Active ammonia excretion across the gills of the green shore crab Carcinus maenas: participation of Na⁺/K⁺-ATPase, V-type H⁺-ATPase and functional microtubules

Non-ionic ammonia (NH₃) and, to a lesser degree, ammonium ion(NH₄⁺) are toxic in most animals. Elevated ammonia levels in low-salinity media disrupt ionoregulatory function in the lobster Homarus americanus (Young-Lai et al., 1991) and the crayfish Pacifastacus leniusculus(Harris et al., 2001). Exposure of the green shore crab Carcinus maenas to 1 mmol l^-1 total ammonia leads to increased ion permeability and salt flux across the gill; higher concentrations reduce both variables(Spaargaren, 1990). In fish,branchial gas exchange and oxidative metabolism are disturbed by excess ammonia (Wilkie, 1997). In mammals, elevated ammonia levels cause mucosal cell damage(Lin and Visek, 1991) and inhibit cyclic-AMP-regulated Cl^- transport across the colon(Prasad et al., 1995). Hydrated ammonium and potassium ions have the same ionic radius and, because of their K⁺-like behavior, ammonium ions affect the membrane potential and excitability of neurons(Cooper and Plum, 1987). An effective ammonia detoxification or excretion system is therefore essential to maintain cellular functions.

Cellular uptake or excretion of non-ionic NH₃ is generally thought to occur by diffusion across the lipid bilayer of cellular membranes,although permeability to NH₃ is much lower than to CO₂(Knepper et al., 1989). Indeed, some plasma membranes of animal cells are relatively impermeable to NH₃ (Burckhardt and Frömter, 1992; Garvin et al., 1987). The entry of charged ammonium ions into animal cells may be mediated by several transport proteins:NH₄⁺-permeable K⁺ channels(Latorre and Miller, 1983),the Na⁺/K⁺/2Cl^- cotransporter(Kinne et al., 1986), the Na⁺/H⁺ exchanger(Kinsella and Aronson, 1981)or a recently described ammonium transporter related to the rhesus protein(Marini et al., 2000). In addition, abundant evidence suggests that NH₄⁺ may be transported actively, utilizing the ubiquitous Na⁺/K⁺-dependent ATPase(Skou, 1960; Towle and Hølleland,1987; Wall,1997).

Cellular excretion of NH₄⁺ may be mediated by an apical Na⁺/NH₄⁺ exchanger, as suggested for mammalian renal proximal tubules (Hamm and Simon, 1990) and the gills of marine teleosts(Randall et al., 1999). In freshwater teleost gills (Wilson et al.,1994) and mammalian inner medullary collecting ducts(Knepper et al., 1989),acidification exterior to the outer apical membrane may induce passive diffusion of non-ionic ammonia by diffusion trapping. Experiments with intact blue crabs (Callinectes sapidus) suggest that ammonia is excreted across the gills by diffusion of non-ionic NH₃ in animals acclimated to sea water (Kormanik and Cameron, 1981) but by Na⁺/NH₄⁺exchange in animals acclimated to low salinities(Pressley et al., 1981). Very little ammonia is excreted in the urine in this species(Cameron and Batterton,1978).

The mechanism by which ammonia crosses the epithelial layer of the excreting tissue is not generally considered, it being assumed that NH₃ and NH₄⁺ diffuse through the cytoplasm in a free state. However, under such conditions, the toxic effects of ammonia could be felt by multiple intracellular targets. In this study, we present the experimental basis for a novel mechanism of active ammonia excretion across a moderately tight epithelium, the gills of the euryhaline shore crab Carcinus maenas. We have previously demonstrated that isolated perfused gills from this animal are capable of transporting ammonia against a concentration gradient under physiologically meaningful conditions (Weihrauch et al., 1998, 1999). Active ammonia excretion across these gills was inhibited by apical amiloride, basolateral Cs⁺ or basolateral ouabain, implicating the participation of an apical Na⁺/NH₄⁺ exchanger, basolateral NH₄⁺-permeable K⁺ channels and basolateral Na⁺/NH₄⁺-dependent ATPase, respectively. We show here that an intracellular V-type H⁺-ATPase and an intact microtubule system are also required for active excretion of ammonia in this system, leading to an experimental model that includes acidification and ammonia-loading of intracellular vesicles, transport of vesicles along microtubules and exocytosis. Ultrastructural and molecular evidence support the model.

Shore crabs Carcinus maenas L. were obtained from a fisherman in Kiel Bay, Germany, or were collected from the intertidal region of the Mount Desert Island Biological Laboratory in Salsbury Cove, Maine, USA. They were maintained at 15°C in recirculating biologically filtered aquaria containing diluted seawater of 10‰ salinity and were fed cleaned squid or beef heart twice weekly.

Gills were prepared and perfused according to previously described methods(Siebers et al., 1985; Weihrauch et al., 1998) at a flow rate of 0.135 ml min^-1. Transepithelial potential differences(PD_te) were monitored using a millivolt meter (Keithley,type 197) to evaluate the quality of the preparation. Only gills generating an initial and continuously negative PD_te (-7.1±1.0 mV, N=32; mean ± S.E.M.) were employed.

The morphology of the phyllobranchiate gills provides sufficient surface area, roughly 247 cm² g^-1 fresh gill mass(Riestenpatt, 1995), to measure transepithelial ammonia fluxes directly. Excretion rates and ammonia concentrations were measured according to methods described in detail in an earlier publication (Weihrauch et al.,1998). Briefly, unless mentioned otherwise, gills were perfused and bathed in saline (S_standard) initially containing symmetrical amounts of ammonia (100 μmol l^-1 NH₄⁺) at identical pH values. Consequently, all measured net fluxes must be based on active transport mechanisms. The external medium was stirred constantly during each treatment. The concentration of ammonia selected was within the range of hemolymph ammonia levels measured in C. maenas(Durand and Regnault, 1998; Weihrauch et al., 1999). When a constant PD_te had been established (within approximately 30 min), the external bath and the perfusion solution were replaced and measurements were made for 30 min (controls). To continue the experiment with the same gill, the procedure was repeated with modified salines. Following application of a modified saline for 30 min, fluxes of total ammonia over the following 30 min were measured again. 2 ml samples were taken from the bath and internal perfusate after each experimental step.

Total ammonia concentrations (T_Amm) were determined with a gas-sensitive NH₃ electrode (Ingold, type 152303000). The sensitivity of the electrode measurements was approximately ±1.5μmol l^-1 in the T_Amm concentration range 50-100 μmol l^-1. To calculate net excretion, only the ammonia removed from the internal perfusion saline was considered to avoid including metabolically produced ammonia (Weihrauch et al., 1998). An alternative approach would be to measure the appearance of total ammonia in the external bath. However, as we show below,such a measurement would lead to an overestimation of transport rates as a result of ammonia production by the gill itself.

After removing the gills, single gill lamellae were isolated and split into their two halves (Schwarz and Graszynski,1989). In this way, a single epithelial layer covered by an apical cuticle was obtained. Isolated cuticle was prepared by carefully removing the epithelial cells using a smooth rounded metal wire. Split gill lamellae or isolated cuticle were mounted in a modified Ussing chamber, allowing area(0.02 cm²)-specific short-circuit currents(I_sc) and transepithelial resistances(R_te) to be measured. The chamber compartments were continuously superfused with saline at a rate of 0.5 ml min^-1 by means of a peristaltic pump.

To measure PD_te, Ag/AgCl electrodes were connected via agar bridges (3% agar in 3 mmol l^-1 KCl) to the chamber compartments; the separation distance of the preparation was less than 0.1 cm. A second pair of Ag/AgCl electrodes, connected through agar bridges,served as current electrodes to short-circuit the PD_tewith an automatic clamping device (VCC 600, Physiologic Instruments, San Diego, USA). I_sc and R_te were calculated according to Riestenpatt et al.(1996). The transepithelial and transcuticular conductances were calculated as G_te=1/R_te and G_cut=1/R_cut, respectively, where R_cut is the transcuticular resistance.

To assess the effects of NH₄Cl solutions on cuticular I_sc and R_cut, the isolated cuticle was superfused (0.5 ml min^-1) on both sides with a non-physiological ammonia-containing saline (S_NH4Cl) consisting of 248 mmol l^-1 NH₄Cl and 2.5 mmol l^-1 Tris adjusted to pH 7.8 with HCl. A clamp voltage of 10 mV with reference to the apical side was maintained to force transcuticular ion fluxes. Resistances for S_standard and S_NH4Cl were 9.0±0.1Ωcm² (N=6) and 7.1±0.4 Ωcm^-2(N=5) (means ± S.E.M.), respectively.

All data presented in this study were corrected by the resistances of the salines employed. All results are presented as means ± S.E.M. Differences between groups were tested using one-way analysis of variance(ANOVA) and the Newman—Keuls multiple-comparison test. Statistical significance was assumed for P<0.05.

The ionic composition of the salines (S_standard) used to bathe and perfuse isolated gills and to measure transepithelial resistance/conductance in gill half-lamellae contained (mmol l^-1)248 NaCl, 5 CaCl₂, 5 KCl, 4 MgCl₂, 2 NaHCO₃,2.5 Tris and 0.1 NH₄Cl. Immediately before use, 2 mmol l^-1 glucose was added to the basolateral salines in all experiments, and the pH of all salines was adjusted to 7.8 (HCl). Amiloride,bafilomycin A₁, colchicine, cytochalasin D, ouabain and taxol were purchased from Sigma (St Louis, USA). The ammonia standard (0.1 mol l^-1) was obtained from Orion Research Incorporated (Boston, USA). CsCl and all other salts were of analytical grade and were purchased from Merck (Darmstadt, Germany). Thiabendazole was kindly provided by Dr R. Gräf (Munich).

Gills from crabs acclimated to 10‰ salinity were shockfrozen in a high-pressure freezer (Wohlwend Engineering GMBH, Sennwald, Switzerland)(Studer et al., 1989),followed by freeze-substitution in 1% O_SO₄ in acetone and embedding in Epon. Ultrathin sections were stained with uranyl acetate and lead citrate and viewed with a Philips electron microscope at 80 kV.

Total RNA was extracted from gill tissue under RNAse-free conditions(Chomczynski and Sacchi, 1987). Reverse transcription of poly(A) mRNA was initiated with oligo-dT primer and Superscript II reverse transcriptase (Invitrogen). Amplification of a putative vesicle-associated membrane protein (VAMP) cDNA sequence was achieved via polymerase chain reaction (PCR) using degenerate primers based on published VAMP sequences (DiAntonio et al.,1993; Mandic and Lowe,1999). The forward primer had the following composition:5′-CARCARACNCARGCNCARGTNGA-3′; the reverse primer had the following composition: 5′-ATDATCATCATYTTNARRTTYT-3′, designed to produce a 198-nucleotide PCR product. Following separation on agarose gels,the PCR product was extracted from gel slices (Qiagen QiaQuick) and sequenced at the Marine DNA Sequencing Center of Mount Desert Island Biological Laboratory. The partial sequence was submitted to a BLAST search of GenBank(Altschul et al., 1997) and analyzed for open reading frame using DNASIS. The partial VAMP cDNA sequence from C. maenas gill was submitted to GenBank (Accession number AY035549).

In preliminary experiments, gills were perfused and bathed in saline from which Tris buffer was omitted, to measure possible pH changes in the apical and basolateral media. Under these conditions, the rate of ammonia removal from the internal perfusate was 14.5±1.4 μmol g^-1 fresh mass h^-1 (N=5) (Fig. 1A). However, the rate of appearance of ammonia in the apical bath was 27.6±2.6 μmol g^-1 fresh mass h^-1. The difference between these two values (13.1±3.8 μmol g^-1fresh mass h^-1) was interpreted to represent the apical release of ammonia produced metabolically by the gill itself. Thus, to avoid overestimation of transepithelial ammonia excretion, only the rate of ammonia removal from the internal perfusion medium was used to represent net active ammonia excretion across the gill. During these initial experiments, no significant changes in the pH (▵pH<0.02) of the apical saline (pH 7.8;volume 30 ml) were detected. However, the pH of the recirculated internal perfusion saline decreased slightly from an initial value of 7.8 to 7.71±0.02.

In the ensuing perfusion experiments, all solutions were buffered with 2.5 mmol l^-1 Tris to ensure pH stability under all experimental conditions. Measuring active ammonia excretion over time, the initial net ammonia efflux was 14.7±2.5 μmol g^-1 fresh mass h^-1 (N=5) and was thus within the range measured under Tris-free conditions. Over a period of 3h, the excretion rate decreased slightly but non-significantly by 4.7 μmol g^-1 fresh mass h^-1 (N=5) (Fig. 1B). Control rates of active ammonia excretion with symmetrical ammonia concentrations varied between experiments, ranging from approximately 12 to 26 μmol g^-1 fresh mass h^-1, perhaps reflecting variability within natural populations.

Symmetrical addition of 1 μmol l^-1 bafilomycin A₁,a specific inhibitor of V-type H⁺-ATPase(Bowman et al., 1988), lowered the transbranchial active net ammonia efflux by 66% from 18.3±2.6μmol g^-1 fresh mass h^-1 (control) to 6.3±1.7μmol g^-1 fresh mass h^-1 (N=7; P<0.001). The concentration of 1 μmol l^-1bafilomycin A₁ employed is known to produce maximal inhibition of crustacean gill V-type H⁺- ATPase without affecting the activity of Na⁺/K⁺-ATPase or F₁F₀-ATPase(Putzenlechner, 1994). After washout of the inhibitor, a minor but non-significant recovery of the efflux rate was detected (data not shown). A similar experiment was performed to verify whether the inhibitory effect of bafilomycin A₁ could be augmented by additional application of ouabain, a specific inhibitor of the Na⁺/K⁺-ATPase. The initial control efflux (19±3.8μmol g^-1 fresh mass h^-1) was reduced by application of 1 μmol l^-1 bafilomycin A₁ to 6.3±2.1μmol g^-1 fresh mass h^-1 (N=4; P<0.001) and was further reduced to 0.9±1.4 μmol g^-1 fresh mass h^-1 by subsequent basolateral application of 5 mmol l^-1 ouabain (N=4; P<0.05)(Fig. 2A). A partial but statistically insignificant recovery was measured after washout. It should be noted that the K_i for ouabain in C. maenas(K_i=2.9×10^-4 mol l^-1) and other crustaceans is more than two orders of magnitude higher than the K_i in mammals (Postel et al., 1998; Towle,1984).

In the next series of experiments, the effects of inhibitors of the cytoskeleton on active ammonia excretion were investigated. Basolateral application of 0.2 mmol l^-1 colchicine, an inhibitor of the microtubule system (Wilson and Farrell,1986), led to almost complete inhibition (to 2.3±1.3μmol g^-1 fresh mass h^-1) of the initial control efflux of 26.2±3.9 μmol g^-1 fresh mass h^-1(N=6; P<0.001) (Fig. 2B). Following washout, the efflux recovered significantly to 15.0±5.4 μmol g^-1 fresh mass h^-1. After establishing an outwardly directed ammonia gradient by adding 200 μmol l^-1 NH₄⁺ to the perfusing saline and none to the external bath, colchicine still reduced the initial efflux rate(37.1±3.9 μmol g^-1 fresh mass h^-1) by 74% of the control value to 10.4±4.3 μmol g^-1 fresh mass h^-1 (N=5; P<0.001). After washout, efflux recovered to 72% of the control value (Fig. 2C).

In contrast to drugs affecting the microtubule system, no significant effects on active ammonia excretion were observed when 5 μmol l^-1 cytochalasin D, a specific inhibitor of actin filaments(MacLean-Fletcher and Pollard,1980), was added to the perfusion saline (N=5)(Fig. 2D). The slight and non-significant decrease in the excretion rate from 20.1±2.8 to 16.5±2.7 μmol g^-1 fresh mass h^-1(P>0.05) resembled the control rates over the experimental period(Fig. 1B)(Weihrauch et al., 1998).

In all the above experiments, PD_te was monitored to detect changes in the electrophysiological variables of the gill(Fig. 2). With the exception of the addition of ouabain, which resulted in a reversible decrease in PD_te of 55% (Fig. 2A) due to a disruption of transepithelial Na⁺transport (Siebers et al.,1985), application of the various inhibitors had no significant effects on PD_te, indicating that Na⁺ transport pathways were not affected by the remaining treatments.

For two other blockers of the microtubule system, microtubule hyper-stabilizing taxol (Nogales et al.,1995) and microtubule destabilizing thiabendazole(Davidse and Flach, 1978), a strong inhibitory effect on active ammonia excretion was observed(Fig. 3). Whereas basolateral application of 10 μmol l^-1 taxol led to a decrease in the control rate (12.4±3.2 μmol g^-1 fresh mass h^-1) of 77% to 2.1±1.0 μmol g^-1 fresh mass h^-1 (N=6; P<0.001), basolateral addition of 0.2 mmol l^-1 thiabendazole altered the efflux rate of 11.5±2.5 μmol g^-1 fresh mass h^-1 to an apparent influx of ammonia (2.7±1.8 μmol g^-1 fresh mass h^-1) (N=6; P<0.001), probably as a result of metabolic ammonia production and release across the basolateral membrane. A statistically significant recovery of ammonia efflux was observed after washout of thiabendazole but not of taxol.

To evaluate changes in the electrical variables of the gill epithelium in response to application of the cytoskeleton inhibitors colchicine and cytochalasin D, the highly sensitive transepithelial short-circuit current(I_sc) and transepithelial resistance(R_te) were measured employing the preparation of the split gill lamella mounted in an Ussing-type chamber. Basolateral application of either 0.2 mmol l^-1 colchicine or 5 μmol l^-1cytochalasin D had no significant effect on I_sc(colchicine, 318.8±51.6 μA cm^-2; control,320.2±52.9 μA cm^-2, N=5; cytochalasin D,353.4±61.5 μA cm^-2; control, 358.8±64.9 μA cm^-2, N=4) (Fig. 4). Control values of R_te measured in parallel were not altered following the addition of colchicine (24.1±1.7 Ωcm², N=5) or cytochalasin D (28.3±1.7 Ωcm², N=4) (data not shown).

Previous studies have shown that amiloride, a blocker of epithelial Na⁺ channels and the Na⁺/H⁺ exchanger(Kleyman and Cragoe, 1988),has a strong inhibitory effect on ammonia excretion by isolated crab gills when applied to the apical (cuticle) side of the epithelium(Lucu et al., 1989; Weihrauch et al., 1998). To investigate whether the amiloride-induced reduction in the rate of ammonia excretion may be based on a possible effect on the electrophysiological properties of isolated cuticle (Lignon,1987) in addition to any effect on the epithelial cells themselves, NH₄⁺-dependent I_sc and G_cut of the isolated gill cuticle of C. maenaswere measured in a micro Ussing chamber. As expected in a cell-free system, a transcuticle potential difference (PD_cut) of 0 mV was measured (N=4). Following the imposition of a clamp voltage of 10 mV,a negative I_sc of -5800±1368 μA cm^-2and a G_cut of 683.0±165.2 mS cm^-2 were measured. The detected current is probably the result of NH₄⁺ effluxes, since cuticular permeability in C. maenas has been described to be 100- to 1000-fold smaller for monovalent anions than for monovalent cations(Lignon, 1987). After apical application of various amiloride concentrations (0.001-1 mmol l^-1),dose-dependent inhibition was observed for both I_sc and G_cut (Fig. 5). Linear regression in a Hanes-Woolf plot revealed simple Michaelis-Menten kinetics for I_sc and G_cut. For the NH₄⁺-dependent I_sc, K_ami was 19.1μmoll^-1 and ΔI_max was -4638 μA cm^-2. K_ami for NH₄⁺-dependent G_cut was 20.4μmoll^-1 and ΔG_max was 549.8 mS cm^-2 (Fig. 5). Symmetrical application of 1 μmoll^-1 bafilomycin A₁had no significant effect on cuticular NH₄⁺-dependent I_sc (control, -4294±206 μA cm^-2;bafilomycin, -3984±217 μA cm^-2) and G_cut (control, 454±45 mS cm^-2;bafilomycin, 400±80 mS cm^-2) (N=3).

Because our inhibitor studies indicated important roles for the V-type H⁺-ATPase and microtubules in active ammonia excretion, we sought ultrastructural and molecular evidence that would help to support or refute such possibilities. A previous study showed that the B-subunit of the V-type H⁺-ATPase was distributed throughout the cytoplasm of gill epithelial cells in C. maenas rather than being located specifically in the apical membrane, suggesting that the H⁺-ATPase may be associated with cytoplasmic vesicles(Weihrauch et al., 2001b) as well as with the apical membrane. In the present study, transmission electron microscopy of sections obtained from posterior gills of C. maenasacclimated to 100 μmoll^-1 external ammonia revealed an apparently dynamic system of membrane vesicles and Golgi bodies associated with the apical region of the epithelium(Fig. 6A). We also observed a well-developed microtubule assemblage associated with the apical membrane(Fig. 6B,C). In several sections, we were able to detect apparent interactions between membrane vesicles and the microtubules.

To ascertain whether any of the known exocytotic mechanisms are expressed in C. maenas gill, we attempted to identify one of the expected components, vesicle-associated membrane protein (VAMP, also called synaptobrevin) (Trimble et al.,1988). Using PCR with degenerate primers based on published VAMP sequences, we identified a VAMP-related sequence in a cDNA mixture prepared from C. maenas gill. Translation of the 132-nucleotide fragment revealed an amino acid sequence that is highly homologous to VAMP sequences determined for other invertebrate and vertebrate species(Fig. 7).

Previous investigations on the ammonia excretion mechanism in the gills of C. maenas and other crab species suggested that ammonia from the hemolymph space is transported into the cytosol by basolateral Na⁺/K⁺-ATPase (Lucu et al., 1989; Towle and Hølleland, 1987) and Cs⁺-sensitive channels,probably K⁺ channels (Weihrauch et al., 1998). In isolated gills, active ammonia efflux was only partially (approximately 60%) inhibited by blocking the Na⁺/K⁺-ATPase or by omission of Na⁺, but was almost completely inhibited (>90%) after addition of dinitrophenol(Weihrauch et al., 1998),suggesting that a second, Na⁺-independent ATP-requiring component is involved in ammonia excretion. The present study shows that inhibition of the V-type H⁺-ATPase, which has been identified at the molecular level in C. maenas gill(Weihrauch et al., 2001b),reduces active ammonia net efflux by 66%. The finding that simultaneous application of bafilomycin A₁ and ouabain almost completely blocked active ammonia excretion identified the V-type H⁺-ATPase and the Na⁺/K⁺-ATPase as the two major ATP-requiring participants in the excretion mechanism.

Active ammonia transport across biological membranes could be mediated by a proton pump via two different mechanisms: (i) the proton gradient generated by the pump drives a parallel H⁺/NH₄⁺ exchanger, or (ii) the proton pump generates a partial pressure gradient for NH₃ over the membrane,inducing transmembrane diffusion of gaseous ammonia. The latter has been suggested for ammonia transport across the apical membrane in the gills of freshwater trout (Wilson et al.,1994) since in this tissue the H⁺-ATPase is localized predominantly in the apical membrane(Sullivan et al., 1995). In addition, buffering the external medium with Hepes dramatically inhibited ammonia excretion across fish gill because, under this condition, the partial pressure gradient for NH₃ diffusion across the apical membrane was abolished.

However, in gill epithelial cells of C. maenas, the V-type H⁺-ATPase is distributed throughout the cytoplasm and only faintly detectable in the apical region (Weihrauch et al., 2001b). In the present study, no significant changes in pH in the apical medium were detected during excretion nor was the active ammonia excretion rate altered when a buffer (2.5 mmoll^-1 Tris-HCl) was applied in the apical saline (Fig. 1) (Weihrauch et al.,1998). These results indicate that in C. maenasacidification of the outer apical membrane is not responsible for driving a putative H⁺/NH₄⁺ exchanger nor does it generate a partial pressure gradient for diffusion of non-ionic NH₃across the apical membrane. However, we cannot discount the possibility of unstirred layers between the gill lamellae and thus cannot exclude the possibility of pH gradients immediately external to the cuticle. Indeed, we suspect that the subcuticular space, between the cuticle and the apical membrane, represents a classic unstirred layer. Diffusion of NH₃coupled with transport of H⁺ to produce NH₄⁺could theoretically occur in these unstirred layers to which buffer might not penetrate.

However, we suggest that the proton pump of crab gills, rather than being restricted to apical membranes, may be inserted into the membranes of cytoplasmic vesicles, generating an inwardly directed partial pressure gradient for NH₃ and leading to the accumulation of NH₄⁺ within the vesicles. It has been shown that radioactively labeled methylamine diffuses into acidified vesicles, where it is protonated into its membrane-impermeable ionic form methylammonium(Riejngoud and Tager, 1973). We suggest that a similar mechanism functions in crab gills, where cytoplasmic NH₃ diffuses into vesicles acidified by the V-type H⁺-ATPase and is trapped in the vesicles as NH₄⁺.

The almost complete inhibition of both active and gradient-driven net ammonia excretion by microtubule inhibitors (colchicine, thiabendazole and taxol) indicates a microtubule-dependent ammonia excretion mechanism. In contrast, blocking the actin filaments with cytochalasin D, which causes a small increase in exocytotic and endocytotic movements in frog nephron epithelia (Verrey et al.,1995), had no significant effect on active ammonia excretion across the isolated gill. We suggest that NH₄⁺-loaded vesicles are transported along the microtubule network to the apical membrane,where ammonia is released by membrane fusion and exocytosis. This suggestion is supported by the abundance of Golgi bodies and vesicles in gill epithelial cells (Fig. 6A), by the presence of bundles of microtubules oriented towards the apical membrane and by the endo/exocytotic activities represented by apparent clathrin-coated pits(Fig. 6B,C).

Following application of bafilomycin A₁, colchicine,thiabendazole, taxol and cytochalasin D, PD_te remained unchanged, indicating that changes in ammonia excretion during treatment with these inhibitors were not caused by alterations to the osmoregulatory NaCl uptake machinery or by damage to the integrity of the preparation. The unaltered I_sc and G_te across the split half-lamella during exposure to colchicine also supported our conclusion that a functional microtubule network is necessary for the process of active ammonia excretion but that its inhibition does not affect the electrical variables of the gill epithelium, at least over the short term(Fig. 2).

The possibility of an apical amiloride-sensitive Na⁺/NH₄⁺ exchange across the apical membrane prompted our examination of the role of the cuticle. Amiloride has been shown to have an inhibitory effect on ammonia transport in renal proximal tubules(Knepper et al., 1989),colonic crypt cells (Ramirez et al.,1999), teleost gills (Randall et al., 1999) and crustacean gills(Lucu et al., 1989; Weihrauch et al., 1998). In contrast to vertebrate tissues, crustacean gills are covered with an ion-selective cuticle. Although the conductance of the isolated gill cuticle of C. maenas has been shown to be approximately 10 times higher than that of the combined epithelium plus cuticle(Riestenpatt, 1995), ion selectivity is demonstrable, with the permeability for monovalent anions(Cl^-, HCO₃^-) being 100-1000 times lower than that for Na⁺ (Lignon,1987).

The present study employing the isolated cuticle showed that NH₄⁺-dependent I_sc and G_cut were inhibited by amiloride in a dose-dependent manner. At an amiloride concentration of 100μmoll^-1, a concentration commonly used in investigating Na⁺ and NH₄⁺ fluxes across the gill epithelia of C. maenas (Lucu et al.,1989; Lucu and Siebers,1986; Onken and Siebers,1992; Weihrauch et al.,1998), more than 70% of the cuticular NH₄⁺-dependent I_sc and G_cut were blocked. In a recent electrophysiological study investigating the effects of amiloride on Na⁺ influx across split gill lamellae and isolated cuticle of C. maenas, it was shown that apical amiloride inhibits both the Na⁺-dependent transepithelial I_sc and G_te and also the transcuticular I_sc and G_cut with similar values for K_ami(Onken and Riestenpatt, 2002). These authors suggested that the effects of amiloride were due to an interaction between the diuretic and the outer cuticle and not with transporters in the apical cell membrane or paracellular junctions. They concluded, however, that amiloride may interact directly with cellular transporters in the gills of other crab species(Onken and Riestenpatt,2002).

The values of K_ami for the cuticular NH₄⁺-dependent I_sc and G_cut (approximately 20 μmoll^-1) obtained in our study are approximately 20-fold higher than the values calculated for the Na⁺-dependent I_sc and G_cut(approximately 1 μmoll^-1) employing an identical experimental design. Comparison of the NH₄⁺-dependent G_cut (683±165 mS cm^-2) and the Na⁺-dependent G_cut (583±71 mS cm^-2) (Riestenpatt,1995) showed a higher conductance of the cuticle for NH₄⁺ ions than for Na⁺ ions, confirming earlier measurements (Lignon,1987). We only can speculate that because of the smaller hydrated ionic size of NH₄⁺ (approximately 0.38nm versusapproximately 0.56 nm for Na⁺), blockage of a cation-permeable structure in the cuticle by amiloride is less efficient. However, from these experiments, the presence of a Na⁺/NH₄⁺exchanger in the apical membrane itself cannot be excluded.

On the basis of previous observations and insights gained from the present study, we have constructed a hypothetical model for transbranchial ammonia excretion in C. maenas functioning at physiological ammonia concentrations. In this model (Fig. 8), we suggest that hemolymph ammonia enters the epithelial cell across the basolateral membrane viaNH₄⁺-permeable Cs⁺-sensitive channels and also via the Na⁺/K⁺-ATPase in exchange for Na⁺ (Lucu et al.,1989; Towle and Hølleland, 1987). The nature of the NH₄⁺-permeable channel is unknown, but it may be related to a recently described rhesus-like protein that appears to mediate transfer of NH₄⁺ across cell membranes(Marini et al., 2000). Using reverse transcription and PCR, we have recently identified such a rhesus-like protein in the gills of C. maenas(Weihrauch et al., 2001a).

The pool of ammonia imported from the hemolymph and produced by gill metabolism occurs within the cytoplasm in a pH-dependent equilibrium between NH₄⁺ and NH₃ (pK_Ammonia=9.48)(Cameron and Heisler, 1983). In our working model, we suggest that non-ionic NH₃ diffuses along its partial pressure gradient into intracellular vesicles acidified by a proton pump. Because of the low pH within the vesicles, NH₃ would be converted into its membrane-impermeable ionic form NH₄⁺and therefore trapped in this compartment. Our microtubule inhibitor studies suggest that the NH₄⁺-loaded vesicles may be transported along the microtubule network to the apical membrane, where NH₄⁺ would be released by exocytosis into the subcuticular space. Our demonstration of a vesicle-associated membrane protein(VAMP) sequence in cDNA prepared from C. maenas gill RNA(Fig. 7) shows that at least one component of the exocytotic machinery is expressed in this tissue,providing circumstantial evidence for branchial exocytotic activity.

From the subcuticular space, NH₄⁺ would diffuse along a concentration gradient via amiloride-sensitive structures across the cuticle into the external medium of the gill chamber. At physiologically meaningful outwardly directed ammonia gradients (50-400μmoll^-1), transepithelial diffusion of ammonia is considered to be low, comprising 12-21% of the total efflux(Weihrauch et al., 1998).

In the present report, the proposed mechanism of exocytotic ammonia excretion is supported only by indirect evidence. To investigate its validity more thoroughly, more direct experiments are necessary. Feasible future approaches include measurements of the capacitance as an indicator of the exocytotic activity of the apical membrane under control and high-ammonia conditions (Zeiske et al.,1998), the use of radioactively labeled methylamine/methyl-ammonium as a traceable competitive inhibitor for ammonia movements (Talor et al., 1987)and the use of laser confocal microscopy combined with video-image analysis to trace intracellular pH-labeled compartments(Miller et al., 1994).

The possibility of an exocytotic ammonia excretion mechanism should be considered since, in this situation, toxic ammonia is trapped in vesicles within the cell rather than diffusing through the entire cytoplasm, where it could cause major damage. In aquatic animals facing an inwardly directed ammonia gradient in the natural environment(Weihrauch et al., 1999), the active component of the mechanism would provide protection for the gill epithelial cells and, indeed, for the entire organism against passive NH₄⁺ influxes. Microtubule-mediated transport of ammonia-loaded vesicles and exocytosis at the apical membrane would permit a potent ammonia detoxification mechanism in such organisms without compromising ionic permeability. Whether such a mechanism applies broadly across species is not known. However, some aquatic species, including the South American rainbow crab Chasmagnathus granulatus(Rebelo et al., 1999), the prawn Nephrops norvegicus(Schmitt and Uglow, 1997) and three fish species in the family Batrachoididae(Wang and Walsh, 2000),tolerate high environmental ammonia levels. Among the adaptive mechanisms in these species may be an active ammonia excretion process similar to that described here.

This study was supported by a National Science Foundation grant(IBN-9807539) to D.W.T. and by a Deutsche Forschungsgemeinschaft grant (Si 295/2-3) to D.W. and D.S.