Regulation of NH 4+ transport by essential cross-talk between AMT monomers through the carboxyl-tails

The ammonium transport across plant plasma membranes is facilitated by AMT/Rh-type ammonium transporters, which have homologs also in most organisms. In the root of the plant Arabidopsis thaliana , AMTs have been identified which function directly in the high affinity NH 4+ acquisition from soil. Here we show that AtAMT1;2 has a distinct role, as it is located in the plasma membrane of the root endodermis. AtAMT1;2 functions as a comparatively low affinity NH 4+ transporter. Mutations at the highly conserved carboxyl-terminus of AMTs, including one that mimics phosphorylation at a putative phosphorylation site, impair NH 4+ transport activity. Co-expressing these mutants along with wild type AtAMT1;2 substantially reduced the activity of the wild type transporter. A molecular model of AtAMT1;2 provides a plausible explanation for the dominant inhibition, as the carboxyl-terminus of one monomer directly contacts the neighboring subunit. It is suggested that part of the cytoplasmic carboxyl-terminus of a single monomer can gate the AMT trimer. This regulatory mechanism for rapid and efficient inactivation of NH 4+ transporters may apply to several AMT members to prevent excess influx of cytotoxic ammonium.


Introduction
Ammonium (NH 4 + /NH 3 ) transporters of the AMT/Rh family are identified throughout all domains of life, including archae, bacteria, fungi, plants and mammals (Ludewig et al., 2001;von Wiren and Merrick, 2004). The ammonium transporters (AMTs) from different species appear to have contrasting transport mechanisms, depending on their physiological role (Ludewig, 2006). While the function of AMTs from plants is the net import and accumulation of NH 4 + (Mayer et al., 2006), the Rh glycoproteins, in contrast, appear to conduct NH 3 or facilitate NH 4 + /H + exchange for export and disposal (Westhoff et al., 2002;Zidi-Yahiaoui et al., 2005;Mayer et al., 2006).
The transcripts of at least three AMTs from Arabidopsis are regulated by nitrogen availability in roots (Gazzarrini et al., 1999). In several plant species studied, AMT transcripts are up-regulated by nitrogen limitation. In roots AtAMT1;1 is localized in the plasma membrane of the rhizodermis, cortex and pericycle (Mayer and Ludewig, 2006) and is responsible for about 30% of the ammonium acquisition in Arabidopsis roots (Kaiser et al., 2002). AtAMT1;1 is partially co-localized with AtAMT1;3, which is also expressed in the rhizodermis and cortex and participates in another 30% of NH 4 + uptake of roots (Loque et al., 2006). The residual 40% of ammonium influx in roots appears thus to be carried by AtAMT1;2, which is studied here, and AtAMT2;1 which has been mainly localized to the vasculature, but promoter activity was also identified in the cortex and root tip (Sohlenkamp et al., 2002).
Whether plant AMT transporter activity is fine-tuned by post-transcriptional mechanisms is unknown, but the homologous bacterial AmtB is negatively regulated by reversible binding of the signal transduction P II protein GlnK to AmtB (Coutts et al., 2002). The deletion of part of the cytoplasmic tail impairs that interaction and reduces the transport activity by ~70% (Coutts et al., 2002). A similar regulation of plant AtAMTs is unlikely since the single P II protein from Arabidopsis is located in chloroplasts (Smith et al., 2002). However, the cytosolic carboxyl-(C) terminus has been shown to be important for transport function in AMTs from Lycopersicon esculentum (tomato): the transport by AMTs was inhibited by a specific mutation in that region (Ludewig et al., 2003). An exchange of a conserved glycine by aspartate (G458D in LeAMT1;1 and G465D in LeAMT1;2) abolished transport, although the mutation did not affect AMT localization at the plasma membrane (Ludewig et al., 2003).
The importance of this glycine had been initially identified in the homologous Mep transporters from the yeast Saccharomyces cerevisiae. In the mep1-1 mutation, the corresponding glycine was exchanged to aspartate (G413D) causing inactivation of Mep1p (Marini et al., 1997;Marini et al., 2000). implicated the importance of the cytosolic C-terminus in the transport activity of AMTs, the physiological relevance of these observations remained nebulous.
Identification of the high-resolution structures of the prokaryotic AmtB from E.coli did not clarify how the carboxy-terminus is involved in the transport of NH 4 + (Khademi et al., 2004;Zheng et al., 2004). Although the cytosolic C-terminus has not structurally been resolved, it appeared that this tail was not a central constituent of the pore, questioning its importance for transport activity. The analysis of homologous transporters in the fungal species Candida albicans revealed that the complete deletion of the cytoplasmic C-terminus impaired AMT function, but shorter truncations covering the region around the conserved glycine were functional in ammonium transport. Interestingly, the cytoplasmic C-terminus was essential for ammoniumrelated signalling (Biswas and Morschhauser, 2005).
In this study we propose a physiological molecular mechanism of how the carboxylterminus regulates the activity of AMT transporters. We concentrated on AtAMT1;2, which is the only root-expressed AMT in Arabidopsis that has not been celluarly localized and conflicting data on the methylammonium transport properties by AtAMT1;2 had been published (Gazzarrini et al., 1999;Shelden et al., 2001). Here it is shown that AtAMT1;2 is a comparatively low-affinity NH 4 + transporter which is preferentially localized in the root endodermis. Specific mutations in the conserved Cterminus of AMTs, including an exchange in a putative phosphorylation site, impair transport. Most interestingly, a molecular model shows that part of the carboxy-tail of one monomer attaches to its neighbor; this provides a reasonable explanation for the observed cross-inhibition of functional, co-expressed AMT monomers. It is concluded that this provides a physiological mechanism to effectively prevent excess ammonium influx and toxicity.

Preferential localization of AtAMT1;2 in the plasma membrane of endodermal root cells
The transcripts of several NH 4 + transporters have been identified in roots, including those of AtAMT1;1 and AtAMT1;2 (Gazzarrini et al., 1999). In apical zones of the roots, the GFP-fusion protein with AtAMT1;1 under the control of the endogenous promoter was preferentially identified in epidermal and cortical cell layers ( Fig.1   A 2006). In contrast, the GFP-tagged AtAMT1;2 was preferentially found in the plasma membrane of the root endodermis, when expressed from the endogenous promoter ( Fig.1 D,E,F). A minor fraction of AtAMT1;2-GFP fluorescence was also observed in cortical cells. Trans-cellular apoplastic transport across the endodermal cell layer is blocked by the casparian strip, thus AtAMT1;2 transfers NH 4 + from the apoplast of outer cell layers into the endodermal cytoplasm for further release into the stele.

Transport characteristics of AtAMT1;2 in oocytes
NH 4 + induced large inward currents in AtAMT1;2-expressing oocytes ( Fig.2A). Similar to the results with other AMT transporters, ionic currents elicited by addition of ammonium were exclusively inward, even at positive voltages. The inward current elicited by 1 mM ammonium was larger than the current induced by the equivalent amount of MeA ( Fig.2A). At -100 mV, the concentration needed to achieve half maximal currents was ~140 µM for NH 4 + and ~1.9 mM for MeA + (Fig.2B). The NH 4 + concentration needed to saturate AtAMT1;2 was much higher than for AtAMT1;1, which has a more than 10-fold higher affinity (Mayer and Ludewig, 2006;Wood et al., 2006). It is possible that the different Km values somewhat reflect the apoplastic ammonium concentrations at the rhizodermis and endodermis.
The concentration needed to achieve half maximal currents ("the K m ") was analyzed at each voltage separately and was found to differ. Less ammonium was required to elicit half maximal currents at more negative voltages. This indicates that both, higher ammonium and more negative vollage lead to saturation (Fig.2C). A similar finding has been made with LeAMT1;1 (Ludewig et al., 2002). Assuming a single binding site for NH 4 + the transport and saturation are characterized by the entry of NH 4 + into the pore and its exit, either back into the external medium or to the cytoplasmic side (this corresponds to transport). The entry (and the exit) rate of NH 4 + , and thus the K m , depend on the membrane voltage, as long as the binding site is located within the membrane electric field. The slope of the voltage dependence of the K m was fitted and the fractional electrical distances δ NH4+ = 0.56 and δ MeA+ = 0.26 were obtained.
These values can be interpreted in the way that the binding sites for NH 4 + and MeA + are located 56% and 26%, respectively, inside the membrane electric field, measured from the outside. Interestingly, the steep voltage dependence indicates that NH 4 + enters deeply into the pore and that it has to cross more than half of the membrane electric field to reach this site.

AtAMT-transport activity is inactivated by mutations in the carboxyl-terminus
The sequences of AtAMT1;1 and AtAMT1;2 share high similarity that even extends into the cytoplasmic carboxyl-terminus (Fig.3A). Specific mutations of a highly conserved glycine in this region have been shown to inactivate AMT/MEP transporters from fungi and tomato (Marini et al., 2000;Ludewig et al., 2003;Smith et al., 2003). However, the role of the cytoplasmic tail is unclear and how the mutation inhibits the transport function remains obscure.
The sequences of AtAMT1;1 and AtAMT1;2 are identical over a strech of 17 amino acids in the carboxyl-terminus (Fig.3A). Within that sequence, the conserved and functionally indispensable glycine is identified (Fig.3A). In addition, a partially conserved threonine is recognized, which was phosphorylated in a large-scale screen for plasma membrane phospho-proteins (Nuhse et al., 2004). The phosphorylation of this threonine may suggest that post-translational modifications are involved in the regulation of AMT function. It is worth mentioning that the identified phospho-peptide (ISSEDEMAGMDMpTR) was isolated from Arabidopsis suspension cells that were grown in ammonium-rich media (2 mM) (Nuhse et al., 2004). When grown in nitrogen rich media, the high affinity uptake systems are efficiently shut off in roots (Loque et al., 2006).
The threonine is conserved in a number of AMTs from different plant species and even in some bacterial and archaeal AMTs (Suppl. Fig.1). In several other AMTs, including the prokaryotic AMT-1 from Archaeglobus fulgidus, this threonine is replaced by serine. Other amino acids are found at the equivalent position in most fungal homologs while the carboxy-tail of mammalian Rh glycoproteins fails to show significant conservation with AMTs.
Whether modifications in the relevant threonine (T460 in AtAMT1;1) affect NH 4 + transport was tested by introducing mutations at this position. The threonine was replaced by aspartate which is negatively charged at physiological pH and may mimic phosphorylation. The exchange in AtAMT1;1 inactivated NH 4 + transport, as determined by growth assays of triple−mep∆ yeast expressing the T460D mutant construct (Fig.3B). By contrast, the mutation of T460 to the uncharged alanine, which is the corresponding residue in AtAMT1;5, yielded a functional transporter (Fig.3B). 9 functional (Fig.4A). In addition, the mutant in which the adjacent glycine (G468) was exchanged to aspartate was also non-functional (Fig.4A). The equivalent mutation inactivates fungal homologs (Marini et al., 2000;Monahan et al., 2002) and two AMTs from tomato (Ludewig et al., 2003). By contrast, mutants having a partially conserved serine at position 461 (Suppl. Fig. 1) exchanged to alanine or aspartate were both functional (Fig.4A).

Equivalent mutations in
The mutations at positions 461, 468 and 472 did not affect the localization pattern of the GFP-tagged AtAMT1;2 proteins expressed in yeast, suggesting that targeting of the membrane proteins was unaffected (Fig.4B). This resembles the properties of the dominant glycine to aspartate mutant in yeast (Marini et al., 2000) and tomato (Ludewig et al., 2003).
The transport rates of the AtAMT1;2 wild type and the mutants T472D, T472A and G468D were also quantified using 14 C-methylammonium ( 14 C-MeA) transport assays in yeast. The T472D and G468D mutants had no residual transport activity when expressed in yeast (Suppl. Fig.3A). In contrast, robust 14 C-methylammonium uptake was observed for the AtAMT1;2 wild type and only slightly reduced uptake for the T472A mutant (Suppl. Fig.3B). Taken together, a specific mutation that mimics phosphorylation at the putative phosphorylation site in the carboxyl-terminal tail of AMTs abolishes NH 4 + transport.

A molecular model of AtAMT1;2 predicts carboxyl-terminal interactions
The crystal structures of prokaryotic AMT homologs were used to generate an AtAMT1;2 homology model. The highest resolution structures of E.coli AmtB (1U7G) and A.fulgidus AMT-1 (2B2F) were taken as templates (Khademi et al., 2004;Andrade et al., 2005). The structures aligned well within the core transmembrane region (Fig.5). Major deviations in the AtAMT1;2 homology model were restricted to the external and internal loops that connect the transmembrane helices. Interestingly, a surprisingly precise alignment and structural model was obtained from the highly conserved cytosolic C-terminus, which forms a structure that includes two short helices (CH1 & CH2) (Fig.5A). The C-terminus was fully ordered and resolved in the AfAMT-1 structure; it aligns with the cytoplasmic face of the same subunit, but also forms hydrogen bonds with the adjacent monomer in the trimer. The same fold is observed in the AtAMT1;2 model; the carboxyl-terminus of the neighboring chain B (blue) is positioned on top of the cytoplasmic surface of chain A (red) (Fig.5). The structure of AfAMT-1 and the model suggest that subunit interactions by the carboxyl-terminus are of functional importance. A close inspection of the position of the putative phosphorylation site threonine 472 identified tight packing with its neighbors, including two residues from the M7-M8 linker of the adjacent subunit (residues G323 and H324). The side chain was accessible from the cytoplasm, which opens the possibility for its modification (Suppl. Fig.2). By contrast, the serine 461 was positioned further away from the adjacent subunit at the beginning of CH1.
Interestingly, glycine 468, which when mutated to aspartate inactivates AMTs from many species, is located in the hinge between helices CH1 and CH2 and is adjacent to T472 (Suppl. Fig.2). It is obvious that a side chain larger than in glycine cannot be accommodated at that position. Any exchange to another amino acid must disrupt the helix-loop-helix structure and the interaction with adjacent residues. Similarly, in AfAMT-1, the oxygen of the corresponding backbone glycine forms two hydrogen bonds with the serine that corresponds to the threonine in AtAMTs. The first hydrogen bond is formed from the glycine (G379:O) to the backbone amino group of serine (S383:N) and the second with the hydroxy group of the serine side chain (S383:OH). Whether modifications in a single subunit also affect proper functioning of co-assembled monomers was tested using co-expression of mutant and wild type in Xenopus oocytes.

NH 4 + transport by AtAMT1;2 mutants in oocytes
The large magnitude of the currents elicited by ammonium in AtAMT1;2-expressing oocytes allowed a reliable and quantitative comparison with currents by mutant transporters (Fig.6A). Consistent with the non-functionality of the mutants T472D and G468D in yeast, injection of equal amounts of cRNA did not lead to detectable NH 4 + currents. The currents from oocytes expressing these mutants were indistinguishable from water injected and non-injected controls (Fig.6A). In contrast, NH 4 + currents were detected in T472A mutant-expressing oocytes, but these were of almost 10-fold lower magnitude compared to the AtAMT1;2 wild type.

Cross-inhibition by co-expressed non-functional monomers
Further analysis was done by co-expression of equal amounts of mutant and wild type cRNA. Doubling the amount of injected cRNA roughly doubled the NH 4 + current by AtAMT1;2 (Fig.6B). However, co-expression of equal amounts of cRNA from wild type and mutant G468D drastically reduced the NH 4 + current to below that of AtAMT1;2 expressed alone. This is consistent with data from tomato AMTs, where the corresponding mutant also inhibited transport by wild type monomers (Ludewig et al., 2003). NH 4 + transport by AtAMT1;2 is reduced to 15% by the co-expressed mutant T472D (Fig.6B). Assuming equal processing and stability of the wild type and mutant proteins, a binomial distribution of wild type/mutant subunits within the trimeric complexes is expected. If a single mutant monomer is sufficient to inactivate the entire trimer, only the trimers consisting completely of wild type monomers will be active and only 12.5% of residual current will remain.
The mutant T472A which elicited less current than the wild type also partially inhibited the NH 4 + current by the co-expressed AtAMT1;2. The current was, however, larger than in co-injections of wild type and non-functional mutants. Despite the reduced NH 4 + transport by the T472A mutant, the residual activity was sufficient to restore yeast growth to the wild type level (Fig.4A). Taken together, the data suggest that the correct carboxy-terminal fold is required for AtAMT activity and that disruption of the carboxyl-tail of a single monomer will inactivate the entire AMT trimer.

Weak temperature dependence of AtAMT1;2
If gating and conformational coupling between co-assembled monomers occurs, larger conformational rearrangements might be involved. We tested that hypothesis by measuring the temperature-dependence of the NH 4 + currents. A weak temperature-dependence was observed for AtAMT1;2 currents (Q 10 = 1.5), which is consistent with a diffusion controlled transport process and minimal conformational rearrangements during transport (Fig.7A).
In contrast, a steeper, but still weak temperature dependence (Q 10 = 1.9) was measured in the partially active T472A mutant, suggesting that some minor conformational rearrangements between active and inactive states may occur. The apparent activation enthalpy was 50 kJ/mol (compared to 33 kJ/mol in AtAMT1;2) ( Fig.7A).

Discussion
Nutrients, such as ammonium, are mostly absorbed at the epidermis (rhizodermis) and move symplastically through the cortex to the stele. However, nutrients may enter the symplasm later in cortical and endodermal cells, but the casparian strip provides a major barrier for further apoplastic movement. required to prove whether conformational changes occur within the cytoplasmic parts of AMTs. The cytosolic TM5-TM6 and TM9-TM10 linkers were partially disordered and varied within the different crystal structures of EcAmtB, which may indicate some flexibility of these peptides (Zheng et al., 2004). Similarly, the structure of the carboxyl-terminus of EcAmtB (1U7G) was disordered which, however, may simply have resulted from the histidine-tag that was attached to AmtB for purification.
EcAmtB interacts with the regulating GlnK protein, which negatively regulates transport. Interestingly, a deletion in the C-terminus reduces the activity of AmtB to roughly 30% and impairs the binding of GlnK to AmtB (Coutts et al., 2002). Very recently, the structure of AmtB in complex with GlnK was resolved (Conroy et al., 2007;Gruswitz et al., 2007). This structure revealed that the carboxy-tail of EcAmtB adopts the identical fold as in AfAMT-1 and in the AtAMT1;2 model structure. The effects of mutations in the C-terminal region on EcAmtB function were also recently investigated (Severi et al., 2007). Remarkably similar conclusions regarding potential interactions between AmtB monomers and the possible role of the C-terminal region have been drawn. Two classes of mutants were identified, either with residual (~25%) activity, or essentially inactive. These two mutant classes were interpreted to reflect two distinct states of the carboxy-terminus and hence EcAmtB (Severi et al.,

2007).
A weak temperature dependence corresponding to a Q 10 of 1.5 was observed for the wild type currents. A Q 10 value of ~1.3-1.6 is characteristic for diffusion limited processes but enzymatic reactions and large conformational changes in proteins are frequently associated with a higher Q 10 of 2-3 (Liu et al., 1996). The observed Q 10 value is compatible with minor conformational changes during transport and a channel-like NH 4 + transport mechanism. However, minor conformational changes that would be required in a H + coupled NH 3 co-transporter mechanism (= net NH 4 + transport) cannot be excluded. An even lower temperature dependence of transport has been reported for EcAmtB (Javelle et al., 2005) and the structurally related Rh glycoproteins (Zidi-Yahiaoui et al., 2005).
When compared with the wild type AtAMT1;2, the transport activity of the T472A mutant was stronger reduced in oocytes than in yeast. It is possible that other cytosolic factors differently regulate wild type AtAMT1;2 activity in these systems; e.g.
one might speculate that AtAMT1;2 is partially phosphorylated in yeast, but not in oocytes. Furthermore, the reduced transport activity of T472A correlated with a larger activation enthalpy which corresponds to a Q 10 of 1.9. It is likely that the overall mechanism of NH 4 + conduction is not altered. However, it was evident from the structure of the carboxyl-terminus (Fig.5) that any exchange in T472 must alter the original fold. We suggest that the active, conducting state is less favored in the mutant T472A and that the equilibrium between the conformations is shifted to the inactive state. This gives a plausible explanation for the reduced current in that mutant, especially at the lower temperatures at which the oocyte experiments were performed. In contrast, protein modifications at the conserved stretch of the Cterminus appear to manipulate the ratio between active and inactive states in the wild type (Fig.7B). If phosphorylation has the same effect as the aspartate mutation, a specific kinase will then be required for inactivation to occur. It is likely that the appropriate kinase for a plant protein is not endogenously expressed in oocytes. This explains the weak temperature dependence of the wild type in oocytes, as it is constitutively active.
Regulation by cytosolic peptides is not unprecedented in membrane transport. For example, the molecular mechanism of inactivation in K + ion channels involves reversible blockage of the conduction pathway by the peptide tail (Hoshi et al., 1990).
Similarly, plant aquaporins are gated by phosphorylation at flexible cytosolic loops.
Aquaporins are tetramers with each monomer forming a solute channel and transition between the open and closed conformations leaves the overall pore architecture intact (Tornroth-Horsefield et al., 2006). Interestingly, interactions between aquaporin monomers affect their overall activity (Fetter et al., 2004).
A schematic model of AMT gating is presented in Fig.7B. The post-translational modification of a single monomer and disruption of the conserved part of its carboxytail fold appear to be sufficient to inactivate its physical neighbors and the entire trimer. This is a much more efficient shut-off than activation mechanism. Its physiological role is likely to minimize excess influx of cytotoxic ammonium; this allows plant roots to rapidly cope with variable levels of nitrogen supply. AtAMT1;1, AtAMT1;2 and AtAMT1;3 are in direct contact with the external root apoplasm. These AMTs are highly conserved in the relevant region, including the threonine. The conservation of the putative phosphorylation site in AMTs from many plant species suggests that the proposed regulation is potentially a feature of many plant AMTs.  conditions. 14 C-MeA uptakes were performed at an optical density of 5 using 100 µM MeA as described (Gazzarrini et al., 1999).  Homology modeling -The primary sequence of AtAMT1;2 was aligned with