Ammonium secretion by Malpighian tubules of Drosophila melanogaster: application of a novel ammonium-selective microelectrode

Insects are predominantly uricotelic, consistent with the Baldwin–Needham hypothesis that terrestrial organisms excrete less-soluble nitrogenous excretion products, such as urea, uric acid or allantoin, as a means of water conservation (Baldwin and Needham, 1934). Ammonia is generally considered to be the most toxic, but also the most soluble, of the nitrogenous wastes (Withers, 1992). However, ammonia is excreted by cockroaches, blowflies and mosquitoes (Mullins and Cochran, 1973; Prusch, 1972; Scaraffia et al., 2005), and significant amounts of ammonium urate are excreted by the desert locust Schistocerca gregaria (Phillips et al., 1994). Some insects can excrete a range of nitrogenous end products simultaneously, including ammonia, urea, uric acid and allantoin (Cochran, 1985), and alterations in diet may shift the relative abundances of these nitrogenous wastes (Baldwin and Needham, 1934).

Ammonia (NH₃) is a weak base that exists in two forms: neutral ammonia (NH₃) and ionic ammonium (NH₄⁺). Throughout this paper, ammonia refers to both NH₃ and NH₄⁺. The pK_a is ~9.3, so that >99% of ammonia is in the form of NH₄⁺ at physiological pH (~7) (Emerson et al., 1975). Because the hydrated radius of NH₄⁺ is similar to that of K⁺, NH₄⁺ may compete with K⁺ for permeation through membrane K⁺ transporters, with consequent effects on transmembrane potentials and intracellular ion concentrations (Martinelle and Häggström, 1993; Moser, 1987). In addition, ammonia may perturb cellular function through effects on extracellular and intracellular pH. There are thus multiple mechanisms of ammonia toxicity, and concentrations of 0.5 to 5 mmol l⁻¹ in the blood or haemolymph are lethal to many invertebrates and vertebrates (Withers, 1992).

Ammonia concentrations have been measured in the haemolymph of only a few insect species. Reported values range from 0.8 mmol l⁻¹ for larvae of the tobacco hornworm, Manduca sexta (Weihrauch, 2006), to 16 mmol l⁻¹ in larvae of the sheep blowfly Lucilia cuprina (Marshall and Wood, 1990). Ammonia is removed from the haemolymph by secretion across the Malpighian tubules and rectum of locusts and the rectum of larval blowflies (Marshall and Wood, 1990; Prusch, 1974; Stagg et al., 1991; Thomson et al., 1988). For example, Malpighian tubules of the desert locust, Schistocerca gregaria, secrete fluid containing 5 mmol l⁻¹ ammonium, fourfold above the concentration in the haemolymph (Stagg et al., 1991). Adults of the yellow fever mosquito, Aedes aegypti, excrete ammonia following ingestion of blood or ammonia-rich solutions (Scaraffia et al., 2005), although the tissues responsible are unknown. In addition, serotonin-stimulated Malpighian tubules of Rhodnius prolixus isolated in saline containing ammonium as the sole cation secrete fluid at 40% of rate seen in (Na, K)-replete saline, suggesting that NH₄⁺ can be a substrate for K⁺ or Na⁺ transporters in the basolateral and apical membranes (Maddrell, 1969). Data from multiple studies thus suggest that insect Malpighian tubules may secrete ammonium.

Insects that feed on decaying, protein-rich food likely ingest ammonia derived from ammonifying bacteria and/or fungi through a process known as ammonification (Payne, 1973). In addition, larval stages may also be exposed to and/or mature in a high ammonia environment, which may provide selective pressure for ammonia tolerance. For instance, ammonia concentrations in the diet of Drosophila cultures increase from ~10 mmol l⁻¹ after 4 days to ~30 mmol l⁻¹ after 20 days (Borash et al., 1998). Whole-body ammonia content of larvae reared on dietary medium containing 370 mmol l⁻¹ ammonia was three times above that of larvae fed a control diet (Borash et al., 2000). Survival during exposure to high concentrations of ammonia in the diet suggests that both larvae and adults are tolerant of the toxic effects of ammonia. In addition, larvae fed on increasing doses of NH₄Cl (~0.25–0.5 mol l⁻¹) showed the greatest fitness when assayed on ammonia-supplemented food (Borash et al., 2000). These data suggest that such larvae increase the capacity to detoxify or excrete ammonia. Drosophila larvae may thus provide a useful model system for studies of ammonia transport by the excretory organs.

One challenge for such studies is the difficulty of measuring ammonia concentrations in the small droplets of fluid secreted by Drosophila tubules set up in the Ramsay fluid secretion assay. A single Malpighian tubule may secrete ~20 nl of fluid over a period of 45 min, but conventional colourometric or enzymatic techniques for measuring ammonia concentrations require sample volumes of 20 μl or more. Ion-selective microelectrodes are suitable for measurement of ion concentrations in nanolitre volumes, and we have previously measured concentrations of Na⁺, K⁺, Ca²⁺and H⁺ in fluid secreted by Drosophila tubules (O'Donnell and Maddrell, 1995). However, current ammonium-selective microelectrodes are based on liquid membranes containing the antibiotic nonactin and are subject to interference from both K⁺ and Na⁺. Nonactin-based microelectrodes are approximately four times more selective for NH₄⁺ compared with K⁺ and ~100 times more selective for NH₄⁺ compared with Na⁺. Detection limits of 0.1 mmol l⁻¹ NH₄⁺ in crayfish saline containing 5.4 mmol l⁻¹ K⁺ have been reported, but the limit in a simulated intracellular saline containing 194 mmol l⁻¹ K⁺ is ~5 mmol l⁻¹ (Fresser et al., 1991). Given that fluid secreted by Drosophila tubules contains ~120 mmol l⁻¹ K⁺ and ~30 mmol l⁻¹ Na⁺ (O'Donnell and Maddrell, 1995), nonactin-based electrodes are of limited use for analysis of NH₄⁺ secretion by Malpighian tubules.

Use of electrodes based on nonactin in physiological solutions typically requires compensation for interference from both Na⁺ and K⁺. Studies with macroelectrodes have shown that membranes containing polyvinyl chloride and a novel ionophore based on a 19-membered crown compound 2,5,12,15,22,26-hexaoxaheptacyclo[25.4.4.4^6.114^16.21.0^1.27.0^6.11.0^16.21]-tritetracontane (TD19C6) have much greater selectivity towards sodium than do membranes based on nonactin (Suzuki et al., 2000). Thus, use of NH₄⁺-selective microelectrodes based on this compound offers the possibility that a complex calculation to correct for the effects of interfering ions could be reduced to one in which only interference from K⁺ needs to be considered. In this paper, we describe both the development of a novel ammonium-selective microelectrode based on TD19C6 and an improved method of correcting for the effects of interference by physiological levels of potassium. We then use these microelectrodes to assess haemolymph ammonia concentrations and ammonia secretion by isolated Malpighian tubules of Drosophila melanogaster.

Stocks of the Oregon R strain of Drosophila melanogaster Meigen 1830 were maintained at room temperature (21–23°C) on medium containing (in g): 100 sucrose, 18 agar, 1 potassium dihydrogen orthophosphate, 8 sodium potassium tartrate tetrahydrate, 0.5 NaCl, 0.5 MgCl₂, 0.5 CaCl₂, 0.5 ferric sulphate, 50 dry active yeast and 7.45 ml l⁻¹ 10% tegosept in EtOH and 10 ml l⁻¹ acid mix (11 parts H₂O, 10 parts propionic acid and 1 part of 85% o-phosphoric acid diluted with 1 litre of deionized water). Ammonia-rich medium was prepared by addition of NH₄Cl to the control diet.

Ion concentrations were measured in droplets of fluid secreted by Malpighian tubules set up in a modified Ramsay assay (Dow et al., 1994). Briefly, pairs of Malpighian tubules connected by the common ureter were removed from third instar D. melanogaster larvae. Dissections were performed in saline containing (in mmol l⁻¹): 117.5 NaCl, 20 KCl, 2 CaCl₂, 8.5 MgCl₂, 10.2 NaHCO₃, 8.6 HEPES, 20 glucose and 10 glutamine. Saline pH was titrated to pH 7.0 with NaOH. Salines containing ammonia were prepared by equimolar substitution of NH₄Cl for NaCl.

Preliminary measurements with K⁺-selective and Na⁺-selective microelectrodes indicated that the haemolymph of third instar D. melanogaster larvae contained 26.5±0.6 mmol l⁻¹ K⁺ and 34.9±0.8 mmol l⁻¹ Na⁺ (N=20), similar to values determined previously (Naikkhwah and O'Donnell, 2011). Malpighian tubules were therefore bathed in ‘haemolymph-like’ Drosophila saline, which contained 25 mmol l⁻¹ K⁺ and 40 mmol l⁻¹ Na⁺. In experiments using saline that mimicked the Na⁺ and K⁺ concentrations of haemolymph, ionic strength was maintained by equimolar substitution of NH₄⁺ by N-methyl-d-glucamine (NMDG). Two stock solutions contained (in mmol l⁻¹): 29.8 NaCl, 25 KCl, 82.7 NMDG Cl, 2 CaCl₂, 8.5 MgCl₂, 10.2 NaHCO₃ and 8.6 HEPES, and 40 NH₄Cl; 29.8 NaCl, 25 KCl, 42.7 NMDG Cl, 2 CaCl₂, 8.5 MgCl₂, 10.2 NaHCO₃ and 8.6 HEPES, respectively. Salines containing 0–40 mmol l⁻¹ NH₄⁺ were prepared from mixtures of the two stock solutions. All salines were titrated to pH 7.0.

Each pair of isolated tubules was transferred into a 20 μl bathing saline droplet held in a shallow well in a Sylgard®-lined Petri dish filled with paraffin oil. One tubule remained in the saline droplet and the other was pulled out of the droplet and wrapped around a 0.15 mm diameter steel pin so that the secreted fluid emerged from the cut end of the common ureter (Dow et al., 1994). Secreted fluid droplets were removed after 45 min and the diameter (d) was recorded using an eye-piece micrometer. Secreted droplet volume was calculated as π d³/6 and secretion rate (nl min⁻¹) was calculated by dividing droplet volume by the time (min) over which it had formed.

K⁺-selective microelectrodes were prepared using a cocktail based on the K⁺ ionophore valinomycin (Potassium ionophore I – cocktail B, Fluka, St Louis, MO, USA). Na⁺-selective microelectrodes were prepared using a cocktail based on 10% (w/v) sodium ionophore X, 0.25% Na tetraphenylborate and 89.75% o-nitrophenyl octyl ether (NPOE) (Messerli et al., 2008). Ammonium-selective microelectrodes were based on cocktails that contained different proportions of TD19C6, lipophilic salt and solvent, as described in the Results.

Micropipettes were pulled from borosilicate glass capillaries using a vertical puller (Narishige, Tokyo, Japan) and lightly silanized with 0.2 μl dichlorodimethylsilane placed in a 10 cm glass Petri dish and inverted over batches of 14 micropipettes on a hot plate at 200°C. The low level of silanization rendered the glass sufficiently hydrophobic to retain the hydrophobic ionophore cocktail and prevent its displacement by capillary rise of aqueous solutions, but not so hydrophobic as to allow capillary rise of paraffin oil into the tip when the microelectrode was used for measurements of fluid droplets collected in the Ramsay assay. Silanized micropipettes were backfilled with 100 mmol l⁻¹ NH₄Cl for ammonium microelectrodes, 150 mmol l⁻¹ NaCl for sodium microelectrodes, and 150 mmol l⁻¹ KCl for potassium microelectrodes and the reference microelectrodes. A column of ionophore cocktail ~200 μm long was drawn into the tip of the microelectrode using negative pressure applied from a 25 ml syringe connected to the back of the microelectrode through plastic tubing and a MEH900S microelectrode holder (WPI, Sarasota, FL, USA). Microelectrodes were conditioned in solutions containing 15 mmol l⁻¹ of the ion of interest for ~30 min prior to their use. Micromanipulators were used to insert ion-selective and reference microelectrodes into 20 μl droplets of calibration solutions under paraffin oil. The tip of each ion-selective microelectrode was broken to a diameter of ~5 μm by touching it against the shank of the reference microelectrode. Tip breakage reduced the 95% response time for the ion-selective microelectrodes, as described in the Results. Slopes of K⁺-selective and Na⁺-selective microelectrodes were determined in calibration solutions of constant ionic strength (150 mmol l⁻¹), similar to that of the saline and the fluid secreted by the Malpighian tubules (Linton and O'Donnell, 1999). K⁺-selective and Na⁺-selective microelectrodes were calibrated in mixtures of NaCl and KCl in which the sum of Na⁺ and K⁺ concentrations was 150 mmol l⁻¹. Slopes of K⁺-selective microelectrodes were 54 mV between 150 and 15 mmol l⁻¹ K⁺ and 58 mV between 15 and 1.5 mmol l⁻¹ K⁺. Slopes of Na⁺-selective microelectrodes were 62 mV between 150 and 15 mmol l⁻¹ K⁺ and 54 mV between 15 and 1.5 mmol l⁻¹ Na⁺. Slopes of NH₄⁺-selective microelectrodes based on TD19C6 are reported in the Results.

Third instar D. melanogaster larvae were rinsed with deionized water on an absorbent pad and allowed to dry for at least 5 min. Using forceps, the cuticle of each larva was torn under paraffin oil and haemolymph was immediately drawn into a pipette and expelled under paraffin oil in another dish for measurements using ISMEs.

Samples of diet that initially contained 0, 30 or 100 mmol l⁻¹ NH₄⁺ were collected from regions of the culturing vials in which ~100 larvae had been feeding for 8 days. Samples of the diet were placed under oil for measurements with K⁺-selective and NH₄⁺-selective microelectrodes.

The regression equation for the line fitting logK_NH4,K to [NH₄⁺] was then used to calculate the appropriate value of K_NH4,K for use in Eqn 3 when analyzing samples of haemolymph or fluid secreted by Malpighian tubules of Drosophila larvae.

Data are presented as means ± s.e.m. for the indicated number (N) of measurements. Statistical significance of differences was evaluated by one-way ANOVA. Significance of unpaired differences between measured and known ammonium concentrations, for each [NH₄⁺], was evaluated by the Student's t-test. Differences were considered significant if P<0.05.

Preliminary measurements indicated that NH₄⁺-selective microelectrodes based on a cocktail containing the solvent bis-(1-butylpentyl) adipate (BBPA) had slower response times but slightly higher selectivity compared with electrodes made with NPOE. All subsequent microelectrodes were therefore made using ionophore cocktails containing BBPA. The cocktail that had the highest selectivity contained 1% of the ionophore TD19C6, 0.25% of the lipophilic salt potassium tetrakis[3,5-bis-(trifluoromethyl) phenyl] borate (KtktFlMePB) and 98.75% BBPA. For microelectrodes with tip diameters of ~5 μm, the 95% response time of NH₄⁺-selective microelectrodes under oil was slower (~7 s) than that of the K⁺-selective or Na⁺-selective microelectrodes (~1 s). Slopes of ≥50 mV decade⁻¹ for NH₄⁺-selective microelectrodes down to 0.1 mmol l⁻¹ NH₄⁺ in the presence of 150 mmol l⁻¹ NaCl (Table 1) were maintained for >24 h. In Drosophila saline containing up to 2 mmol l⁻¹ K⁺, slopes ≥50 mV decade⁻¹ were observed down to 1 mmol l⁻¹ NH₄⁺. In K⁺-free Drosophila saline, concentrations of NH₄⁺ ≥0.1 mmol l⁻¹ could be resolved. In general, increasing concentrations of K⁺ in Drosophila saline of up to 20 mmol l⁻¹ progressively reduced the slope of the ammonium microelectrode (Table 1).

A previous study determined selectivity coefficients for the ionophore TD19C6 incorporated into macroelectrodes containing polyvinyl chloride (Suzuki et al., 2000). However, selectivity coefficients are not always similar in macroelectrodes and microelectrodes, or in liquid-membrane electrodes that do not contain polyvinyl chloride. We therefore measured selectivity coefficients for both the NH₄⁺-selective and K⁺-selective microelectrodes used in our experiments (Fig. 1). Selectivity coefficients are typically presented in log form (e.g. logK_NH4,K). A selectivity coefficient of −1 thus indicates that the electrical potential corresponding to a given [NH₄⁺] will also be produced by a 10-fold higher concentration of K⁺. Positive values of the selectivity coefficient indicate that the ionophore is more selective for the interfering ion than for the principal ion.

Potential interfering ions of similar charge to those of NH₄⁺ and K⁺ were chosen along with ions found commonly in biological fluids. Although methylammonium produced the greatest interference (K_NH4,CH3NH3=−0.63), levels of this compound are negligible in biological fluids. Among physiological ions, K⁺ produced the greatest interference (logK_NH4,K=−0.94; Fig. 1A). Selectivity coefficients for interfering ions using a cocktail based on the classical NH₄⁺-selective ionophore nonactin (Fresser et al., 1991) are shown for comparison in the right column of Fig. 1A. The selectivity coefficients for both K⁺ (−0.6) and Na⁺ (−2.0) indicate greater interference of these ions on nonactin-based microelectrodes relative to the novel NH₄⁺-selective microelectrode based on TD19C6. Na⁺, the predominant cation in extracellular fluids, does not interfere with NH₄⁺-selective microelectrodes based on TD19C6 to any significant degree (logK_NH4,Na=−3.58).

In general, physiological ions interfered less with the potassium-selective microelectrode (Fig. 1B). Selectivity coefficients of the K⁺-selective microelectrode for several ions have been determined previously and are presented in the right column of Fig. 1B (Ammann et al., 1987). The value of logK_K,NH4 was −1.76, as shown in the left column (Fig. 1B). As a result, interference due to equimolar NH₄⁺ would increase apparent K⁺ concentrations above the actual value by 1.74%. In summary, these results suggested that the novel NH₄⁺-selective microelectrode based on TD19C6 suffered from significant K⁺ interference, whereas the K⁺ microelectrode is highly selective for K⁺ versus all other potential interfering ions in extracellular fluids such as haemolymph or the fluids secreted by the Malpighian tubules.

Isolated Malpighian tubules bathed in ammonia-free haemolymph-like Drosophila saline secreted fluid that contained 124±3.1 mmol l⁻¹ K⁺ (N=11) and 6.7±0.9 mmol l⁻¹ Na⁺ (N=13). When tubules were bathed in haemolymph-like Drosophila saline containing 30 mmol l⁻¹ NH₄⁺, there was a significant reduction in secreted fluid K⁺ (113.6±3.9 mmol l⁻¹, N=15; P<0.05) but no significant change in secreted fluid Na⁺ (10.0±1.7 mmol l⁻¹, N=12). Competition between K⁺ and NH₄⁺ for transport by the Malpighian tubule is examined further in the concluding paragraph of the Results section.

The concentrations of K⁺ in the haemolymph and fluid secreted by the Malpighian tubules were well above the levels that produce significant interference for NH₄⁺-selective microelectrodes. Use of salines containing physiological levels of K⁺ thus required correction for this interference. K⁺ concentrations in all samples of haemolymph or secreted fluid were therefore measured with K⁺-selective microelectrodes and the corrected NH₄⁺ concentration was calculated by subtracting the product of the measured K⁺ concentration and the selectivity coefficient K_NH4,K (Eqn 3). We initially used the value of K_NH4,K determined by the separate solutions method (10^−0.94). For example, if [NH₄⁺]_uncorrected was 7 mmol l⁻¹ in a solution containing 17.5 mmol l⁻¹ K⁺, then the corresponding [NH₄⁺]_corrected was: 7–(17.5)(10^−0.94)=5 mmol l⁻¹.

We measured ammonium concentrations in solutions containing 150 mmol l⁻¹ K⁺, a level expected to be the maximum found in fluid secreted by insect Malpighian tubules (Maddrell and Klunsuwan, 1973). However, when a single value for the selectivity coefficient (logK_NH4,K =−0.94) was used to correct for K⁺ interference, a consistent absolute error of approximately 3 mmol l⁻¹ NH₄⁺ below the known concentration was found across all the test solutions over the range 0–40 mmol l⁻¹ (Fig. 2).

The results shown in Fig. 2 indicated that use of a single selectivity coefficient determined using the separate solutions method resulted in significant errors in correcting for K⁺ interference. We therefore used Eqn 4 to determine values of logK_NH4,K at each ammonium concentration in solutions resembling both haemolymph (25 mmol l⁻¹ K⁺; Fig. 3A) and fluid secreted by Malpighian tubules (120 mmol l⁻¹ K⁺; Fig. 3B). The relationship between logK_NH4,K and [NH₄⁺] was then fitted by linear regression. In both linear regressions, logK_NH4,K values declined with increasing ammonium concentrations. Hereafter, the method used in this paper to correct for the effects of K⁺ interference on NH₄⁺ microelectrodes is referred to as the variable selectivity coefficient method (VSCM).

We first tested the VSCM by preparing haemolymph-like solutions containing 25 mmol l⁻¹ K⁺ and secreted fluid-like solutions containing 120 mmol l⁻¹ K⁺. We then compared the known [NH₄⁺] and the [NH₄⁺] measured by the microelectrode and corrected for K⁺ interference. There were no significant differences between [NH₄⁺]_corrected and [NH₄⁺]_known for all concentrations over the ranges 0–20 and 0–40 mmol l⁻¹ NH₄⁺ (Fig. 4A,B, supplementary material Table S1). We were particularly interested in measuring NH₄⁺ at low concentrations in the presence of high concentrations of K⁺. Importantly, although the selectivity coefficients for the secreted fluid-like solutions were measured in a single set of solutions containing 120 mmol l⁻¹ K⁺, there was no significant difference between known and measured NH₄⁺ concentrations between 0 and 5 mmol l⁻¹ when the K⁺ concentration was increased to 140 mmol l⁻¹ (Fig. 4D, supplementary material Table S1). When the K⁺ concentration was reduced to 100 mmol l⁻¹, the difference between known and corrected NH₄⁺ concentrations over the range 0–5 mmol l⁻¹ was small (0.3–0.4 mmol l⁻¹) but statistically significant (Fig. 4C, supplementary material Table S1).

A previous study has indicated that the NH₄⁺ concentration of the diet in which Drosophila larvae are reared increases progressively to ~30 mmol l⁻¹ after 21 days (Borash et al., 1998). We measured the concentrations of K⁺ and NH₄⁺ in diets that initially contained 42.2±0.7 mmol l⁻¹ K⁺ (N=30) and 0, 30 or 100 mmol l⁻¹ NH₄⁺. After 8 days, the K⁺ concentration had declined to 29.4±0.8 mmol l⁻¹ (N=30); there were no significant differences in K⁺ concentrations in the diets with initial NH₄⁺ concentrations of 0, 30 or 100 mmol l⁻¹ and the data were therefore pooled. Uncorrected NH₄⁺ concentrations measured in the three diets after 8 days were corrected for K⁺ interference using selectivity coefficients (logK_NH4,K) calculated from Eqn 4. After the larvae fed for 8 days on diets that initially contained 0, 30 and 100 mmol l⁻¹ NH₄⁺, corrected NH₄⁺ concentrations in samples of the diets were 21.7±1.0, 52.1±0.6 and 137.7±1.8 mmol l⁻¹, respectively (N=10).

There was no significant change in NH₄⁺ concentrations in the haemolymph of larvae that developed for 8 days in media that initially contained 0, 30 or 100 mmol l⁻¹ NH₄Cl (Fig. 5). Potassium concentrations in the haemolymph increased slightly (<10%) but significantly in haemolymph of larvae reared on diet that initially contained 100 mmol l⁻¹ NH₄Cl. Taken together, these results indicate that D. melanogaster larvae maintain haemolymph NH₄⁺ and K⁺ concentrations in response to high levels of dietary ammonia.

We hypothesized that a possible mechanism underlying the tolerance to dietary ammonia may be active clearance of ammonium from the haemolymph by secretion of ammonia into the lumen of the Malpighian tubules. Lumenal pH in Malpighian tubules of larvae of the related species Drosophilia hydei is 7.1 (Bertram and Wessing, 1994), suggesting that NH₄⁺ comprises the vast majority (>99%) of ammonia in the secreted fluid. When Malpighian tubules were bathed in haemolymph-like saline containing 25 mmol l⁻¹ KCl and 40 mmol l⁻¹ NaCl, the rate of fluid secretion was unaffected by increasing concentrations of NH₄⁺ of up to 40 mmol l⁻¹ (Fig. 6A). Mean fluid secretion rate (0.21 nl min⁻¹) was lower than the rate of ~0.35 nl min⁻¹ for Malpighian tubules of larvae bathed in Drosophila saline containing 20 mmol l⁻¹ K⁺ and 137 mmol l⁻¹ Na⁺ (Naikkhwah and O'Donnell, 2011). Ammonium concentrations in the secreted fluid did not differ significantly from those in the bathing salines in all treatments, although they trended (P=0.06 to 0.09) towards concentrations slightly below those in the bath over the range 20–40 mmol l⁻¹ NH₄⁺. K⁺ concentration in the secreted fluid declined slightly in saline containing 40 mmol l⁻¹ NH₄⁺ (Fig. 6B). In saline containing similar concentrations of K⁺ (25 mmol l⁻¹) and NH₄⁺ (30 mmol l⁻¹), the secreted fluid contained a fourfold higher concentration of K⁺ than NH₄⁺, indicating that the epithelial transporters have a preference for K⁺. Fluxes of NH₄⁺ and K⁺ (pmol min⁻¹) shown in Fig. 6C were calculated as the product of fluid secretion rate (nl min⁻¹) and secreted fluid ion concentration (mmol l⁻¹). NH₄⁺ flux increased with increasing concentrations of NH₄⁺ in the bathing saline, whereas potassium flux was independent of [NH₄Cl]. The mean potassium flux (23.6 pmol min⁻¹) was higher than ammonium flux in salines containing 0–40 mmol l⁻¹ NH₄⁺. In aggregate, these results suggest that Malpighian tubules transport ammonium but that potassium transport is unaffected by ammonium at concentrations up to 30 mmol l⁻¹.

This study has shown that the use of an improved NH₄⁺ ionophore, in conjunction with a novel method for correcting for the interfering effects of K⁺ and simultaneous measurement of K⁺ concentration, permits precise measurement of NH₄⁺ concentration in nanolitre samples of haemolymph or fluid secreted by isolated Malpighian tubules. We believe that the methods used in this paper may have relevance for studies of NH₄⁺ transport in other epithelia, such as perfused vertebrate kidney tubules.

The primary advantage of NH₄⁺-selective microelectrodes based on TD19C6 rather than nonactin is much greater selectivity for NH₄⁺ relative to Na⁺. The value of the selectivity coefficient (logK_NH4,Na) of −3.58 means that TD19C6-based microelectrodes are 3800 times more selective for NH₄⁺ relative to Na⁺. As a result, variations in Na⁺ concentration of up to 10 mmol l⁻¹ alter the uncorrected NH₄⁺ concentration determined from TD19C6-based microelectrode measurements by only 3 μmol l⁻¹.

The selectivity of microelectrodes based on TD19C6 for NH₄⁺ relative to K⁺ (logK_NH4,K=−0.94) slightly exceeds that for microelectrodes based on nonactin [logK_NH4,K=−0.6 (Ammann, 1986)]. Furthermore, the accuracy with which NH₄⁺ concentration can be measured is much improved by using the variable selectivity coefficient method to determine the value of K_NH4,K over a range of NH₄⁺ concentrations. In solutions mimicking the composition of fluid secreted by Drosophila Malpighian tubules, there was no significant difference between known and measured NH₄⁺ concentrations over the range (0–5 mmol l⁻¹) when K⁺ concentration was varied between 120 and 140 mmol l⁻¹. The error was 0.3–0.4 mmol l⁻¹ over the range 0.1–5 mmol l⁻¹ NH₄⁺ when the K⁺ concentration was lowered to 100 mmol l⁻¹. Taken together, these findings indicate that the variable selectivity coefficient method based on measurements in 120 mmol l⁻¹ K⁺ allows precise measurement of NH₄⁺ even when K⁺ concentration varies substantially. Similarly, in fluid mimicking the composition of haemolymph, there was no significant difference between known and measured NH₄⁺ concentrations over the range 0–20 mmol l⁻¹ NH₄⁺. Our methods thus permit more precise measurement of NH₄⁺ than can be achieved with nonactin-based microelectrodes, for which the detection limit in the presence of 194 mmol l⁻¹ K⁺ is ~5 mmol l⁻¹ (Fresser et al., 1991). NH₄⁺-selective microelectrodes based on nonactin are suitable, however, for use in freshwater, where the concentrations of the interfering ions K⁺ and Na⁺ are low (Rahaman-Noronha et al., 1996; Shih et al., 2008). It is also worth noting that although we used seven or eight different NH₄⁺ concentrations to determine the regression lines used in the variable selectivity coefficient method, re-analysis of Fig. 3 indicates that three to four NH₄⁺ concentrations bracketing the range of interest would provide a sufficiently high value for the regression coefficient (r²>0.95, data not shown)

We have used NH₄⁺-selective microelectrodes based on TD19C6 to analyze both haemolymph samples collected from larval D. melanogaster and the fluid secreted by their Malpighian tubules. The larvae maintain the concentration of NH₄⁺ in their haemolymph at ~1 mmol l⁻¹ even when feeding for 8 days on diets in which the initial concentration of NH₄⁺ was 100 mmol l⁻¹. Maintenance of haemolymph ammonia concentrations at this level is presumably adaptive for larvae feeding on initially ammonia-free diets in which the concentration of NH₄⁺ rises progressively to ~30 mmol l⁻¹ (Borash et al., 1998). Several dipteran insects are known to contain high concentrations of ammonia in the haemolymph. Larvae of the sheep blowfly L. cuprina, for example, are exposed to elevated ammonia levels as they develop in urine-soaked wool, proteinaceaus sera and blood from the lesions produced by the larvae. Their haemolymph contains 13–23 mmol l⁻¹ ammonia as the NH₄⁺ concentration in the rearing medium is varied between 3 and 560 mmol l⁻¹ (Marshall and Wood, 1990). The haemolymph of black flies (Simulium venustum) also contains high levels (5 mmol l⁻¹) of ammonia (Gordon and Bailey, 1974).

Our data indicate that the tubules of larval D. melanogaster secrete fluids containing concentrations of NH₄⁺ equal to those in the bathing saline over the range 0.1–10 mmol l⁻¹. Given that the transepithelial potential of the Malpighian tubules of the Oregon R strain of D. melanogaster is lumen positive by ~30 mV (O'Donnell et al., 1996), NH₄⁺ is transported against an electrochemical gradient, indicating that some form of active transport is present. Our results suggest that competition between K⁺ and NH₄⁺ is not simple. Secreted fluid ammonium concentrations equal those in the bath at ammonium concentrations approximately in the physiological range, then trend towards concentrations slightly below those in the bath above 20 mmol l⁻¹ ammonium. This pattern may indicate more than one mechanism for ammonium secretion. For example, amiloride-sensitive cation:proton exchangers that are energized by the V-ATPase mediate ammonia transport across the apical membrane in the midgut of larval M. sexta (Blaesse et al., 2010). It will thus be of interest in future studies to examine whether cation:proton exchangers mediate transport of ammonium as well as K⁺ and Na⁺ into the lumen of Malpighian tubules, or whether there is a separate transport pathway selective for NH₄⁺ that is most effective over a specific range of bath ammonium concentrations.

Our measurements can also be used to estimate the contribution of the Malpighian tubules to clearance of NH₄⁺ from the haemolymph. Haemolymph contains ~550 pmol of NH₄⁺ based on the concentration of ~1 mmol l⁻¹ NH₄⁺ (Fig. 5) and the volume of haemolymph in third instar larvae of ~0.55 μl (Naikkhwah and O'Donnell, 2011). Malpighian tubules bathed in saline whose ionic composition is based on the larval haemolymph secrete NH₄⁺ at the rate of 0.65 pmol min⁻¹ in the presence of 1 mmol l⁻¹ NH₄⁺ (Fig. 6). The four tubules thus secrete at the combined rate of 156 pmol h⁻¹. The rates of NH₄⁺ transport by the Malpighian tubules are thus sufficient to clear the haemolymph content of NH₄⁺ in ~3.5 h. These estimates suggest that the Malpighian tubules play an important role in clearance of ammonia from the haemolymph, although other epithelia such as the hindgut may also contribute.

Although we have examined ammonium secretion by Malpighian tubules of D. melanogaster, in part because the larvae may ingest high concentrations (>10 mmol l⁻¹) of ammonia in the diet (Borash et al., 1998), mosquitoes also excrete ammonia following ingestion of ammonia-rich fluids or blood (Scaraffia et al., 2005). The techniques developed in this paper will be useful for assessing the role of the Malpighian tubules in ammonia excretion by mosquitoes and other blood-feeding insects.

The authors are grateful to Dr. C. M. Wood for helpful discussions.

Selectivity coefficients are usually reported as logK_NH4,K, so it is necessary to convert them as in 10^logKNH₄,K and use the arithmetic value of the selectivity coefficient in the equations above.