The Amt/Mep/Rh family of ammonium transport proteins (Review)

Abstract

The Amt/Mep/Rh family of integral membrane proteins comprises ammonium transporters of bacteria, archaea and eukarya, as well as the Rhesus proteins found in animals. They play a central role in the uptake of reduced nitrogen for biosynthetic purposes, in energy metabolism, or in renal excretion. Recent structural information on two prokaryotic Amt proteins has significantly contributed to our understanding of this class, but basic questions concerning the transport mechanism and the nature of the transported substrate, NH₃ or , remain to be answered. Here we review functional and structural studies on Amt proteins and discuss the bioenergetic issues raised by the various mechanistic proposals present in the literature.

• Ammonium transport protein
• methylammonium permease
• rhesus proteins
• transport assays
• protein crystallography

Introduction

Nitrogen is an essential part of all biological macromolecules, but more than 99% of this element on earth are found as atmospheric dinitrogen, N₂[1]. In spite of its abundance, most organisms are unable to metabolize this molecule, due to the extraordinary kinetic stability of the N–N triple bond (ΔG⁰' = − 946 kJ×mol⁻¹) [2]. Only a distinct class of microorganisms, the diazotrophs, is able to reductively fix dinitrogen into ammonia (NH₃), in a unique reaction catalyzed by the enzyme nitrogenase [2], [3]. A second important source of ammonia is the reductive assimilation of nitrate (), achieved by the successive action of a nitrate and a nitrite reductase in many bacteria and plants. The reason for the particular emphasis on ammonia among all the basic chemical modifications of nitrogen is that only in this reduced state can the element be assimilated into biological molecules [4–6]. This occurs predominantly via either the glutamine synthetase/glutamate synthase pathway (GS/GOGAT) or through glutamate dehydrogenase (GDH), both eventually yielding glutamine, from which other amino acids can be synthesized through transamination reactions. Its exposed role in nitrogen metabolism thus makes ammonia an important nutrient whose availability is often the limiting factor for the production of biomass, requiring for example the extensive use of fertilizers in modern, industrial agriculture [3].

In natural environments, uptake and storage of ammonia are essential for survival, and not surprisingly, organisms have developed various, intricate ways to assure their nitrogen supply. While animals largely rely on the uptake of nitrogen-containing amino acids through their diet, some plants harness the nitrogen-fixing ability of diazotrophic bacteria by providing them with highly specialized organs, the root nodules. Within their protected, anoxic environment with carbon and metabolic energy being provided by the host plant, the symbiotic bacteria use their highly oxygen-sensitive nitrogenase system to produce ammonia that can then be taken up by the plant for its own biosyntheses.

In its free form, ammonia is a hydrophobic gas that can permeate biological membranes with a permeability coefficient in the order of 10⁻³ cm·s⁻¹, a value close to that of H₂O [7–9]. It has initially been suggested that – similar to other metabolically important gases, such as O₂, N₂ or H₂ – a specific transmembrane transport system may not be required. However, NH₃ is in protonation equilibrium with the charged ammonium cation () at a pK_a of 9.25, such that at a typical physiological pH value of ∼7.5, about 99% will be in the membrane-impermeable, cationic form. During the 1970s and early 1980s, several lines of evidence gave rise to the assumption that ammonia/ammonium uptake into cells should be an active process occurring through dedicated membrane proteins [7], [10], [11]. An increasing number of publications reported the existence of concentration gradients across cell membranes in various microorganisms [12–15] and evidence for ammonium uptake systems was presented. These would be required at low external concentrations of ammonium, where passive transport would not suffice to satisfy metabolic requirements [14], [16], [17].

In 1994, the first sequences for ammonium transport proteins were presented for MEP1 from Saccharomyces cerevisiae[18] and Amt-1;1 from Arabidopsis thaliana[19], followed by the first complete prokaryotic Amt sequence from Corynebacterium glutamicum[20]. Today, a large amount of available sequence data outlines a distinct protein family with a high degree of conservation, comprising the ammonium transport proteins (Amt) of bacteria, archaea and plants as well as the methylammonium permeases (Mep) from yeast, and the Rhesus proteins (Rh) of animals.

The Amt/Mep/Rh Family

The nomenclature of proteins within the ammonium transport protein family may seem confusing, as many systems and most published functional assays focus on the transport properties of methylammonium (MA, ), an alternative but presumably non-physiological substrate. Because the detection and precise quantification of ammonia is often difficult, most groups turned to monitoring the accumulation of radioactively labelled ¹⁴C-MA, following protocols established by Segel and co-workers [10]. Uptake of MA showed strong, competitive inhibition by , and it has thus been inferred that the latter is the natural substrate of both Amt and Mep proteins [10–13]. Sequences of both groups are highly similar, and in 1997 the Rhesus blood group proteins were identified as a further line of homologues [21], [22]. In the following we will use the term Amt proteins to refer to the entire family of Amt/Mep/Rh proteins. All genes encode for polypeptides of approximately 400 − 500 amino acids and hydropathy analyses consistently predict 11 − 12 transmembrane helices. They all function as uptake systems for ammonium into the cytoplasm or into a storage vacuole within the plant cell. Of particular interest is the discovery of amt genes in bacteria growing by anaerobic oxidation of ammonia (anammox), such as Kuenenia stuttgartiensis[23] and by nitrification of ammonia to nitrite such as Nitrosomonas europaea[24]. In both cases, reduced nitrogen is not used for biosyntheses but for energy metabolism.

The Rhesus blood group antigens of mammalian erythrocytes had long been recognized for their immunogenic characteristics and involvement in haemolytic diseases of infants when they were identified as Amt homologues. They are components of the Rh complex, originally proposed to be a heterotetramer of two RhAG glycoproteins and the non-glycosylated RhCE or RhD subunits [25]. In a more recent publication, a trimeric quaternary structure similar to Amt proteins has been proposed [26], but final evidence remains to be presented. RhAG alone has been shown to rescue an ammonium-uptake-deficient phenotype in yeast, while at the same time conferring resistance to MA [22], [27]. The genetic inability to express Rh proteins in erythrocytes is the cause of a syndrome termed Rh_null disease [28]. Its symptoms include various membrane defects and an altered cation transport across the membrane, leading to fragile red blood cells and predisposing the patients for haemolytic anaemia. Moreover, the presence of other Rhesus glycoproteins (e.g., RhBG) in mammalian kidney and liver tissues [29] is consistent with the need to excrete ammonia for acid-base homeostasis [30], [31].

The structure of Amt proteins

It was a significant step forward for research on Amt proteins when Merrick and co-workers showed in 2002 that the AmtB protein of Escherichia coli could be overexpressed and purified from the bacterium in large amounts [32]. They predicted AmtB to form stable trimers in the membrane that would remain intact during the purification procedure and eventually presented a low-resolution model obtained by cryo electron microscopy [33]. In parallel, structural information on two Amt proteins was obtained by X-ray crystallography. AmtB from E. coli was solved independently by two groups [34], [35], followed by Amt-1 from the hyperthermophilic archaeon Archaeoglobusfulgidus[36]. At resolutions of 1.4 Å [34], 1.54 Å [36] and 1.8 Å [35], these structural models were of exceptional quality for integral membrane proteins and provided a detailed picture of their molecular architecture.

In all three-dimensional models, the tertiary structure shows a highly conserved fold with 11 transmembrane α-helices (Figure 1). The sequence of AmtB includes a twelfth, N-terminal helix that was processed and removed in the mature protein, such that both AmtB and Amt-1 have an N-out/C-in topology. As seen previously in other membrane transporters such as the ClC chloride channel [37] or lactose permease [38], the Amt monomers have an internal, pseudo-twofold symmetry with its axis in the plane of the membrane that relates helices I-V to helices VI-X. The additional, eleventh helix has a length of more than 48 Å and crosses both symmetry-related halves, presumably adding significantly to the stability of the protein (Figure 1). The structures of AmtB and Amt-1 are highly similar in their transmembrane parts, but only in Amt-1 the entire protein was ordered in the crystal, including the conserved C-terminal region that is functionally relevant both in plant Amts and in their bacterial and archaeal homologues, where it determines the interaction with the regulatory P_II proteins (see below). Recent work on A. thaliana Amt1;1 confirmed the important role of the C-terminal region and showed that it can act as an allosteric switch for transport events between monomers [39].

Figure 1.  Stereo representation of the monomer of Amt-1 from A. fulgidus (PDB accession code). Eleven α-helices cross the membrane. The protein is displayed in cartoon representation, with the protein chain coloured from blue at the N-terminus on the extracellular side (top) to red at the C-terminus on the cytoplasmic side (bottom).

The crystal structures show a putative substrate recruitment site in each monomer on the extracellular side of the trimer (Figure 2A, 2B). A cavity in the surface of Amt-1 is lined by two conserved residues, Trp 137 and Ser 208, and its bottom is formed by the side chain of Phe 96. A globular electron density peak in this cavity has been modelled as a water molecule in Amt-1, forming a short hydrogen bond of 2.65 Å to the O_γ atom of Ser 208. This position is a promising binding site for , as a positively charged ion here would be able to engage in cation-π interaction with the indole moiety of Trp 137 in addition to the hydrogen bond to Ser 208. Not only could this mode of interaction explain the observed high affinity, but the combination of H-bonding and cation-π interaction would also exclude similarly sized and abundant molecules such as H₂O (not a cation) or K⁺ (no H-bonding interactions). Direct experimental evidence for this binding site has not yet been obtained and an assignment based solely on crystallographic data is impossible, as water and ammonia/ammonium have identical numbers of electrons. The putative binding site shows an electron density peak at identical positions in the presence and absence of , and we have therefore modelled this peak as water in Amt-1 (Figure 2A).

Figure 2.  Substrate recruitment and transport in Amt-1. (A) A putative recruitment site for is located on the extracellular side of each Amt-1 monomer, where a water molecule was found in the model (red dot). The combination of hydrogen-bonding to Ser 208 and cation-π-interaction with Trp 137 provides selectivity. (B) Trimer of Amt-1 seen from the extracellular side. While each monomer is able to transport substrate, the trimeric arrangement is essential for interaction with the regulatory P_II protein (Figure 3). (C) A channel crosses the Amt-1 monomer, starting immediately below the putative substrate recruitment site at Typ 137 and Ser 208, extracellular side above, cytoplasmic side below. (D) Detail of (C). Substrate is blocked from passing through the channel by the side chains of Phe 96 and Phe 204, but once this obstacle is passed a channel leads directly to the cytoplasmic side of the membrane. It is lined exclusively by hydrophobic residues, with the exception of the conserved histidines His 157 and His 305.

In spite of the highly conserved, trimeric quaternary structure of Amt-1, substrate transport occurs through separate pathways within each monomer, passing in between the two pseudo-symmetry-related halves. Below the extracellular binding site, a distinct channel crosses the protein (Figure 2C), but is blocked by two conserved residues, Phe 96 and Phe 204, implying that substrate passage is not possible without some degree of structural rearrangement within the protein. The putative binding site is located directly above Phe 96, and immediately below the constriction an exit pathway begins, leading to the cytoplasmic side (Figure 2C, 2D). Pressurization experiments with the noble gas xenon have shown this part of the channel to be hydrophobic by nature [36], with the exception of two conserved histidine residues, His 157 and His 305. They are arranged with nearly coplanar imidazole rings, hydrogen-bonded through their N_δ atoms. The pseudo-twofold axis of the monomer passes right between these residues, as they are situated in corresponding positions in helices V and X, respectively. First mutational studies have underlined the significance of this motif, although the position of His 157 is found to be replaced by glutamate in some fungal Amt proteins [40]. The observation of this predominantly hydrophobic channel has given rise to the assumption that NH₃ rather than the more hydrophilic cation may be the transported substrate of Amt proteins [34]. Within the channel, three weak electron density peaks have been observed only in the presence of ammonium in one AmtB structure [34], both with and without ammonium in the other [35], and not at all in A. fulgidus Amt-1 [36]. Their nature and relevance thus remains to be confirmed.

Regulation of Amt proteins

Beside their transport function, Amt proteins have been reported to participate in the regulation of nitrogen metabolism by acting as ammonium sensors [41], [42]. In a majority of prokaryotic organisms, the amt gene is found directly associated in an operon with a second gene encoding a protein belonging to the regulatory P_II family, GlnB or its homolog GlnK (the respective products of the glnB and glnK genes [43]. In order to avoid ambiguous nomenclature, the name GlnK has been proposed for all Amt-linked P_II proteins and will be used in the following [44]. P_II proteins have also been found in higher plants, where they are encoded in the nucleus, but located in chloroplasts [45]. Several crystal structures of P_II proteins are available [46–52]. In all cases these proteins are stable trimers, whose most prominent feature is a protruding loop in each monomer, the T-loop, which is functionally relevant and also the site of regulation by covalent modifications [44]. In a most recent study, crystal structures of the P_II protein GlnK1 from Methanococcus janaschii have been obtained with and without the effectors ATP and 2-oxoglutarate [53]. These structures show a dramatic difference in the conformation of the T-loop in the presence and absence of Mg-ATP. While the entire loop region is extended and mildly disordered in a complex with ADP, the coordination of Mg-ATP between two monomers of the GlnK1 trimer changes the T-loop conformation such that it adopts a more compact position, closer to the protein surface. Only in this conformation is a binding site for 2-oxoglutarate observed on top of the compacted T-loop. A high ADP/ATP ratio signals a low energy level of the cell and it is assumed that under such conditions biosyntheses are down-regulated and uptake of ammonium is not desired. Yildiz et al. therefore propose that GlnK1 binds to its Amt protein in a conformational state similar to the one observed in the presence of ADP [53].

First evidence for a direct association of GlnK to the Amt proteins of E. coli, Azotobacter vinelandii[54], and C. glutamicum[55] was obtained in response to an ammonium pulse. The result is a direct blockage of Amt-mediated transport when nitrogen is no longer limiting or the accumulation of high levels of ammonium in the cell needs to be avoided [41]. Complex formation was shown to be reversible, with dissociation occurring as the cellular nitrogen levels decrease. Based on the crystal structure of A. fulgidus Amt-1 that included the entire C-terminus and a homology model of A. fulgidus GlnK-1 we proposed a docking model for the Amt/P_II complex, in which the regulatory protein binds directly to the cytoplasmic face of the transporter, its T-loops inserting into the substrate channels of Amt-1, blocking them efficiently [36]. The M. jannaschii GlnK1-ADP complex supported this model, suggesting that the extended T-loops would reach even deeper into the transporter trimer [53]. Direct confirmation of this docking mode came through two parallel publications of the E. coli AmtB/GlnK complex, one of which was obtained by co-crystallization of both protein components [56] while in the other the complex was kept intact during purification (Figure 3A) [57]. A striking feature of the interaction of both proteins is the deep insertion of a conserved arginine residue, Arg 47 in E. coli GlnK, into the substrate channel (Figure 3B). This residue is fixed by hydrogen bonds, resulting undoubtedly in blockage of substrate transport.

Figure 3.  The complex of E. coli AmtB with GlnK. (A) Side view of the complex (PDB accession-# 2NS1) with the AmtB trimer shown in surface representation and GlnK as cartoon. In each trimer, one monomer is shown in colour. (B) Detail of the blocking of the cytoplasmic exit channel of AmtB (green) by GlnK (blue). The side chain of Arg 47 inserts into the channel, and Tyr 51, the site of uridylylation though GlnD, is in close proximity: (not shown).

The regulation of nitrogen metabolism has been intensively studied in various species, in particular in E. coli and related enteric bacteria [44], [58–62]. As nitrogen is assimilated to form the amino acid glutamine, the intracellular pools of glutamine and 2-oxoglutarate are proposed to directly reflect the nitrogen supply status of the cell [44], [63]. Regulation of ammonium uptake occurs on the genetic level, through controlling the expression levels of the amt genes [44], [64] and on the protein level, by direct action of P_II proteins on the function of the transporters and enzymes involved, as seen for example in the "switch-off" reaction of nitrogen fixation, a P_II-controlled ADP-ribosylation of the nitrogenase iron protein that down-regulates the energetically demanding nitrogenase enzyme as soon as ammonium becomes available [65], [66].

Among archaea, only the nitrogen-fixing methanogens have been subject to closer investigation [67]. As in other organisms, P_II proteins obviously play a central regulatory role, but due to the absence of all known, covalent modification sites in the T-loop it can be assumed that other mechanisms exist to control the activity of proteins such as the ammonium transport proteins or glutamine synthetase. Genes for Amt proteins have also been found to occur in more than one copy in many organisms, such as in yeast [18], tomato [68], or Methanosarcina acetivorans[69]. The euryarchaeon A. fulgidus possesses three independent amt genes for Amt proteins, designated amt-1, amt-2, and amt-3, each being individually followed by a gene for a P_II protein. Such cases suggest a variability in the regulation of expression [70] as well as in substrate affinity [71], [72].

Bioenergetic considerations on ammonium transport

Initial models concerning the mechanism of transmembrane ammonium transport assumed an active process leading to intracellular accumulation of . Kleiner stated that the mere finding of intracellular ammonium accumulation strongly supports active uptake systems, and that the pH gradient (acidic outside, alkaline inside) across the energized, cytoplasmic membranes of bacteria should promote depletion of ammonium rather than its accumulation if permeation was exclusively a passive process [11], [17] (see below). Thus the loss of ammonia due to passive diffusion across the membrane may be a significant problem for many organisms, for which they compensate by expressing Amt proteins as active (re)importers. This was referred to as cyclic retention, a process that would require a secondary active transporter, most likely as cotransport of NH₃/H⁺[7]. Such a transport system would be electrogenic, dependent on the proton-motive force and also dependent on the membrane potential. Indeed, this behaviour was found in Mep proteins of yeast [18], [71] and Amt proteins of A. thaliana[19], [73]. Voltage-clamp experiments on the tomato protein LeAMT1;1 [68] and the RhBG glycoprotein [74] reconstituted in Xenopus laevis oocytes further supported active uptake, albeit without being able to distinguish between a symport of NH₃/H⁺ or uniport of , due to the identical net charge of both processes.

An alternative mechanistic proposal originates from the work of Kustu and co-workers, who did not observe accumulation of MA in the cytoplasm of E. coli[75] and Salmonella typhimurium[76]. These in vivo studies indicated the Amt protein to function as a facilitator for NH₃ permeation when the external concentration of ammonium is low and led to the suggestion to use the term ammonia channel rather than ammonium transporter. As such, Amt proteins were said to constitute the first characterized class of transmembrane gas channels [76]. Recently this hypothesis gained further support from the first in vitro assays done with purified AmtB reconstituted into proteoliposomes [34]. Upon addition of ammonium to the outside medium, a rise of internal pH was seen, consistent with the notion that ammonia translocation over the membrane is followed by uptake of protons inside the vesicle, converting NH₃ into . Together with the observations that the presence of ammonium in the crystallization buffer did not induce any change in the protein structures of either AmtB or Amt-1 and that the substrate channel is predominantly hydrophobic by nature, the view of Amt proteins as passive channels for NH₃ seemed to become widely accepted [77].

However, an ammonia facilitator mechanism cannot account for the observed currents in voltage-clamp experiments or the findings of intracellular ammonium accumulation. In addition it poses a very fundamental, bioenergetic problem: Since the action of Amt proteins is mainly required at low substrate concentrations, when the amounts of free NH₃ are negligible, the proposed translocation would then require a deprotonation of on the extracellular side and reprotonation in the cytoplasm (Figure 4A). Effectively, this would result not in passive transport of NH₃, but rather in a net antiport of and H⁺. In organisms using a proton gradient for energy conservation, the direction of proton transport would be against this gradient, and as a consequence the uptake of ammonia would lead to the generation of a proton motive force. More realistically, the direction of transport would reverse and the proton gradient would cause the removal of ammonia from the cytoplasm. Both possibilities clearly challenge the hypothesis of passive ammonia transport.

Figure 4.  Mechanistic scheme of Amt transport. (A) At physiological pH values ammonium will be almost exclusively in the cationic form, such that translocation of NH₃ implies extracellular deprotonation and intracellular reprotonation. Uniport of NH₃ must therefore be considered a net antiport of versus H⁺. (B) Electrogenic transport as observed in some plant AMTs and Rhesus proteins is either a net uniport of or a symport of NH₃/H⁺.

A possible solution to explain both the observed currents in voltage-clamp experiments and the seemingly passive uptake of NH₃ into liposomes may be found in a mechanistic model that links the transport of NH₃ to a symport of H⁺. In total this would result in a net uniport of , but after binding of substrate to the transporter, a deprotonation would occur with both ammonia and protons following separate – but coupled – pathways. In the presence of a proton motive force the result would be an electrogenic, secondary active transport while in its absence passive conductance of NH₃ may be observed (Figure 4B).

Conclusions and outlook

Recent years have seen exciting developments towards our understanding of the function of Amt proteins. High-quality structural information has become available and functional data have been collected through a multitude of techniques. Yet the results are inconclusive in respect to some of the most fundamental questions on Amt transport. Based on the crystal structures, theoretical calculations have been presented, but all were following the pre-assumption of passive transport. Ammonium has been predicted to enter the protein in its charged form and get deprotonated upon reaching the histidine pair deep inside the membrane, but the further fate of the lone proton has not yet been investigated [78].

Within a diverse family of transport proteins it is not without precedent to find highly similar proteins that work by very different mechanisms. The ClC family of chloride channels/transporters for example features both active and passive transport systems with nearly identical three-dimensional structures [37], [50], [79] and a similar situation seems possible for Amt proteins considering the fundamental differences in cellular environment found, e.g., in soil bacteria, hepatocytes, or plant root nodules. The question whether facilitated diffusion of NH₃ is bioenergetically reasonable is, however, of a fundamental nature. It has been argued that all free ammonium within a cell may immediately be converted to glutamine such that a flowing equilibrium would allow for passive influx even at low external ammonium concentrations [76], but any mechanism coupled to the action of glutamine synthetase would still lead to intracellular consumption and extracellular generation of protons and seems thus energetically unlikely.

Further experiments will be required to closely examine the functionality of Amt proteins in vivo and in vitro and recent structural data have provided an ideal starting point and inspiration for their design.

Acknowledgements

The authors gratefully acknowledge financial support from the European Community 6th framework program (Marie Curie) and the DFG Emmy-Noether-Programme to S.L.A.A. and from the EMBO Young Investigator Programme to O.E.