Modes of Ammonia Transport Across the Gill Epithelium of the Dogfish Pup (Squalus Acanthias)

Fish generally excrete ammonia as the major product of protein and nucleic acid catabolism (see reviews by Kormanik & Cameron, 1981; Cameron & Heisler, 1985; Evans, 1985; Evans & Cameron, 1986). Both in teleosts and in elasmobranchs, renal efflux of ammonia is vanishingly small compared with branchial loss (e.g. Evans, 1982).

The mechanisms of ammonia transport across the fish branchial epithelium are somewhat controversial (see Evans & Cameron, 1986, for a recent review), because one cannot experimentally separate the fluxes of NH₃ and NH₄⁺, and manipulations of internal or external solutions are difficult in vivo and may alter ammonia fluxes secondary to perturbations of other transport steps. For example, alteration of external or internal ammonia concentrations can alter pH (assuming relatively high NH₃ permeance) and therefore putative Na⁺/H⁺ exchange, rather than Na⁺/NH₄⁺ exchange directly (e.g. Cameron & Heisler, 1983); addition of the inhibitor amiloride to external solutions can alter Na⁺/H⁺ exchange and therefore NH₃ gradients rather than Na⁺/NH₄⁺ exchange (e.g. Kirschner et al. 1973); and removal of external Na⁺ can reduce paracellular cationic permeability rather than inhibit cationic exchange systems (e.g. Zadunaisky, 1984).

Despite these uncertainties, recent studies indicate that ammonia extrusion by the freshwater rainbow trout (Salmo gairdneri) is predominantly via non-ionic diffusion of NH₃, with a variable role played by Na⁺/NH₄⁺ exchange, depending on the experimental conditions (Cameron & Heisler, 1983; Wright & Wood, 1985). It also appears that branchial H⁺/NH₄⁺ exchange (external H⁺ for internal NH₄⁺) may play a role in catfish (Ictalurus punctatus) exposed to extremely high external ammonia concentrations (Cameron, 1986). Interestingly, ionic diffusion of NH₄⁺ appears to play a major role in ammonia extrusion by the marine teleosts, Myoxocephalus octodecimspinosus (Longhorn sculpin) and Opsanus beta (Gulf toadfish) (Goldstein et al. 1982), but basolateral Na⁺/NH₄⁺ exchange also seems to be important in the toadfish (Claiborne et al. 1982).

Thus, these rather sketchy and often contradictory data indicate that ionic diffusion of NH₄⁺ may be important in ammonia efflux from marine teleost species, but that non-ionic diffusion of NH₃ and/or Na⁺/NH₄⁺ exchange probably predominates in freshwater species. The structural bases for these apparent differences are unknown, but ion permeability is much higher in marine teleosts than in freshwater species (e.g. Evans, 1979). It may be that diffusion of NH₄⁺ is a pathway for ammonia extrusion across the marine teleost gill epithelium simply because of a relatively high cationic permeability.

Modes of ammonia transport across the gill epithelium have rarely been studied in elasmobranchs, and may be especially interesting since these fish, despite their marine habitat, maintain a branchial cationic permeability of the same low order as that of the freshwater teleosts (Evans, 1979). The only published data supporting a role for Na⁺/NH₄⁺ exchange in the elasmobranchs (Payan & Maetz, 1973; Evans, 1982) suffer from the problems of secondary effects noted above, and our work on the little skate (Raja erinacea) could not demonstrate Na⁺/NH₄⁺ exchange (Evans et al. 1979). We have succeeded in perfusing the isolated head of the spiny dogfish pup (Squalus acanthias;Evans & Claiborne, 1983) and the present study was therefore an attempt to define more carefully the usefulness of this preparation and delineate and quantify various potential modes of ammonia transport across the gill epithelium of this species. The putative modes examined were: basolateral (ouabain-sensitive) and apical (amiloride-sensitive) Na⁺/NH₄⁺ exchange, basolateral (bumetanide-sensitive) Na⁺+NH₄⁺+2Cl ⁻ cotransport, non-ionic diffusion of NH₃, and ionic diffusion of NH₄⁺ (see Evans & Cameron, 1986, for a review of the data supporting the presence of these transport systems in the fish branchial epithelium).

Near-term pups (40 – 70 g including the yolk sac) were removed from pregnant female spiny dogfish (Squalus acanthias) caught by gill nets in Frenchman Bay, Maine, and were either maintained in running sea water (12 – 15°C) at the Mount Desert Island Biological Laboratory (MDIBL) or were air-freighted to Gainesville in insulated containers, and maintained at 12±1°C in a constant-temperature cold-room. Experiments were either performed at MDIBL at 12-15°C using water-jacketed perfusion chambers [with irrigation solutions cooled via thermoelectric cold plates (Model TCP-2, Thermoelectrics Unlimited, Inc.)], or in the cold-room in Gainesville. The isolated, perfused pup head was prepared as before (Evans & Claiborne, 1983), except that: the dorsal aorta was not cannulated; afferent pressures were recorded on a Gilson Duograph; and the seawater irrigate (100 ml) was recirculated with a Masterflex pump. Irrigation and perfusion flow rates were maintained at 10mlmin⁻¹ and approx. 650 μl min⁻¹, respectively. The perfusate (elasmobranch Ringer’s solutions: ERS) was bubbled with 1% CO₂ in air, and formulated as by Forster et al. (1972), with the addition of 80 mmol I⁻¹ trimethylamine oxide and 10⁻⁷moll⁻¹ adrenaline (L-hydrochloride). Preliminary studies (see Results) indicated that the pup head was haemodynamically sensitive to concentrations as low as 10⁻⁸moll⁻¹ adrenaline, and spontaneous ventilation was usually prompted at 10⁻⁷moll⁻¹ (unpublished results). Moreover, Butler et al. (1978, 1986) have found that resting levels of adrenaline are about 10⁻⁸moll⁻¹ in Scyliorhinus canicula and reach nearly 10 times that level during repeated burst exercise or hypoxia. In addition, in the perfused head of Salmo gairdneri viability is increased and lamellar oedema is avoided when 10⁻⁸– 10⁻⁶moll⁻¹ adrenaline is added to the perfusate (Part et al. 1982; Perry et al. 1984b). Finally, Ellis & Smith (1983) have suggested that lamellar oedema might also be associated with the lack of spontaneous .ventilation of perfused gill preparations.

An earlier study (Evans & Claiborne, 1983) determined that the perfused shark pup head was sensitive to high concentrations (10⁻⁵moll⁻¹) of adrenaline, so, as part of an effort to examine more carefully the integrity and viability of this perfused-head system, we tested the haemodynamic sensitivity of the gill vasculature to levels of adrenaline more closely approximating those found in vivo (see above). In these experiments the adrenaline concentration of the perfusate was sequentially increased from 10⁻⁸ to 10⁻⁶ mol I⁻¹, after an initial control period with adrenaline-free perfusate.

To ensure that perfusion did not alter the gill morphology, as has been described for both the trout and eel (Anguilla australis) head preparations (Part et al. 1982; Ellis & Smith, 1983; Bornancin et al. 1985), gill morphology was examined by scanning electron microscopy. Arches were removed from non-perfused and perfused heads and fixed for at least 24h in 2% glutaraldehyde in 0·1 moll⁻¹ cacodylate-HCl buffer (pH 7·4) adjusted to the proper osmolarity (approx. 1000mosmoll⁻¹) with sucrose. Fixed tissues were rinsed in distilled water, dehydrated through a graded ethanol series, critical-point dried in liquid CO₂, mounted on a specimen stub, coated with a gold-palladium alloy, and viewed on a Hitachi Model S-415A scanning electron microscope at an acceleration voltage of 15 kV.

Rates of ammonia transport across the perfused gill were measured by adding a known quantity of NH₄C1 to the perfusate and monitoring the rate of appearance of ammonia in the recycled, seawater irrigation solution. Perfusate and irrigate total ammonia concentrations were determined chemically by the phenol-hypo-chlorite method of Solorzano (1969). Actual perfusate ammonia concentrations were measured in each experiment because breakdown products from the urea (350 mmol I⁻¹) and colloidal osmolyte polyvinylpyrrolidone (PVP, 3%) generated perfusate ammonia concentrations in the range of 200–400 μnol I⁻¹ in nominally ammonia-free perfusate. Control experiments consisted of three, 20-min efflux periods separated by approx. 10min during which the irrigation solution was changed. At the end of this ‘wash’ period, a time zero sample (5 ml) of the irrigate was taken and the next 20-min efflux period was started. At the end of 20min, a second 5-ml sample was taken, and the rate of ammonia efflux, in μmol 100g⁻¹ h⁻¹, was calculated from the difference in ammonia concentration of the T₀vs T₂₀ samples, corrected for the declining volume of the irrigation medium and the wet mass of the intact fish (measured before preparation of the perfused head). In other experiments (see below) the perfusate was also changed during the wash period between successive efflux periods so that the serosal side of the gill epithelium was in contact with the new perfusate for 10 min before the start of a new efflux period.

The role of basolateral (serosal) Na⁺+NH₄⁺+2Cl⁻ cotransport and basolateral Na⁺/NH₄⁺ exchange was investigated by perfusing the head with ERS containing 5×l0⁻⁵moll^−t bumetanide during the second flux period (after an initial control efflux), and then ERS containing bumetanide plus 10⁻⁴moll⁻¹ ouabain in the third period. In another series of experiments, the efflux of ammonia via apical (serosal) Na⁺/NH₄⁺ exchange was monitored, after an initial control flux period, by adding ouabain to the ERS in the second period, and 10⁻³ mol I⁻¹ amiloride (in the presence of perfusate ouabain) to the irrigate in the third period. When amiloride was added, serosal ouabain was present to obviate potential effects of amiloride on basolateral Na⁺, K⁺-activated ATPase (see Discussion). Amiloride at such a high concentration is nearly insoluble in sea water so it was first dissolved in an equivalent of 1 moll⁻¹ HC1, added to 1 ml of hot (90–100°C) distilled water and then to 500 ml of hot sea water while stirring. The solution was then brought to the correct pH (approx. 81) by addition of 1moll^−lNaOH, and cooled slowly. Control analyses of ammonia in the presence of 10⁻³moll⁻¹ amiloride indicated that the drug inhibited the phenol-hypochlorite reaction producing an apparent 30% decline in measured ammonia concentration. This value was subsequently used to correct the measured ammonia concentrations in the irrigate solutions containing 10⁻³moll⁻¹ amiloride.

The relative roles of non-ionic diffusion of NH₃vs ionic diffusion of NH₄⁺ were quantified using an experimental protocol similar to that of Goldstein et al. (1982) where the perfusate NH₃ and NH₄⁺ concentrations were manipulated by specific alterations of the perfusate pH and ammonium chloride concentrations. It is obvious that diffusion of NH₃ will be down gradients of partial pressure, rather than of concentration. However, for comparative purposes (see Table 4) we have equated concentration with partial pressure, assuming constant solubility of the gas in the experimental solutions. Perfusate pH values were adjusted by altering HCO₃⁻ concentrations (2–20 mmol I⁻¹ NaHCO₃). Actual perfusate and irrigate ammonia concentrations were determined for each experiment (see above) and the pH determined in the same samples to the nearest 0·01 on a Radiometer PHM84 pH meter and Corning Model 150 pH meter, respectively. Both pH electrodes were maintained at 12 °C in a refrigerated incubator (Maine) or the cold room (Gainesville), and were calibrated with buffers in saline solutions with ionic concentrations similar to either Ringer’s solution or sea water, respectively. Perfusate and irrigate NH₃ vs NH₄ concentrations were calculated by the Henderson-Hasselbalch equation, using a pK of 9·75, according to the calculations of Cameron & Heisler (1983). Perfusate contained 5 × 10⁻⁵ mol I⁻¹ bumeta-nide plus 10⁻⁴moll⁻¹ ouabain in these experiments to inhibit basolateral uptake steps (i.e. Na⁺+NH₄⁺+2Cl⁻ cotransport and Na⁺/NH₄⁺ exchange) which could potentially play a role in the NH₄⁺-stimulated ammonia efflux. Amiloride was not added to the irrigate in these experiments because the perfusate NH₃ concentration never exceeded that used in the control (amiloride) experiments (see above), which determined that amiloride had no effect on ammonia transport (see Results). Since we found that the NH₃ permeability of the pup gill epithelium was relatively high (see Results), we corrected for changes in the NH₃ gradient across the gill produced in the NH₄⁺-stimulation experiments, in order to separate NH₃ from NH₄⁺ effects. Similar corrections were not necessary in the NH₃-stimulation experiments because of the relatively low NH₄⁺ permeability of the pup gill, and the relatively small changes in the NH₄⁺ gradients.

To assess the magnitude of ammonia loss via structural leaks, and as another test of the structural integrity of the perfused pup head preparation, most experiments included a final perfusion period with 2·3×l0⁻⁴moll⁻¹ erioglaucine (acid blue no 9, M_r783) in the ERS. The efflux of dye over the 20-min flux period was determined by measuring its concentration in T₀ and T₂₀ samples of the irrigation solution at 532 nm with a Beckman model 34 spectrophotometer. Percentage leak per 20 min was computed by comparing the change in absorbance of these samples of the experimental solutions with a standard curve prepared by adding known volumes of perfusate containing erioglaucine to 50ml of irrigation solution.

Since the flux of NH₄⁺ could be affected by the transepithelial potential (TEP) across the gill epithelium, and as a further test of the integrity of this preparation, we monitored the TEP in a separate series of experiments in which perfusate and irrigate substitutions were made which mimicked those used in the ammonia efflux experiments. The TEP was measured via 2% agar/3 mol I⁻¹ KC1 bridges placed in the irrigation solution surrounding the gills, and in the perfusate leaving the head. Stable potentials were only possible with the irrigation and perfusate flow momentarily stopped. The measuring bridges were connected to calomel electrodes in 3 mol I⁻¹ KC1, and the potential between the electrodes was measured with a Keithley model 616 digital multimeter. Asymmetry and tip potentials, measured by placing the bridges into beakers containing irrigate and perfusate connected with a separate agar-KCl bridge, were subtracted from the TEPs measured across the gills. In another series of experiments, we measured the TEPs across the gills when the seawater irrigation solution was replaced with Na⁺-free or Cl⁻-free artificial seawater solutions (see Claiborne & Evans, 1984, for solution formulation).

Adrenaline was manufactured by ESI Pharmaceuticals, amiloride and ouabain were obtained from Sigma Chemical Co., and bumetanide was kindly supplied through Dr Rolf Kinne by Leo Pharmaceutical Products.

Experimental results are given as means ± standard error (TV) and statistical differences between experimental means were determined by Student’s t-test (using paired data when appropriate). Calculations of slopes for NH₃-stimulated and NH₄⁺-stimulated ammonia effluxes, and corrections for changes in the NH₃ gradient in the latter experiments were performed using Multiplan (Microsoft) on an Apple He computer.

The surface morphology of representative gills taken from unperfused and perfused pups is shown in Fig. 1 and it is apparent that the gill lamellae of this near-term foetus are somewhat thicker than those found in the adults (De Vries & De Jager, 1984). Earlier studies (Evans et al. 1982; Kormanik & Evans, 1986) have demonstrated that the pups are capable of prolonged survival in sea water (either intra utero or outside the mother) and display normal blood ionic and acid-base parameters, so it is clear that these gills are functional in gas exchange and in ion and pH regulation. Comparison of the unperfused with the perfused gill indicates no obvious structural abnormalities (e.g. oedema) even after perfusion periods of approximately 2–3 h.

The concentration-response curve for the sensitivity of the perfused pup head branchial vasculature to adrenaline is shown in Fig. 2. Our experimental protocol does not permit accurate determination of the actual EC50 of this response, but it is clearly below 10⁻⁷moll⁻¹, especially since 10⁻⁵moll⁻¹ adrenaline reduces the afferent pressure to some 60% of the control (Evans & Claiborne, 1983).

The transepithelial electrical potential (TEP) measured between the perfusate and the irrigate of the perfused pup head was +l·9±0·6mV (N = 6; perfusate relative to irrigate) during irrigation with sea water. Subsequent irrigation with sea water in which the Na⁺ had been replaced with choline depolarized the TEP to –1·0 ±0·5 mV (N = 6, P<0·01 compared with sea water, SW). When the head was irrigated with Cl⁻-free sea water (benzensulphonate) the TEP repolarized to +5·7 ±0·5 mV (N = 6; P<0·01 compared with SW and Na⁺-free SW). Since Na⁺ and Cl⁻ concentrations in the sea water were approximately equal (≈ 500 mmol I⁻¹) it appears that Cl⁻ conductance is greater than Na⁺ conductance across the perfused shark pup head.

The erioglaucine ‘leak’ for the 53 perfused heads used in this study was 0·66 ±0·12% of the perfusate for 20min, indicating that structural leaks (and erioglaucine permeability) were vanishingly small across this preparation. In the current study, only two perfused heads displayed erioglaucine leaks greater than 3%, and data from these were not included in the results.

At the end of each experiment the gills were removed and examined under a dissecting microscope at 250× for completeness of perfusion. This examination was facilitated by the presence of the blue, erioglaucine dye. Perfusion was usually complete, but in a very few experiments the tips of a few lamellae remained red due to trapped erythrocytes. Thus perfusion was considered to be greater than 99%.

To establish the consistency of efflux over the time period utilized in subsequent experiments, we monitored ammonia efflux during three, sequential 20-min periods separated by the usual 10-min wash period (Table 1). Efflux fell slightly (17%; P<0·02) in the second period, but did not decline further in the third period. These perfused heads also displayed a relatively stable afferent pressure over the course of the 90-min experiment (Table 1).

The roles of basolateral Na⁺+NH₄⁺+2Cl⁻and Na⁺/NH₄⁺transport

To examine whether basolateral Na⁺+NH₄⁺+2Cl⁻ and Na⁺/NH₄⁺ transports are involved in ammonia efflux across the pup gill we monitored the effect of addition of bumetanide and ouabain + bumetanide to the perfusate after an initial, control flux period (Table 2). Whereas bumetanide inhibited the efflux by 34% (P<0·05), subsequent addition of ouabain in the presence of bumetanide did not inhibit the ammonia efflux further. Bumetanide did not affect the afferent pressure, but ouabain, in the presence of bumetanide, did increase the pressure slightly above the control period (+4·3 ± 1·6 mmHg; 1 mmHg = 133·3 kPa; Table 2). In an effort to quantify the potential effect of such an increase in afferent pressure, we examined the relationship between control ammonia fluxes and measured afferent pressures in heads which had not been treated with either bumetanide or ouabain (Fig. 3). There appeared to be a slight negative relationship, but the correlation was not significant (P> 0·10). Even if it were, a 4 mmHg change in pressure would only result in a fall in ammonia efflux of less than 1 μmol 100 g⁻¹ h⁻¹, certainly too small to be significant in the present experiments. These data are consistent with the proposition that slight changes in afferent pressure in the perfused head are not associated with significant changes in ammonia efflux.

To examine the importance of apical Na⁺/NH₄⁺ exchange, we added ouabain to the perfusate during the second flux period of another set of experiments, and amiloride to the irrigate (in the presence of perfusate ouabain) during the third period. Neither treatment affected the ammonia efflux from the perfused pup head (Table 3). Treatment with ouabain once again increased the gill resistance (Table 3), as did amiloride, but the final afferent pressures were still within the control range displayed in Fig. 3.

Increasing the perfusate NH₃ concentration, exclusive of changes in the perfusate NH₄⁺ concentration, stimulated the total ammonia efflux significantly, with a slope of 0·56±0·121100g⁻¹ h⁻¹ [μmol 100g⁻¹h⁻¹ (μmoll⁻¹)⁻¹; N = 22] (Fig. 4). Specifically increasing the perfusate NH₄⁺ concentration also stimulated the total ammonia efflux (Fig. 5), but millimolar concentrations were necessary (note the abscissa) so that the slope of the NH₄⁺-stimulated ammonia efflux was 0·52 ± 0·15×10⁻³1100g⁻¹h⁻¹ (N=8), less than 0·1% of the slope of the NH₃-stimulated ammonia efflux. Clearly, the NH₃ permeability of the pup gill is significantly greater than the NH₄⁺ permeability. Neither manipulation of perfusate ammonia concentration affected the afferent perfusion pressures (data not shown).

None of the perfusate or irrigate manipulations in the experiments detailed above affected the TEP across the gills (data not shown).

Perfusion of the pup head for as much as 3 h does not affect the gross lamellar structure (Fig. 1). Thus, this preparation does not suffer from the obvious structural abnormalities described for the eel, A. australis, holobranch (Ellis & Smith, 1983), or the rainbow trout, S. gairdneri, head perfused with adrenaline-free Ringer’s solution (Part et al. 1982). More recently, it has been shown that the addition of at least 10⁻⁶moll⁻¹ adrenaline and bovine serum albumin (0·2% and 2%, respectively) maintained normal lamellar anatomy for 30–45min in the perfused trout head (Perry et al. 1984b; Bornancin et al. 1985), but these manipulations are not necessary with the perfused pup head.

The gill resistance of the perfused pup head is quite sensitive to the presence of adrenaline in the perfusate (Fig. 2), with an EC50 in the range of in vivo concentrations described for intact, exercising or hypoxic adult sharks (e.g. Butler et al. 1978, 1986). Since the perfused pup head retains functional cranial nerves, specific concentration-response curves for adrenaline are probably impossible to obtain because of reflex vasoconstriction subsequent to hormone-mediated vasodilation. Indeed, in most cases, we observed partial reversal of the adrenaline-mediated fall in gill resistance within minutes after adding a given concentration of the catecholamine to the perfusate (unpublished observations). We added 10⁻⁷ mol I⁻¹ adrenaline to the perfusate in subsequent experiments because it was the lowest concentration that usually elicited spontaneous ventilation by the head.

The transepithelial electrical potential (TEP) measured across the pup head was 6·3 mV more positive than the –4·4 mV measured in the intact pup (Evans et al. 1982). Only a very few measurements of TEP across perfused heads have been published. Claiborne & Evans (1984) found that the TEP across the head of the marine teleost M. octodecimspinosus was equivalent to that of the intact fish (+7·7 mV vs +7·2 mV), but Perry & Wood (1985) found that the TEP across the perfused trout head was –1·3 mV when perfused with irrigate containing 1·29mequivl⁻¹ Ca²⁺ and Kerstetter et al. (1970) measured +6 mV across the intact fish under similar (1 mequivl⁻¹) external Ca²⁺ conditions. One might argue that the TEPs measured across the perfused heads are better estimates of the true potentials than those measured in vivo via bridges in the peritoneal cavity and external medium; indeed, our unpublished observations indicate that changes in ionic composition of the irrigate produce nearly instantaneous changes in the TEP across the perfused head, but quite slow (minutes) TEP changes across the intact fish. However, if one assumes that the in vivo TEP values are the controls, then the data in Fig. 3 are consistent with the proposition that, like Myoxocephalus octodecimspinosus (Claiborne & Evans, 1984), the perfused pup head maintains a lower Cl⁻ conductance than normal. This proposition is supported by our finding that the depolarization produced by removal of Na⁺ from the irrigate was 2·9 mV, compared with 2·6mV in vivo; but the hyperpolarization produced by Cl⁻ removal was only 3·8 mV, compared with 13·5 mV in vivo (Evans et al. 1982). Despite this uncertainty about the ‘normality’ of the relative Cl⁻ and Na⁺ conductances across the perfused pup head, it is clear that measurements of the TEP are possible, and can be used to determine if ionic flux changes under other experimental conditions are secondary to changes in the electrochemical gradients for ionic species (see below).

Our preliminary experiments with the perfused pup heads determined that simple measurement of postbranchial perfusate outflow vs prebranchial inflow was sufficiently variable to preclude accurate estimation of the degree of structural leak of this preparation. Our data using erioglaucine to measure the sum of structural leak and erioglaucine permeability indicate that, under most circumstances (53 of 55 perfused heads), the leaks were minimal, less than 1%. Assuming that the actual branchial permeability to erioglaucine is zero, and an average perfusate ammonia concentration of approximately 400μmol I⁻¹, a perfusion flow of 650μmin⁻¹, and an animal mass of 50g, an erioglaucine leak of 0·7% per 20min would produce an apparent ammonia efflux of only 0·2μmol 100g⁻¹ h⁻¹, vanishingly small compared with the fluxes described below. It is therefore clear that the perfused pup head is structurally ‘tight’, and any leak of perfusate cannot account for more than approximately 1% of the measured ammonia efflux.

The data in Table 1 demonstrate that the ammonia efflux from the perfused pup head is relatively constant throughout the entire 90 min of perfusion. In addition, the gill resistance is stable because the afferent perfusion pressures do not change over this period. The control ammonia efflux from the perfused pup head is significantly higher than that described for the in vivo pup (2–10 μmol 100 g⁻¹ h⁻¹; Evans et al. 1982). Since structural leaks cannot account for this relatively high ammonia flux (see above), it appears most likely that it is secondary to increased gill surface area and/or increased gill permeability to ammonia. There are no data on the in vivo extent of lamellar perfusion in the elasmobranchs, but data from trout indicate that in vivo perfusion in this species is only of the order of 60%, but increases to 75% when adrenaline is added (Booth, 1979). Adrenaline also stimulates lamellar recruitment in the catfish, I. punctatus (Holbert et al. 1979). Visual inspection of our perfused heads indicated virtually complete replacement of blood with perfusate so it is likely that at least some of the increased ammonia efflux from the perfused head is secondary to supranormal lamellar perfusion and, hence, increased gill surface area. The presence of significant ammonia in perfusate without added NH₄C1 (secondary to contamination from urea and PVP) precluded a determination of the role of branchial cell deamination vs perfusate clearance in total ammonia efflux.

Although it has been generally accepted that ouabain-sensitive Na⁺, Reactivated ATPase can bind NH₄⁺ at the K⁺ site (e.g. Stekhoven & Bonting, 1981; Mallery, 1983), it is only recently that a sensitivity to NH₄⁺ at the K⁺ site of the bumetanide (furosemide)-sensitive Na⁺+K⁺+2C1⁻¹ cotransporter has been described (Kinne et al. 1986 a, b; O’Grady et al. 1987) in vesicles isolated from the shark rectal gland and medullary thick ascending limb of the rabbit kidney. In fact, Good et al. (1984) demonstrated a furosemide-sensitive ammonia reabsorption in the thick ascending limb, but suggested that changes in the TEP secondary to this inhibition (e.g. Greger & Schlatter, 1984) could have altered diffusive movements of NH₄⁺, rather than transport of the cation via the cotransporter. Ammonia is transported across the pup gill epithelium, at least in part, via a basolateral, bumetanide-sensitive system (Table 2), consistent with the proposition that a basolateral Na⁺+NH₄⁺+2C1⁻ transporter is involved. However, the lack of a basolateral effect of ouabain (Tables 2 and 3) indicates that Na⁺, K⁺-activated ATPase is not directly involved. Neither inhibitor affects the gill resistance (Table 2) or TEP (data not shown) enough to obscure a direct effect on transport. Our finding that ammonia transport across the pup gill is insensitive to ouabain does not corroborate our previous data (Claiborne et al. 1982) on the perfused head of the marine teleost, O. beta. In addition, these earlier studies showed that increasing the perfusate K⁺ concentration also inhibited ammonia efflux. Ouabain can inhibit the cotransport system indirectly by abolishing the Na⁺ electrochemical gradient which drives the cotransport system (e.g. Frizzell et al. 1979; Greger & Schlatter, 1984), and perfusate K⁺ could also compete with NH₄⁺ for the cotransporter. However, our recent unpublished data indicate that ammonia transport across the O. beta gill is indeed sensitive to perfusate ouabain, but insensitive to perfusate bumetanide, so it appears that basolateral Na⁺/NH₄⁺ but not Na⁺+NH₄⁺+2C1⁻ transport is involved in this marine, teleost species.

Our experiments do not, of course, allow estimates of the relative affinity of the Na⁺+NH₄⁺+2C1⁻ cotransporter in the pup gill for NH₄⁺ vs K⁺. Kinne et al. (1986b) found relative affinities for the cotransporter in the thick ascending limb of the rabbit kidney for NH₄⁺:K⁺ of 1:5. The relative perfusate concentrations in our experiments were approx. 1:10 (400μmoll⁻¹ NH₄⁺ vs 6 mmoll⁻¹ K⁺). If we make the simplifying assumption of similar affinities in the dogfish pup gill and the rabbit thick ascending loop, it appears that NFL₄⁻ has only a 1 in 50 chance of binding to the cotransporter in our experiments. One might suggest, therefore, the bumetanide-sensitive NH₄⁺ transport is merely ‘sloppiness’ in the transport system, rather than a specific mode of extrusion. Data to be discussed below are consistent with this conclusion.

Our finding of a bumetanide-sensitive carrier on the basolateral aspect of the pup gill is of much more general interest than merely ammonia transport. Despite the rectal gland’s obvious role in salt secretion (Evans, 1979), various experiments have shown that osmoregulation continues in elasmobranchs which have had the rectal gland ligated or removed (Chan et al. 1967; Haywood, 1975; Evans et al. 1982), supporting the proposition for another site for salt secretion. Cells have been described in the gill epithelium of elasmobranchs (Doyle & Gorecki, 1961; Laurent & Dunel, 1980; Laurent, 1984) which are similar to the chloride cells found in teleosts. Since chloride cells in teleosts are generally considered to be the site of active secretion of Cl⁻via a basolateral Na⁺+K⁺+2C1⁻ cotransporter (Zadunaisky, 1984), it is not surprising that we have found a bumetanide-sensitive transporter in the gill epithelium of the dogfish pup. Nevertheless, it is the first physiological evidence for this putative carrier in an elasmobranch epithelium. A study of the effects of perfusate bumetanide on Na⁺ and Cl⁻ efflux from the perfused pup head would be most interesting.

Previous studies on teleosts have demonstrated that ammonia efflux from intact fishes is at least partially dependent upon external Na⁺ (Evans, 1977, 1982) and Payan (1978) found that ammonia efflux from the perfused trout head was also inhibited by removal of Na⁺ from the irrigate, implicating apical Na⁺/NH₄⁺ exchange in ammonia extrusion. However, Evans & Cameron (1986) have recently pointed out that inhibition of apical Na⁺/H⁺ exchange, with a concomitant reduction in the NH₃ diffusion gradient, could account for these results. Moreover, removal of apical Na⁺ has been shown to inhibit the paracellular, diffusive pathway for Na⁺ across the chloride-cell-rich opercular epithelium of the teleost Fundulus heteroclitus (Zadunaisky, 1984). So, it is quite possible that removal of external Na⁺ can affect diffusive, rather than carrier-mediated, movements of ammonia. Interestingly, removal of external Na⁺ has only a slight, and non-reversible, effect on ammonia efflux from intact dogfish pups (Evans, 1982), and no effect on ammonia efflux from the intact skate, Raja erinacea, (Evans et al. 1979), despite complete, and reversible, cessation of acid efflux in Na⁺-free artificial sea water in both species. These data are certainly consistent with the supposition that, not only is apical Na⁺/NH₄⁺ exchange not involved in elasmobranch ammonia extrusion, but that extrusion is also not affected by gradients produced by apical Na⁺/H⁺ exchange or by the control of paracellular, cation-selective diffusive pathways by external Na⁺.

Amiloride (5 × 10⁻⁵-10⁻⁴ mol I⁻¹) is effective in inhibiting ammonia efflux when added to the medium external to the trout (Kirschner et al. 1973; Payan, 1978; Wright & Wood, 1985), but is not effective (at 10⁻⁴moll^−l) in inhibiting ammonia efflux from the skate, Raja erinacea, in sea water (Evans et al. 1979). Since amiloride has been shown to be effective (at such high concentrations) in inhibiting Na⁺/H⁺ exchange, which can also include Na⁺/NH₄⁺ exchange (Kinsella & Aronson, 1981; Aronson, 1985), it appears from these data that the trout, but not the skate, is able to transport NH₄⁺via an apical Na⁺/NH₄⁺ exchanger. However, once again, one cannot rule out indirect effects of amiloride via changes in the diffusional gradient for NH₃ produced by inhibition of Na⁺/H⁺ exchange. Amiloride experiments have another complicating factor: amiloride may be a competitive inhibitor for Na⁺ transport via channels or Na⁺/cation exchange, so extremely high concentrations of amiloride (approx. 10⁻⁴–10⁻³moll⁻¹) must be used in Ringer’s solutions (e.g. Henderson et al. 1987; Moran, 1987) or sea water (e.g. Knakal et al. 1985) which contain approximately 150 and 500mmoll⁻¹ Na⁺, respectively. At these concentrations amiloride has been shown to block paracellular cation channels (Balaban et al. 1979) and even basolateral Na⁺,K⁺-activated ATPase (Soltoff & Mandel, 1983), both of which may play a direct or indirect role in ammonia efflux (present study; Evans & Cameron, 1986).

Given the potential inhibition of basolateral Na⁺,K⁺-activated ATPase by the high concentration of amiloride used in our experiments, we perfused the gills first with ERS containing ouabain, and then ouabain plus amiloride, in an attempt to cancel out basolateral events. Changes in the gill resistance in these experiments were slight, and ouabain did not, in itself, inhibit ammonia transport (Table 3). After correction for the inhibition of the phenol-hypochlorite reaction, it can be seen that 10⁻³moll⁻¹ amiloride applied to the apical surface of the pup gill does not inhibit ammonia efflux (Table 3). This is certainly consistent with the conclusion that apical Na⁺/NH₄⁺ exchange does not play a measurable role in ammonia transport across the gill of the shark pup, although our data cannot exclude the possibility that 10⁻³moll⁻¹ amiloride was ineffective because of competition with the 500 mmol I⁻¹ Na⁺ concentration in the irrigate.

Since we can find no evidence for transport of NH₄⁺via Na⁺/NH₄⁺ exchange across the apical membrane we are unsure of the mechanism whereby the NH₄⁺ which enters the basolateral aspect of the branchial cell via Na⁺+NH₄⁺+2C1⁻ cotransport (see above) leaves from the apical aspect of the cell into the sea water. NH₄⁻ has a hydrated radius similar to that of K⁺ (Kieland, 1937), and NH₄⁺ has been shown to traverse K⁺ channels in some tissues (see Zeiske & Van Driessche, 1983), so it is certainly possible that the NH₄⁺ which enters the cell on the basolateral aspect leaves via putative potassium channels either on the apical surface directly into the sea water, or on the basolateral surface and subsequently via the paracellular pathway, which may be conductive to NH₄⁺ (see below).

Until very recently it was generally assumed that, because of its lipid solubility, NH₃ was able to diffuse freely across biological membranes (e.g. Pitts, 1973; Valtin, 1983). However, recent determinations of the solubilities of NH₃ in lipid solvents such as chloroform and olive oil (J. N. Cameron, G. A. Kormanik & L. Goldstein, unpublished data), have confirmed a much older study which determined a chloroform: water partition coefficient of NH₃ of only 0·04 (Bell & Field, 1911), far below that of substances considered to be lipid-soluble. Nevertheless, Cameron & Heisler (1983) demonstrated that the net movement of ammonia followed the NH₃ partial-pressure gradients across the trout gill. Interestingly, in an earlier study of ammonia transport across the perfused heads of two marine teleosts (O. beta and M. octodecimspinosus), Goldstein et al. (1982) could find no evidence for NH₃ permeance, but specific NH₃ gradients were not measured, and our subsequent study of at least O. beta has demonstrated a significant NH₃ permeability (D. H. Evans & K. J. More, in preparation). Goldstein et al. (1982) described a significant NH₄⁺ permeance across the gills of both teleost species, and this ionic permeability to NH₄⁺ has been corroborated in more recent studies on the turtle bladder (Schwartz & Tripolone, 1983) and the rabbit renal proximal straight tubule (Garvin et al. 1987).

The present experiments indicate that the pup gill is permeable to both NH₃ and NH₄⁺ (Figs 4, 5). The actual site(s) of these permeabilities is unknown, but may be paracellular because of the charge on NH₄⁺ and possible low lipid-solubility of NH₃ (see above). Using the slopes of the respective experiments we can calculate that the NH₃ permeability is some 1100 times the NH₄⁺ permeability. If we assume that the functional surface area of the perfused pup gill is of the same order as the structural surface area of the gill of the adult of the same species (3·7 cm² g⁻¹; Hughes & Morgan, 1973), we can calculate the respective ‘apparent’ permeabilities and compare them with the few similar calculations for other epithelia (Table 4). It is clear that the apparent permeability of the pup gill to NH₃ is significantly below that described for the rabbit proximal straight tubule, but of the same order as that found for the turtle bladder. The relative NH₄⁺ permeability of the pup gill is only 10% that of the turtle bladder and only 1% that of the proximal tubule. One is tempted to correlate these relative permeabilities of the rabbit and turtle tissues with the evidence that the turtle bladder is a ‘tight’ epithelium, whereas the proximal tubule is considered to be ‘leaky’ (e.g. Fromter & Diamond, 1972; Erlij & Martinez-Palomo, 1978). Intact sharks certainly have a very low rate of Na⁺ efflux, especially when compared with marine teleosts, and therefore their gill epithelium is generally considered to be rather impermeable to ions (Evans, 1979). Our data on apparent ammonia permeabilities support this supposition, but firm statements are only possible when true, functional surface areas are known for complex epithelia like those of the gill.

The data in Fig. 5 were generated with ouabain and bumetanide in the perfusate to obviate any stimulation of basolateral transport events by the increased perfusate NH₄⁺ concentration. In four experiments, the perfusate did not contain either transport inhibitor and the slope of the NH₄⁺-stimulated ammonia efflux was 0·67 ± 0·21 ×10³1 100g⁻¹ h⁻¹, not significantly different from that described for the slope of the NH₄⁺-stimulated ammonia efflux when basolateral transport steps were inhibited. This indicates that increasing the perfusate NH₄⁺ concentration by approximately 10-fold did not stimulate the bumetanide-sensitive cotransporter, supporting the notion that the bumetanide-sensitive component may be merely sloppiness in the transporter, rather than a directed, controlled transport of NH₄⁺ (see above).

Given the data presented here, and assuming an ammonia pK of 9·75 for elasmobranch Ringer’s solution at 12°C (Cameron & Heisler, 1983), we can calculate the relative roles played by the various, putative transport pathways across the shark pup gill epithelium. At the pH of the perfusate (= shark blood ≈ 7-8) the concentration ratio of the two species is 0·011:1·0 (NH₃:NH₄⁺). Since the actual relative efflux is the product of the relative permeability and the relative concentration (or partial pressure for a gas), the actual ratio of the effluxes for NH₃:NH₄⁺ is 12:1, despite the 1100-fold difference in their respective permeances across the gill epithelium. Since bumetanide inhibited the ammonia efflux by 17% (below the second control period), we can calculate that the other 83% of the total efflux is diffusion of NH₃ and NH₄⁺. Fig. 6 presents a summary of these calculations and indicates that, despite the dominance of non-ionic diffusion of NH₃, basolateral Na⁺+NH₄⁺+2C1 cotransport and ionic diffusion of NH₄⁺ play measurable roles in the transport of ammonia across the gill epithelium of the pup of Squalus acanthias.

This research was supported by NSF PCM 8302621 to DHE. Portions have been published in abstract form in the Bulletin of the Mount Desert Island Biological Laboratory, and in Federation Proceedings. The technical assistance of Craig Hooks, Steven Robbins, Andrew Evans, Karl Weingarten and John Payne during various phases of this study is greatly appreciated. Figures were drawn by Daryl Harrison.