A Cl⁻ Cotransporter Selective for Nh₄ ⁺ over K⁺ in Glial Cells of Bee Retina

Abstract

There appears to be a flux of ammonium (NH₄ ⁺/NH₃) from neurons to glial cells in most nervous tissues. In bee retinal glial cells, NH₄ ⁺/NH₃ uptake is at least partly by chloride-dependant transport of the ionic form NH₄ ⁺. Transmembrane transport of NH₄ ⁺ has been described previously on transporters on which NH₄ ⁺ replaces K⁺, or, more rarely, Na⁺ or H⁺, but no transport system in animal cells has been shown to be selective for NH₄ ⁺ over these other ions. To see if the NH₄ ⁺-Cl⁻ cotransporter on bee retinal glial cells is selective for NH₄ ⁺ over K⁺ we measured ammonium-induced changes in intracellular pH (pH_i) in isolated bundles of glial cells using a fluorescent indicator. These changes in pH_i result from transmembrane fluxes not only of NH₄ ⁺, but also of NH₃. To estimate transmembrane fluxes of NH₄ ⁺, it was necessary to measure several parameters. Intracellular pH buffering power was found to be 12 mM. Regulatory mechanisms tended to restore intracellular [H⁺] after its displacement with a time constant of 3 min. Membrane permeability to NH₃ was 13 μm s⁻¹. A numerical model was used to deduce the NH₄ ⁺ flux through the transporter that would account for the pH_i changes induced by a 30-s application of ammonium. This flux saturated with increasing [NH₄ ⁺]_o; the relation was fitted with a Michaelis-Menten equation with K _m ≈ 7 mM. The inhibition of NH₄ ⁺ flux by extracellular K⁺ appeared to be competitive, with an apparent K _i of ∼15 mM. A simple standard model of the transport process satisfactorily described the pH_i changes caused by various experimental manipulations when the transporter bound NH₄ ⁺ with greater affinity than K⁺. We conclude that this transporter is functionally selective for NH₄ ⁺ over K⁺ and that the transporter molecule probably has a greater affinity for NH₄ ⁺ than for K⁺.

INTRODUCTION

Although transmembrane transport of ammonium in animals has been studied, mainly in the mammalian kidney, there are two well-established cases of fluxes of ammonium from neurons to glial cells in nervous tissue. In vertebrate brain, where glutamate is the main neurotransmitter, the uptake of glutamate by astrocytes followed by its amination to glutamine, which is returned to the neurons and deaminated, implies a flux of ammonium (Benjamin and Quastel 1975; Hassel et al. 1997). In bee retina, the main metabolic substrate of the neurons (photoreceptors) is alanine formed by amination of pyruvate in the predominant glial cells (“outer pigment cells”). The alanine is transferred to the photoreceptors and deaminated to pyruvate and the tissue releases ammonium (Tsacopoulos et al. 1994, Tsacopoulos et al. 1997b; Coles et al. 1996).

Uptake of ammonium into cells can be monitored continuously, but indirectly, by measuring the changes in intracellular pH (pH_i) that it causes. Ammonium has a pK_a of ∼9.2 in water (Sillén 1964) so that at physiological pH (in the range 6.5–7.5) a fraction in the order of 1% is in the neutral NH₃ form. Nearly all cell membranes are permeable to NH₃ (but see Singh et al. 1995), so, when ammonium is applied outside a cell, NH₃ diffuses into it, combines with H⁺, and tends to raise pH_i (Jacobs 1940). In contrast, in astrocytes cultured from neonatal mouse, application of ammonium lowers pH_i because there is an influx of NH₄ ⁺ whose effect on pH_i outweighs the effects of NH₃ fluxes (Nagaraja and Brookes 1998). The glial cells in slices of bee retina also take up NH₄ ⁺ (Coles et al. 1996), an observation that has been confirmed and extended on bundles of glial cells freshly dissociated from adult retinas (Marcaggi et al. 1999). Application of ammonium causes a fall in pH_i that requires the presence of external Cl⁻ and is blocked by loop diuretics such as bumetanide (Marcaggi et al. 1999). These observations suggest that NH₄ ⁺ enters the glial cells by cotransport with Cl⁻ on a transporter with functional similarities to the cation–chloride cotransporters present on many types of cells. The transport on the bee glial cells is not blocked in the absence of Na⁺ (Marcaggi et al. 1999), indicating that the transport is of the K⁺-Cl⁻ class rather than the Na⁺-K⁺-2Cl⁻ class (see Race et al. 1999).

Several cases have been described of cation-chloride cotransporters, particularly in kidney, being able to transport NH₄ ⁺ in the place of K⁺, although with a lower affinity (Kinne et al. 1986). However, in plant roots, transporters are known that are selective for NH₄ ⁺ over K⁺ (e.g., Kaiser et al. 1998) so such selectivity is a demonstrated biological possibility. We have found that uptake of NH₄ ⁺ by the transporter in bee retinal cells is only moderately affected by external [K⁺]. This suggested that the transporter might be the first to be described in an animal cell that is selective for NH₄ ⁺ over K⁺ and prompted us to make a quantitative estimate of its selectivity.

Influx of NH₄ ⁺ into a cell is generally associated with transmembrane fluxes of NH₃ (Boron and De Weer 1976; see Fig. 2 C), so the relation between changes in pH_i (ΔpH_i) and NH₄ ⁺ flux (F_NH4) is complex. We tackled the question of the NH₄ ⁺/K⁺ selectivity in two stages. First, we deduced F_NH4 from ΔpH_i for relatively brief applications of ammonium. This required accurate absolute measurements of pH_i and measurement of several other parameters: membrane permeability to NH₃, intracellular buffering power, and the kinetics of pH_i regulation. Use of this “cell model” showed a functional selectivity for NH₄ ⁺ over K⁺. We then recorded pH_i responses to longer and more complex NH₄ ⁺ application protocols. By simulating these responses with a standard minimal model for a cotransport process, to which we added competitive inhibition, we estimated the NH₄ ⁺ and K⁺ affinities of the transporter molecule.

Figure 2.

Effect of extracellular ammonium application on pH_i of an isolated bundle of glial cells. (A) The ammonium-induced acidification was slightly inhibited by K⁺ or Rb⁺. For each application of 2 mM NH₄ ⁺, [K⁺]_o was either maintained at its baseline value of 10 mM, changed as indicated, or replaced by 50 mM Rb⁺. (B) The response to a 5-min application of 2 mM NH₄ ⁺ can be divided into five phases, which correspond to different patterns of fluxes (C). This cell had a fairly acid baseline pH_i (≈7.17) so that Phase 1 was prominent. (C) Schemes of fluxes corresponding to four of the phases indicated in B. (Phase 1) Inward flux of NH₃ is greater than ∼1% of inward flux of NH₄ ⁺. The maintenance of intracellular equilibrium NH₄ ⁺ ↔ NH₃ + H⁺ consumes H⁺ ions. (Phase 2) NH₄ ⁺ transmembrane gradient is still inward, while [NH₃]_i slightly exceeds [NH₃]_o. H⁺ ions are shuttled into the cell. (Phase 3) Extrusion of H⁺ ions by pH regulatory mechanisms equals inward flux of NH₄ ⁺. (Phase 4) This phase is approximately the inverse of Phase 1. Phase 5 (not shown) consists almost entirely of H⁺ efflux. (D) Scheme of transmembrane fluxes during ammonium exposure. pH_i changes result from three transmembrane fluxes, F_NH4, F_NH3, and net H⁺ flux through pH regulatory processes (F_reg) (see text).

MATERIALS AND METHODS

Intracellular pH (pH_i) in bundles of glial cells dissociated from the retina of the drone (male) Apis mellifera was measured by techniques developed from those described in Marcaggi et al. 1999. One record is shown (see Fig. 9 D) from an intracellular microelectrode recording of glial membrane potential in a slice of retina prepared and superfused with oxygenated Cardinaud solution, as described previously (e.g., Coles et al. 1996). Unless otherwise stated, results are given as mean ± SD and the two-tailed paired t test was used to determine P values. Errors of quotients were estimated by the calculus of errors (Abramowitz and Stegun 1965).

Figure 9.

Inhibition of inward NH₄ ⁺ flux by external K⁺. (A) Comparison of responses to 2 and 5 mM NH₄ ⁺ in 0 and 10 mM K⁺. For each response, NH₄ ⁺ was applied for 30 s; when applied in 0 K⁺, K⁺ was removed from the superfusate 15 s earlier. (B) Double inverse plot of mean ΔpH_i vs. [NH₄ ⁺]_o (n = 6). Straight lines passing through mean data points intersect the abscissa at [NH₄ ⁺]_o = 5.02 mM (0 K⁺, ○) and 6.90 mM (10 mM K⁺, •). (C) Double inverse plot of ⁱⁿF_NH4 vs. [NH₄ ⁺]_o. ⁱⁿF_NH4 is the mean value during the NH₄ ⁺ application calculated from ΔpH_i of the six experiments of B using the relation of Fig. 8 E. Straight lines passing through mean data points intersect the abscissa for [NH₄ ⁺]_o = 5.16 mM (0 K⁺, ○) and 7.09 mM (10 mM K⁺, •). (D) Intracellular recording from a retinal slice with a microelectrode in the glial compartment. Increasing [K⁺] from 10 to 20 mM in the superfusate induced a depolarization of the cell membranes. When the same increase of [K⁺] was made after superfusing the slice with 5 mM barium for a few minutes, the depolarization was undetectable for at least the first 45 s.

Dissociation Procedure and Loading of the Cells

Bees were obtained from A. Dittlo (Villandraut) or J. Kefuss (Toulouse, France) and maintained on sugar water. A slice of drone head ∼500-μm thick was cut with a razor blade. The slice was incubated for 40 min in a 1.5 ml Eppendorf tube containing 1 ml oxygenated Cardinaud solution (see below) to which had been added 2 mg trypsin (T- 4665; Sigma-Aldrich). The slice was washed in Cardinaud solution lacking Ca²⁺ and Mg²⁺ and the retinal tissue dissected out and triturated. 150 μl of cell suspension was placed in the perfusion chamber (see below) whose floor consisted of a microscope cover slip coated with poly-l-lysine. The cells were allowed to settle for 10 min and then exposed to the acetoxymethyl ester of 2′,7′-bis(2-carboxyethyl)-5(6)-carboxyfluorescein (BCECF-AM) (Molecular Probes, Inc.) at a concentration of 10 μM for 40 min.

Measurement of Fluorescence

The chamber was placed on the stage of an inverted microscope (Diaphot; Nikon) equipped with a 40× objective, photomultiplier detection, and dual wavelength excitation at 440 and 495 nm switched by liquid crystal shutters, as described in Coles et al. 1999. The stimulating light intensity was attenuated so that fluorescence from a bundle of loaded glial cells excited at 440 nm gave a signal of ∼10,000 photon counts s⁻¹, which remained stable for several hours. Dark noise plus autofluorescence from a bundle of unloaded cells was <2,500 photon counts s⁻¹ for both excitation wavelengths and it was checked that this fluorescence was not affected by ammonium superfusion (n = 4). This background fluorescence was automatically taken into account in the in situ pH calibration of each cell bundle (see below). The excitation pattern was usually: 440 nm, 100 ms; off, 20 ms; 495 nm, 600 ms; off 20 ms. To minimize the noise of the ratio, the signal resulting from excitation near the isosbestic point (440 nm) was averaged over several minutes before the PC computer calculated the ratio using a program available from Jean-Louis Lavie (University Bordeaux, Bordeaux, France).

Solutions

The standard perfusion solution contained (mM): 200 NaCl, 10 KCl, 4 MgCl₂, 2 CaCl₂. pH was buffered with 10 mM MOPS hemisodium salt and set to 6.90 with HCl. Osmolality was adjusted to 685 mOsm with mannitol (∼240 mM). The salt components, the pH, and the osmolarity of this solution are similar to those measured in vivo (Cardinaud et al. 1994). For other pHs, PIPES, MOPS, or HEPES were used for solutions of pH 6.20–6.50, 6.90–7.50, or 7.70, respectively. Most other variants were obtained by equimolar replacement of NaCl (or by increasing [NaCl] when [KCl] was reduced). Chloride-free solutions were made by replacing Cl⁻ by an equivalent quantity of gluconate and increasing Ca²⁺ to 8 mM to counteract the chelating effect of gluconate (Kenyon and Gibbons 1977). To test the sensitivity of the responses to ammonium to changes in the osmolarities of solutions, osmolarity was intentionally increased or decreased by 5% (∼34 mOsm) by changing the concentration of mannitol. Such a change in osmolarity had in itself a barely detectable effect on the emission ratio and no detectable effect on the pH response to 2 mM ammonium (n = 3). Therefore, in some experiments, salts such as NH₄Cl were simply added to solutions to final concentrations up to 5 mM without a compensatory reduction in [NaCl].

Perfusion System

To be able to make sufficiently rapid solution changes without detaching the cells from the floor of the chamber, we developed a perfusion chamber with no eddy currents. A factor that appeared to be important was the presence of a curved junction between the floor and the wall of the channel (Fig. 1 A). Solutions were gravity fed and selected by computer-controlled solenoid valves whose outflows passed through fine tubes at ∼30 μl s⁻¹ into a common pathway to the chamber. It was found that mixing of solutions was negligible. We obtained a measure of the speed of the solution change in the chamber by recording the change in fluorescence during a switch from standard solution to one containing 1 μg liter⁻¹ fluorescein (Fig. 1 B). The change in pH_i measured with BCECF in response to propionate or trimethylamine (TMA) was nearly as fast (see Fig. 4 B and 5 A). The change in fluorescein fluorescence was well described by an exponential; for flow rates used in experiments with cells, the mean time constant was: 5.4 ± 1.9 s (± SD, n = 16), and this exponential was used to describe the changes in extracellular concentration in our numerical models.

Figure 1.

Speed of the solution changes. (A) The perfusion chamber. A channel was milled in a polymethacrylate slab and a glass cover slip was glued on the bottom. Note the rounded edges of the channel. (B) Speed of solution change. The data points show the change in fluorescence observed through the 40× microscope objective focused on cells on the floor of the chamber during a 120-s perfusion change from standard solution to one containing 1 μg liter⁻¹ fluorescein. The time scale was displaced so that zero coincides with the beginning of the fluorescence change. The solid lines represent the best fits of data points obtained by regressions with simple exponentials {1 − exp(−t/τ₁) and exp[−(t − 120)/τ₂]} whose time constants, τ₁ and τ₂, are given in the figure.

Figure 4.

Direct measurement of P_NH3. (A) A bundle of cells was held with a pipette ∼50 μm above the floor of the chamber and perifused for 30 min in pH_o 6.20 to reduce pH_i. The cells were then perifused with 0 Cl⁻ solution containing bumetanide 500 μM for 1 min before 10 mM NH₄ ⁺ _o was substituted for Na⁺ _o. (B) A few minutes after the response of A, the same bundle attached to the pipette was perifused with 10 mM propionate in the same conditions. The propionate-induced pH_i change gave a lower limit for the speed of the change in solution at the cell membrane.

Calibration of pH_i Measurements

We initially used two techniques for calibrating pH_i. To estimate the shape of the curve that gives pH_i as a function of I₄₄₀/I₄₉₅, the cell membranes were made permeable to H⁺ with nigericin so that pH_i varied with pH_o (Thomas et al. 1979). At the end of each of 11 experiments, the cells were superfused with 130 mM K⁺ Cardinaud solution. Perfusion was stopped and nigericin was added to the chamber to a final concentration of 10 μM. After 10 min, cells were quickly superfused with 130 mM K⁺ Cardinaud solutions at different pHs. It was found that the effect of the nigericin persisted so that it was unnecessary to include it in the calibration solutions: there was no significant difference between calibration curves obtained with solutions containing 5 μM nigericin (n = 4) and those obtained without (n = 7). The value of I₄₉₅/I₄₄₀ corresponding to pH 6.84 ([I₄₉₅/I₄₄₀]_6.84) was estimated by linear interpolation for each of the 11 data sets. I₄₉₅/I₄₄₀ was described by the equation of Boyarsky et al. 1988 ():

1

The values obtained for the constants were 6.93 ± 0.03 for pK and 0.991 ± 0.022 for b (n = 11). The advantage of this procedure is that calibration for each experiment is reduced to obtaining the fluorescence ratio corresponding to pH_i 6.84. This ratio was obtained by superfusing the cells with 2 mM NH₄ ⁺ at pH_o 6.90, a procedure that we found to give a pH_i ≈ 6.84 (see Fig. 3, A–D).

Figure 3.

Absolute measurement of pH_i during application of 2 mM NH₄ ⁺. (A) Recording of BCECF fluorescence ratio from a bundle of glial cells. Different mixtures of propionate (Prop) and trimethylamine (TMA) were applied during prolonged application of 2 mM ammonium, first with pH_o = 6.90, and then with pH_o = 7.30. (B) In experiments like that of A, pH_i was calculated from responses to application of two mixtures of propionate and TMA and ascribed to the time midway between the two applications (e.g., arrows in A and B correspond to the same point). The data points shown are for an ammonium concentration of 2 mM and the solid line is a linear regression through them. The dashed line simply illustrates that the mean initial pH_i was 7.3. (C) Summary of results for the pH_i measured ∼10 min after application of 2 mM ammonium at pH 6.90. The results for different pairs of weak acids and weak bases are shown separately; for the pair propionate/TMA, a possible dependence on pH_i was examined by comparing the mean values for cells with an initial baseline pH_i < 7.10 (“acid”) and those with baseline pH_i > 7.10 (“alkaline”). Bars show SDs. (D) Absolute pH_i reached during 2 mM NH₄ ⁺ application at pH_os 6.50 (n = 7), 6.90 (n = 6), 7.30 (n = 6), and 7.70 (n = 7). The solid line is a linear regression through the data points (R = 0.996), vertical bars represent ±SD, horizontal bars represent the precision of the adjusted pH_o of the solutions. (E) Recording showing pH_i reached in nigericin 10 μM pH_o 6.90 compared with pH_i reached in NH₄ ⁺ 2 mM pH_o 6.90.

Absolute Measurement of pH_i

The null method of Eisner et al. 1989 was used. Let Δ^apH_i be the change of pH_i that would have been produced by superfusion with a concentration ^aC of a weak acid AH + A⁻ and Δ^bpH_i that for a concentration ^bC of a weak base BH⁺ + B. Assuming that the diffusion of the neutral form (AH or B) and its re-equilibration with the charged form in the cell are rapid compared with pH_i regulatory mechanisms, then:

2a

2b

where β_i is the buffering power. Let ΔpH_i be the net pH_i change produced by a simultaneous application of a concentration ^aC of AH + A⁻ and ^bC of BH⁺ + B. ΔpH_i = Δ^apH_i + Δ^bpH_i.

When ΔpH_i ≈ 0, it follows from and that:

3

To determine the ratio ^bC/^aC for which ΔpH_i ≈ 0, two pairs of concentrations (^aC; ^bC1) and (^aC; ^bC2), which gave rise to ΔpH_i1 and ΔpH_i2 were applied successively. The desired ^bC was then estimated from:

4

This method is most accurate when ^bC/^aC = 1 and hence when pH_i = pH_o; we were able to bring pH_i close to pH_o by applying NH₄ ⁺ (see Fig. 3, A–D).

Comparison of the Permeabilities of the Neutral Forms of a Weak Base and a Weak Acid

We choose a weak acid AH/A⁻ whose pK_a = ^apK_a < 5 so that at pH_o ∈ [6; 8] its total concentration C_a = [AH]_o +[A⁻]_o ≈ [A⁻]_o. We choose a weak base BH⁺/B whose pK_a = ^bpK_a > 9 so that at pH_o ∈ [6; 8] its total concentration C_b = [BH⁺]_o + [B]_o ≈ [BH⁺]_o. We set C_a = C_b and find the pH_o (∈ [6; 8]) for which the initial inward transmembrane flux of B (F_B = P_B × [B]_o) is equal to that of AH (F_AH = P_AH × [AH]_o):

5

For pH_i ∈ [6; 8], the initial rate of pH_i change induced by the weak acid is −F_AH/β_i since, in the cell, most of AH dissociates to form A⁻ + H⁺; similarly, the initial rate of pH_i change induced by the weak base is F_B/β_i. Thus, if one of the permeabilities is known, the other permeability can be deduced from the value of pH_o for which the initial direction of the pH_i change during the application of the mixture of the weak base and the weak acid reverses (F_B = F_AH).

Online Supplemental Material

The arguments leading from the observed changes in pH_i to the properties of the transporter molecule involve a model of transmembrane fluxes in the cell (essentially that used by Marcaggi et al. 1999) and a multistate model of a hypothetical NH₄ ⁺-Cl⁻ cotransporter with K⁺ inhibition. Details of these models and their analysis are available online at http://www.jgp.org/cgi/content/full/116/2/125/DC1

RESULTS

In agreement with Marcaggi et al. 1999, pH_i in bundles of glial cells superfused with solution at the physiological pH of 6.90 had values up to ∼7.55 (e.g., see Fig. 7). More acid pH_is (<7.0) were encountered in bundles that were visibly damaged or whose pH_i recovered only slowly from an acid load. In slices of bee retina, mean pH_i measured in glial cells selected for their negative membrane potentials has been reported as 7.31 (Coles et al. 1996). For this reason, and also because the amplitude of pH_i responses of isolated bundles to NH₄ ⁺ application correlated positively with pH_i (Marcaggi et al. 1999), bundles with pH_i > 7.1 were usually selected, except for some experiments on NH₃ permeability for which a more acid baseline pH_i was advantageous (e.g., see Fig. 4).

Figure 7.

pH_i changes as a function of ammonium concentration (for 30-s applications in 0 K⁺). (A) Typical recording of pH_i responses to various NH₄ ⁺ concentrations. Cells were normally superfused with 10 mM K⁺ standard solution that was switched to a 0 K⁺ solution 15 s before each NH₄ ⁺ application. The response to 10 mM propionate (near the end of the experiment) gives a lower limit for the rapidity of the solution changes. On the same cells, after a delay of ≈20 min (“Δt*”), removal of extracellular K⁺ with no ammonium application had no detectable effect on pH_i. (B) Superposition of pH_i responses to NH₄ ⁺ (from A) on a shorter time scale. Dotted lines for 0.5 and 1 mM NH₄ ⁺ are the second (control) responses. The delay between activation of the solenoid valve (t = 0) and the start of the changes in pH_i was ∼10 s. Note that during the NH₄ ⁺ application, pH_i fell more rapidly for 10 mM NH₄ ⁺ than for 20 mM, but that the peak reached after the application (Phase 4) was greater for 20 mM. (C) Calculation of δpH_i/δt(NH₄ ⁺) for each NH₄ ⁺-induced acidification. Mean slopes were measured between 15 and 35 s after the onset of the NH₄ ⁺ application by linear regression. Taking into account the delay for arrival of solutions, this interval corresponds to the last 20 s of the NH₄ ⁺ applications. (D) Mean values for ΔpH_i([NH₄ ⁺]_o) (peak change in pH_i, open symbols) and δpH_i/δt([NH₄ ⁺]_o) (filled symbols) from four experiments similar to that of A. Bars represent ± SEM. The best Michaelis-Menten fits of ΔpH_i([NH₄ ⁺]_o) for [NH₄ ⁺]_o = 0.5, 1, 2, 5, 10, and 20 mM (solid line) and of δpH_i/δt([NH₄ ⁺]_o) for [NH₄ ⁺]_o = 0.5, 1, 2, and 5 mM (dotted line) are shown.

The Ammonium-induced Decrease in pH_i Is Inhibited by a High Concentration of K⁺ _o

Fig. 2 A illustrates how 2 mM ammonium applied for 30 s to an isolated bundle of bee retinal glial cells at the measured physiological pH_o of 6.90 (Cardinaud et al. 1994) causes a decrease in pH_i, indicating entry of NH₄ ⁺. Marcaggi et al. 1999 have reported that this acidification requires external Cl⁻ (but not Na⁺) and is inhibited by bumetanide (at 100 μM) and by piretanide, properties of the family of K⁺-Cl⁻ cotransporters. Between applications of NH₄ ⁺ in Fig. 2 A, the external K⁺ concentration ([K⁺]_o) was at its normal physiological value of 10 mM (Cardinaud et al. 1994), and it was maintained at this value during the second and fifth applications of NH₄ ⁺. For the first NH₄ ⁺ application, [K⁺]_o was reduced to 1 mM, which slightly increased the acidification, and for the third application it was increased to 50 mM. Although increasing [K⁺]_o to 50 mM for up to 2 min in the absence of ammonium caused only negligible changes in pH_i (n = 6; not shown), 50 mM K⁺ _o reduced the ammonium-induced acidification to about half that in the presence of 10 mM K⁺ _o. Rb⁺, an ion that can replace K⁺ in many transport processes, produced a greater inhibition that we did not investigate further (fourth ammonium application). Hence, Fig. 2 A suggests that inward transport of NH₄ ⁺ is inhibited by K⁺ _o and Rb⁺ _o (as would be expected if NH₄ ⁺ and K⁺ (and Rb⁺) competed for the same transporting site, for example). More interestingly, the inhibition may be relatively weak: increasing [K⁺]_o from 10 to 50 mM only halved the pH_i response to 2 mM ammonium, suggesting that the transport may be selective for NH₄ ⁺ over K⁺. To quantify this selectivity from measurements of pH_i, it was necessary to have a scheme of the transmembrane fluxes of NH₄ ⁺ and NH₃ and to determine parameters relating these fluxes to changes in pH_i.

Parameters to be Determined

Fig. 2 B shows a response to a longer (5 min) application of ammonium on an expanded time scale. This response can be divided into five phases that can be explained by the schemes of Fig. 2 C (see also Boron and De Weer 1976; Marcaggi et al. 1999). Initially, pH_i increases because of the predominant effect of the rapid entry of NH₃, which combines with H⁺ (Phase 1). Since the equilibrium of the reaction NH₃ + H⁺ ↔ NH₄ ⁺ is so far to the right at pH near 7, it is sufficient that the inward flux of NH₃ exceeds ∼1% of the inward flux of NH₄ ⁺. As the ratio [NH₄ ⁺]_i/NH₃]_i ∝ [H⁺]_i, Phase 1 is expected to be greater for cells with acid pH_i. This was actually the case: Phase 1 was detected only for cells with baseline pH_i < 7.2. When the NH₃ concentrations approach equality on each side of the cell membrane, there is still an inward NH₄ ⁺ gradient because [H⁺]_o > [H⁺]_i. Then the NH₃ flux becomes outward while NH₄ ⁺ continues to enter the cells and release H⁺ ions (Phase 2). A steady state is reached (Phase 3) when the production of H⁺ ions equals their extrusion by pH regulatory processes. When extracellular ammonium is suddenly removed, intracellular ammonium exits the cell faster in the NH₃ form than in the NH₄ ⁺ form, so NH₄ ⁺ dissociates to form NH₃ and there is a rebound acidification (Phase 4), followed by a slower return to baseline as proton equivalents are pumped out of the cell (Phase 5).

Marcaggi et al. 1999 showed that a simple mathematical model based on Fig. 2 D could simulate the main features of the experimental records, but used parameters that were only roughly estimated. We have now made more precise measurements of the following parameters required by the model: the absolute values of baseline pH_i, the buffering power (β_i), the NH₃ permeability (P_NH3), the pH_i regulation rate (characterized by a time constant τ_reg), and the Cl⁻ concentration gradient that can help drive NH₄ ⁺ into the cell.

Absolute Determination of pH_i During Application of NH₄ ⁺

The precise value of pH_o − pH_i in the presence of external ammonium (the plateau phase) is related to the force driving NH₄ ⁺ across the membrane and is our main motivation for seeking an accurate measure of pH_i. Marcaggi et al. 1999 calibrated their measurements by applying nigericin, but it has been shown that this technique can give systematic errors (Nett and Deitmer 1996; Boyarsky et al. 1996). To determine the absolute value of pH_i during NH₄ ⁺ perfusion, we applied a weak acid and a weak base simultaneously as described by Eisner et al. 1989 (see materials and methods). Fig. 3 A shows a typical experiment. NH₄ ⁺ was first applied at pH_o 6.90 and then at pH_o 7.30. At each plateau phase, ΔpH_i (10 mM propionate, 10 mM TMA) and ΔpH_i (10 mM propionate, 5 mM TMA) were in opposite directions, and we estimated by linear interpolation the concentration of TMA that would have given no change in pH_i when applied with 10 mM propionate (). The absolute value of pH_i was then calculated by . The method assumes that intracellular pK_a equals extracellular pK_a and that the membranes are relatively impermeable to the charged forms of the weak acid and base. This latter assumption was confirmed by the observation that, during applications of propionate (n = 21; not shown) or TMA (see Fig. 5 A), recovery of pH_i was slow and could be fully accounted for by pH regulatory processes. Since pH_i during the plateau phase depends partly on pH_i regulatory processes (Fig. 2 C 3), short NH₄ ⁺ applications at the beginning and end of the experiment were made to check that the rates of recovery remained approximately the same. From 12 experiments, as in Fig. 3 A, pH_i was calculated to be 6.844 ± 0.017 (±SD, n = 12) after an 8-min application of NH₄ ⁺ with pH_o = 6.90. This pH_i was significantly less than pH_o (P < 0.0001) and remained so for at least 35 min (Fig. 3 B). To see whether the result depended on the specific weak acid and weak base used, we used other weak acid/weak base couples. In experiments similar to that of Fig. 3 A, the estimated difference pH_o − pH_i after >10 min of perfusion with 2 mM ammonium was not significantly different when the following couples were used: propionate/TMA; propionate/MA; acetate/TMA and caproate/TMA (Fig. 3 C). To see whether the value of (pH_o − pH_i) reached during the plateau phase was related to the baseline pH_i, we compared cells with a baseline pH_i < 7.1 with those with pH_i > 7.1 (Fig. 3 C, first two columns). The difference in the mean values of (pH_o − pH_i) during the plateau phase was not significant. In contrast, as illustrated in Fig. 3 A, the level of the plateau did indeed depend strongly on the pH_o at which the NH₄ ⁺ was applied. Absolute values of pH_i estimated during superfusion with 2 mM NH₄ ⁺ at pH_o 6.500 ± 0.005, 6.900 ± 0.005, 7.300 ± 0.005, and 7.700 ± 0.005 are plotted in Fig. 3 D and show a very precise linear correlation with pH_o such that (pH_i − 6.142) = 0.9264 × (pH_o − 6.142). In later experiments, we calibrated the measurements of pH_i simply by superfusing the cells with 2 mM ammonium for at least 8 min and using this relation.

Figure 5.

Indirect measurement of NH₃ permeability. (A) Direct measurement of P_MA. The pH_i change induced by 10 mM methylamine (MA) (thick line) was compared with that induced by 10 mM trimethylamine (TMA) (thin line) and to the change in fluorescence on switching to a solution containing 1μg liter⁻¹ fluorescein (dotted line). The microscope was focused on the same bundle for the three recordings. (B) Relative cell membrane permeabilities to neutral forms of propionate (Prop) and MA. A bundle of cells was superfused with 2 mM NH₄ ⁺ so that pH_i was maintained at a value close to pH_o (Fig. 3 D). When 10 mM Prop + 10 mM MA was applied at pH_o 7.2, entry of the neutral form of Prop initially predominated, but as [Prop]_i increased, entry of MA began to predominate and pH_i started to increase. The initial change in pH_i reversed for a value of pH_o between 7.40 and 7.60. (C) Relative cell membrane permeabilities to NH₃ and to the neutral form of Prop. A mixture of 5 mM NH₄ ⁺ and 5 mM Prop was applied to cells superfused with a 0 Cl⁻ solution at pH_o 7.10 (thick trace) or 7.00 (thin trace). The two recordings were from the same bundle of cells. At pH_o 7.10, the initial change in pH_i was an increase, while at pH_o 7.00 it was a decrease.

Intracellular Buffering Power

The H⁺ ions released into (or taken up from) the cytoplasm as a consequence of the transmembrane fluxes of NH₄ ⁺ and NH₃ affect pH_i according to the relation ΔpH_i = ΔQ/β_i, where ΔQ is the quantity of H⁺ ions/U volume and β_i is the intracellular buffering power (see Roos and Boron 1981). We estimated β_i by applying the weak acid propionate (as in Figure 1 of Marcaggi et al. 1996); the change in pH_i reached its maximum very rapidly compared with the time course of pH_i recovery in the presence of propionate, and we did not attempt to block pH_i regulation (compare Szatkowski and Thomas 1989). Mean β_i was 12.2 ± 2.9 mM (n = 11) and we took 12 mM for the model.

To see if β_i varied markedly with pH_i, we shifted pH_i by applying NH₄ ⁺ at various pH_os. The results and the analysis, which is complicated by the effects of the NH₃/NH₄ ⁺ system, are given in Marcaggi 1999; the conclusion is that β_i is effectively constant in the range 6.7–7.3.

Permeability to NH₃

We estimated NH₃ permeability (P_NH3) from measurements of pH_i under conditions in which entry of NH₄ ⁺ was blocked so that changes in pH_i were due only to the inward flux of NH₃. We have previously shown that NH₄ ⁺ does not enter through barium-sensitive K⁺ channels, the major cationic conductance in these cells, and also that NH₄ ⁺ entry is totally blocked by bumetanide or by removal of external chloride (Marcaggi et al. 1999). We therefore applied ammonium in Cl⁻-free solutions to which, in some cases, bumetanide had been added, and measured the rate of change of pH_i (in the alkaline direction) induced by NH₃ entry into the cells.

With the cell bundles adhering to the floor of the perfusion chamber, we failed to find a molecule causing a 10–90% pH_i change faster than the one produced by ammonium: perhaps the change of solution at the cell membrane (0–20 μm from the floor of the chamber) was not fast enough for this measurement. To expose cells to faster solution changes, we caught hold of bundles of cells with a 3-μm tip diameter pipette and carried them 50–100 μm up from the floor of the chamber. To increase the time resolution of the rapid initial slope of the pH change, we measured the fluorescence ratio with faster switching of the excitation wavelengths (>3 Hz). To reduce delays due to diffusion, we applied ammonium at a high concentration (10 mM) but at an acid pH (6.50) so that [NH₃]_o was low but benefited from facilitated diffusion (Engasser and Horvath 1974). To increase the NH₃-induced ΔpH_i, baseline pH_i was reduced (to ∼6.80) by perfusing the cells for 30–60 min with solution buffered at pH_o 6.20. After a 1-min perfusion with 0 Cl⁻ + 0.5 mM bumetanide, 10 mM ammonium was applied at pH_o 6.50 (Fig. 4 A). In these conditions, the 10–90% pH_i change induced by 10 mM propionate (Fig. 4 B) was twice as fast as the one induced by ammonium, showing that the speed of solution change at the cell membrane did not significantly limit the influx of NH₃.

Although pH_i was ∼7.00 during the ammonium application, while pH_o was 6.50, no slow pH_i decrease was observed, as would have been the case if the membranes had had some permeability to NH₄ ⁺. This confirms that NH₄ ⁺ pathways were insignificant in these conditions.

The slope of the pH_i change was measured at 50% of the ΔpH_i response, where it is known that [NH₃]_o has reached its final concentration, since the effect of propionate is 90% at this time (Fig. 4). Since, for pH_i ∼ 7.00, [NH₄ ⁺]_i > 100 × [NH₃]_i, then β_i × δpH_i/δt = δ [NH₄ ⁺]_i/δt ≈ δ ([NH₄ ⁺]_i + [NH₃]_i)/ δt = F_NH3 × ^S/_V, where F_NH3 is the NH₃ transmembrane flux and ^S/_V is the ratio of membrane surface to intracellular volume in which the ammonium is distributed. The ratio of the surface to the total cell volume has been estimated to be ≈1.2 μm⁻¹ (Marcaggi et al. 1999), but the ammonium will be present almost entirely in the water phase that occupies 0.775 of the total volume of the tissue (Coles and Rick 1985). Taking this factor for the water content of the glial cells gives an estimated effective ^S/_V of 1.55 μm⁻¹. Knowing β_i, we could then calculate the transmembrane flux of NH₃ from the rate of change of pH_i: F_NH3 ≈ ^V/_S × β_i × δpH_i/δt. From this flux, we found P_NH3 = 14.7 ± 2.9 μm · s⁻¹ (n = 6) in 0 Cl⁻ + 0.5 mM bumetanide, which was not significantly different from the value in 0 Cl⁻ only (n = 5), showing that bumetanide did not further inhibit NH₄ ⁺ entry.

Holding up the cells with a pipette will have introduced some stress in the cell membrane, which may have modified its permeability. To check that NH₃ permeability is the same for cells plated on the bottom of the chamber (the conditions used for the other experiments), we also determined P_NH3 by an indirect method. Methylamine (MA; CH₃NH₃ ⁺/CH₃NH₂) is a weak base with a pK_a that is high (≈10.6; Robinson and Stokes 1959) compared with that of ammonium (≈9.2). Because of this pK_a difference, at pH < 8 (at which charged forms are preponderant), if [CH₃NH₃ ⁺] + [CH₃NH₂] = [NH₄ ⁺] + [NH₃], then [CH₃NH₂] < 0.04 × [NH₃]. It follows that if P_CH3NH2 (P_MA) is not far different from P_NH3, for equal concentrations of MA and ammonium applied, F_CH3NH2 << F_NH3. This is why P_MA can be measured directly even with a slow speed of solution change at the cell membrane. Fig. 5 A illustrates the pH_i response to 10 mM MA compared with the pH_i response to 10 mM TMA (pK_a ≈ 9.6). The speed of the pH_i change induced by TMA was far faster than the one induced by MA, showing that the speed of the solution change was fast enough for measurement of P_MA, which was found to be 27.4 ± 8.1 μm s⁻¹ (n = 8). Once this permeability was known, it was possible to deduce the permeability of propionate by ascertaining the initial direction of the pH_i change induced by a simultaneous application of 10 mM propionate and 10 mM MA. To avoid too great a variation of the net ΔpH_i during this simultaneous application, we used a condition in which pH_i ≈ pH_o, which was obtained by including 2 mM ammonium in the superfusate (Fig. 3 D). As illustrated in Fig. 5 B, the initial direction of the pH_i change reversed for 7.40 < pH_o < 7.60 (n = 4). Taking the mean of this range (pH_o ≈ 7.50 ± 0.10) gives x = P_MA/P_prop ≈ 3.16 ± 1.46 according to . So, P_prop = P_MA/x ≈ 8.67 ± 6.57 μm s⁻¹.

The same protocol was used to estimate P_NH3 from the now known P_prop, the simultaneous application being done in 0 Cl⁻ to prevent the entry of NH₄ ⁺. As illustrated in Fig. 5 C, the initial direction of the pH_i change reversed for 7.00 < pH_o < 7.10 (n = 4), which gives x = P_NH3/P_prop ≈ 1.02 ± 0.24 according to . So P_NH3 = x × P_prop ≈ 8.84 ± 8.78 μm s⁻¹.

In conclusion, the two methods of estimation of P_NH3 gave values not significantly different. The standard deviation obtained with the second method was increased by the successive approximations so we give more weight to the value obtained with the first method and conclude that P_NH3 is in the range of 7–19 μm s⁻¹; we take the value 13 μm s⁻¹ for the model.

pH Regulation

When pH_i falls below its baseline value, pH regulatory mechanisms tend to restore it by extruding H⁺ ions. To quantify the kinetics of this regulation, we acid loaded the cells by exposure to ammonium and analyzed the recovery (Roos and Boron 1981; Thomas 1984). Fig. 6 A shows the recoveries of pH_i in a single bundle of cells after initial displacements of various amplitudes induced by applications of ammonium at various concentrations. For each ammonium application, the recovery was analyzed for 8 min starting 45 s after the end of the application («t _f») to allow for the rebound acidification (Phase 4). The plot is semilogarithmic, the ordinate being ln([H⁺]_i − [H⁺]_∞), where [H⁺]_∞ was the baseline [H⁺]_i at rest. Linear regressions showed that the recoveries were exponential irrespective of the initial displacement, and had slopes (= −1/τ_reg) that were not systematically different.

Figure 6.

pH_i recovery from ammonium-induced acid loads. (A) Analysis of recoveries from acidifications induced by applications of 0.5, 1, 2, 5, and 10 mM NH₄ ⁺ on the same bundle of glial cells. ln([H⁺]_i − [H⁺]_∞) was plotted against t − t _f, where t _f is 45 s after the end of the NH₄ ⁺ superfusion. From linear regressions of the data points, τ_reg was calculated as being (min): 2.74 (correlation coefficient, R = 0.761), 3.96 (R = 0.971), 3.29 (R = 0.978), 3.68 (R = 0.996), and 3.75 (R = 0.998) for the recoveries from the increasing acidifications induced by 0.5, 1, 2, 5, and 10 mM NH₄ ⁺. (B) τ_reg, calculated as in A, as a function of ΔpH_i at the beginning of the recovery. The data are from 17 cell bundles (each represented by a different symbol) for which at least three different [NH₄ ⁺]_o were applied. For each cell bundle, a linear regression was calculated and a line with the mean of their slopes [(0.3 min (pH unit )⁻¹] is shown passing through the barycenter of the points.

Values of the time constant τ_reg for 17 bundles of cells for which at least three different ammonium concentrations were tested were plotted as a function of the initial pH_i displacement, ΔpH_i(NH₄ ⁺) (Fig. 6 B). Linear regression of τ_reg[ΔpH_i(NH₄ ⁺)] confirmed that τ_reg was independent of the pH [mean slope of 0.3 ± 1.8 min (pH unit)⁻¹; n = 17]. We conclude that despite considerable variability, τ_reg was approximately constant irrespective of the initial pH_i displacement with a mean value of 3.0 ± 1.1 min (n = 17). We therefore described the pH_i regulation by :

6

with τ_reg = 3 min.

Driving Force

The flux rate of a Cl⁻ cotransporter will depend in part on [Cl⁻]_o and [Cl⁻]_i, and we will use values of these concentrations in the transporter model of Fig. 10 A (see online supplemental material). [Cl⁻]_o being known, we attempted to estimate [Cl⁻]_i. Measurements in slices of bee retina with ion-selective microelectrodes have shown that in the glial cells Cl⁻ (and also K⁺) are at close to electrochemical equilibrium (Coles et al. 1986, Coles et al. 1989). Hence, [Cl⁻]_i/[Cl⁻]_o could be deduced approximately from V_m. We did not succeed in measuring V_m in the isolated bundles of glial cells directly (by electrode techniques) and used an indirect method. We argued that if the membranes were made permeable to H⁺ then [H⁺]_i would be determined by V_m. We applied the H⁺/K⁺ exchanger, nigericin, in solutions with normal [K⁺]_o and observed the resulting change in pH_i (Fig. 3 E). The minimum pH_i reached during nigericin was compared with pH_i at the plateau phase induced by 2 mM NH₄ ⁺, the mean difference being 0.02 ± 0.05 (n = 7; Fig. 3 C). It follows that the mean value of pH_o − pH_i at the maximum of the nigericin-induced pH_i change was ∼0.07 (Fig. 3 C). On the assumption that H⁺, being now in equilibrium with K⁺, was distributed passively across the membrane, V_m was −4 mV. Since nigericin is not perfectly selective for H⁺ (Pressman et al. 1967; Margolis et al. 1989), it probably depolarized the membranes somewhat, as suggested by the slow increase in pH_i during nigericin (Fig. 3 E). Thus, the true V_m is probably more negative than −4 mV and [Cl⁻]_o/[Cl⁻]_i = exp(−V_m F/RT) > 1.18.

Figure 10.

pH_i responses to stepwise increases in [NH₄ ⁺]_o are accounted for by a model of the transport process. (A) Kinetic scheme of the cotransport of Cl⁻ and NH₄ ⁺ or K⁺. The unloaded transporter molecule is symbolized by X; o or i indicate the position of the transporter at the external or internal side of the cell membrane. Binding ions Cl⁻, NH₄ ⁺, and K⁺ are not shown in this scheme for visual simplicity. The three binding constants K _c, K _m, and K _i for Cl⁻, NH₄ ⁺, and K⁺ are unaffected by the side to which the transporter faces. Two kinetic constants, k and g, describe the transit steps of the unloaded and loaded transporter. (B) Experimental response to stepwise increases in [NH₄ ⁺]_o. (C) Simulated response of the cell model of Fig. 2 D including the transporter model of A to stepwise increases in [NH₄ ⁺]_o with affinity for NH₄ ⁺, K _m = 5 mM (continuous line) and 20 mM (dashed line). K _i was 15 mM.

Concentration Dependence of the pH_i Changes Induced by 30-s Applications of Ammonium

To record the responses to increasing concentrations of NH₄ ⁺ in the absence of external K⁺, we superfused the cells in 0 K⁺ for 15 s before and during each NH₄ ⁺ application (Fig. 7 A). Repeated exposure to high [NH₄ ⁺] appeared to lead to impairment of pH_i regulation and, for 7 of 11 experiments, pH_i did not recover from the acidification induced by 10 mM NH₄ ⁺. Measurements were therefore made only on the records from the four experiments for which pH_i recovered from 10 mM NH₄ ⁺ and for which the response to subsequent control application of 0.5 or 1 mM NH₄ ⁺ was closely similar to the initial response (in the record of Fig. 7 A, a final application of 20 mM NH₄ ⁺ was made). To make sure that the effect of NH₄ ⁺ was not rate limited by the speed of the solution change (as was probably the case in the previous study by Marcaggi et al. 1999), we checked that application of propionate gave a more rapid pH_i change (Fig. 7 A). Data were analyzed only for experiments in which the time for the 10–90% propionate-induced pH_i change was <15 s. The last part of Fig. 7 A shows that 0 K⁺ alone did not affect pH_i on the time and pH scales of these experiments.

Fig. 7 B shows the NH₄ ⁺ responses from the record of Fig. 7 A on a shorter time scale. The time of onset of the response to propionate (not shown) indicated that in this experiment there was a dead time of ∼5 s between the switching of the electromagnetic valves and the arrival of a new solution at the cell membrane. The slope of the NH₄ ⁺-induced pH_i change (δpH_i/δt) was measured before the rebound (Phase 4), between 15 and 35 s after the valves were actuated. δpH_i/δt(NH₄ ⁺) was calculated by linear regression as shown in Fig. 7 C. δpH_i/δt(NH₄ ⁺) increased with [NH₄ ⁺]_o in the range 0.5–10 mM NH₄ ⁺; but for 20 mM NH₄ ⁺, although the total pH_i change induced by NH₄ ⁺ [ΔpH_i(NH₄ ⁺)] continued in every case to increase, in three of the four experiments, δpH_i/δt for 20 mM NH₄ ⁺ was less than for 10 mM, as in the example shown in Fig. 7A–C. Mean data from the four experiments are shown in Fig. 7 D. ΔpH_i([NH₄ ⁺]_o) was well fitted by a Michaelis-Menten curve ΔpH_i([NH₄ ⁺]_o) = a[NH₄ ⁺]_o/(b + [NH₄ ⁺]_o) (R = 0.993) with half saturation, b, for [NH₄ ⁺]_o = 6.62 ± 0.57 mM (n = 4). But δpH_i/δt([NH₄ ⁺]_o) could only be fitted by a Michaelis-Menten curve for [NH₄ ⁺] ≤ 5 mM (R = 0.926), half saturation being at [NH₄ ⁺]_o = 3.37 ± 0.98 mM (n = 6) (Fig. 7 D). Because the relation between NH₄ ⁺ transport and pH_i changes is indirect, the value of [NH₄ ⁺]_o that half saturates pH_i changes does not necessarily correspond to the one that half saturates transport of NH₄ ⁺. To deduce the flux of NH₄ ⁺, we had recourse to a mathematical model.

Dependence of NH₄ ⁺ Flux on [NH₄ ⁺]_o

Three transmembrane fluxes determine pH_i during and after application of ammonium (Fig. 2 C). Of these, we have a phenomenological description of the pH_i regulation (F_reg in Fig. 2 D), and we assume that the flux of NH₃ (F_NH3 in Fig. 2 D) results from simple diffusion (Fick's law). To deduce the flux of NH₄ ⁺ through the cotransporter (F_NH4 in Fig. 2 D) from the changes in pH_i, we use the model of Fig. 2 D, expressed mathematically in the supplemental material. From the measurements described above, values for parameters of the model were: β_i = 12 mM, pH_i = 7.4, τ_reg = 3 min, and P_NH3 = 13 μm s⁻¹. The surface-to-volume ratio, ^S/_V, with its attendant uncertainty, was used to calculate P_NH3, but cancels out in the calculations.

As a first step, a constant inward F_NH4 (ⁱⁿF_NH4) was imposed for 30 s, with [NH₄ ⁺]_o (+ [NH₃]_o) set to 2 mM. The resulting δpH_i/δt was calculated 15 s after the onset of the imposed ⁱⁿF_NH4 and plotted against ⁱⁿF_NH4 for various P_NH3 (7, 13, and 19 μm s⁻¹; Fig. 8 A). Increasing P_NH3 increased δpH_i/δt, but only slightly, showing that P_NH3 is not a major rate-limiting factor.

Figure 8.

Use of the cell model (Fig. 2 D) to derive inward F_NH4([NH₄ ⁺]_o) from measured pH_i changes induced by brief applications of NH₄ ⁺. (A) Constant inward F_NH4 (ⁱⁿF_NH4) was imposed on the model for 30 s. δpH_i/δt, calculated 15 s after onset, was plotted against ⁱⁿF_NH4. Data are shown for simulations with [NH₄ ⁺]_o set to 2 mM (to fix [NH₃]_o) and P_NH3 = 7, 13, and 19 μm s⁻¹. (B) As in A, δpH_i/δt was plotted versus ⁱⁿF_NH4 for simulations with [NH₄ ⁺]_o = 0.5 mM (triangles), 1 and 2 mM (diamonds), and 5, 10, and 20 mM (circles). P_NH3 = 13 μm s⁻¹. Because increasing [NH₄ ⁺]_o increases [NH₃]_o, δpH_i/δt(ⁱⁿF_NH4) is smaller for higher [NH₄ ⁺]_o. (C) Simulations in which an ⁱⁿF_NH4 was imposed for 30 s and followed by an outward F_NH4 (^outF_NH4). ⁱⁿF_NH4 was set to 6.65 mM min⁻¹ and [NH₃]_o was fixed by setting [NH₄ ⁺]_o = 2 mM; this gave δpH_i/δt = 0.44 pH unit min⁻¹, equal to the mean measured pH_i change induced by 2 mM NH₄ ⁺ in 0 K⁺ for cells with baseline pH_i ≈ 7.4. At t = 30 s, F_NH4 switched instantaneously from ⁱⁿF_NH4 to maximum ^outF_NH4, ^outF_{NH4 max}, and decreased to zero as [NH₄ ⁺]_i decreased to zero (see text). Simulations for ^outF_{NH4 max} equal to 0, −6.65, and −20 mM min⁻¹ show that the rebound acidification after 30 s decreased when ^outF_{NH4 max} increased. (D) Plot of ΔpH_i(ⁱⁿF_NH4) measured as baseline pH_i (7.4) minus the minimal pH_i reached during the rebound acidification after 30 s of influx ⁱⁿF_NH4. Simulations for 0.5 mM (triangles) or 20 mM (circles) NH₄ ⁺ _o (+NH_3o) show that [NH₃]_o has little effect on ΔpH_i(ⁱⁿF_NH4). ^outF_{NH4 max} = 0. (E) ⁱⁿF_NH4([NH₄ ⁺]_o) calculated from δpH_i/δt([NH₄ ⁺]_o) (•) and ΔpH_i([NH₄ ⁺]_o) (○) of Fig. 7 D. The points were fitted by Michaelis-Menten curves (R = 0.963 and 0.994, respectively) with apparent constants K ′_m equal to 5.9 ± 1.3 and 7.8 ± 0.7 mM.

A similar simulation, still using an imposed ⁱⁿF_NH4, was then performed in the presence of various [NH₄ ⁺]_o (+ [NH₃]_o). Increasing [NH₄ ⁺]_o increased [NH₃]_o, reduced outward F_NH3, and, as expected, reduced δpH_i/δt(ⁱⁿF_NH4) (Fig. 8 B). It is clear that the experimental result in which δpH_i/δt was smaller for an application of 20 mM ammonium than for 10 mM (Fig. 7) does not necessarily imply that inward F_NH4(20 mM NH₄ ⁺) < inward F_NH4(10 mM NH₄ ⁺). We also note that since the relation of δpH_i/δt to ⁱⁿF_NH4 is curved (Fig. 8A and Fig. B), δpH_i/δt vs. [NH₄ ⁺]_o will saturate more rapidly than will ⁱⁿF_NH4 vs. [NH₄ ⁺]_o.

The pH_i peak reached after withdrawal of external ammonium must depend both on [NH₄ ⁺]_i at the end of the ammonium application, and on the effluxes of NH₄ ⁺ and of NH₃ after withdrawal. To start modeling this, we considered the case of a 30-s application of 2 mM extracellular NH₄ ⁺ with a constant ⁱⁿF_NH4 (6.65 mM min⁻¹ in Fig. 8 C). After removal of extracellular NH₄ ⁺, the concentration gradient of NH₄ ⁺ is outwards. We tested the simplest reasonable assumption, which is that outward F_NH4 = ^outF_NH4 ∝ [NH₄ ⁺]_i. With no loss of generality, this can be written: ^outF_NH4 = ^outF_{NH4 max} × ([NH₄ ⁺]_i/[NH₄ ⁺]_{i max}), where [NH₄ ⁺]_{i max} is the intracellular NH₄ ⁺ concentration reached at t = 30 s and ^outF_{NH4 max} is an initially arbitrary constant corresponding to the maximum transient ^outF_NH4. As illustrated in Fig. 8 C, the rebound acidification on removal of extracellular NH₄ ⁺ is maximal for zero ^outF_NH4 and decreases with increasing ^outF_NH4. Let ΔpH_i(ⁱⁿF_NH4) be the total pH_i change induced by a 30-s ⁱⁿF_NH4 followed by an ^outF_NH4 defined as above. From the experimental data, the mean ratio ΔpH_i/(δpH_i/δt) measured from 30-s applications of 2 mM NH₄ ⁺ in 0 K⁺ _o was 0.60 ± 0.12 min (n = 10); the closest approach to this in Fig. 8 C is 0.56 min for ^outF_{NH4 max} = 0, which we accept as an approximation. ΔpH_i(ⁱⁿF_NH4) is very little affected by [NH₄ ⁺]_o (still for an imposed ⁱⁿF_NH4; Fig. 8 D), much less so than is δpH_i/δt (Fig. 8 B). Thus, the inverse operation of estimating inward F_NH4([NH₄ ⁺]_o) is better done from ΔpH_i([NH₄ ⁺]_o) than from δpH_i/δt([NH₄ ⁺]_o). By comparing experimental ΔpH_i([NH₄ ⁺]_o) (Fig. 7 D) and simulated ΔpH_i(ⁱⁿF_NH4) (Fig. 8 D), we calculated ⁱⁿF_NH4([NH₄ ⁺]_o) (Fig. 8 E). The points were well fitted by a Michaelis-Menten equation of the form ():

7

The constant, K ′_m, corresponding to half saturation of inward F_NH4, was 7.8 ± 0.7 mM. Variant analyses {from δpH_i/δt([NH₄ ⁺]_o or using Lineweaver-Burke plots} all gave lower values, down to 4.9 mM (Marcaggi 1999. We conclude that K ′_m = 7.8 mM is a conservative estimate of the affinity (an upper limit for K ′_m) of the transporter for NH₄ ⁺ in these experimental conditions.

Functional Selectivity for NH₄ ⁺ over K⁺

Having established the dependence of ⁱⁿF_NH4 on [NH₄ ⁺]_o (for 30-s applications of ammonium), we then extended the approach to analyze the inhibitory effect of K⁺. Fig. 9 A shows an experiment in which cells were superfused for 30 s with NH₄ ⁺ in 0 or 10 mM K⁺. ΔpH_i([NH₄ ⁺]) from six such experiments is shown plotted with double inverse scales as a function of [NH₄ ⁺]_o in Fig. 9 B. Using the model, as described above, a value of ⁱⁿF_NH4([NH₄ ⁺]) was deduced for each measurement of ΔpH_i([NH₄ ⁺]) and a second inverse plot was made (Fig. 9 C). This plot suggests that the inhibition was competitive since straight lines passing through the data points intersect near the ordinate axis (same ⁱⁿF_{NH4 max}).

Fig. 9 D illustrates how K⁺ depolarizes these glial cells. In this record, from a glial cell in a retinal slice, the depolarization is greatly damped by electrical coupling between the cells and the slowness of the increase in [K⁺] in the extracellular clefts (Coles and Orkand 1983). But in isolated cell bundles, the depolarization might be greater and in some way affect NH₄ ⁺ uptake. We therefore used Ba²⁺, which blocks the depolarization for at least 45 s after the application of K⁺ (Fig. 9 D), to study the effect of K⁺ on ΔpH_i(NH₄ ⁺) in the absence of changes in membrane potential. In confirmation of Marcaggi et al. 1999, Ba²⁺ (at 5 mM) had in itself no effect on ΔpH_i(NH₄ ⁺) (n = 5; not shown). Nor did it have a significant effect on the inhibition of ΔpH_i(NH₄ ⁺) produced by raising K⁺ to 20 mM (n = 11; not shown). Hence, the depolarization is unlikely to be responsible for the inhibition of NH₄ ⁺ transport by extracellular K⁺.

To quantify the inhibitory effect of K⁺, we calculated an apparent inhibitory constant, K ′_i, defined by: K ″_m = K ′_m (1 + [K⁺]_o/K ′_i), where K ′_m is the Michaelis-Menten constant estimated above from responses to NH₄ ⁺ in 0 K⁺ and K ″_m is the constant estimated from the responses to NH₄ ⁺ in 10 mM K⁺. K ′_i was found to be 26.7 mM, which is greater than K ′_m (≤7.8 mM). Variant analyses also gave K ′_i > K ′_m (see Marcaggi 1999).

Affinities of the Transporter Molecule for NH₄ ⁺ and K⁺

In the previous two sections, we established the dependence of the mean inward F_NH4 on [NH₄ ⁺]_o during 30-s (brief) applications of ammonium, and the inhibition of this flux by [K⁺]_o. We now describe the changes in pH_i under more varied conditions; notably, longer applications of NH₄ ⁺. These more complex responses impose additional constraints on the interpretation of the underlying processes and allow us to test whether the transport can be described by a standard minimal kinetic model of membrane cotransport to which we add competition by K⁺ for the NH₄ ⁺ binding site (Fig. 10 A). As explained by Sanders et al. 1984, the kinetic behavior of a cotransporter can be accounted for by models with different orders of binding of ions; we have chosen to consider binding first of Cl⁻ then K⁺ or NH₄ ⁺. This and other assumptions implicit in the model of Fig. 10 A, their justification where possible, and the techniques used to deduce the parameter values from experimental responses to NH₄ ⁺ are given in the online supplemental material. Using the experimental results obtained so far, the model suggested that K _m, the binding affinity for NH₄ ⁺, was 6–8 mM and K _i, the binding affinity for K⁺, was in the range 10–20 mM, values that we now refine.

In the experiment of Fig. 10 B, [NH₄ ⁺]_o was increased in steps, each lasting 8 min. The level of the plateau phase (Phase 3 in Fig. 2 B) rose no further for [NH₄ ⁺]_o > 5 mM (n = 4). Fig. 10 C shows simulated responses to the same protocol of stepwise increases in [NH₄ ⁺]_o for a cell containing the transporter model of A with K _i = 15 mM and K _m = 5 mM (continuous trace) or 20 mM (dashed trace). It is seen that the time course of pH_i, particularly for the step change [NH₄ ⁺]_o from 5 to 10 mM, is better simulated with K _m = 5 mM; i.e., with K _m < K _i.

Increasing [K⁺]_o in the Presence of NH₄ ⁺

Fig. 11 A illustrates the inhibitory effect on NH₄ ⁺ transport of increasing [K⁺]_o during the plateau phase induced by a long application of NH₄ ⁺. An increase in [K⁺]_o from 10 to 50 mM rapidly increased pH_i by 0.092 ± 0.012 pH unit in 2 min in 2 mM NH₄ ⁺ (n = 5) and by a greater amount, 0.115 ± 0.022 pH unit in 20 mM NH₄ ⁺ (n = 5). The difference is significant with P = 0.01. This observation raised the question of whether the inhibition by [K⁺]_o was purely competitive.

Figure 11.

Effects of increasing [K⁺]_o from 10 to 50 mM on the plateau phase pH_i during superfusion with 2 mM or 20 mM NH₄ ⁺. (A) Typical recording. At the beginning and end of the experiment, the cells were superfused with 2 mM NH₄ ⁺ at pH_o 7.50 to provide a pH_i calibration (see Fig. 3 D). (B) Simulated effect of increasing [K⁺]_o. In the model, [K⁺]_i was kept constant when [K⁺]_o was increased. The cell model included the transporter of Fig. 10 A (see online supplemental material for full list of parameter values). K _m was 7 mM and K _i was 10, 15, or 20 mM (superposed traces), the greatest inhibition being for K _i = 10 mM. (C) Simulation with the model modified so that transport of NH₄ ⁺ was inhibited by extracellular K⁺ in a noncompetitive manner. The transporter of Fig. 10 A was modified by removing the effect of increasing [K⁺]_o on the binding step XClo → XClKo (this modification amounts to suppressing this binding step with no modification of the other parameters) and by the addition of an inhibitory step XClNH₄o → XClNH₄Ko, whose affinity constant for K⁺ _o, K _K, was 5, 10, or 15 mM (superposed traces), the greatest inhibition being for K _K = 5 mM.

Simulations were performed with the transporter model of Fig. 10 A and the protocol of Fig. 11 A. K _m was set to 7 mM. Simulations with K _i = 10, 15, and 20 mM (Fig. 11 B) show that inhibition by a 2-min increase in [K⁺]_o from 10 to 50 mM differed from the experimental record in three aspects. First, the inhibition in 20 mM NH₄ ⁺, although larger than the inhibition in 2 mM NH₄ ⁺, was not as markedly larger as in the experiments. Second, the increases in pH_i induced by rises in [K⁺]_o were slower than the experimental ones. Third, after returning to 10 mM K⁺, the small rebound acidification present in the experimental records was not reproduced. A transporter model in which inhibition by extracellular K⁺ was noncompetitive (Fig. 11 C, legend) corrected these failings, but excessively so. We did not attempt to fit the experimental data more precisely since our transporter model is highly simplified, but these comparisons to simulations do suggest that inhibition by extracellular K⁺ may be partly noncompetitive.

DISCUSSION

Sensitivity to loop diuretics and external chloride (Marcaggi et al. 1999) indicate that the cotransporter studied belongs to the electroneutral cation-chloride cotransporter family (Haas and Forbush 1998). Despite the electroneutrality of the process and the simultaneous flux of NH₃ across the membrane (Marcaggi et al. 1999), we have quantified the transport of NH₄ ⁺ on the cotransporter after first measuring several parameters (buffering power, P_NH3…) that link the transmembrane flux of NH₄ ⁺ to changes in pH_i.

Parameters for the Cell Model: pH_i, β_i, P_NH3

The null method of pH measurement used on the isolated bundles of glial cells showed that many bundles had pH_is at least as alkaline as those measured with pH microelectrodes in slices of retina (mean: 7.31; Coles et al. 1996), in agreement with the generally alkaline pH reported in many kinds of glial cells (see Deitmer and Rose 1996). β_i [12.2 mequiv (pH unit · liter)⁻¹] is close to the value of 10.4 measured in snail neurons (Szatkowski and Thomas 1989). Our estimate of P_NH3 (13 μm s⁻¹) is well within the large range of values reported for biological membranes, which range from 108 μm s⁻¹ or higher in erythrocytes (Klocke et al. 1972; Labotka et al. 1995) to undetectably small at the apical membranes of colonic crypt cells (Singh et al. 1995). This variation appears in part to be correlated inversely with the density of proteins in the membrane: P_NH3 is high in protein-free artificial membranes (Antonenko et al. 1997) and low in membranes of urinary bladder, which are densely packed with uroplakins (Chang et al. 1994). P_NH3 has not, to our knowledge, been determined for cells of nervous tissue other than the bee retinal glial cells, so we do not know if our value is typical. From experiments similar to that of Fig. 6 B, we compared the permeabilities of the membrane to various neutral lipophilic compounds. We found that the permeabilities to the neutral forms of the amines TMA, MA, and ammonium or the carboxylic acids caproate, propionate, and acetate were greater the greater the hydrophobic part of the molecule (-CH₂- groups); i.e., P_TMA > P_MA > P_NH3 and P_caproate > P_propionate > P_acetate. Thus, it appears that the relative permeability of the cell membrane to these nonelectrolytes depends more on the hydrophobicity of the molecule than on its size, in accordance with Overton's rule (Overton 1899).

Membrane Potential, Cl⁻ Gradient, and pH Regulation

A major difference, potentially important for certain cell functions, between the glial cells in the isolated bundles and those on which published results were obtained in slices of bee retina, is the apparent membrane potential. On the assumption that application of nigericin caused H⁺ to distribute across the membrane with the same passive distribution as K⁺, we concluded that mean V_m in the bundles was −4 mV. Support for a small V_m is given by the observation (Marcaggi et al. 1999) that in dissociations of the kind used here, rhodamine 123 selectively labeled photoreceptor cells. Since rhodamine 123 tends to partition preferentially into negatively charged compartments, this observation is compatible with the isolated bundles of glial cells having a membrane potential, V_m, much smaller than that of the photoreceptors. Despite the smallness of V_m, the glial cell bundles were able to regulate their pH_i and to recover from repeated acid loads, although slightly more slowly than the recovery from a stimulus-induced acidification of glial cells in slices (Coles et al. 1996). And in electrically functioning retinal slices, a wide range of glial cell membrane potentials have been recorded (−10 to −75 mV) with little apparent consequence for homeostasis of extracellular ions or metabolism (Bertrand 1974; Coles et al. 1986; our unpublished observations). The mechanism of pH regulation in the bee glial cells is unknown and a mechanism not dependent on an ionic gradient is conceivable, as reported in C6 glial cells (Volk et al. 1998).

Transporters of the cation-Cl⁻ family are normally electroneutral, and the effect of ammonium on glial cell V_m in bee retinal slices is compatible with electroneutral transport (Coles et al. 1996), so V_m is expected to have no direct effect on the thermodynamics of the NH₄ ⁺ transport. However, since Cl⁻ is distributed approximately passively (Coles et al. 1989), the concentration gradient is much greater in glial cells in vivo and it is predicted that uptake of NH₄ ⁺ would be more effective than in the isolated bundles.

NH₄ ⁺/K⁺ Selectivity of the Transporter

Until now, the few studies of competition between K⁺ and NH₄ ⁺ for inward transport into animal cells on transporters have reported a selectivity for K⁺ (Kinne et al. 1986; Cougnon et al. 1999). It has, however, been proposed that NH₄ ⁺ transport is a physiologically significant process, notably in kidney cells (Good 1994) and in salivary acinar cells (Evans and Turner 1998). We have shown that for brief applications of ammonium in the millimolar range, the Cl⁻-dependent transport in bee retinal glial cells is functionally selective for NH₄ ⁺ over K⁺. Further, a minimal numerical model of the transport process in which NH₄ ⁺ competes for a transporting site with an affinity approximately twice that for K⁺ accounted for the main features of the pH_i responses not only for brief applications of ammonium but also for more complex protocols. Since K⁺ is the physiological cation whose ionic radius is closest to that of NH₄ ⁺ (Robinson and Stokes 1959) and whose permeation through channels is most similar to that of NH₄ ⁺ (Hille 1992), it is unlikely that the transporter has as high an affinity for any other major physiological ion, and we conclude that it is selective for NH₄ ⁺.

Reported values for K _m (K⁺) calculated for K⁺ influx by Cl⁻-dependent transport into erythrocytes are 55 mM (sheep; Delpire and Lauf 1991) and 140 mM (human; Kaji 1989). These values are higher than the K _i calculated in this study (10–27 mM) on the assumption (supported by the Lineweaver-Burke plots of Fig. 9B and Fig. C) that the inhibition is purely competitive. However, if, as suggested by Fig. 11, the inhibition is partly noncompetitive, then the K _i for the competitive component will be higher and closer to the values for erythrocytes. Not only does the transporter on the bee retinal glial cell have a lower affinity for K⁺ than for NH₄ ⁺, but preliminary results suggest that even the K⁺ that is bound may not be transported rapidly (Marcaggi and Coles 1998).

Possible Advantages of Glial Uptake of Ammonium in the NH₄ ⁺ Form

We have shown that ammonium enters bee retinal glial cells overwhelmingly in the NH₄ ⁺ form. It is striking that this is also the case for mammalian astrocytes (at least those cultured from neonatal mice), although, in contrast to the bee glial cells, the NH₄ ⁺ entry into cultured astrocytes appears to occur mainly through Ba²⁺-sensitive channels (Nagaraja and Brookes 1998; P. Sartor and P. Marcaggi, unpublished data). Entry of ammonium into cells is favored in two ways if it crosses the membrane as NH₄ ⁺ instead of as NH₃. First, at physiological pHs, the majority of the ammonium is in the NH₄ ⁺ form. Second, the entry can be coupled to a gradient. In the case of bee retinal cells in vivo, this is the Cl⁻ concentration gradient, and in cultured mouse astrocytes it is the electrical potential gradient.

A major ammonium-consuming process in bee retinal glial cells is the conversion of pyruvate to alanine (Tsacopoulos et al. 1994, Tsacopoulos et al. 1997a). As a substrate, NH₄ ⁺ will contribute to the regulation of the reactions (Tsacopoulos et al. 1997a,Tsacopoulos et al. 1997b). In addition, NH₄ ⁺ allosterically activates phosphofructokinase (Lowry and Passoneau 1966; Sugden and Newsholme 1975), an effect that, in mammals, may contribute to the coupling of glutamate release by neurons to glycolysis in astrocytes proposed by Pellerin and Magistretti 1994 (see also Magistretti et al. 1999).

In the case of bee retinal glial cells, the ammonium consumption can be summarized by the reaction: CH₃-CO-COO⁻ + NH₄ ⁺ + NADH + H⁺ → CH₃-CHNH₃ ⁺-COO⁻ + H₂O + NAD⁺.

Since this reaction consumes H⁺, pH_i is better conserved if ammonium is supplied in the NH₄ ⁺ form. In astrocytes, the pathways of energy metabolism are still a matter of debate (see, e.g., Demestre et al. 1997), but there, too, the proportion of ammonium that enters as NH₃ or NH₄ ⁺ will affect pH homeostasis in the brain.