Active excretion of ammonia across the gills of the shore crab Carcinus maenas and its relation to osmoregulatory ion uptake

Transcription

1 J Comp Physiol B (1998) 168: 364±376 Ó Springer-Verlag 1998 ORIGINAL PAPER D. Weihrauch á W. Becker á U. Postel á S. Riestenpatt D. Siebers Active excretion of ammonia across the gills of the shore crab Carcinus maenas and its relation to osmoregulatory ion uptake Accepted: 11 March 1998 Abstract The mechanism of transbranchial excretion of total ammonia of brackish-water acclimated shore crabs, Carcinus maenas was examined using isolated, perfused gills. Applying physiological gradients of NH 4 Cl (100± 200 lmol á l )1 ) directed from the haemolymph space to the bath showed that the e ux of total ammonia consisted of two components. The saturable component (excretion of NH 4 ) greatly exceeded the linear component (di usion of NH 3 ). When an outwardly directed gradient (200 lmol á l )1 ) was applied, total ammonia in the perfusate was reduced by more than 50% during a single passage of saline through the gill. E uxes of ammonia along the gradient were sensitive to basolateral dinitrophenol, ouabain, and Cs + and to apical amiloride. Acetazolamide (1 mmol á l )1 basolateral) or Cl ) - free conditions had no substantial e ects on ammonia ux, which was thus independent of both carbonic anhydrase mediated ph regulation and osmoregulatory NaCl uptake. When an inwardly directed gradient (200 lmol á l )1 ) was employed, in ux rates were about 10-fold smaller and una ected by basolateral ouabain (5 mmol á l )1 ) or dinitrophenol (0.5 mmol á l )1 ). Under symmetrical conditions (100 lmol á l )1 NH 4 Cl on both sides) ammonia was actively excreted against the gradient of total ammonia, which increased strongly during the experiment and against the gradient of the partial pressure of NH 3. The active excretion rate was reduced to 7% of controls by basolateral dinitrophenol (0.5 mmol á l )1 ), to 44% by basolateral ouabain (5 mmol á l )1 ), to 46% by Na + -free conditions and to 42% by basolateral Cs + (10 mmol á l )1 ), indicating basolateral membrane transport of NH 4 via the Na+ / K + -ATPase and K + -channels and a second active, D. Weihrauch (&) á D. Siebers á U. Postel á S. Riestenpatt Biologische Anstalt Helgoland (Zentrale), Notkestr. 31, D Hamburg, Germany W.Becker Zoologisches Institut und Museum, UniversitaÈ t Hamburg, Martin-Luther-King-Platz 3, D Hamburg, Germany apically located, Na + independent transport mechanism of NH 4. Anterior gills, which are less capable of active ion uptake than posterior gills, exhibited even increased rates of active excretion of ammonia. We conclude that, under physiological conditions, branchial excretion of ammonia is a directed process with a high degree of effectiveness. It even allows active extrusion against an inwardly directed gradient, if necessary. Key words Excretion ammonia á Ouabain á Carcinus á crab Abbreviations AZ Acetazolamide á CA carbonic anhydrase á DNP 2,4-dinitrophenol á FW fresh weight á Kt transport constant á P NH3 partial pressure of non-ionic ammonia á np NH3 gradients of P NH3 á PD te transepithelial potential di erence á T Amm total ammonia á TRIS Tris-(hydroxymethyl)- aminomethan á V max maximum rates of excretion Introduction As in the vast majority of aquatic animals, most crustaceans excrete their metabolic nitrogenous end products largely as ammonia, regardless of whether they occupy marine, fresh-water, or terrestrial habitats (Kormanik and Cameron 1981; Claybrook 1983). The main site of N-excretion in crustaceans is the gill epithelium, while antennal and maxillary glands play a minor role (Regnault 1987). In spiny lobsters Jasus edwardsii and blue crabs Callinectes sapidus less than 2% of total excreted ammonia is eliminated via the urine (Binns and Peterson 1969; Cameron and Batterton 1978). Usually the term ammonia or total ammonia (T Amm ) is used to describe the mixture of nonionic ammonia (NH 3 ) and ammonium ions (NH 4 ). In solution, both forms of ammonia exist in a ph-dependent equilibrium. The relationship of the concentrations of NH 3 and NH 4 can be calculated using the Henderson-Hasselbalch equation for ammonia:

2 365 ph ˆ pk log NH 3Š NH 4 Š 1 At 20 C and a concentration of 250 mmolál )1 NaCl, pk amounts to 9.48 (Cameron and Heisler 1983). Using this pk and a physiological ph of 7.8, approximately 2% of total ammonia is present as nonionic NH 3. However, the higher lipid solubility of NH 3 makes it more di usible through phospholipid bilayers. In crab gills the permeability ratio for NH 3 and NH 4 indicates that NH 3 is approximately 53 times more permeable (Cameron and Heisler 1983). The ionic form of ammonia can also permeate biological membranes, although in di erent ways. Due to its hydrophilic properties NH 4 can pass lipid bilayers only very slowly. The transport of NH 4 proceeds via membrane-associated carrier proteins or, as reported by Wilkie (1997) for marine teleosts, via the highly cationand ammonium-permeable shallow tight junctions between chloride or accessory cells. Kormanik and Cameron (1981) reported that ammonia excretion of sea water adapted blue crabs Callinectes sapidus occurs mainly by di usion of nonionic NH 3. Other authors have obtained experimental evidence for at least partial excretion of ammonia in its ionic form, NH 4 (Pressley et al. 1981; Lucu et al. 1989; Siebers et al. 1995). Experiments with intact crabs Cancer pagurus revealed a permeability of the gills for ionic NH 4 (Kormanik and Evans 1984). In other studies on Callinectes sapidus, a correlation of ammonia excretion with Na + absorption was found (Pressley et al. 1981). The same result was obtained by experiments using the chinese crab Eriocheir sinensis (Pe queux and Gilles 1981), the shrimp Macrobrachium rosenbergii (Armstrong et al. 1981) and the shore crab Carcinus maenas (Lucu et al. 1989). Experimental studies employing membrane vesicles of gill epithelia (Towle and Hùlleland 1987) and isolated, perfused gills (Lucu et al. 1989) indicated that NH 4 ions can substitute for K+ ions in activation of the ouabain-sensitive Na + /K + -ATPase. In the apical membrane the presence of an amiloride-sensitive Na + / NH 4 antiporter translocating NH 4 from the epithelial cell to the ambient medium in exchange for Na + was reported by Pressley et al. (1981) for the crab Callinectes sapidus and by Lucu et al. (1989) and Siebers et al. (1995) for the shore crab Carcinus maenas. Previous studies in this laboratory have shown that excretion of T Amm across the gill epithelium of the shore crab proceeds at least partially via some of the transporting proteins operative in osmoregulatory NaCl uptake. The following investigations were undertaken to obtain further information about the mode of excretion of total ammonia in the euryhaline shore crab Carcinus maenas. Of both forms of ammonia comprising T Amm,NH 3 has been long recognized as being the most toxic. For example, sh tolerate relatively high concentrations of NH 4, but su er severe damage from micromolar concentrations of NH 3 (Rice and Stokes 1975). At toxic internal levels ammonia is detrimental to the metabolism of glutamine-glutamate (Stryer 1991), branchial gas exchange in salmonids (Burrows 1964), oxidative metabolism (Arillo et al. 1981) and to the central nervous system (Cooper and Plum 1987). Therefore, excretion of ammonia must also be considered in the context of reducing NH 3 -toxicity. Materials and methods Crabs Shore crabs (C. maenas L.) were obtained from a sherman in Kiel Bay (Baltic Sea). In the laboratory adult males of approximately 5± 7.5 cm carapace width were maintained in 200 l aquaria (ca. one crab per 20 l). In order to avoid moulting, light periods were reduced to 8 h per day. The water was continuously aerated and ltered over gravel, and the temperature was kept at 16 C. Before experimental use the crabs were acclimated for at least 1 month in dilute seawater (10&). The crabs were fed 3 times a week with small pieces of bovine heart. Preparation and perfusion of gills The crabs were killed by destroying the ventral ganglion using a spike, which was pressed through the ventral side of the body wall. The carapace was lifted and the gills were removed. The gills were perfused according to Siebers et al. (1985) with a ow rate of ml á min )1. During perfusion, transepithelial potential differences (PD te ) were monitored using a millivolt meter (type 197, Keithley, Cleveland, USA), connected with the perfusate and the bath solutions by means of two Ag/AgCl electrodes (type 373-S7, Ingold, Frankfurt/Main, Germany). PD te was measured in individual gills to control the success of the preparation and the e ects of changing the composition of the bath or perfusion solution. Previous experiments employing NaCl salines symmetrically in the bath and in the haemolymph space of the gills have shown PD te of about )5to)8 mv (haemolymph side negative) in ion-transporting posterior gills 7±9 and about )3 to)4 mv in the less transporting anterior gills 4±6. When a constant PD te was established (within approximately 30 min) the external bath and the perfusion solution were replaced, followed by a sampling period of 30 min (controls). In order to quantify the small concentrations of T Amm originating from gill metabolism the experimental periods were prolonged to 1 h in this experiment. Samples of 2 ml were taken from the bath (original volume 30 ml) and from the perfusate after passage through the gill. In order to continue the experiment with the same gill the procedure was repeated with modi ed salines (changes of the ionic composition or addition of inhibitors of ionic transport or energy metabolism). Following application of a modi ed saline for 0.5 h uxes of T Amm during the following hour were measured again in 2 ml samples taken from the bath and the perfusate. There was no indication of changes in the output of endogenous ammonia with time (see Fig. 1). In the experiments on active excretion of ammonia across the gill epithelium, concentration changes in T Amm measured in the bath and in the perfusate were used for calculations of uxes. In the experiments on e ux of ammonia across the posterior gills along varying gradients, concentrations of T Amm were measured in the bath at perfusate concentrations between 800 and 4800 lmol á l )1 and as decreases in the perfusate at perfusate concentrations between 25 and 400 lmol á l )1. The samples were analysed for their content of T Amm on the day of experimentation or were sealed and immediately frozen at )70 C with measurement of ammonia taking place within 2 days. At the end of the experiment the gill was cut above the clamp, dried under light pressure between two sheets of soft paper (Kleenex) and weighed.

3 366 Determination of ammonia Concentration of T Amm in magnetically stirred samples was determined using a gas sensitive NH 3 electrode (Ingold, type ) connected to a digital ph meter (Knick, type 646). In order to transform total ammonia into NH 3, 30 ll of an alkaline solution (1.36 mol á l )1 trisodiumcitrate dihydrate, 1 mol á l )1 NaOH) were added to the 2-ml sample directly before measurement. For the calculation of the concentration of T Amm a calibration curve with de ned solutions of ammonia (3 á 10 )6 ±4 á 10 )4 mol NH 4 á l)1 ) prepared from an ammonia standard solution (Orion, no ) in saline was used. The electrode potentials (mv) measured were plotted against the logarithms of the concentrations of originally dissolved T Amm of the the calibration solution. The calibration curves obtained were straight lines with a high degree of accuracy, as evidenced by regression coe cients of or, in a few cases, The millivolt values of the samples were calculated as T Amm -concentrations, using the lin-log equation of the calibration line. Calibration solutions and samples were measured under the same conditions (identical ionic strength, a sample volume of 2 ml, temperature of 12 C, identical stirring conditions). Fluxes (J) were expressed in lmol á gfw )1 á h )1 and calculated according to J ˆ C beg C end V t FW where C beg is the concentration of ammonia in the sample (lmol á l )1 ) at the beginning of the experiment, C end is the concentration of ammonia in the sample at the end of the experiment (lmol á l )1 ), V is the volume of the external bath or perfusate (ml), t is the sampling period (h) and FW is the fresh weight of the gill (g). The drift of the electrode over time was small, usually less than 5% per hour, however, the drift was linear over time. Therefore a standard sample (0.1 mmol NH 4 á l)1 ) was included after every tenth measurement. From the values measured a time-dependent increment (positive or negative) was calculated for the correction of every sample. The sensitivity of the electrode measurements thus amounted to 1 lmol T Amm á l )1 in the concentration range 4±50 lmol á l )1 and to ca. 1.5 lmol T Amm á l )1 in the concentration range 50±100 lmol á l )1. Calculations of NH 3 partial pressures As reviewed by Wilkie (1997) for sh gills, ammonia excretion patterns follow blood-to-water NH 3 partial pressure gradients (DP NH3 in both freshwater and marine environments. In order to obtain information on the ammonia excretion pattern in the shore crab, DP NH3 was calculated for all experimental conditions. Using the pk of 9.48 (see introduction), the concentration of T Amm, and the ph-value measured concentrations of NH 3 were calculated according to Eq. 1. Partial pressures of NH 3 (P NH3 ), considering it as a dissolved gas, were calculated according to Eq. 3: P NH3 ˆ NH 3 Š=a 3 ; where a is the solubility coe cient and [NH 3 ] the concentration of NH 3. For our calculations we used the value a ˆ mmol á l )1 á torr )1 for plasma at 20 C (Cameron and Heisler 1983). DP NH3 was calculated from the di erences in (P NH3 between external and internal media. Salines and chemicals The haemolymph-like saline (control solution) used as perfusate and bathing solution contained (mmol á l )1 ): 248 NaCl, 5 CaCl 2,5 KCl, 4 MgCl 2, 2 NaHCO 3, 2.5 Tris-(hydroxymethyl)-aminomethan, and had a ph of 7.8. At the beginning of each experiment ph values of the perfusion and bathing solutions were controlled and, if necessary, readjusted. In addition, ph was measured in the internal and external saline at the end of each experiment. In order to keep changes in ph due to dissolution of NH 4 Cl small the physiological bu ering with bicarbonate was increased by addition of 2.5 mmol á l )1 TRIS. In order to analyse the e ects of TRIS added to the bicarbonate bu er, some speci ed experiments were carried out with salines bu ered only with bicarbonate. For replacing Na + or Cl ) -ions in the salines, choline-cl or Na-gluconate were used in equimolar concentrations. Sodium-free salines contained 2 mmol á l )1 KHCO 3 instead of NaHCO 3. Acetazolamide (AZ) and amiloride were purchased from Sigma (St. Louis, USA), the ammonia standard (0.1 mol á l )1 ) was obtained from Orion Research Incorporated (Boston, USA), choline chloride was obtained from Aldrich (Steinheim, Germany), Mg-Dgluconate and ouabain were obtained from Fluka (Buchs, Switzerland), CsCl, 2,4-dinitrophenol (DNP) and all other chemicals were of analytical grade and purchased from Merck (Darmstadt, Germany). Calculations The transport constants (K t ) and the maximum rates of excretion of total ammonia along varying gradients of ammonia directed from the haemolymph to the external medium (V max ) (Fig. 2A) were calculated by assuming the linear portion observed at higher gradients of T Amm as a di usive component. Shifting the linear portion to the concentration of zero (gradient-driven di usion does not occur when no concentration gradient is present) permits the subtraction of the di usive, linear portion from the measured e uxes of T Amm. The resulting, obviously saturable, component (Fig. 2A, line c) was linearized (Fig. 2B) using the Hanes-Woolf-plot (LuÈ thje 1990) to obtain transport constants and maximum excretion rates of the saturable process. All values are presented as means SEM. Di erences between groups were tested with the paired StudentÔs t-test. Statistical signi cance was assumed for P < Results Transbranchial potential di erences In order to monitor the proper physiological functioning of isolated, perfused gills during the experiments, the PD te between the bath and the perfusion solution was measured. Only gills generating an initial and continuously negative PD te were employed. Variations in the concentrations of ammonia in the bath and perfusion solutions did not signi cantly a ect the PD te. Measurements of the uxes of ammonia under control conditions showed that there was no correlation between the rates of the transbranchial uxes of T Amm and the PD te. Production and release of ammonia by posterior gills In order to measure uxes of total ammonia across the gill epithelium along varying gradients, it is necessary to know the contributions to these uxes by the production and release of ammonia from gill metabolism. Therefore no ammonia was added to the salines at the beginning. Total liberation of ammonia by the gill epithelium was lmol á gfw )1 á h )1 (n ˆ 4) (Fig. 1). Of this, 81.5% was released to the apical and 18.5% to the basolateral side. Since the ph of the salines remained constant at 7.8 the DP NH3 across the epithelium remained negligibly small. Using Na + -free salines total liberation of ammonia was similar ( lmol á g FW )1 á h )1, n ˆ 4);

4 367 3 lmol á l )1 (Fig. 1). Internal concentrations that had been decreased as a result of e uxes can thus only be underestimated by this gure. E ux of ammonia across the posterior gills along varying gradients Fig. 1 Production and release rates of ammonia by the gill epithelium of Carcinus maenas in controls and under Na + -free conditions. Gills were bathed and perfused with identical salines without ammonia. Data represent means SEM of four observations (P < 0.01) (FW fresh weight) however, the portion of ammonia transported to the basolateral side increased signi cantly (P < 0.01) to 35.1% of the total release (Fig. 1). Another experiment showed that the addition of 2 mmol á l )1 glucose to the perfusion solution reduced the liberation of total ammonia signi cantly to 68.5% (n ˆ 9; P < 0.01) of controls, without signi cant changes in the proportion of e uxes of ammonia to the apical and basolateral side (data not shown). In the experiments employing a xed gradient of T Amm directed from perfusate to bath (Table 1) the contributions of the production and release of ammonia by the gill to changes in T Amm on either side are di cult to quantify and represent a factor of uncertainty. We have, however, no reason to assume that the production and release of total ammonia by the gill had been increased. In order to avoid overestimation of ux rates of ammonia, only the concentration changes detected in the perfusate were used for ux calculations. The portion of T Amm due to release of metabolic ammonia into the perfusate is considered not to exceed approximately By applying varying concentrations of NH 4 Cl (25±4800 lmol á l )1 ) in the perfusion saline without NH 4 in the external bath, it was possible to determine e ux rates of total ammonia across the gill epithelium along the respective gradients. The gradient-driven excretion of total ammonia was composed of a saturable and a linear component (Fig. 2A, a). The non-saturable, linear component (r ˆ 0.998; Fig. 2B) dominated at unphysiologically high gradients between the haemolymph and external medium (1600±4800 lmol á l )1 ), the saturable component (Fig. 2A, c) dominated at the lower physiological conditions. The saturable curve plotted in a Hanes-Woolfdiagram exhibited a regression of A K t of 301 lmol á l )1 NH 4 and V max of 104 lmol NH 4 á g FW )1 á h )1 were calculated from this plot (Fig. 2B). Measurements of the PD te indicate that the ion transport processes in the gills were not damaged by unphysiologically high concentrations of ammonia (data not shown). At the beginning of the experiment the ph of the perfusion solutions and in the bath was kept constant at 7.8. After each experiment ph of the bath and the perfusion solution was measured. This allowed us to calculate P NH3 in both saline solutions before and after the experiment and the respective DP NH3, which are the driving force for di usive translocation of NH 3. During the experiments (0.5 h) the ph of the perfusate changed slightly by 0.08 units. At low concentrations ph increased from 7.80 at the beginning to ca (Fig. 3a). Along with increasing T Amm ph decreased steadily to reach nal values of In the bath ± obviously due to the bu ered saline and the comparatively high volume of 30 ml ± the ph remained constant at at all concentrations applied. Measured excretion rates of total ammonia were plotted against calculated DP NH3 at Table 1 Percentage inhibition of the excretion of total ammonia across the gill epithelium along a gradient of 200 lmol. l )1 NH 4 Cl directed from the haemolymph side to the external bath (DNP 2,4 -dinitrophenol) Manipulation Concentration of inhibitor (mmol. l )1 ) Reduction of ammonia e ux (%) Acetazolamide b Cl ) a,b ± Cl ) a,b/ouabain b ±/ DNP b Ouabain b Ouabain b /Cs + b 5/ Cs + a Amiloride a a manipulation on the apical side of the gill epithelium b manipulation on the basolateral side of the gill epithelium Number of observations (n)

5 368 Fig. 2A E ux rates for total ammonia (T Amm :NH 3 /NH 4 ) across isolated, perfused gills along various gradients directed from the haemolymph into the external bath (a excretion of ammonia: data represent means SEM of at least six observations; b linear component; c saturable component). B Hanes-Woolf-plot of data obtained from the saturable component of mean-e uxes to calculate the transport constant K t and the maximum release rate V max the beginning and at the end of the experiment (Fig. 3b). A ux of NH 3 driven by DP NH3 should be linear over time. This linearity was, however, only observed at higher DP NH3 ; indicating that at lower DP NH3 another mode of excretion other than di usion of NH 3 must be operating. Flux of ammonia across the posterior gills along a xed gradient: e ects of dinitrophenol, ouabain, Cs +, amiloride, acetazolamide and substitution of Cl ) Based on the concentrations of ammonia measured in the haemolymph of C. maenas ( lmol á l )1, 1 day after feeding, n ˆ 21), the posterior gills immersed in control solutions were perfused with the same saline but containing in addition 200 lmol á l )1 NH 4 Cl. Measuring T Amm of the perfusion medium and the perfusate after passage through the gill showed that Fig. 3A,B E ux rates for T Amm across isolated, perfused gills along various gradients directed from the haemolymph into the external bath as shown in Fig 2. At the beginning the ph was 7.8 in the internal and the external saline. A Changes in ph of the perfusate after the experimental period of 0.5 h B Fluxes of total ammonia plotted against the gradients of P NH3 before (closed circles, r of the linear component ˆ 0.998) and after the experiment (open circles, r of the linear component ˆ 0.986) T Amm had been reduced from 200 lmol á l )1 to a mean concentration of 90.9 lmol á l )1. This result shows that the original T Amm of the perfusion solution was reduced by more than 50% during a single passage of saline through the gill. Using the perfusion rate of 8 ml á h )1, the FW of the gills (between 18 and 40 mg), and the measured reductions of T Amm in the perfusate of individual gills, an e ux of T Amm from the basolateral to the apical side of the epithelium (J b->a ) of lmol á g FW )1 á h )1 (n ˆ 32) was calculated. Mean e ux rates decreased sligthly by lmol á g FW )1 á h )1 (n ˆ 3) when measurements were continued up to 4 h under control conditions (data not shown). Addition of 0.5 mmol á l )1 DNP, an inhibitor of oxidative phosphorylation and a H + -shunt reagent, to the perfusion saline reduced the e ux of total ammonia (J b->a ) along the gradient of 200 lmol á l )1 NH 4 directed from the haemolymph side into the external bath from lmol á g FW )1 á h )1 (controls) to lmol á g FW )1 á h )1 (n ˆ 6; P < 0.01) (Table 1). Basolateral application of 5 mmol á l )1

6 369 ouabain, a potent and speci c inhibitor of Na + /K + - ATPase (Skou 1965), decreased the e ux (J b->a ) from (controls) to lmol á gfw )1 á h )1 (n ˆ 7; P < 0.01). The e ux rates further decreased to lmol á g FW )1 á h )1 (P < 0.01) when, in addition to ouabain, 10 mmol á l )1 of Cs +, an inhibitor of K + -channels (Van Driessche and Zeiske 1980; Riestenpatt 1995), was added to the perfusion saline (Table 1). Apical application of Cs + had no substantial e ect on e ux rates [controls: lmol. g FW )1 á h )1,Cs + apical: lmol á gfw )1 á h )1 ; (n ˆ 4; P < 0.05)]. Addition of 0.1 mmol á l )1 amiloride, a potent inhibitor of Na + /H + antiporters in crustacean epithelia (Shetlar and Towle 1989; Ahearn et al. 1990; Ahearn 1996), resulted in a reduction of ef- ux rates of T Amm from lmol á gfw )1 á h )1 (controls) to lmol á gfw )1 á h )1 (n ˆ 12; P < 0.01) (Table 1). High activities of carbonic anhydrase (CA) were found in the gill tissue of C. maenas (BoÈ ttcher and Siebers 1993). In order to investigate the e ects of AZ, a potent inhibitor of CA (Maren 1977), 1 mmol á l )1 AZ was added to the perfusion saline. Compared to controls J b->a of total ammonia along the gradient of 200 lmol á l )1 NH 4 directed from the perfusion solution to the bath was slightly reduced from to l mol á gfw )1 á h )1 (n ˆ 6; P < 0.05) (Table 1). Previous investigations have shown that active osmoand ionoregulatory ion absorption across the posterior gills of the shore crab strictly proceeds in a coupled mode of Na + and Cl ) (Onken and Siebers 1992). The question as to whether or not ammonia excretion is directly coupled to active ion uptake was tested by symmetrical substitution of Cl ) by gluconate. Using this saline no active ion uptake occurs (Riestenpatt et al. 1996). When Cl ) ions in the perfusion and the bathing solution were replaced by gluconate, e ux rates of ammonia decreased slightly from to lmol á g FW )1 á h )1 (n ˆ 8; P < 0.01). Additional basolateral application of 5 mmol á l )1 ouabain to the perfusion solution further reduced the e ux rates to lmol á g FW )1 á h )1 (P < 0.01) (Table 1). In ux of ammonia across the posterior gills In order to investigate gradient-driven ux rates of ammonia from the external bath to the haemolymph space, a gradient of NH 4 Cl (200 lmol á l )1 ) from the bath to the perfusion solution was applied, which resulted in an in ux of ammonia (J a->b ) of lmol á gfw )1 á h )1 (n ˆ 9). Compared with the previously measured e ux rates (J b->a ˆ lmol á gfw )1 á h )1 ) in ux rates of ammonia were approximately 10-fold smaller than e uxes (Fig. 4). Additional basolateral application of 5 mmol á l )1 ouabain had no signi cant e ect on the in ux of ammonia (in- ux ouabain lmol á gfw )1 á h )1 ; n ˆ 9). The Fig. 4 Rates of in ux and e ux for total ammonia across isolated, perfused gills of C. maenas along a gradient of 200 lmol. l )1 NH 4 Cl. The white bar represents e uxes (haemolymph side: 200 lmol. l )1 NH 4 Cl;bath:0lmol. l )1 NH 4 Cl); the black bar represents in uxes (haemolymph side: no NH 4 Cl; bath: 200 lmol. l )1 NH 4 Cl). Data represent means SEM of nine observations (P < 0.01) same negative e ect was found after basolateral addition of 0.5 mmol á l )1 DNP (in ux controls : lmol á g FW )1 á h )1 ; in ux DNP : lmol á gfw )1 á h )1 ; n ˆ 3) (data not shown). Active excretion of ammonia across the gill epithelium: e ects of DNP, ouabain, Cs + and of substitution of Na +, and active e ux of ammonia across anterior and posterior gills In previous experiments ux of ammonia was measured using concentration gradients of ammonia between the haemolymph and the external bath. Under these conditions, gradient-driven ux of NH 3 and NH 4 and active net movement of NH 4 mediated by energy dependent processes cannot be distinguished. In a series of experiments 100 lmol á l )1 of NH 4 were added to both the perfusion saline and the external bath. Measurements of the concentration of T Amm in the perfusion solutions and the external bath show that ammonia is actively excreted across the gill epithelium. Since the perfusate passes the vessels of the gill only once, the reduction of T Amm in the perfusate after passage through the gill is a direct measure of the proportion of ammonia (%) excreted from the perfusate across the gill into the bath. Under these symmetrical conditions, 66.7% of T Amm in the perfusate was eliminated during a single passage and an active net e ux of lmol á gfw )1 á h )1 (n ˆ 7) was calculated. At the beginning of the experiment T Amm in the bath and the perfusate was 100 lmol á l )1 without a DP NH3. After 0.5 h of experiment T Amm in the bath had increased to lmol. l )1. T Amm in the perfusate was reduced from 100 to lmol á l )1. Increases in the bath and decreases in the perfusates were calculated for a standard gill of 30 mg FW. In this experiment an additional release of ammonia ( lmol á gfw )1 á h )1 ) generated from the gill

7 370 metabolism and excreted mainly into the bath can be calculated. During the experiment P NH3 changed from 46.9 to 53.9 ltorr in the bath and from 46.9 to 15.7 ltorr in the perfusate. DP NH3 therefore changed from zero to 38.2 ltorr and was directed into the perfusate. This result shows that the gill has excreted ammonia against a large gradient of T Amm increasing from zero to 82 lmol. l )1 during the experiment and an opposing DP NH3 increasing from zero to 38.2 ltorr. In contrast to partial inhibition by DNP of ammonia excretion along an outwardly directed gradient (see above), an almost complete reduction of active net e ux across the gill epithelium was observed using symmetrical concentrations of ammonia (100 lmol á l )1 NH 4 ) in the bath and in the perfusion solution. E uxes (J active, b->a ) under control conditions amounted to lmol á g FW )1 á h )1 and were reversibly blocked to lmol á gfw )1 á h )1 (n ˆ 5; P < 0.001) following basolateral addition of 0.5 mmol á l )1 DNP (Fig. 5a). Results on the e ux of ammonia along an outwardly directed gradient (see above) showed that the Na + /K + - ATPase plays an important role in the excretion of ammonia across the posterior gills of C. maenas. Following the addition of 5 mmol á l )1 ouabain to the perfusion solution the active net e ux in the symmetrical presence of 100 lmol á l )1 NH 4 was reduced from lmol á gfw )1 á h )1 (controls) to (n ˆ 7; P < 0.001) (Fig. 5b). When we applied Na + - free salines prepared by isoionic replacement of Na + by choline, active net e ux of NH 4 decreased from lmol á gfw )1 á h )1 (controls) to lmol á g FW )1 á h )1 (n ˆ 3; P < 0.05) (Fig. 5c). Basolateral application of 10 mmol á l )1 CsCl inhibited active net e ux (J active, b->a ) of ammonia from (controls) to lmol á gfw )1 á h )1 (n ˆ 5; P < 0.01). Additional enrichment of the perfusion solution with 5 mmol á l )1 ouabain increased the inhibitory e ect of Cs +. Net e uxes of ammonia nally decreased to lmol á gfw )1 á h )1 (n ˆ 5; P < 0.001) (Fig. 5d). In order to obtain information about the presence of active excretion of ammonia in the di erent gills of the shore crab, gills 4±9 were symmetrically exposed to 100 lmol á l )1 NH 4 Cl. As shown in Fig. 6, anterior gills 4±5 are also capable of active excretion of ammonia. The excretion rates were even larger than those observed in anterior gills 7±9. Fig. 5a±d E ects of di erent inhibitors and of omission of sodium on active excretion of ammonia across the gill epithelium of C. maenas. At the beginning of all experiments the perfusate and the external bath contained 100 lmol. l )1 NH 4 Cl. a Basolateral application of 0.5 mmol. l )1 2,4-dinitrophenol (DNP; n ˆ 5, P < 0.001); b basolateral application of 5 mmol. l )1 ouabain (n ˆ 7, P < 0.001); c symmetrical Na + -free conditions (n ˆ 3, P < 0.05); d basolateral applicationof10mmol. l )1 Cs + and Cs mmol. l )1 ouabain (n ˆ 5, P < 0.01). Data represent means SEM (i basolateral, e apical) Fig. 6 Active excretion of ammonia across anterior gills 4±6 and posterior gills 7±9 and transepithelial potential di erences. At the beginning of all experiments the perfusate and the external bath contained 100 lmol. l )1 NH 4 Cl. Data represent means SEM (n ˆ 5 for all gills except gill 6 and 9 where n ˆ 4)

8 371 Active excretion of ammonia across posterior gills: utilization of salines bu ered without TRIS Using salines bu ered with 2 mmol. l )1 bicarbonate alone (omitting TRIS) ux rates of ammonia and changes of P NH3 were measured under symmetrical application of 100 lmol á l )1 ammonia at the beginning of the experiment. After the 0.5 h experiment T Amm in the bath had increased from 100 to lmol. l )1. T Amm in the perfusate was reduced from 100 to lmol. l )1 (n ˆ 6) (Fig. 7A). Increases in the bath and decreases in the perfusates were calculated for a standard gill of 30 mg FW. Taking into consideration the reduction of ammonia in the perfusate a net excretion of ammonia from the haemolymph space into the bath of lmol á g FW )1 á h )1 (n ˆ 6) was calculated. Measurements of ammonia in the bath showed an additional increase of ammonia that must have resulted from the N-metabolism of the gill itself and amounted to lmol á g FW )1 á h )1 Fig. 7A,B Active excretion of ammonia across posterior gills 7 and 8 using symmetrical salines bu ered only by 2 mmol. l )1 bicarbonate [omitting Tris-(hydroxymethyl)-aminomethan]. At the beginning of all experiments the perfusate and the external bath contained 100 lmol. l )1 NH 4 Cl (n ˆ 6). Data represent means SEM. A Concentrations of T Amm in the bath and in the perfusate before and after an experimental period of 0.5 h. B Flux rates of T Amm indicating active net e uxes from the perfusate into the bath (Gill release calculated proportion of T Amm resulting from the N-metabolism of the ionocytes and liberated mainly into the bath) (Fig. 7B). During the experiment P NH3 changed from 46.9 to ltorr in the bath and from 46.9 to ltorr in the perfusate. DP NH3 therefore changed from 0 to ltorr and was directed into the perfusate. This result shows that the gill has excreted ammonia against a large concentration gradient of T Amm increasing from 0 to 57.6 lmol. l )1 during the experiment, and an opposing DP NH3 that increased from 0 to 31.6 ltorr. Discussion E ux of ammonia across the posterior gills along varying gradients Release of T Amm along varying gradients directed from the haemolymph to the external medium was composed of a saturable and a non-saturable component. The saturable component indicates that transport of the ionic form of ammonia (NH 4 ) proceeds via a nite number of transporting proteins. The non-saturable component most probably shows simple di usion of non-ionic NH 3, dependent on the partial pressure gradient. Simple diffusion of NH 3 under physiologically meaningful concentration gradients of T Amm comprised only a small portion of total N-e ux, amounting to approximately 16% at a concentration gradient of T Amm of 100 lmol á l )1, and 18% at a gradient of 200 lmol á l )1. Under physiological concentration gradients branchial N-excretion in Carcinus maenas is thus considered to proceed mostly as NH 4. These ndings are in line with the results of the in ux experiments (Fig. 4). In contrast, Kormanik and Cameron (1981) showed that branchial N-excretion in sea water-adapted blue crabs Callinectes sapidus occurs mainly as non-ionic NH 3.In Carcinus maenas excretion of NH 3 exceeded the rates of the saturable component only at unphysiologically high concentration gradients of T Amm (larger than ca. 1.5 mmol á l )1 ) between haemolymph and ambient medium. The results di er from the ndings by Lucu et al. (1989) who reported only saturable e ux of T Amm across isolated, posterior gills of the shore crab without any linear component. The tendency of slight increases in the ph of the perfusate from 7.80 to ca at low perfusate concentrations, followed by a steady decrease to nal values of ca along with increasing perfusate concentrations, can be explained by considering the NH 3 /NH 4 equilibrium of T Amm: NH 4 H 2O, NH 3 H 3 O 4 At low perfusate concentrations of T Amm the preferable outward transport of NH 4 may result in alkalinization of the perfusate, while at higher perfusate concentrations the high di usion rate of NH 3 may result in acidi cation. The changes in ph due to transport of ammonia probably became visible because of limitations in the capacity of ph regulation by the gill (Siebers et al. 1994).

9 372 E uxes of ammonia across posterior gills along a xed gradient Application of a near-physiological 200 lmol á l )1 gradient of NH 4 Cl directed from the haemolymph to the external side of the gill resulted in e ux rates indicative of a high degree of e ectiveness in removing ammonia from the haemolymph: T Amm in the perfusate was reduced by more than 50% during a single passage through the gill. Involvement of Na + /K + -ATPase is obvious from the nding that uxes of ammonia along the 200 lmol á l )1 gradient was reduced by 52% following basolateral addition of ouabain (Fig. 8). The e ux rates further decreased to reach a total reduction of 73% when, in addition to ouabain, 10 mmol á l )1 Cs + was added to the basolateral side (Table 1). The results suggest that apart from Na + /K + -ATPase, Cs + -sensitive K + -channels located in the basolateral membrane (Riestenpatt et al. 1996), which do not discriminate between K + and NH 4, also play a role in the translocation of NH 4 from the haemolymph across the basolateral membrane into the epithelial cell (Fig. 8). Application of Cs + in the external medium reduced e ux rates for ammonia by only 12%, a result implying that apical Cs + -sensitive K + -channels are not involved in N-excretion. Ouabain sensitivity of ammonia uxes across the gills has also been shown in the chinese crab Eriocheir sinensis (Pe queux and Gilles 1981) and in the shore crab Carcinus maenas (Lucu et al. 1989). In comparison to controls, addition of 0.5 mmol á l )1 DNP to the perfusion saline reduced the e ux of T Amm along its gradient by 55%, implying that the process depends on provision of energy. The nding that the addition of both ouabain and DNP resulted in an incomplete reduction of ammonium e ux by only approximately 55% suggests that the remaining 35% of e ux (an assumed maximum gure when considering ca. 20% passive nonionic di usion) is driven by an ammonium gradient across the apical membrane, generated from NH 4 translocation into the epithelial cell via basolateral K + -channels. We assume that the gradient across the apical membrane drives the ux out o the epithelial cell via transporting structures, which have gone unidenti ed until now. Apical amiloride (0.1 mmol. l )1 ) reduced the e ux of ammonia along its gradient by 55%, a result allowing the assumption that amiloridesensitive cation antiporters are responsible for the exit of ammonia across the apical side of the epithelial cell (Fig. 8). Following the addition of 0.3 mmol amiloride to the ambient medium, excretion rates of ammonia of intact blue crabs Callinectes sapidus were inhibited by 63% in specimens adapted to a salinity of 17& and by 67% in specimens adapted to 35& sea water (Pressley et al. 1981). Also, isolated non-perfused gills of the blue crab excreted 45% (acclimated to 17& salinity) and 40% (acclimated to 35& salinity) less ammonia after the addition of amiloride. Based on the correlation between amiloride sensitivity of Na + -in ux and ammonia excretion in the blue crab, the authors proposed the presence of a Na + /NH 4 -exchanger in the apical membrane. With respect to published inhibition constants for amiloride of 200 lmol. l )1 of the K + /H + -antiporter in the apical membrane of the midgut of the tobacco hornworm Manduca sexta (Wieczorek et al. 1991) and of 280 lmol. l )1 of the 2Na + /H + -antiporter in the apical membrane of branchial epithelial cells of the shore crab (Shetlar and Towle 1989), we propose that the incomplete amiloride-induced reduction of the e uxes of ammonia observed in our experiments resulted from the relatively low doses of the drug administered. However, one cannot exclude the possibility that the release of ammonia that could not be inhibited by amiloride may have proceeded via amiloride-insensitive apical cationtranslocating structures. An amiloride-sensitive cation pore of low selectivity for monovalent cations including NH 4 was identi ed by S. Riestenpatt (personal communication) in the isolated branchial cuticle of the shore crab. The possibility that this pore is a potential translocation site for NH 4 cannot be excluded.. l )1 Relation between nitrogen excretion and osmoregulatory ion transport, excretion of respiratory CO 2 and ph regulation of the haemolymph in posterior gills Fig. 8 Preliminary model of ammonia excretion across the gill epithelium of the shore crab C. maenas (1 ouabain sensitive Na + / K + -ATPase, 2 Cs +) sensitive K + channel, 3 routes of NH 3 di usion, 4 unknown active translocation mechanism of NH 4, 5 amiloridsensitive cation-transporting structure of the cuticle) Replacement of Cl ) ions by gluconate in the perfusion and the bathing solution resulted in slight decreases of e ux rates of ammonia of 24% (Table 1). Since active translocation of NaCl depends strictly on the presence of

10 373 Na + and Cl ) ions in the perfusate and the bathing solution (Onken and Siebers 1992), this result implies that only a small portion of the excretion of ammonia along its gradient is coupled to the process of ion absorption. Additional basolateral application of ouabain to the perfusion solution reduced the e ux by 51% (Table 1), a gure already known from application of ouabain alone. The data obtained from application of acetazolamide and omission of Cl ) indicate that excretion of ammonia is not necessarily coupled to active osmoregulatory ion absorption in spite of the the fact that it utilizes some of the proteins (basolateral Na + /K + -ATPase, basolateral K + -channel, apical amiloride sensitive cation-transporting structures; see Table 1) that play a role in this process. In spite of the well established role of CA in CO 2 excretion (BoÈ ttcher et al. 1995) and ph regulation (Siebers et al. 1994), this enzyme seems not to be involved in active ion uptake across the posterior gills of the shore crab. Only small or no e ects of AZ were observed on PD te (Siebers et al. 1994) and on the in ux of sodium or chloride (BoÈ ttcher et al. 1991) in isolated perfused gills, and on the short circuit current measured in isolated gill half lamellae (Onken and Siebers 1992). In contrast to the ndings in the shore crab, Henry and Cameron (1982) considered the extensive salinity modi cations of speci c CA activity in posterior gills of hyperregulating blue crabs Callinectes sapidus, and deduced that one of the potential roles of the enzyme is osmoregulatory (see also Henry 1988). In the chinese crab Eriocheir sinensis AZ is a strong inhibitor of active ion transport (Onken et al. 1995). In this crab, active Na + -independent Cl ) -uptake is driven by an apical V-type H + -pump and proceeds via apical Cl ) /HCO 3 ) -exchange (apical H + -pump and anion exchange in concerted action with cellular carbonic anhydrase) and basolateral Cl ) channels (Onken et al. 1995). In the shore crab the coupled in ux of sodium and chloride is energized by only one pump, the Na + /K + -ATPase, and the in ux of these ions across the apical membrane is considered to proceed via apical Na + /K + /2Cl ) - cotransport (Riestenpatt et al. 1996). Basolateral application of the comparatively high concentration of AZ (1 mmol. l )1 ) reduced the e ux of ammonia along a 200 lmol á l )1 gradient by only 18% (Table 1). Considering that e ux along the gradient was reduced by approximately 10% per hour under control conditions (see results) this nding shows that the majority of T Amm excretion does not depend on the carbonic anhydrase mediated processes of excretion of respiratory CO 2 (BoÈ ttcher et al. 1995) or ph regulation of the haemolymph (Siebers et al. 1994). In these experiments on isolated perfused gills the e ects of AZ (0.1 mmol. l )1 ) on ph regulation and excretion of CO 2 were observed within approximately 10 min following its application, implying that 30 min preincubation of perfused gills with 1 mmol á l )1 AZ (this paper) was su cient for drug equilibration. In ux of ammonia If the ux of ammonia exclusively proceeded as nonionic NH 3 -di usion, it can be anticipated that in ux and e ux will equal each other when the gradients of T Amm and P NH3 directed from internal to external medium, and vice versa, are identical. However, compared with the e uxes of ammonia along a 200 lmol á l )1 gradient directed from the basolateral to the apical side of the epithelium (J b->a ˆ lmol á gfw )1 á h )1 ), the in uxes proceeding along an equal gradient in the opposite direction (J a->b ˆ lmol á g FW )1 á h )1 ) amounted to only approximately 10% of e uxes (Fig. 4). Due to the insensitivity of in ux to ouabain and DNP this 10% is considered as nonionic di usion of NH 3, which does not proceed via ouabain-sensitive Na + / K + -ATPase nor on DNP-sensitive provision of energy. In addition, the results of the in ux experiment show that ux of ammonia across the gills of the shore crab is highly directed. It has been shown by Cameron (1986) for intact specimens of the crab Callinectes sapidus that rapid net in ux of ammonia occurs at high external concentrations of T Amm (1 mmol á l )1 ). Since our in ux experiments were conducted on isolated gills at comparatively low external concentrations we cannot exclude the possibility that at high external concentrations of T Amm high net in uxes may also occur. Production and release of ammonia by the posterior gills The posterior gills of Carcinus maenas are multifunctional organs responsible for acid-base regulation (Siebers et al. 1994; BoÈ ttcher et al. 1995), N-excretion (Lucu et al. 1989; Siebers et al. 1995) and osmoregulatory ion uptake (Siebers et al. 1985; Riestenpatt et al. 1996). These processes require energy produced by cell metabolism. In order to quantify the production and release of ammonia by the posterior gills no ammonia was added to any of the salines at the beginning of the experiments. Of the T Amm liberated by the gill epithelium (Fig. 1) 81.5% was transported to the apical and 18.5% to the basolateral side. Application of Na + -free salines increased the liberation of ammonia to the basolateral side to 35.1%, probably as a result of deactivation of Na + /K + -ATPase. Following Na + -omission, approximately two-thirds of total liberation of ammonia still proceeded to the apical side. In the absence of Na + this high amount of ammonia excreted to the external medium cannot proceed via an apical Na + / NH 4 exchanger, which was considered to operate in this process by Lucu et al. (1989). Addition of 2 mmol á l )1 glucose to the perfusion solution reduced the liberation of total ammonia signi cantly to 68.5% of controls without signi cant changes in the pattern of e ux of ammonia in the apical and basolateral direction. This result indicates that even in the presence of su cient glucose approximately two-thirds of metabo-

11 374 lism is still accounted for by amino acids or other N- containing molecules. Active excretion of ammonia across the gill epithelium In contrast to experiments dealing with gradient-driven uxes of ammonia, 100 lmol á l )1 of NH 4 Cl were added to both the perfusion saline and the external bath to characterize the potential presence of active mechanisms of net e ux (J active, b->a ) of ammonia across the gill epithelium. Even under these symmetrical conditions, 69.3% of T Amm originally contained in the perfusate was eliminated during a single passage through the gill, a result showing active net excretion of NH 4. The gills of the shore crab seem to counteract the in ux of toxic ammonia with active excretion when the external concentrations are elevated up to or higher than internal levels. As shown for Callinectes sapidus this capacity may be limited when external concentrations rise to higher levels of approximately 1 mmol ál )1 (Cameron 1986). It is presently not known whether the capability of active excretion of ammonia observed in Carcinus maenas is also present in other crustacean species. Dependence on metabolic energy is shown by the e ects of DNP, which after basolateral addition nearly reduced the process completely (by 93%; Fig. 5). Dependence of active excretion of ammonia on Na + /K + - ATPase is obvious from sensitivity to ouabain inhibiting it by 56% with regard to controls (Figs. 5, 8). Under Na + -free conditions active extrusion of ammonia was reduced to 54% of controls. It is worth mentioning that blocking of the access to energy by DNP resulted in a nearly complete reduction of this process, but inhibition of Na + /K + -ATPase and omission of Na + decreased active extrusion by one-half. This nding allows the assumption that besides Na + /K + -ATPase a second active process, which is independent of the presence of Na + ions, is involved in active excretion of ammonia. This process is not known at present (Fig. 8). Involvement of basolateral K + -channels (Fig. 8) in this mechanism is evident from its sensitivity to Cs + administered on the haemolymph side of the epithelium, which inhibited active ammonia extrusion by 58%. Additional application of ouabain reduced the active process by 86%, a nding showing that the entrance of NH 4 from the haemolymph space into the epithelial cell almost entirely proceeds via Na + /K + -ATPase and K + -channels. With respect to basolateral entrance into the epithelial cell, excretion of ammonia along its gradient equals active extrusion. We therefore consider the second active component to be located on the apical side (Fig. 8). Active e ux of ammonia across anterior and posterior gills Of the nine gills located on each side of the cephalothorax of the shore crab the posterior gills are assumed to play the dominant role in active ion uptake (Siebers et al. 1982; Siebers et al. 1987; see also Towle 1981 for other crustaceans) and carbonic anhydrase-dependent ph regulation (BoÈ ttcher and Siebers 1993; BoÈ ttcher et al. 1995). In anterior gills the capability of active ion uptake is markedly reduced. This is also obvious from the transepithelial potential di erences shown in Fig. 6. Interestingly, this pattern was not found when considering active excretion of ammonia. The capacity for active excretion of ammonia was even more pronounced in the anterior gills 4±6 (gills 1±3 were too small for perfusion). This observation is consistent with the assumption that active excretion of ammonia across the gills of the shore crab does not depend on the coupled active transport of NaCl, though it utilizes some of the membrane proteins (Na + /K + -ATPase and basolateral K + -channel) operating in it. With respect to large sodium in ux rates (approximately 800 lmol á g FW )1 á h )1 ) (Siebers et al. 1987), active excretion rates of ammonia are small (10±20 lmol á g FW )1 á h )1 ). Thus it can be assumed that the lower activities of Na + / K + -ATPase in anterior gills (Siebers et al. 1987) are also su cient for the translocation of ammonia from the haemolymph space into the interior of the ionocyte. Active excretion of ammonia across posterior gills: utilization of salines bu ered without TRIS It may be argued that the di usion of ammonia gas across the gill depends on the protonation of NH 3 to NH 4 in order to maintain the small but signi cant P NH 3 gradient. In a saline more heavily bu ered by 2.5 mmol ál )1 TRIS those protons may be less available, so NH 3 may build up in the external boundary layer and potentially retard di usion. In order to analyse the transport pattern under more physiological conditions, rates of active excretion of ammonia across the posterior gills were additionally measured in the absence of TRIS, using only bicarbonate bu ering (Fig. 7). Under these conditions active excretion rates of ammonia and changes in P NH3 in the bath and the perfusion solution were nearly similar to the ndings measured during bicarbonate/tris bu ering. So, even under more physiological conditions ammonia was actively excreted against a gradient of T Amm and P NH3. Conclusions The N-excretory pattern detected in the gills of the shore crab is a highly exible system allowing excretion of ammonia over a wide range of concentration gradients of T Amm, which may vary with respect to internal and environmental levels. N-excretion under physiological conditions mainly consisted of carrier-mediated e ux of NH 4. These e ux mechanisms utilize some of the transport proteins playing a role in active osmoregulatory ion uptake (Fig. 8). Because of the toxicity of NH 3,