The Distribution of Ammonia and H⁺ Between Tissue Compartments in Lemon Sole (Parophrys Vetulus) at Rest, During Hypercapnia and Following Exercise

There is a substantial body of knowledge on the mechanisms of ammonia excretion across the gills of various fish (e.g. Maetz & Garcia-Romeu, 1964; Maetz, 1973; Evans, 1977; Payan, 1978; Cameron & Heisler, 1983; Wright & Wood, 1985), but little is known about ammonia stores in body tissues of fish. Ammonia is a weak base that exists in solution as ammonia (NH₃) and ammonium ions (NH₄⁺). (The term ammonia or T_amm will be used to indicate the total ammonia concentration, while NH₄⁺ and NH₃ will refer to ammonium ion and non-ionic ammonia, respectively.) With a pK of 9·7 at 10°C, the majority of ammonia in body fluids (pH = 6·5–8·0) will be in the ionized form, as NH₄⁺. Movement of ammonia between tissue compartments is thought to be primarily dependent on NH₃ partial pressure gradients ⁠, because biological membranes are highly permeable to NH₃ (Castell & Moore, 1971; Klocke, Andersson, Rotman & Forster, 1972; Boron, 1980; Lockwood, Finn, Campbell & Richman, 1980; see also Good & Knepper, 1985). Despite the greater permeability of NH₃, there is evidence that NH₄⁺ electrochemical gradients also play a role in ammonia transfer across cell membranes (Thomas, 1974; Boron & DeWeer, 1976; Aickin & Thomas, 1977). Transfer of ammonium ions will depend on the diffusive permeability of the membrane for NH₄⁺, the availability of ion carriers in the membrane, and their affinity for NH₄⁺. A significant diffusive permeability to NH₄⁺ is now recognized in many biological membranes (Schwartz & Tripolone, 1983; Arruda, Dytko & Withers, 1984; Evans & Cameron, 1986). Ammonium ions can substitute for K⁺ in the Na⁺/K⁺/Cl⁻ cotransporter (Kinne, Kinne-Saffran, Schutz & Scholermann, 1986) andin the Na⁺,K⁺-ATPase pump (e.g. Post & Jolly, 1957; Skou, 1960; Robinson, 1970; Aickin & Thomas, 1977; Sorensen, 1981; Kurtz & Balaban, 1986), and for H⁺ in the Na⁺/H⁺ ion exchange mechanism (e.g. Kinsella & Aronson, 1981; Aronson, 1983). Also, NH₄⁺ can replace K⁺ in nerve K⁺ channels (Binstock & Lecar, 1969; Hille, 1973). Thus, it appears that transfer of ammonia between tissue compartments may involve both NH₃ and NH₄⁺ movements.

In a closed system at equilibrium, the distribution of ammonia across biological membranes will be determined by the pH gradient across the membrane, as long as the effective permeability to NH₃ is much greater than that to NH₄⁺, and there is no chemical binding of the species. However, the greater the effective NHt⁴” permeability, the greater will be the influence of the electrical gradient on the equilibrium distribution. Living animals, especially ammoniotelic fish, are certainly not closed systems at equilibrium. Nevertheless, the assumption has often been made that the distribution of ammonia is largely a function of the intracellular to extracellular pH gradient, in both mammals (e.g. Visek, 1968; Pitts, 1973; Meyer, Dudley & Terjung, 1980; Mutch & Banister, 1983; Remesy, Demigne & Fafournoux, 1986) and fish (Randall & Wright, 1987; Dobson & Hochachka, 1987).

In a recent study, Wright, Wood & Randall (1988) investigated the distribution of ammonia between plasma and red cells of rainbow trout, Salmo gairdneri, and found that ammonia was passively distributed according to transmembrane pH gradients under normal conditions both in vitro and in vivo. Red cells are unusual compared with many other tissues in that the hydrogen ion distribution is passive and follows the membrane potential over a range of pH values (Heming et al. 1986; Heming, Randall & Mazeaud, 1987; Nikinmaa, Steffensen, Tufts & Randall, 1987; Lassen, 1977). Thus, it was impossible with the trout data to determine whether the distribution of ammonia was solely dependent on pH gradients or if electrical gradients were also important. To distinguish between the two it is necessary to look at tissues other than red cells, where H⁺ is not passively distributed (see Roos & Boron, 1981). If ammonia movements across cell membranes are solely dependent on gradients, that is, NH₄⁺ movements are negligible, then one would expect the distribution of ammonia to follow transmembrane pH gradients. However, if NH₄⁺ is permeant, then one would expect the ammonia distribution to follow electrochemical NH₄⁺ gradients. In the present study, we investigated the distribution of ammonia and H⁺ across cell membranes in white muscle, heart and brain, as well as red blood cells. Measurements were made at rest, to establish the situation under conditions as close to steady state as possible, and under two treatments designed to perturb the normal acid-base status by very different mechanisms - respiratory acidosis (hypercapnia) and metabolic acidosis (exhaustive exercise). Lemon sole, Parophrys vetulus, were chosen for these experiments as flatfish have a discrete white muscle mass uncontaminated by pink or red fibres, in contrast to salmonids (Mosse, 1979), and because their blood pH is somewhat lower (by about 0·3 units) than in trout. Instructive comparisons could therefore be made with our previous study on ammonia distribution in trout blood (Wright et al. 1988), as well as with a previous theoretical analysis of ammonia distribution in trout muscle (Randall & Wright, 1987).

Lemon sole (Parophrys vetulus) weighing 292 ±14 g (mean ± IS.E.M.) were collected by two 15-min otter trawls in Barkley Sound, British Columbia. Fish were held at the Bamfield Marine Station in a sandy bottomed, outdoor, fibreglass tank supplied with flow-through sea water (pH approx. 7·8, salinity approx. 31 %o, temperature 11 °C) for at least 7 days before experimentation. Sole were not fed during captivity, so that variations in body ammonia levels due to feeding history would be eliminated (Fromm, 1963; Brett & Zala, 1975).

Sole were fitted with caudal artery cannulae (PE 50) after anaesthetization in MS 222 (1:15 000), as described by Watters & Smith (1973). A small amount of oxytetracycline-HCl, a general antibiotic, was placed in the caudal incision before the wound was closed with surgical silk. After surgery, the sole were placed in small plastic chambers (approx. 81), covered with sand, and allowed to recover in flowing sea water for 72 h prior to experimentation. To re-establish resting conditions, flatfish require relatively long recovery periods and the availability of sand in which to bury (Wood, McMahon & McDonald, 1979).

Three experimental conditions were studied, (i) A control in which fish were subjected to unaltered, aerated sea water (pH 7·8) prior to sampling, (ii) A second group of fish were placed in a seawater bath equilibrated to approximately 1 % CO₂ (pH 6·7) for 30 min before sampling to alter rapidly the extracellular and intracellular pH via high levels (respiratory acidosis), (iii) The third group of fish were transferred to a large plastic tank and chased for 20min. The aim of subjecting fish to exhaustive exercise was to induce a metabolic acidosis in both extracellular and intracellular compartments. These fish were allowed to recover for 30 min prior to sampling.

To determine intracellular pH (pHi) by the DMO (5,5-dimethyl-2,4-oxazoli-dinedione) distribution technique (Waddell & Butler, 1959), fish were injected with 1mlkg⁻¹ of 5 μCiml⁻¹ [¹⁴C]DMO (New England Nuclear, specific activity 50mCimmol⁻¹) and 20 μCiml⁻¹ of the extracellular marker, [³H]mannitol (New England Nuclear, specific activity 27·4mCi mmol⁻¹) in CortLond saline (adjusted to 160 mmol l⁻¹ NaCl) approximately 12 h prior to sampling (Milligan & Wood, 1986a,b).

At sampling, 2 ml of blood was withdrawn from each fish and replaced with an equal volume of saline. Blood was immediately analysed for pHe, pHi, plasma and whole blood total carbon dioxide content ⁠, and haematocrit (Het). Samples were appropriately fixed and stored for later determination of whole blood and plasma ammonia concentration (T_amm), whole blood lactate ([lactate]) and haemoglobin ([Hb]) concentrations and plasma and red cell water content. Less than 5 min after blood withdrawal, fish were removed from the water and the spinal cord was quickly severed with a scalpel. Epaxial white muscle samples were excised and a thin slice of tissue was immediately frozen with freeze-clamp tongs, cooled by immersion in dry ice chips. The frozen samples were stored on dry ice for later determination of intracellular T_amm and [lactate], whereas the unfrozen tissue samples were later analysed for [¹⁴C]DM0 and [³H]mannitol. The heart ventricle was then quickly removed. Part of the ventricle was freeze-clamped for subsequent analysis of intracellular T_amm, while the remaining tissue was analysed for [¹⁴C]DMO, [³H]mannitol and water content. Brain tissue was then collected after cutting through the cranium and, as before, part of the tissue was immediately frozen and later assayed for intracellular T_amm, whereas the rest of the tissue was reserved for determination of [^l4C]DMO, [³H]mannitol and water content. The entire tissue excision and freeze-clamping procedure took approximately 1-–2 min. At the end of this procedure, large samples of white muscle tissue were taken for water content analysis.

Arterial pHe was measured immediately upon collection using a Radiometer microelectrode (type E5021) maintained at 11 °C and linked to a Radiometer PHM71 acid-base analyser. Plasma and whole blood were measured on 50-μ1 samples by the technique of Cameron (1971), using a Radiometer E5036 CO₂ electrode and the same acid-base analyser. Plasma and [HCO₃⁻] were calculated by standard manipulations of the Henderson-Hasselbalch equation and tabulated values for pK ′ and (Boutilier, Heming & Iwama, 1984). Red cell pHi was directly measured by the freeze-thaw technique of Zeidler & Kim (1977) using the same microelectrode assembly. The freeze-thaw method was used instead of the DMO method to measure red cell pHi, because the former technique is a more direct method for pHi determination of isolated cells, it is a simpler method, and there are no significant differences between red cell pHi values from the two techniques in fish blood (Milligan & Wood, 1985). Plasma and red cell water content were calculated by drying to constant weight in an oven at 100°C.

Lactate levels were assayed enzymatically (L-lactic dehydrogenase/NADH method, Sigma reagents) after deproteinizing 100 μl of whole blood in 200 μl of 8 % HClO₄, or homogenizing 100 mg of white muscle in 1 ml of 8 % HC1O₄ with a glass homogenizer (see Turner, Wood & Clark, 1983). Muscle intracellular [lactate] was expressed per litre of intracellular fluid (ICF) as outlined for muscle intracellular T_amm below (equation 9). Haematocrit was determined by centrifugation (5000g for 5 min) and [Hb] using the cyanmethaemoglobin method of Blaxhall & Daisley (1973) using Sigma reagents.

Intracellular muscle, heart and brain T_amm were determined in the following manner. Frozen tissue samples were weighed, deproteinized in iced 8 % HC1O₄ (1:20 dilution), and homogenized (Tekmar Tissumizer with microprobe head). Samples were centrifuged and the supernatant was extracted and neutralized with saturated Tris buffer. The enzymatic assay technique was identical to that described above for blood, except that muscle tissue extract provided the reference medium for internal standards. (There were no differences between internal standards made up in muscle, heart and brain tissue extracts.)

Muscle, heart and brain levels of [³H] and [¹⁴C] radioactivity were measured by digesting 50 –150 mg of tissue in 2 ml of NCS tissue solubilizer (Amersham) for 1-2 days until the solution was clear. Extracellular fluid [³H]mannitol and [¹⁴C]DMO activity were determined in the same manner, except 100 μl of plasma was added to 2 ml of tissue solubilizer. The tissue digests were neutralized with 60 μl of glacial acetic acid and 10ml of fluor (OCS; Amersham) was added. To decrease chemiluminescence, all samples were kept in the dark overnight before being counted on a liquid scintillation counter. Dual-label quench correction was performed using quench standards prepared from each type of tissue and the external standard ratio method (Kobayashi & Maudsley, 1974), as used by Milligan & Wood (1985, 1986a,b).

Tissue [NH₃], [NH₄⁺] and levels were calculated as described previously using equations 3, 4 and 5, respectively.

To determine whether the distribution of ammonia between extracellular and intracellular compartments in various tissues was according to the pH gradient, tissue pHi was predicted from the measured ammonia distribution using equation 6, assuming intracellular [NH₃] - plasma [NH₃], These predicted pHi values were compared with the measured values determined from the DMO distributions. A predicted intracellular T_amm could also be calculated using equation 7, again assuming intracellular [NH₃] = plasma [NH₃].

Data are presented as means ± 1S.E.M. (N). Student’s unpaired t-test was employed to evaluate the significance of differences between mean values (P<0·05).

The acid-base status and haematological variables of the three experimental groups are given in Tables 1 and 2. pHi was by far the highest in brain with a value approximately equal to pHe (Table 1B). Red cells, white muscle and heart ventricle all maintained a pHe-pHi gradient of 0·3–0·4 units. Hypercapnia resulted in lower pHi values in all tissues compared with the values in control fish, with greater decreases in brain and white muscle (approx. 0·3 units) than in red cells or heart (approx. 0·2 units). The acidifying effect of exhaustive exercise was much more selective, with pHi significantly lower only in the working muscle, whereas the acid-base status of the other tissues was unaffected.

Muscle was the only tissue in which there were changes in fluid volume distribution between treatments. Following exhaustive exercise, ECFV was significantly lower (control, 0·0997 ± 0·0113, exercise, 0·0599 ± 0·0090 ml g⁻¹) and ICFV (control, 0·7174 ± 0·0129, exercise, 0·7552± 0·0097 mlg⁻¹) higher in white muscle, suggesting a net water flux into the cells, probably in response to the osmotic effect of the increased intracellular [lactate] (Table 1A).

The distribution of ammonia between plasma and tissues under control, hypercapnia and exercise conditions is presented for red cells in Table 3, white muscle in Table 4, heart ventricle in Table 5, and brain in Table 6. In hypercapnic fish, plasma ammonia levels were significantly lower than in the control group by about 25 %, although tissues showed generally smaller, non-significant reductions. In exercised fish, intracellular ammonia levels were elevated by about 55 % over control levels, an effect which was significant in all tissues except brain. Plasma ammonia levels were not significantly altered.

Red cell ammonia levels were consistently higher than plasma levels, resulting in ammonia concentration ratios of 0·22–0·29 (plasma-to-red cell, Table 3). Red cell pHi predicted from the plasma-to-red cell ammonia distribution was not significantly different from measured pHi in the control experiment, but during hypercapnia and exercise, predicted pHi was significantly less than measured pHi (Table 3). NH₃ was in equilibrium between plasma and red cells at rest (Fig. 1), but not during either of the extracellular acidoses (hypercapnia and exercise, Fig. 1). The gradient from red cell-to-plasma during hypercapnia was 10μTorr and following exercise about 30μTorr. Our calculations of levels assume an equilibrium between NH₃ and NH₄⁺ in the red cell as well as the plasma, because the NH₃→NH₄⁺ reaction is considered to be instantaneous (<50ms, Stumm & Morgan, 1981). If there is a NH₃ gradient from red cell to plasma and if H⁺ is passively distributed across red cell membranes, as commonly believed (Lassen, 1977; Heming et al. 1986; Nikinmaa et al. 1987), then there must also be an electrochemical gradient for NH₄⁺. As red cell NH₃ levels are elevated, NH₄⁺ will also be elevated as NH₃ molecules combine with H⁺ to form NH₄⁺. In hypercapnia and exercise regimes, therefore, there was a net diffusional gradient for both NH₃ and NH₄⁺ out of the red cell. This is illustrated by the values of and E_H+ calculated from the Nernst equation (equation 13) (Table 3). and E_H+ were similar under control conditions; however, during both acidotic treatments, E_H+ became less negative as expected (e.g. Lassen, 1977) whereas became more negative.

Ammonia distribution ratios between muscle and plasma (control ratio = 0·03, Table 4), heart and plasma (control ratio = 0·02, Table 5) and brain and plasma (control ratio = 0·06, Table 6) were considerably lower than those between red cells and plasma (ratios >0·20, Table 4). Ammonia was clearly not distributed among tissue compartments according to the H⁺ distribution. In all tissues and treatments, intracellular pHi calculated from the plasma-to-tissue ammonia distribution was significantly less (by 1·0–2·0 units) than pHi calculated from the [¹⁴C]DMO distribution (Tables 4–6). Also, predicted tissue T_amm was significantly less than measured tissue T_amm in each treatment (Tables 4–6). The tissue intracellular-to-extracellular gradients (Fig. 2) were an order of magnitude greater than red cell-to-plasma gradients (Fig. 1). As stated above for red cells, our calculations of P_NH3 assume an equilibrium between NH₃ and NH₄⁺ in each tissue compartment. The difference between red cells and other intracellular compartments, however, is that H⁺ distribution is not passive (Roos & Boron, 1981). Hence, NH₃ gradients can exist in the absence of NH4⁺ electrochemical gradients, in a situation where NH₄⁺ is passively distributed across the membrane, but H⁺ is not (see Discussion). Control P_NH3 gradients were smaller between muscle and plasma (410 μTorr) than between heart (630μTorr) or brain (640μTorr) and plasma. In all tissues, P_NH3 was significantly reduced during hypercapnia, largely because of intracellular acidosis, although intracellular T_amm did decrease slightly in all tissues. There were no significant changes in PNH₃ gradients after exercise (Fig. 2), although T_amm gradients from tissue to plasma were greatly increased in all tissues (Tables 4–6).

The equilibrium potentials for NH4⁺ (E_NH4+) across muscle, heart and brain cell membranes were calculated from the Nemst equation and are presented in Tables 4, 5 and 6, respectively. These values were 2–3 times larger than those in red blood cells (see Table 3). There were no significant differences in E_NH4⁺ between treatments.

The accuracy of red cell T_amm calculations depends largely on Hct, increasing with increasing Hct (equation 1). For instance, in control fish, the error in red cell Tamm ^may be as great as 7 % if Hct is out by 1 %, at the mean Hct value of 11 %. A 7 % error in measured red cell T_amm would not alter the conclusion that ammonia is distributed according to the H⁺ distribution across red cell membranes at rest. The significance of this problem diminishes in the hypercapnia (Hct = 16 %) and exercise (Hct = 17 %) treatments, where the error in red cell T_amm is at most 4 % for a 1 % digression in Hct. The potential for error in measured red cell T_amm calculations in the hypercapnia and exercise experiments is small relative to the large discrepancy between measured and predicted red cell T_amm. The potential for error in measured intracellular T_amm for muscle, heart and brain tissue (equation 9) is even smaller; for instance, for a 1 % digression in ICFV, the error in tissue T_amm is less than 2%. Thus, ICFV is not a critical variable in the calculation of intracellular T_amm.

The accuracy and limitations of the [¹⁴C]DMO/[³H]mannitol technique for intracellular pH measurements in fish have been assessed in detail previously (Milligan & Wood, 1985, 1986a,b) and will not be discussed in the present paper.

Resting blood acid-base parameters in lemon sole were very similar to those measured by McDonald, Walker, Wilkes & Wood (1982) on the same species. Hypercapnia resulted in a plasma acidosis that was typical of other saltwater species (Conger conger, Toews, Holeton & Heisler, 1983; Scyliorhinus stellaris, Heisler, Weitz & Weitz, 1976) when compared over the same period of time. The metabolic acidosis incurred after exhaustive exercise in lemon sole was typical of other benthic, sluggish fish swimming to exhaustion, with very low blood [lactate] despite high muscle [lactate] and a considerable blood metabolic acid load (Wood & Perry, 1985). In other fish, however, a respiratory component (⁠ elevation) often persists for 1–2h following exercise, which was not seen here. Hence, if lemon sole did suffer a respiratory acidosis in addition to the metabolic acidosis following exercise, full recovery must have occurred within 30min.

White muscle, brain and red cell pHi in resting lemon sole were identical to the values reported for another flatfish, the starry flounder (Platichthys stellatus), using the same DMO and freeze-thaw techniques (Wood & Milligan, 1987), although heart pHi was approximately 0·2 units lower. Exhaustive exercise in lemon sole resulted in a less marked intracellular acidosis in white muscle, and no acidosis in heart or brain tissue, in contrast to the effects on the starry flounder. This difference undoubtedly reflects the lack of a respiratory component in lemon. sole relative to the large elevation in flounder at a comparable time after exercise. Thus, in the sole, the intracellular acidosis was limited to the working muscle itself and was completely metabolic in nature, as signalled by the large elevation in [lactate] white muscle. We are aware of no comparable studies on the intracellular responses to hypercapnia in fish over this time period.

The results show that ammonia is distributed according to the pH gradient across red cells at resting blood pH values. These findings agree with our earlier study on in vitro and in vivo trout red cells (Wright et al. 1988) and a study on human red cells (Bone, Verth & Lambie, 1976). If NH₄⁺ transfer across red cell membranes is negligible, and ammonia distribution is solely dependent on gradients, then one would expect the red cell-to-plasma ammonia distribution to follow the H⁺ distribution, as for other weak acids and bases with impermeant ion forms. Red cells, however, are unusual in that H⁺ is passively distributed across the membrane according to the membrane potential (Lassen, 1977; Heming et al. 1986; Nikinmaa et al. 1987), unlike the situation in other tissues (see Roos & Boron, 1981). Therefore, exactly the same distribution would occur if the membrane was permeable only to NH₄⁺, and the ammonia distribution was set entirely by the membrane potential. We conclude that ammonia is passively distributed between red cells and plasma under control conditions, but we cannot determine wHcther ammonia flux across red cell membranes depends solely upon gradients or if NH₄⁺ electrochemical gradients are also important.

Ammonia is not distributed across red cell membranes according to the pH gradient during hypercapnia and exercise in sole, which agrees with data on trout red cells in vitro and in vivo during hypercapnia (Wright et al. 1988) and avian red cells in vitro over the pH range 7–8 (Bone et al. 1976). Bone et al. (1976) proposed that ammonia and DMO would distribute differently between the cytosol and the acidic nucleus, resulting in a discrepancy between pHi calculated from the DMO distribution versus the ammonia distribution. Roos & Boron (1981), however, reanalysed their data and found that intracellular compartmentalization could not fully account for the discrepancy between the two calculated pHi values. Wright et al. (1988) discussed three possible explanations why ammonia accumulation in red cells might be out of equilibrium with the pH gradient during an acidosis: intracellular ammonia synthesis, nonsteady-state conditions, and active uptake of NH₄⁺ by the cell membrane Na⁺,K⁺-ATPase pump. All three possibilities were experimentally eliminated in the in vitro situation, though it was not possible to perform similar tests in vivo. However, because of the similarity between the in vitro and in vivo data, the authors suggested that other factors must be involved, one of which could be active NH₄⁺ uptake by a mechanism other than Na⁺,K⁺-ATPase.

We now propose that one other explanation should also be considered. Red cell-to-plasma ammonia gradients during hypercapnia and exercise may be due to binding of ammonia to proteins (Barker, 1968) in the red cell. It seems unlikely, however, that differential binding would occur, that is, binding under acidotic conditions (hypercapnia and exercise) but not at control pH values. Furthermore, there is evidence that ammonia binds in some mammalian intracellular compartments (see below), but we are unaware of any reports of ammonia binding in red cells.

In summary, ammonia is passively distributed according to the plasma-to-red cell H⁺ distribution in blood at resting pH values, but not during an extracellular acidosis. We suggest that red cell T_amm levels are maintained above those expected from pH considerations by an active NH₄⁺ uptake mechanism other than Na⁺,K⁺-ATPase or, possibly, by binding of ammonia to intracellular proteins.

Ammonia was not distributed between plasma and muscle, heart or brain tissue according to the H⁺ distribution either at rest, during hypercapnia, or following exercise in lemon sole. These findings are consistent with studies on mammalian hepatocytes where, at physiological pH and ammonia concentrations, intracellular ammonia levels are much higher than predicted from the H⁺ distribution (Sainsbury, 1980; Remesy et al. 1986). Ammonia also accumulates in mammalian brain tissue (Hindfelt, 1975; Benjamin, 1982) and colon (Bown et al. 1975) in concentrations above those expected from extracellular ammonia and pH considerations. These studies and the present results indicate that ammonia distribution across tissue membranes cannot be described by the classic diffusion trapping model (Visek, 1968; Pitts, 1973), which predicts intra-to extracellular ammonia distribution according to pH gradients, assuming that only NH₃ is transferred across cell membranes and that NH4⁺ transfer is negligible. If we assume that NH4⁺ permeability is significant, then the distribution of ammonia at equilibrium will be influenced by the membrane potential (E_M) as well as the pH gradient. The greater the permeability to NH4⁺, the greater the influence of E_M.

Similar analyses have been performed with measured and predicted ammonia distributions in heart ventricle and brain, with the same conclusions, i.e. pNH₃/pNH₄⁺ is low and the distribution is almost entirely a function of E_M.

Calculated for resting heart ventricle was −90 ±3 mV (Table 5), in agreement with Jaeger’s (1962) measured E_M of −84 mV in Leuciscus rutilus ventricle, whereas calculated for resting brain tissue was —69 ±3 mV (Table 6), in close accord with the generally accepted value of -70 mV for nervous tissue (Hille, 1984). The similarity between calculated E_NH4+ values for muscle, heart and brain tissue and published E_M values implies that ammonia is passively distributed according to E_M in these tissues at rest, and that the permeability of tissue membrane to NH₄⁺ is significant.

It is possible that NH₄⁺ permeability is not as great as is suggested by this analysis and that intracellular binding of ammonia accounts for the high intracellular T_amm values. Wanders, Hoek & Tager (1980) presented evidence that a large fraction of intracellular ammonia was bound to mitochondrial protein in rat heart tissue, whereas smaller fractions of ammonia were associated with protein in skeletal muscle and brain. Thus, deproteinization of tissues during the extraction method would liberate this protein-bound ammonia and would result in an overestimation of free ammonia in the cytosol. We have calculated mean cellular Tamm levels in heart, muscle and brain of the sole, assuming that (1) the values given by Wanders et al. (1980) for bound ammonia in mitochondria of rat are valid for fish, (2) the mitochondrial volume is between 9 and 13 % (Hoek, Nicholls & Williamson, 1980), (3) the cytosol constitutes the major portion of the remaining volume, and (4) the T_amm levels in the cytosol are determined by extra-to intracellular pH gradients. T_amm levels calculated with these assumptions are only 8–13 % of our measured values. Hence, even if some fraction of cellular ammonia binds to mitochondrial proteins in fish, it cannot account for the high T_amm levels measured in the present study.

Resting levels of ammonia in sole white muscle measured in this study are between published values for cod, Gadus morhua (Fraser et al. 1966) and carp, Cyprinus carpio (Driedzic & Hochachka, 1976), but about 4-8 times higher than resting levels in rainbow trout (Dobson & Hochachka, 1987; T. Mommsen & P. Hochachka, unpublished data). In a recent review, Randall & Wright (1987) analysed the distribution of ammonia between trout white muscle and plasma, at rest and following exhaustive exercise (data taken from Mommsen & Hochachka, 1988; Milligan & Wood, 1986b). The analysis was based on the principle that ammonia was distributed across muscle cells according to the H⁺ distribution, assuming that muscle cells were essentially impermeable to NH₄⁺. From their calculations, ammonia was distributed according to the H⁺ distribution in trout, at rest, but not following exercise, where NH₃ gradients existed between muscle and blood. It seems odd that the ammonia distribution in resting trout white muscle apparently follows the H⁺ distribution, while in sole it does not. Possible explanations include the differences in acid-base status and muscle fibre composition alluded to in the Introduction. Another problem with comparing the trout and sole studies is that the trout muscle T_amm levels (Mommsen & Hochachka, 1988) and pH_i values (Milligan & Wood, 1986b) were obtained from two separate studies (see Randall & Wright, 1987). Before any firm conclusions can be drawn, muscle T_amm levels and pHi values must be determined on the same group of trout.

Following exercise, muscle ammonia levels increased by approx. 55 % compared with control values. An increase in white muscle ammonia levels with exercise is consistent with other studies on fish (Suyama, Koike & Suzuki, 1960; Fraser et al. 1966; Driedzic, 1975; Driedzic & Hochachka, 1976; Dobson & Hochachka, 1987) and is due to deamination of the adenylate pool (Driedzic & Hochachka, 1976; Dobson & Hochachka, 1987). In mammalian muscle, it is recognized that ammonia production and glycolysis may be functionally coupled in several ways. Ammonia may enhance glycolysis by activating phosphofructokinase, H⁺ production by glycolysis may in turn enhance ammonia production by activating AMP deaminase, and ammonia production (as NH₃) may be important in buffering the H⁺ produced by glycolysis (Lowenstein, 1972; Mutch & Banister, 1983; Dudley & Terjung, 1985). The quantitative importance of the latter has recently been discounted by the detailed study of Katz, Broberg, Sahlin & Wahren (1986) on working human leg muscle, because ammonia production was less than 4% of lactate production. However, in the lemon sole, 30min after exhaustive exercise, ammonia accumulation in white muscle (approx. 3·3 mmol l⁻¹, Table 4) was 20% of lactate accumulation (approx. 16·3 mmol l⁻¹, Table 1). Identical stoichiometry was reported by Dobson & Hochachka (1987) in trout white muscle after exhaustive exercise. Thus, in ammoniotelic teleost fish, deamination/gly-colysis coupling may be more important than in mammals, especially in limiting intracellular acidosis.

Ammonia levels also increased in heart (approx. 55 %) and brain (approx. 45 %) tissue following strenuous exercise, although it is unlikely that ammonia production can account for this. Heart is predominantly an aerobic tissue and therefore, unlike white muscle, the deamination of adenylates during exercise would be negligible. In nervous tissue, ammonia is known to be extremely toxic (see Randall & Wright, 1987), and high ammonia production rates are therefore improbable. In addition, it is doubtful that increased body ammonia levels after exercise are related to reduced ammonia excretion, because in other teleosts, ammonia excretion either increases or remains unchanged after strenuous exercise (Holeton, Neumann & Heisler, 1983; Heisler, 1984; Milligan & Wood, 1986a, 1987). We propose that elevated ammonia levels in heart and brain tissue after exercise are simply due to the passive uptake of ammonia from the increased plasma ammonia pool.

With hypercapnia there were no significant changes in muscle, heart and brain ammonia stores, although mean levels decreased in all tissues, as well as in the plasma. A reduction in total body ammonia stores may have been due to a decrease in ammonia production and/or an increase in ammonia excretion rates. The decrease in water pH (7·8 to 6·7) caused by hypercapnia would certainly favour increased diffusive loss of NH₃ across the gills. Ammonia excretion rates have been reported to increase with hypercapnia in some teleosts (trout, Lloyd & Swift, 1976; Perry, Malone & Ewing, 1987; carp, Claiborne & Heisler, 1984) and decrease in others (Conger conger, Toews et al. 1983; Opsanus beta, Evans, 1982). Based on these studies, it is difficult to predict how ammonia excretion would be affected by hypercapnia in lemon sole. Metabolism may be suppressed by high levels in cell culture (Folbergrova, MacMillan & Siesjo, 1972) and in whole animals (Bickler, 1986). It is likely, therefore, that a component of decreased T_amm levels in tissues is caused by suppressed metabolism during hypercapnia.

Our analysis of the distribution of ammonia between extracellular and intracellular compartments strongly suggests that muscle, heart and brain cell membranes are permeable to NH₄⁺. We assume that NH₃ permeability is also significant, because NH₃:NH4⁺ permeability ratios for other tissues are at least 5:1 (Castell & Moore, 1971; Bown et al. 1975; Kleiner, 1981; Schwartz & Tripolone, 1983; Arruda et al. 1984; Evans & Cameron, 1986). If this is indeed the case, why do P_NH, gradients (200-1000μTorr) exist between muscle, heart and brain tissues and plasma (Fig. 2)? Fig. 4 describes a model of NH₃ and NH₄⁺ movements across a hypotHctical tissue where NH₄⁺ is passively distributed according to E_M, intracellular NH₃ and NH₄⁺ are in equilibrium (interconversion reaction rate for NH₃→ NH₄⁺< 50 ms, Stumm & Morgan, 1981), and ammonia production is insignificant (Fig. 4A). NH₃ gradients will form in the intracellular compartment as NH₄⁺ enters the tissue and dissociates to form NH₃ and H⁺. The direction of this reaction will always be towards the formation of NH₃ and H⁺, because intracellular H⁺ levels are lower (i.e. pHi higher) than expected for a passive distribution (Roos & Boron, 1981). The rate of NH₃ ‘back diffusion’ out of the tissue will depend on the magnitude of the gradient and the NH₃ permeability (pNH₃), but NH₄⁺ will diffuse back into the intracellular compartment down the electrochemical gradient. The result of this cycle will be a net transfer of H⁺ to the intracellular compartment. The net effect would be an acidification of the intracellular compartment and alkalinization of the extracellular compartment, were there not active H⁺ extrusion mechanisms which keep [H⁺], low and well out of electrochemical equilibrium under steady-state conditions. Thus, the uptake of NFL increases the load on any H⁺ transporter.

With ammonia production by muscle during exercise, it seems likely that PNH₃ and NH₄⁺ electrochemical gradients would form between muscle and plasma (Fig. 4B). If ammonia is produced as NH₃, then intracellular ammonia will raise pHi by trapping H⁺ (from glycolysis) with an almost 1:1 stoichiometry. Ammonia may diffuse out of the cell as either NH₃ or NH₄⁺. Ammonia leaving as NH₃ will not affect the H⁺ budget of the intracellular compartment. However, ammonia leaving as NH₄⁺ would act as a mechanism to export H⁺, which will also help to regulate intracellular pH.

In summary, the results of this study show that the distribution of ammonia between plasma and muscle, heart and brain tissue in lemon sole is not dependent on the H⁺ distribution, indicating that the cell membranes of these tissues have a significant permeability to NH₄⁺. The ammonia distribution appears to follow the membrane potential at rest; calculated values generally agree with published values for E_M in these tissues. During hypercapnia and following exercise, NH₄⁺ as well as NH₃ gradients may exist between tissue and plasma, particularly in tissues where ammonia production is stimulated, such as white muscle after exhaustive exercise.

We gratefully acknowledge the excellent technical assistance of Steve Munger, helpful discussions with Dr M. J. O’Donnell, Dr T. Mommsen and Brent Thomson, and the hospitality and support of the staff of the Bamfield Marine Station. We would also like to thank the Vancouver Public Aquarium for their generous donation of two flatfish. This study was supported by NSERC grants to CMW; EPA and NSERC grants to DJR; and an NSERC postgraduate scholarship to PAW.