in cells in the filament epithelium of freshwater rainbow trout and seawater wrasse (Edwards et al., 1999). We have observed a similar labelling pattern in tilapia (Oreochromis mossambicus) to that reported by Edwards et al. (1999), but with the NHE2 antibody (Wilson, et al., 2000).

Thus, it would appear that the distribution of the Na⁺/H⁺exchanger in the mudskipper is unique amongst teleost fishes. In the mammalian kidney medulla thick ascending limb, although both isoforms are also present, it is the NHE3 isoform that plays a greater role in ammonium efflux (Paillard, 1998). Recently, Claiborne et al. (1999) have reported the presence of βNHE-like and NHE2-like isoforms in the gill of the sculpin Myoxocephalus octodecimspinosus using reverse transcriptase/polymerase chain reaction (RT-PCR). In the marine fish Opsanus beta, Evans et al. (1989) found no evidence for apical Na⁺/NH₄⁺exchange. Instead, they found that total ammonia excretion could be accounted for by non-ionic NH₃ (57 %) and paracellular ionic NH₄⁺(21 %) diffusion and basolateral Na⁺/K⁺-ATPase activity (22 %). However, their data could be interpreted as indicating that the Na⁺/K⁺-ATPase and Na⁺/H⁺exchange mechanisms are operating in series. Randall et al. (1999) found an increased plasma Na⁺concentration in P. schlosseri challenged with high external ammonia levels, which is consistent with the proposed mechanism of Na⁺/NH₄⁺exchange.

The distribution of carbonic anhydrase in the branchial MR cells is similar to that reported by Lacy (1983) in MR cells from the opercular epithelium of seawater-adapted killifish Fundulus heteroclitus. The sensitivity of J_Amm to carbonic anhydrase inhibition indicates that intracellular CO₂ hydration maybe important in providing H⁺for NH₃ protonation to maintain P_NH gradients across the basolateral membrane and in providing NH₄⁺for apical Na⁺/NH₄⁺exchange. From the lack of sensitivity of J_Amm to ouabain under conditions of low external ammonia concentration (Randall et al., 1999), it would appear that the facilitated movement of NH₄⁺across the basolateral membrane is not important. Amiloride, however, reduced the rate of ammonia excretion in animals in water without any additional ammonia added, indicating that apical Na⁺/NH₄⁺exchange may be taking place even when ammonia levels in the water are low.

Immunoreactivity for the V-ATPase could be demonstrated in the apical crypts of MR cells. However, the role of the V-ATPase in ammonia excretion is not clearly defined in the light of the absence of a pH-dependent boundary-layer effect (Fig. 2) and in vivo experiments using KNO₃, which had a delayed effect on J_Amm (Randall et al., 1999). The V-ATPase activity observed may only be involved in vesicle acidification (Harvey, 1992).

It was initially thought that the CFTR could be used as a marker for MR cells involved in Cl⁻secretion, a function normally ascribed to this cell type in seawater teleost fishes (chloride cells; Marshall and Bryson, 1998; Singer et al., 1998). In the initial investigation of the fine structure of the branchial epithelium, few chloride cell/accessory cell complexes were found, so it was thought that few of the MR cells were actually involved in Cl⁻elimination (Wilson et al., 1999). However, CFTR-like immunoreactivity was associated with the apical crypts in almost every branchial MR cell, and it may be that all MR cells are involved in Cl⁻elimination. This would be the conclusion if the function of the CFTR were limited to Cl⁻conductance; however, the CFTR-like anion channel is also involved in the regulation of other apical channels as well as the conductance of other anions, notably HCO₃⁻(Poulsen et al., 1994). The HCO₃⁻excretion into the lumen of the duodenum (Hogan et al., 1997) and airways

(M. C. Lee et al., 1998) of mammals is mediated by a CFTR-like anion channel. This latter function becomes quite intriguing. Assuming that CO₂ hydration catalyzed by intracellular carbonic anhydrase provides H⁺for NH₃ protonation to produce NH₄⁺for the Na⁺/H⁺exchanger, then a bicarbonate ion is left behind. HCO₃⁻must exit the cell across either the basolateral or apical membrane; however, basolateral exit would result in a base load during ammonia excretion. Apical exit facilitated by the CFTR-like channel would be acid–base neutral and would account for the lack of a boundary-layer pH effect. It also explains why J_Acid is generally approximately zero and is not affected by the inhibition of J_Amm with acetazolamide or amiloride (J. M. Wilson, unpublished observations).

Fig. 10 is a model that summarizes our data. The elimination of ammonia by P. schlosseri involves a Na⁺/H⁺exchanger and carbonic anhydrase and is not dependent on boundary-layer pH effects. Na⁺/K⁺-ATPase plays a role in the elimination of ammonia only against an ammonia gradient. Although a V-ATPase and Na⁺/K⁺/2Cl⁻cotransporter are also present, their roles in ammonia elimination are not certain. Immunolocalization studies indicate the presence of all these proteins in the branchial mitochondria-rich cells. It therefore seems reasonable to conclude that these cells are the sites of active elimination of ammonia mediated by ion-transport proteins.

Financial support for this work was provided by NSERC (Canada) and City University of Hong Kong research grants to D.J.R. J.M.W. was supported by UBC graduate fellowships. We would like to thank P. Laurent (CEPE-CNRS), J. Hoffman (IMBC-CNRS) and G. K. Iwama (UBC) for generously providing laboratory space and use of equipment.