AMT (ammonium transporter)/Rh (Rhesus) ammonium transporters/channels are identified in all domains of life and fulfil contrasting functions related either to ammonium acquisition or excretion. Based on functional and crystallographic high-resolution structural data, it was recently proposed that the bacterial AmtB (ammonium transporter B) is a gas channel for NH3 [Khademi, O'Connell, III, Remis, Robles-Colmenares, Miercke and Stroud (2004) Science 305, 1587–1594; Zheng, Kostrewa, Berneche, Winkler and Li (2004) Proc. Natl. Acad. Sci. U.S.A. 101, 17090–17095]. Key residues, proposed to be crucial for NH3 conduction, and the hydrophobic, but obstructed, pore were conserved in a homology model of LeAMT1;1 from tomato. Transport by LeAMT1;1 was affected by mutations of residues that were predicted to constitute the aromatic recruitment site for NH4+ at the external pore entrance. Despite the structural similarities, LeAMT1;1 was shown to transport only the ion; each transported 14C-methylammonium molecule carried a single positive elementary charge. Similarly, NH4+ (or H+/NH3) was transported, but NH3 conduction was excluded. It is concluded that related proteins and a similar molecular architecture can apparently support contrasting transport mechanisms.
- ammonium transport
- ammonium transporter/Rhesus (AMT/Rh)
- homology modelling
- ion channel
There is an ongoing debate as to whether some or all AMT (ammonium transporter)/Rh (Rhesus) proteins function as NH3 channels or as NH4+ transporters . Based on the recently obtained high-resolution structure of a bacterial homologue, AmtB (ammonium transporter B) from Escherichia coli, a channel-like transport of NH3 in AMT/Rh proteins has been proposed [2,3]. Ammonium-mediated vesicular alkalinization by reconstituted AmtB , and studies with bacterial strains compromised in ammonium assimilation and transport are in accordance with NH3 transport . However, mechanistic interpretations from a static structure should be taken with care. Though probably within the range of experimental error, some MeA (methylammonium) accumulation by AmtB was detected, which was only to be expected if AmtB transports NH4+. MeA accumulation would also be expected if, hypothetically, the ion occasionally ‘slipped’ through the channel/transporter . Additional in vitro and in vivo experiments, including flux measurements against ammonia gradients and/or current recordings, are required to establish the NH3 transport mechanism of AmtB. According to its transcriptional and post-transcriptional regulation, the physiological role of AmtB appears to be the uptake of ammonium under low ammonium concentrations, but the NH3 gradient is unfavourable for uptake under most conditions .
The transport mechanism proposed for AMT/Rh proteins from the study of the X-ray structures involves NH4+ recruitment at the external pore mouth close to the aromatic residues Phe103, Phe107 and Trp148, de-protonation and conduction as for NH3 [2,3]. Hallmarks of the AMT/Rh family are two histidine residues that line the central hydrophobic pore. It has been proposed that NH4+ cannot pass these histidines and thus no NH4+ transport and no ionic currents would be expected in AmtB. An electroneutral NH3 transport by AMT/Rh homologues is compatible with our results on human RhBG (Rh B glycoprotein) and RhCG (Rh C glycoprotein). The studies suggest that ammonium transport by these proteins is electroneutral when heterologously expressed in oocytes and yeast [7,8], in accordance with results from other groups .
In contrast, specific, saturable, voltage-dependent, pHo (extracellular pH)-independent NH4+ and MeA+ currents have been recorded from two tomato AMT protein homologues expressed in oocytes [5,6]. There is little doubt that ammonium uptake in plants occurs at least partially as NH4+, since electrogenic ammonium transport in plants is well established . Ionic currents are in conflict with the proposed NH3 channel mechanism in AMT/Rh proteins, but NH4+ currents might represent an ion ‘slippage’, while the major transport form is NH3. It has not been tested if NH3 is transported through LeAMT1;1.
The surprising similarity of an LeAMT1;1 homology model to AmtB triggered a further analysis of NH4+ and MeA+ currents mediated by LeAMT1;1. Although NH4+ and MeA+ currents were recorded, no ‘slippery’ gas (NH3) transport was detected. It was additionally found that an aromatic residue at the outer pore mouth is involved in high-affinity recruitment of NH4+. LeAMT1;1 may thus function either as an NH4+ channel (or ‘uniporter’) or, physiologically equivalent, as an NH3/H+ co-transporter.
Plasmid constructs, preparation and injection of oocytes
The plasmid pOO2-LeAMT1;1 and oocyte handling have been described in previous studies [5,7]. Mutations were introduced into LeAMT1;1 by using the QuikChange® method from Stratagene and the complete coding region was verified by sequencing.
Electrophysiological measurements, radiotracer uptake and pHi (intracellular pH) changes
Recordings were performed as described in . Means±S.E.M. are given. There was no special selection applied, but studies of ammonium transport in oocytes may be hampered by endogenous ion channels that are activated by high ammonium concentrations such as 10 mM [10,11]. Currents in the oocytes used in the present study were insensitive to ammonium and MeA at concentrations of 1 or 10 mM respectively [5,12], and care was taken to verify that these endogenous ammonium-inducible currents were not present in the oocyte batches used. Oocytes used in our laboratory had a background conductance of ≤0.9 μS (between 0 and −100 mV) and background currents as low as approx. −80 nA at −140 mV. This background current is approximately half that of the endogenous current in oocytes used in the laboratories that have recorded and investigated NH4+-induced currents [11,13]. Interestingly, seasonal variations were reported for the endogenous amine-inducible current . Although activation of such endogenous oocyte currents at lower ammonium concentrations has been reported , we and others have observed that ionic currents from oocytes are insensitive to ammonium at ≤1 mM at physiological pH ([5–7,12,14]; results not shown). Although the nature of this discrepancy and the reason for natural variance in oocyte currents is unclear, Xenopus oocytes with reduced or absent endogenous MeA+ and NH4+ currents have the great advantage that measurement of ammonium transport can be performed with minimal contamination from the endogenous ammonium-induced currents. The absence of endogenous MeA+ and NH4+ currents in oocytes used in the present study suggests that internal pHi will respond differently to ammonium concentration than in previous studies on oocytes with large NH4+ currents (see Figure 5) [10,11]. Oocytes expressing large endogenous ammonium-activated currents had been reported to acidify slowly when measured with pH-sensitive electrodes [11,13]. As NH4+ conductance was absent from the oocytes used in our study, native oocytes were alkalinized upon addition of 10 mM ammonium (see Figure 5).
In the cases where currents ‘induced by ammonium’ are given; these were obtained by subtracting the current in the absence of ammonium (background) from the current in the presence of ammonium at each voltage. NH3 and NH4+ concentrations were determined from total ammonium concentration using the Henderson–Hasselbach equation: log10 [NH3]/[NH4+]=pH−9.25. A pKa of 10.66 was used for calculations involving MeA. The concentration dependence of current/uptake was fitted using the following equation: I=Imax/(1+Km/c), where Imax is the maximal current at saturating concentration, Km is the substrate concentration permitting half-maximal currents, and c is the experimentally used concentration. [14C]MeA uptake was determined as in .
pHi changes were monitored using the pH-sensitive dye BCECF [2′,7′-bis-(2-carboxyethyl)-5(6)-carboxyfluorescein] and were calibrated as described in . Briefly, oocytes were incubated in uptake buffer (ND96, containing 5 μM BCECF acetoxymethyl ester, from a 50 mM stock in DMSO) for 30 min protected from light. Fluorescence of BCECF-loaded oocytes placed in a recording chamber and superfused with the standard solution (±NH4Cl) was monitored using an inverted Leica fluorescence microscope at maximal rate (5 s) using a spinning wheel system with excitation at 410±30 and 470±20 nm and emission at 535±20 nm. In some experiments, oocytes were voltage-clamped during fluorescence recording.
The primary sequence of LeAMT1;1 was aligned with AmtB using Clustal W, followed by manual inspection of functionally important residues. The alignment was identical with the alignment in Khademi et al.  for the first transmembrane helix (see Supplementary Figure at http://www.BiochemJ.org/bj/396/bj3960431add.htm). Structural fitting was done using MODELLER 7v7 (available at http://salilab.org/modeller/), a well-established structure prediction program . The structures were evaluated using the WHATIF program (http://biotech.ebi.ac.uk:8400/) and PROCHECK (http://www.biochem.ucl.ac.uk/∼roman/procheck/procheck.html). The LeAMT1;1 model used had 90.5% residues in the most favoured regions in a Ramachandran plot (compared with 92.7% in the AmtB structure). Only nine amino acids were identified that had ‘bad’ stereochemistry. All of these were located outside of the core structure of the transmembrane helices. Further evaluation was carried out by energy minimization of the monomers of AmtB (PDB accession code 1U7G; ) and the LeAMT1;1 model for 10 ps using NAMD2.5 (http://www.ks.uiuc.edu/Research/namd/), using the Amber7 force field. The stability of the structure was used as a measure for the quality and was assessed by measuring the rmsd (root mean square deviation) between backbone residues of the structures. Minor attention was given to reconstruction of loops, as these are expected to minimally influence the core structure and transport mechanism. The pore diameter was accessed using the program HOLE2 . Monomeric bovine AQP1 (aquaporin 1; PDB accession code 1J4N) was analysed in parallel.
LeAMT1;1 expression in yeast
The open reading frames of LeAMT1;1 and a translational LeAMT1;1–GFP (green fluorescent protein) fusion were cloned into plasmid pDR196 and heat-shock-transfected in the ura− Saccharomyces cerevisiae wild-type strain (23344c) or the ammonium transporter-defective yeast strain (31019b; ΔΔΔmep1;2;3) . The nitrogen-deficient growth medium for strain 31019b was YNB (1.7 g/l yeast nitrogen base and 5 g/l glucose; Difco), supplemented with 3% (w/v) glucose and appropriate NH4Cl concentrations. The medium was adjusted to pH 5.5 using 50 mM Mes buffer. The MeA toxicity medium contained YNB, 3% glucose, 0.1% arginine and 125 mM MeA. Growth of yeast was not affected by expression of the different constructs under non-selective conditions.
Homology model of LeAMT1;1
To identify the molecular determinants for differences in the transport mechanism between AMT/Rh homologues, a homology model of LeAMT1;1 with AmtB (PDB code 1U7G) as template was constructed. Both proteins are 25% identical and 44% similar within the protein core (excluding the N- and C-termini). The most stable model (see the Experimental section) was obtained by using a slightly modified alignment from that published by Khademi et al. , which is given as a Supplementary Figure (http://www.BiochemJ.org/bj/396/bj3960431.add.htm) and only differs in the first transmembrane region. A ribbon representation of the structural backbone of AmtB and the model of LeAMT1;1 is shown in Figures 1(A) and 1(B). Major deviations are seen in the extrahelical loops, consistent with the low similarity between these regions. The model was less stable than AmtB after a minimization for 10 ps, which reduces ‘bad’ contacts within the protein (Figure 1C). Three monomers preferably assembled in a trimeric form (Figure 1E).
AmtB does not form an ‘open’ channel-like structure as e.g. AQP channels. Several AQPs transport ammonia in addition to water [18,19], which justifies the comparison between these protein classes. The substrate pore of AQP monomers is illustrated for bovine AQP1 (PDB code 1J4N)  using the program HOLE2, a program that identifies putative ionic and solute pores in proteins based on the van der Waals radii of amino acid residues close to the pore . The AQP1 pore constriction does not go below 0.8 Å (1 Å=0.1 nm) (Figure 1D) , but the solute pathway in AmtB is blocked by residues Phe107 and Phe215 and partially by Phe31 (Figures 1C and 1F). In accordance with this finding, preliminary comparative steered molecular dynamics studies did reveal solute flux in AQPs, but not in AmtB (M. Dynowski and U. Ludewig, unpublished work). The fundamental difference appears to be important, since the proposed channel-like mechanism was recently illustrated by a misleading ‘open-pore’ cartoon-like model . Minimal conformational changes at least are necessary to obtain a channel-like open pore structure. Another possibility is that transport-cycle-dependent conformational changes are involved in ammonium transport.
The pore analysis using HOLE2 did not reveal a continuous solute pathway in the LeAMT1;1 homology model. Most residues, if not all, that have been suggested to be crucial for NH4+ recruitment and NH3 conduction in the occluded pore of AmtB are identical or similar in LeAMT1;1. The residues Phe103, Phe107, Trp148 and Ser219 in AmtB have been predicted to participate in the recruitment and co-ordination of the NH4+ ion; the equivalent residues in LeAMT1;1 are at positions Tyr133, Phe137, Trp178 and Ser263 respectively. The aromatic residues Phe107 and Phe215 block the pore in AmtB; these residues are at positions Phe137 and Phe259 in LeAMT1;1. The model has a similar hydrophobic pore to AmtB, with two conserved residues at positions His207 and His374 that line the putative pore lumen (Figure 1G). Although the model cannot be a substitute for a crystal structure, its similarity to the NH3-transporting AmtB is striking.
The aromatic nature of two residues forming the ion-binding cavity (Tyr133 and Trp178 in LeAMT1;1) is conserved between bacterial and plant homologues, but not in Rh glycoproteins. In Rh proteins, non-polar small neutral residues occupy these positions (isoleucine and leucine in RhCG). These residues might thus be responsible for the different substrate affinity that is observed in Rh glycoproteins [7,8]. The overall quality of the LeAMT1;1 model was tested by mutational analysis of the residues Tyr133 and Trp178 that are putatively involved in substrate affinity.
Ammonium transport by LeAMT1;1 pore mutants
Initial attempts to functionally express EcAmtB (AmtB from E. coli) in oocytes failed (results not shown), so we concentrated on the function of LeAMT1;1. Residues Tyr133 and Trp178 were exchanged with the respective residues found in RhCG (Y133I and W178L). LeAMT1;1 and, to a lesser extent, the two mutants functionally complemented the ammonium transporter-deficient yeast strain ΔΔΔmep1;2;3 with ammonium as the sole nitrogen source (Figure 2). The mutations did not largely alter the expression pattern as assessed by confocal microscopy of translational C-terminal GFP fusions in yeast. A cell-to-cell variability in fluorescence levels was observed in homogeneous yeast cultures when analysed with the identical microscopy settings at low magnification (Figure 2E). LeAMT1;1 and the mutants showed the same subcellular localization when analysed at higher magnification (Figure 2E).
At 125 mM, MeA is toxic to wild-type yeast and expression of Rh glycoproteins has been shown to rescue growth on toxic MeA. Rh glycoproteins probably export uncharged MeA along its chemical gradient, even though Rh glycoproteins are saturated well below these high MeA concentrations . Both mutants and LeAMT1;1 showed increased sensitivity to MeA, which is most simply explained by MeA+ import by LeAMT1;1 and the mutants (Figure 2D). In all assays, the W178L mutant displayed less transport activity than the Y133I mutant.
Altered saturation of NH4+ transport in an LeAMT1;1 mutant
For a more detailed analysis, the ionic currents elicited by the mutant transporters were assayed using the two-electrode voltage clamp in oocytes. No significant [14C]MeA uptakes and currents were detected with W178L, precluding further analysis. In contrast, the mutant Y133I displayed small, but reliable currents on treatment with ammonium. Currents were more than 10-fold smaller than in the wild-type (Figure 3), in accordance with its reduced overall transport rate or reduced functionality compared with wild-type LeAMT1;1 in the yeast-growth experiments. Importantly, the half-maximal current, which is a measure that is independent of the expression level, was saturated at very low concentrations. In contrast with wild-type (Km=12 μM), the Km for NH4+ in the Y133I mutant was only 1 μM at −120 mV. Similarly, MeA+ currents were saturated at a half-maximal concentration of 584 μM at −120 mV in LeAMT1;1, while the small currents of the mutant were half-maximal at 4.5 μM (Figure 3).
Charge coupling of MeA+ transport in LeAMT1;1
The ratio of charged versus uncharged solute transport was assayed using the analogue MeA labelled at its carbon ([14C]MeA). Transport and charge translocation were monitored simultaneously in native and LeAMT1;1-expressing oocytes from the same batch. LeAMT1;1-expressing oocytes incubated in 1 mM [14C]MeA accumulated 9.0±1.5 pmol of [14C]MeA (H314C-NH3+/H314C-NH2)·min−1. During incubation, oocytes had a membrane potential of +4±1 mV. The accumulation rate can be converted using Avogadro's constant into molecules of MeA·min−1. If each molecule transported is charged (H314C-NH3+ carries one elementary charge), this corresponds to an ionic current of −14.4±2.4 nA. At the measured resting potential of +4 mV, MeA induced −12.5 nA inward current (Figure 4), suggesting that roughly 100% of MeA is taken up as charged H314C-NH3+. The MeA+/MeA-induced current was specific to LeAMT1;1-expressing oocytes, was time-independent over several minutes and was strongly inwardly rectifying. Taking the error bar into account, the maximal slippage of uncharged H314C-NH2 must be below 10%. Identical results were obtained with three independent batches of oocytes; one of these experiments is shown in Figure 4. The results show that H314C-NH3+ is preferentially transported but, at this resolution, the possible transport of H314C-NH2 cannot be excluded (Figure 4).
Ionic currents and acidification by LeAMT1;1
To determine if a residual flux of NH3 is maintained by LeAMT1;1, current and pHi recordings were combined. Current–voltage relationships from LeAMT1;1-expressing oocytes are shown in Figure 5. The open circles represent a current–voltage relationship from LeAMT1;1-expressing oocytes in the absence of MeA+ and the closed circles show the same after addition of 200 μM ammonium (Figure 5A). The ammonium-induced current was inward and the reversal potential of the total currents shifted to more positive membrane potentials. No pHi changes upon 500 μM ammonium treatment were observed in BCECF-loaded, LeAMT1;1-expressing oocytes and in water-injected controls (Figure 5B). Although NH4+ transport is almost saturated at 500 μM in LeAMT1;1 (Figure 3) , 10 mM ammonium was also used to compare the results with older studies on native oocytes that used these excess ammonium concentrations. Only negligible NH4+ influx is expected from non-voltage-clamped, LeAMT1;1-expressing oocytes at depolarized potentials (compare the current–voltage relationship in Figure 5A), but significant NH4+ influx was observed in voltage-clamped oocytes at a negative membrane potential. The negative membrane potential mimics the in vivo situation of LeAMT1;1 in tomato root epidermal cells .
The pHi of water-injected and LeAMT1;1-expressing oocytes clamped to −80 mV is shown in Figures 5(C) and 5(D). Ammonium at 500 μM induced −150 nA current in LeAMT1;1 and a persistent acidification, in accordance with NH4+ influx, but in opposition to significant NH3 flux. This acidification was not observed in water-injected control oocytes. It is necessary to recall that at the cytosolic pHi of 7.4, the pHi is highly sensitive to NH3 influx, as almost every incoming NH3 will bind H+ to form NH4+. This situation is distinct from NH4+ transport: most of the influxed NH4+ will remain as NH4+ and only a minor fraction (∼1.5%) will dissociate into NH3 and H+, and thus acidify the cytosol.
The pHi changes were not due to possible oocyte to oocyte variability, as shown in a continuous recording from a single LeAMT1;1-expressing oocyte in Figure 5(D). Low ammonium concentration induced acidification and a −170 nA current during voltage clamping, while no acidification was seen when the membrane potential was then allowed to change freely. Interestingly, the voltage-clamped LeAMT1;1 oocytes that acidified with NH4+ uptake did not recover the pHi to the baseline level after the withdrawal of ammonium. This suggests that ammonium that passively exits the oocyte as NH3 is reabsorbed as NH4+ and thus the cytosolic ammonium concentration and pHi are maintained.
Prediction of the function of proteins based on their crystal structure is an exceptionally elegant method, but we cannot yet distinguish a channel from a transporter that couples conformational changes with transport on the basis of the crystal structure. The structure of the bacterial AMT/Rh homologue AmtB does not show a clearly open transmembrane pore (Figure 1), which would be required for a channel but is unlikely in any well-coupled exchanger. This is also true for the model of LeAMT1;1, which shows that key residues are highly conserved between these proteins. Thus, even in AmtB, conformational changes or side chain movements are required to allow solute flow. If, however, AmtB undergoes a conformational change that is not detected by the crystallographic approach, then hydrogen bond acceptors may become available for interaction with ammonium.
Similar to the case of AMT/Rh proteins, the X-ray structures of a prokaryotic homologue of CLC (chloride channel) anion channels did not reveal its mechanism, which was later identified as rapid 2Cl−/H+ antiport . Although the pathway for anions is clearly apparent from the atomic CLC structures, it has not yet been resolved how H+ is transported via CLCs. Similarly, although residues involved in sugar/H+ co-transport via LacY transporters had long been known, the mechanism of H+ co-transport was not readily apparent from the structure . Crystal structures provide a static view, and even if structures look alike in the presence or absence of substrate, this does not exclude the existence of other conformations, since the protein may preferably crystallize in one state. Solute transport in proteins needs to be viewed as dynamic, but crystallographic structures are, by nature, static. A full understanding of the transport mechanism will only result from a resourceful combination of structural and functional biochemical approaches.
The three-dimensional model of LeAMT1;1 was successfully used to predict residues involved in NH4+ recruitment (or binding). Although no attempt is made to derive mechanistic conclusions from the homology model and the structure of the hydrophobic pore of LeAMT1;1, the molecular modelling may be useful in guiding future experimental work on this transporter/channel class. A very similar AMT/Rh structure has recently been reported for a thermophilic archaeon , which emphasizes that AMT/Rh proteins adopt a similar structure. The functional data for LeAMT1;1 in Figures 4 and 5 show that LeAMT1;1 is a net NH4+ transporter (or ‘channel’). Based on the AmtB structure and the hydrophobic pore in the LeAMT1;1 model, one may speculate that the molecular mechanism is NH3/H+ co-transport. If NH4+ dissociates into NH3 and H+ during passage, an unresolved pathway would exist for H+. This had already been discussed  and explains the superb selectivity of NH4+ transporters against alkali cations . NH4+ transport is equal at external pHo 5.5 and 8.5  and exceeds that of NH3, which suggests that LeAMT1;1 can transport against NH3 gradients.
Despite similar key residues in its structure, AmtB apparently uses a different transport mechanism to LeAMT1;1, as it transports at least some NH3, shown by the rapid alkalinization of AmtB vesicles in ammonium solution . It is currently impossible to exclude the possibility that AmtB is sloppier and passes more NH3 together with NH4+. It remains to be shown that AmtB dissipates NH3 gradients and thus acts as a channel. Even if AmtB does transport 95% in the form of NH4+ and only 5% in the form of NH3, alkalinization would be expected due to a net inward NH3 flux (at physiological pHi total ammonium is ∼98.5% NH4+ and ∼1.5% NH3).
The data on LeAMT1;1 presented here are in accordance with previous data  and argue against the speculation that LeAMT1;1 currents may be endogenous to oocytes . The saturation of currents was altered by more than 10-fold (NH4+) and 100-fold (MeA+) in LeAMT1;1 mutants and differed from endogenous ammonium-induced oocyte currents . Transport of NH4+ is not restricted to LeAMT1;1, since other plant homologues have now been identified in our laboratory that also induce saturable NH4+ and MeA+ currents when expressed in oocytes . Similar to the plant transporters, the Saccharomyces homologues (MEP1–MEP3; where MEP1 is methylammonium permease 1 from S. cerevisiae) are involved in re-uptake and accumulation of lost ammonium . Micro-organisms and plants maintain a near neutral cytoplasmic pHi, but the acidic pHo often requires the acquisition of NH4+/NH3 against an NH3 concentration gradient.
Small conformational changes that may represent ‘gating’ of AmtB have already been documented in the different crystal structures [2,3], but more structures showing different conformations of AMT/Rh proteins are required to distinguish between channel and transporter-like mechanisms. In any case, the functional results suggest that the classical concept of clear discrimination between channels and transporters is diminishing on a molecular level: highly similar proteins from the AMT/Rh family may function as equilibrative ‘gas’ or ion channels/transporters.
We thank Binghua Wu (ZMBP, Tübingen University) for initial help with the constructs, Petra Neumann (ZMBP, Tübingen University) for excellent technical assistance and Felicity deCourcy (ZMBP, Tübingen University) for critically reading this paper before submission. Support from the Deutsche Forschungsgemeinschaft to U.L. is acknowledged.
Abbreviations: AMT, ammonium transporter; AmtB, ammonium transporter B; AQP1, aquaporin 1; BCECF, 2′,7′-bis-(2-carboxyethyl)-5(6)-carboxyfluorescein; CLC, chloride channel; GFP, green fluorescent protein; MeA, methylammonium; pHi, intracellular pH; pHo, extracellular pH; Rh, Rhesus; RhCG, Rh C glycoprotein; rmsd, root mean square deviation
- The Biochemical Society, London