Research Article

Mechanism of Ammonia Transport by Amt/MEP/Rh: Structure of AmtB at 1.35 Å

See allHide authors and affiliations

Science  10 Sep 2004:
Vol. 305, Issue 5690, pp. 1587-1594
DOI: 10.1126/science.1101952


The first structure of an ammonia channel from the Amt/MEP/Rh protein superfamily, determined to 1.35 angstrom resolution, shows it to be a channel that spans the membrane 11 times. Two structurally similar halves span the membrane with opposite polarity. Structures with and without ammonia or methyl ammonia show a vestibule that recruits NH 4+/NH3, a binding site for NH 4+, and a 20 angstrom–long hydrophobic channel that lowers the NH 4+ pKa to below 6 and conducts NH3. Favorable interactions for NH3 are seen within the channel and use conserved histidines. Reconstitution of AmtB into vesicles shows that AmtB conducts uncharged NH3.

The transport of ammonia/ammonium is fundamental to nitrogen metabolism throughout all domains of life (14). In what follows, Am refers to (NH3 + NH 4++), and MA to (CH3NH2+ CH3NH3). The so-called ammonia transporters (Amt proteins), whose paralogs in yeast are called methylammonium/ammonium permeases (MEP proteins) (5, 6) are typically >420 amino acids in length and are assembled as trimers of proteins (7). The transport rates of most Amt/MEP proteins can be conveniently measured by transport of radioactive 14C–labeled MA (8).

Plant, bacterial Amt, and yeast MEP family members show a range of concentration dependence for Am conductance up to “high affinity,” as reflected in their ability to grow on very low concentrations (<5 μM) of ammonium salts as sole nitrogen source. The “response Km” (the concentration that evokes half-maximal conductance) for Am is ∼5 to 10 μM for MEP1 and 1 to 2 μM for MEP2. MEP3 is of much lower affinity, Km ∼ 2mM (6). amtB and glnK located on the same operon, are cotranscribed in bacteria (9) and archaea, and GlnK plays a dynamic role in the regulation of nitrogen uptake (10) by binding to AmtB (11, 12). In humans, the Rhesus (Rh) family of proteins, both erythroid (RhAG, RhD, and RhCE) and nonerythroid (RhCG, RhBG, and RhGK), also share conservation with the Amt/MEP family throughout their sequence (13) (Fig. 1).

Fig. 1.

The amino acid sequence alignments of AmtB/MEP/Rh homologs from E. coli (AmtB_E. coli), A. aeolicus (AmtB_AQFX), Neurospora (AMT_Neurosp), Saccharomyces cerevisiae (MEP2_Sacch), Lycopersicon esculentum (LeAMT1_Lycop), Arabidopsis thaliana (AMT_Arabid), Caenorhabditis elegans (AMT1_Celeg), and human Rh factors (RhBG, RhCG, and RhAG). The numbering is that of E. coli AmtB. Conserved amino acids are in white in red-filled rectangles. Similar residues are in red surrounded by blue lines. The signal sequences in E. coli and A. aeolicus are underlined. Residues that line the lumen of the channel are labeled with asterisks above. Eleven transmembrane helices are identified by helical motifs above the sequence, labeled by transmembrane helix numbers M1 to M11.

Our structural analysis is based on crystals grown in the absence or presence of Am or MA. It reveals a recruitment vestibule for cations such as NH 4+ or neutral NH3, a site that can bind CH3NH3+ or NH4+ using π-cation interactions, and a hydrophobic channel that incorporates NH3 using weak interactions with C-H hydrogen bond donors. Our assays carried out on reconstituted vesicles show that the protein conducts NH3 because addition of Am salts outside raises internal pH. This is therefore the first structure of a transmembrane channel family that can conduct unhydrated molecules that in isolation would be gaseous.

Structure Determination

We cloned two orthologs of AmtB, from Aquifex aeolicus (AmtB_AQFX) (14) and Escherichia coli (AmtB_Ecoli) (4, 7). Twenty amino acids were excised from the N terminus of AmtB_AQFX and 22 were excised from AmtB_Ecoli, as established by matrix-assisted laser desorption/ionization mass spectroscopy (MALDI-MS) and N-terminal amino acid sequencing. This and the prediction of signal peptide cleavage sites through neural network approaches (15) led to the identification of these regions as signal sequences, which are removed upon insertion of the proteins into the cell membrane (Fig. 1). Crystals of AmtB_Ecoli, grown in the presence of 25 mM AmSO4 at pH 6.5, diffract to a resolution of 1.35 Å (Table 1). Crystals of AmtB_AQFX diffracted to 4.5 Å and have not yet been pursued.

Table 1.

Crystallographic statistics to 1.35 Å resolution. Crystals of SeMet AmtB (used for all crystallography) were in space group P63. Data were collected at the ALS beamline 8.3.1, with a CCD detector (Quantum 4), and integrated and scaled with DENZO, SCALEPACK (Native), and MOSFILM (MAD data). Phases were calculated with CCP4. After solvent flattening and phase extension to 2.0 Å, the model was refined with CNS to 1.8 Å and SHELX to 1.35 Å.

Data set With 25 mM AmSO4 With 100 mM MASO4 Without Am/MA SeMet MAD remote SeMet MAD peak
Cell dimensions (a Å, c Å) 96.5, 94.6 95.7, 94.6 95.8, 94.7 95.8, 94.7 95.8, 94.7
Wavelength (Å) 1.1159 1.1159 1.1159 1.0080 0.9797
Resolution (Å) 1.35 1.85 2.0 2.5 2.5
Multiplicity 14.3 13.3 19.6 7.3 6.6
Completeness %View inline 97.3 (90.4) 99.7 (99.3) 87.4 (48.3) 99.7 (99.6) 99.7 (99.6)
I/σ 20.4 (1.6) 19.3 (1.9) 25.8 (2.0) 10.6 (3.3) 9.4 (2.8)
Rsym (%) 5.8 (62.1) 6.4 (80.0) 6.4 (69.5) 11.4 (81.0) 12.5 (87.0)
Phasing power 0.7 0.34
Figure of merit 0.265
(after solvent flattening) (0.877)
Rcryst/Rfree (%) 13.3/16.8 19.5/20.5 18.2/20.5
RMSD bond length (Å) 0.007 0.007 0.013
RMSD bond angle (°) 1.3 1.3 1.8
Mean B factor (Å2) 24.6 37.5 43.2
  • View inline* Numbers in parenthesis refer to the high-resolution shell for data refinement (1.45 to 1.40 with AmSO4, 1.92 to 1.85 for MASO4, and 2.07 to 2.00 for without substrate).

  • A Trimer of Three Channels

    AmtB crystallizes as the physiological threefold symmetric trimer of channel-containing proteins (Fig. 2, A and B). The trimer extends ∼65 Å parallel to the threefold axis and is 81 Å in diameter in the plane of the membrane. Eleven transmembrane-spanning α-helices (M1 to M11) form a right-handed helical bundle around each channel. Residues from M1, M6, M7, M8, and M9 of one monomer interact with helices M1, M2, and M3 of the neighboring monomer, with a total interacting surface area of 2716 Å2. As described for other membrane proteins, polar aromatic side chains of residues Tyr62 at the periplasmic (termed extracellular) side and Tyr180, Trp250, and Trp297 at the cytoplasmic side would lie in the membrane-aqueous phase interface. The trimer has a net negative charge of –7.5 (13.5 positive + 21 negative) on the outer surface and a net positive charge of +9 (42 positive + 33 negative) on the cytoplasmic side, as is the trend in membrane proteins.

    Fig. 2.

    Three-dimensional fold of AmtB. (A) Ribbon representation of the AmtB trimer viewed from the extracellular side. In the top monomer, the quasi- twofold axis (vertical, in the plane of the page, and intersecting the threefold axis) relates the left side to the right side. Pairs of corresponding quasi-twofold related helices are shown in the same color. The right monomer has a solvent-accessible transparent surface, colored according to electrostatic potential (red for negative and blue for positive). The three blue and one orange spheres are potential ammonia molecules and an ammonium ion, respectively. (B) A stereoview of the monomeric ammonia channel viewed down the quasi-twofold axis. In this and all subsequent figures, the extracellular side is uppermost. The vertical bar (35 Å) represents the inferred position of the hydrophobic portion of the bilayer. Three NH3 molecules seen only when crystallized in the presence of ammonium sulfates, are shown as blue spheres. (C) The amino acid sequence of AmtB is arranged topologically as in the structure, with helices viewed as if from inside the channel looking away from the threefold axis. The quasi-twofold axis is perpendicular to the center of the figure. Five helices compose each segment, labeled M1 to M5 and M6 to M10. Related helices M1 and M6, M2 and M7, etc., are boxed in similar colors. Side chains of residues in red circles contribute to the substrate-contacting walls of the channel. Residues in blue circles contribute side chains to the inter-monomer contacts that immediately surround just the threefold axis of the trimer, because many oligomeric membrane proteins either need to insulate against passage of alternate molecules there or use this special location for stability, as in the aquaporins, or for conductance as in K+ channels. The deduced location of the cell membrane is illustrated in gray (35 Å) with light gray for the head group region (to 40 Å thickness).

    At the extracellular side, the threefold axis is surrounded by three closely packed copies of just M1 of each monomer, which seal the central axis against passage. Toward the cytoplasmic side, the three M1s (+16° to each other) veer away from the threefold axis to leave an open pocket ∼10 Å across, formed by three copies of M1 and M6. M1 has a kink (22°) in the helix secured by the only cis-proline (Pro26) in AmtB, a residue not conserved in the superfamily. M1 and M6 are not long enough to span the bilayer, consistent with the trimer being the stable physiological quaternary structure. The interfaces between subunits are almost as hydrophobic as the exterior, suggesting that a monomer could be transiently stable in the membrane upon synthesis, before forming trimers. An isolated square planar arrangement of four water molecules, each hydrogen-bonded to each other (average hydrogen bond length 3.0 Å), makes two hydrogen bonds to carbonyl oxygens of Cys56 and Ala102 in an otherwise hydrophobic cavity in the interface.

    An 11-Crossing Membrane Protein with a Quasi-Twofold Axis in the Plane of the Membrane

    We found that AmtB from E. coli and AmtB from A. aeolicus each begin with a signal sequence that is cleaved during synthesis. The topology of AmtB was correctly determined by a careful series of PhoA and LacZ insertions, and the C terminus was shown to be cytoplasmic (4). Within each AmtB channel, M1 to M10 diverge outward from the central plane in a right-handed helical bundle to generate a vestibule on each surface. Their inter helix angles are +26° to –58°. Within the trans-bilayer region, there are three glycine Cα-H...O=C hydrogen bonds (between the Cα hydrogen of glycine and a main chain carbonyl oxygen of a neighboring α-helix) between M1 and M6 (Gly204, Gly211, and Gly34) and one between M8 and M10 (Gly325). Cys109 and Cys56, found only in E. coli AmtB, are close enough to form a disulfide in the transmembrane region.

    The fold is not homologous to any other membrane protein structure previously reported. Although there is no evidence for any gene duplication in the family, the structure of M1 to M10 reflects a quasi-twofold axis in the midplane of the membrane that intersects the trimer threefold axis. It relates the structural context of the M1 to M5 region to that of M6 to M10. M11 is an additional ∼50 Å–long straight helix, inclined at –50° to the perpendicular to the membrane plane that surrounds the lipid accessible side of each monomer (Fig. 2, B and C). This type of structural duplication with opposite polarity with respect to the membrane plane is seen in a number of membrane proteins, including AmtB, GlpF and all aquaporins, the SecY protein of the translocon, and the ClC chloride channel. However, of these, only in the aquaporin family were the vestiges of duplication recognized in the gene sequences (16). The sparse vestiges of duplication now apparent in sequence after the structure imply that they should be identifiable in other proteins when conservation among multiple sequences is taken into account. Because of the opposite polarity of the duplicated segments relative to the membrane, the primordial gene duplication event of such membrane proteins must have occurred before generation of enough functional surroundings to support any transport of molecules from one side of the membrane to the other.

    The Mechanism of Transport or Conductance

    We sought to define any preferred sites for Am or MA and the mechanism for transport or conductance of these molecules by comparison of the structure in the absence of any ammonium derivative, in 25 mM AmSO4 at pH 6.5, and in 100 mM MASO4 at pH 6.5. These concentrations are in ∼1000-fold excess relative to the response Km (Am, Km ∼ 10 μM; MA, Km ∼50 μM). There is no significant overall conformational change of the protein with Am or MA, consistent with AmtB as a channel rather than as an ammonia transporter that would harness alternating conformational states. Therefore, from now on, we refer to this family of molecules as channels.

    The quasi-twofold (i.e., up to down) symmetric channel generated within each monomer by the structural antiparallel duplication begins with an extracellular vestibule. This is followed by one of two most constricted hydrophobic regions (Fig. 3A). The mid-membrane center of the pathway has two in-line histidines followed by the second constricted hydrophobic region and the intracellular vestibule. Between the two hydrophobic constrictions the channel wall is narrow and mostly nonpolar throughout its ∼20 Å length, consistent with conduction of uncharged NH3.

    Fig. 3.

    The ammonia-conducting channel in AmtB. (A) The surface of the lumen of the ammonia channel is colored according to electrostatic potential, after removal of surfaces of helix M9 and parts of M8 and M10 (whose helical backbones are indicated by the cyan line). The positions of two histidines near the three NH3 sites (blue spheres) are shown in yellow and blue stick representation and were not included in the surface electrostatic calculation. Two narrow hydrophobic regions through the channel lie above and below the NH3 positions (the zone within a dashed rectangle). The orange sphere represents an ammonium ion. (B) Stereo views of the CH3NH3+/NH4+ binding site Am1. The electron density map (2Fo-Fc) is contoured at 2σ for the protein (blue), and the (Fo-Fc) CH3NH3+ omit map is contoured at 4.5σ (red) for AmtB in 100 mM MASO4 at pH 6.5. The MA order is 67% occupancy for each (CH3 and NH3) group. Hydrogen bonds between the NH of CH3NH3+ and Oγ of Ser219 and between the Nϵ1H of Trp148 and O of Asp160 are indicated by yellow dashed lines, with bond distance in yellow. (C) As in (B), for AmtB in 25 mM Am SO4 at pH 6.5. Am order is 67% occupancy (6.7 electrons).

    The outer vestibule contains 30 ordered water molecules. Am has similar density to that of H2O; thus, if it replaces H2O, it is difficult to identify unambiguously as Am. The density for MA is clearly distinguished from that of H2O/Am, and the difference can uniquely mark a site. Because the Am and the MA crystals have slightly different unit cells and so were not isomorphous, Fo-Fo maps were not useful; thus, we cite Fo-Fc maps. One such MA site (60% ordered, where this order parameter represents the integrated electron content, referenced to the highest occupied water molecule) is located against both aromatic rings of Trp148 and Phe107 (Fig. 3B). The –NH +3 moiety is a hydrogen bond donor to the Oγ of Ser219. This provides a favorable two π-cation interaction for NH 4+ and for CH3NH +3, stabilized by ring currents, and is accessible to solvent. MA displaces two separate peaks (Am1, 67% ordered) and a water peak in the Am structure (Fig. 3C), very similar to two water peaks in the H2O structure (60% ordered). The order parameter we use primarily reflects movement in the site and secondarily statistical occupation.

    Mutation of conserved Asp160 to Ala160 completely destroys transport of MA, implying that Asp160 plays a key role (12), and Merrick proposed that it might provide a primary binding site for NH 4+. However, there is no peak of density against Asp160 even in the presence of 25 mM AmSO4 or 100 mM MASO4. Conserved Asp160 is a helix-capping residue for M5 (Fig. 4A), being a hydrogen bond acceptor at its Oδ2 from the OγH (2.51 Å) and N-H of Thr165 (2.8 Å) and at its Oδ1 from the N-H of Gly164 (2.7 Å) and the N-H of Gly163 (2.8 Å) at the N-terminal end of M5. Thus, the Asp160 carboxyl orients the carbonyls of Asp160, Phe161, and Ala162 that, along with carbonyls of Ser68, Ser219, Val147, and Trp148 from the outside ends of helices M2, M4, and M6, line the vestibule and make it cation-attracting. Asp160 itself is not accessible to bulk solvent and is conserved across the superfamily, underlying its key role that appears to be primarily structural. Cys326 shields one planar face of Asp160. The indole ring of Trp148 is interposed between Asp160 and the MA site and shields Asp160 from solvent. Trp148 is conserved among the Amt/MEP subfamilies. The lowest polar carbonyl oxygen is for Ala162 at z = 8.5 Å. The first hydrophobic constriction in the channel formed by Trp148, Phe103, Phe161, and Tyr140, all of which are conserved only in Amt's, and everywhere conserved Phe107 and Phe215 suggests possible π-cation interactions for NH 4+ on entry to the channel. The diameter is 1.2 Å; thus, the side chains of Phe107 and Phe215 must move dynamically during any conduction event.

    Fig. 4.

    (A) Electron density (2Fo-Fc) contoured at 1.5σ (blue) for the two-histidine region and surrounding structure, including conserved Asp160 that accepts four short hydrogen bonds (dashed yellow). Additional peaks Am2, Am3, and Am4 seen when crystallized with 25 mM ammonium sulfate are defined in the Fo-Fc omit map at 1.5σ (in red), indicating putative NH3 molecule positions (blue spheres). The hydrogen-bonding network shows interactions between His168 and His318 and NH3 peaks in yellow (distances in red). (B) Stereo view of the two-histidine center of the channel. Surrounding hydrophobic residues are shown in ball and stick representation. The surface representation covers other surrounding amino acids. Three ammonia-dependent sites are shown (blue spheres) with associated distances (dashed yellow line and yellow labels).

    In 25 mM AmSO4, the crystal structure shows three additional peaks (Am2, Am3, and Am4 of order 20%, 15%, and 20%, respectively), not present without AmSO4 adjacent to two quasitwofold, related, conserved imidazoles of His168 and His318 at the center of the narrow hydrophobic channel (Fig. 4, A and B). The partial order indicates that these Am-specific peaks are poorly ordered, consistent with there being no good localizing hydrogen bonds in the channel or with their being occupied alternately with each other, or conceivably partially occupied such that at higher concentration they might be fully occupied. This latter reason seems very unlikely, because Am is already at 1000-fold the Km. At this resolution (1.35 Å), C, N, and O in AmtB are clearly distinguished. The orientation and hydrogen bonding of the imidazoles are unambiguously determined. The OH of conserved Thr273 is a hydrogen donor to the C=OofLeu269 and therefore an acceptor of the hydrogen bond from Nϵ2H of His168. Unprotonated His168 Nδ1 and His318 Nδ1H are fixed by hydrogen bond to each other. They provide two Cϵ1-H hydrogen bond donors (17) to the N of Am2 and the N of Am3 (3.2 Å and 3.4 Å between heavy atoms, respectively), and one Nϵ acceptor for the N-H of Am4. The other surrounding side chains are all hydrophobic. π-cation stabilization is possible at Am2 from side chains of Phe215 and Trp212. In this low dielectric environment, the imidazole Cϵ1-Hs will appear more acidic than imidazole in aqueous solution, because the effective dielectric constant (ϵ) is much lower. Therefore, coulombic attraction forces will increase as 1/ϵ.

    Whereas imidazole nitrogens might act as a hydrogen donor N-H or an acceptor N at the lone pair of electrons on an unprotonated nitrogen, the Cϵ1–Hs can only be donors, easing passage of a molecule that is an acceptor, as is NH3 but not NH 4+. Thus, this signature structure harnesses all of its hydrogen-bonding potential to accommodate the passage of hydrogen bond–accepting molecules. The pKa of NH3 must be lowered to below 6 at sites Am3 and Am4, because peaks clearly reflect NH3 at a pH of 6.5 in the crystals. At some point close to the aromatic constriction, NH 4+ gives up its proton, leaving it predominantly on the side of entry, and NH3 is transported. The cation-selective vestibule, possibly down to the Am2 site, can recruit NH 4+. This and the aliphatic hydrophobic channel compatible with NH3 can explain many of the seemingly inconsistent observations of function.

    Phe31, Tyr32, and Val314 surround the second cytoplasmic constriction. Asp310, the conserved quasi-twofold relative of Asp160, similarly acts as a helix cap for M10. The first exit-peak of density within the channel pathway is hydrogen-bonded to Asp313 and could be close to where NH3 would reacquire a proton on the inside to become NH 4+.

    Ammonia Conductivity by AmtB

    In order to determine the substrate specificity and rates of conductance of AmtB, we adapted a fluorescence-based assay to measure the influx of ammonia into vesicles (liposomes or proteoliposomes) by monitoring the pH-sensitive fluorescence of 5-carboxy fluorescein (CF) (18, 19). Rapid mixing of CF-loaded vesicles with ammonium chloride (0.5 mM or 5 mM) was initiated at pH 6.8. The internal pH rose, reflecting influx of NH3 through the lipids and through AmtB, which subsequently acquires a proton from H2O inside, to give NH 4+ and OH (Fig. 5A).

    Fig. 5.

    Channel conductance in proteoliposomes. (A) The time course of change in pH inside of vesicles containing CF buffered by 20 mM Hepes as detected by fluorescence change. Initial pH was 6.8 both inside and outside. To initiate flux, 5 mM NH4Cl was added externally to protein-free liposomes (+) and to AmtB-containing proteoliposomes (solid triangles) (protein to lipid ratio was 1: 200 by weight) (36). To control for possible osmotic effects, instead of 5 mM NH4Cl, 5 mM NaCl was added to proteoliposomes (solid circles). The dashed lines are exponential fits to the data. (B) Water conductance assessed by concentration-dependant self-quenching of CF-containing liposomes (solid squares) and proteoliposomes (open circles) upon addition of 500 mM sucrose at t = 0. The osmotic change leads to conductance of water through the lipids or through protein. The dashed lines are exponential fits to the data.

    The rate constant of ammonia influx indicated by the rate of pH change, 115.6 ± 13.2 s–1 (average of n = 6 determinations, typical curve fitting error in a single curve), for 5 mM NH4Cl outside is 10-fold faster than for protein-free liposomes, 12.8 ± 0.7 s–1 (n = 6). The rate of pH change for 0.5 mM NH4Cl (one tenth the concentration) is 70 s–1 (60% of the rate). Substitution of 5 mM NH4Cl by 5 mM NaCl did not lead to any detectable change in fluorescence even after 10 s, showing that changes in CF fluorescence are due to ammonia influx and not, for example, to any water efflux due to change in osmolarity. As a further control, ammonia conductivity for proteoliposomes reconstituted with the aquaglyceroporin GlpF was measured at 11.4 ± 0.06 s–1 (n = 6), the same as for liposomes, consistent with there being no leakage due to the reconstitution procedure per se. Thus, when the ammonium concentration drops 10-fold, the rate drops only 40%, generally consistent with the fact that the Km for transport is below 500 μM Am.

    To assess possible water conductivity by AmtB, osmotic permeability of water was measured by concentration-dependent self-quenching of vesicular CF with an aliquot of the same batch of AmtB proteoliposomes (19) (Fig. 5B), and also by light scattering as a monitor of vesicle shrinkage and swelling (20). By the fluorescence assay, when vesicles were mixed with sucrose to a final concentration of 250 mM sucrose outside, water efflux was 7.85 ± 0.08 s–1 (n = 6) for liposomes, versus 8.78 ± 0.07 s–1 (n = 6) for AmtB proteoliposomes, indicating no additional water conductivity in AmtB proteoliposomes, versus ∼150 s–1 for a water channel AqpZ in the proteoliposomes. Thus, AmtB does not conduct water.

    Likewise, the structure of AmtB without (NH4)2SO4 showed no ordered water in between the hydrophobic constricted regions of the channel. Thus, the channel removes water of hydration from NH 4+ and a proton, as NH3 passes through the channel. The channel is narrow and the cost of dehydration of H2O itself is not compensated by oxygen in the lumen of the channel, and no line of waters can be supported within the channel. This implication was also challenged by molecular mechanics simulation with NAMD, which after 2 ns of simulation confirmed that no water enters the channel (21).

    The Mechanism: Recruitment of NH4+ and Filtered Passage of NH3

    The mechanism suggested by the atomic resolution structures, difference maps for Am/MA, and NH3 conductivity in proteoliposomes involves a vestibular recruitment of total Am, a site that can bind NH 4+, and a channel for NH3 that lowers its pKa to <6 (Fig. 6). This mechanism of conductance can reconcile many, but not yet, all data that lead to the seemingly inconsistent proposals of how members of the Amt/MEP family work.

    Fig. 6.

    The deduced mechanism of conductance is summarized. The low electron density for NH3 may represent substitutional interchange or relative freedom of NH3 within the hydrophobic channel. NH3 normally undergoes rapid inversion. This may be impeded against the weak hydrogen bond C-H donors of imidazole.

    Soupene et al. have maintained that the Amt/MEP family conducts NH3 bidirectionally (22, 23), with which our findings are consistent. On the basis of experiments, it has also been proposed that the Amt/MEP family variously transports NH4+ (9, 2426), cotransport 26 and exchanges NH3/H+ (26), and exchanges NH4+/H+ in Rh proteins (27, 28). We attempt to address these data here.

    pH-dependent effects. NH 4+ (or CH3NH3+) becomes less basic in the channel as it becomes progressively desolvated, until it reaches the Am2 position. At Am3, it is deprotonated and becomes uncharged (pKa < 6). Entry of the uncharged NH3 or CH3NH2 species into the hydrophobic channel at pHs above the new pKa could eliminate all dependence of rate on pH and hence any dependence on concentration of the uncharged species in the bulk solution. Correspondingly, the solution equilibrium between NH3 and NH 4+ has little to do with the mechanism, because both species can enter the vestibule to become NH3 at the Am2/3 site. If at all, lower pH might contribute to reduce the rate of Am conductance, not by reducing the level of NH3 in bulk solution, but by opposing proton release from NH 4+ in the vestibule. The vestibule may also have some preference for NH 4+ or NH3 that could be reflected as an indirect effect of pH in either direction; a tendency to recruit NH4+ would favor a small increase in rate at lower pH.

    In Coryne bacterium glutamicum, two ammonia channels are present, Amt and AmtB. MA uptake has an apparent Km (CH3NH3+) of 53 ± 11 μM at pH 6.0 that is unchanged at pH 8.5, leading to the conclusion that NH 4+ is the transported species, because NH3 concentration in bulk solution would change 300-fold over this pH range, whereas NH 4+ concentration is little changed (25, 29). Likewise, the Km for transport by Arabidopsis AMT2 is not pH-dependent between pH 5.0 and 7.5 (30). However, recruitment of Am to the vestibule can be completely pH-independent. Therefore, the inference that NH4+ is conducted is incorrect; all Am will become NH3 as it reaches the Am2/3 sites.

    Effects of pH on transport of Am or MA by the human Rh factor RhAG show that the rate of inward transport of (external) 14CH3NH3+ increases as external pH rises from 5.5 to 8.5 (27), tending to suggest that the neutral species may be conducted, because its concentration increases exponentially as pH is raised. However, a lack of effect of changes in membrane potential on conductance of MA, and a small decrease in conductance as internal pH rises, led to the conclusion that transport was electrically net neutral. Westhoff and colleagues reasoned that an increase in the outward-directed proton gradient might drive transport of ammonium ion, the predominant species in solution, and so suggested that RhAG might be a H+/CH3NH3+ antiporter. However, in light of our conclusions, it is of note that the increase in conductance was much more strongly dependent on the external pH than on internal pH, and it can equally be interpreted as transport of uncharged CH3NH2 as implied by our mechanism, in which the deprotonation of CH3NH3+ in the vestibule might well become easier as pH is raised.

    Soupene et al. show that growth in minimal nitrogen-limiting conditions of Am (≤1 mM) is especially deleterious at pH values below 7, where the effect of AmtB/MEP disruptions are profound (22). The bulk solution concentration of NH3 decreases as pH is lowered, one reason that they propose that the Amt/MEP proteins increase the rate of equilibration of the uncharged species NH3 across the membrane, rather than actively transporting the charged species NH 4+ (23). Although our results are consistent with this conclusion, and with their findings that conductance is bidirectional, we contend that the effect of pH is not by effect on solution concentration of NH3. Any pH dependence could be consistent with a possible pH effect on H+ release from NH 4+ in the vestibule. Consistent with this conclusion, the increase in rates of transport of MA versus pH do not increase by a factor of 10 per unit increase in pH as would be expected for a bulk pH effect, but increase by a factor of ∼2 per pH unit (27).

    Competitive inhibition. Ammonia channels conduct Am and MA but no larger secondary or tertiary amines. However, seemingly inconsistently, dimethylamine and ethylamine were found to inhibit the uptake of CH3NH2 by Amt and AmtB from C. glutamicum, whereas trimethyl- and tetramethyl amine did not (25). Although the channel is too small to accommodate these molecules, competition for the smaller cationic forms can take place at the Am1 site in the vestibule. This inhibition of MA conductance by Am in RhAG is pH-independent, surprisingly because their pKas are different (27). However, we now would expect this pH independence because competition is between total MA and Am.

    Transmembrane potential. Transmembrane potential accentuated conductance by the channel has been noted because Km for MA decreases as transmembrane voltage (negative in the direction of conductance) increases, interpreted as indicating that NH 4+ is conducted (26). However, the electric field will concentrate NH 4+ on the NH 4+ rich side of the membrane and in the vestibule, increasing the local concentration there and so explaining this effect. Thus, Km (apparent) appears to be lower because the local concentration of MA/Am is higher in the vestibule than in the bulk solution.

    Westhoff et al. show by expression of human RhAG in oocytes that the rate of conductance of MA remains constant even when membrane potential is modulated from –35 mV to –9 mV (inside the oocyte), concluding that transport is not electrogenic (27). This could be consistent with an electrically neutral NH 4+/H+ antiport mechanism as these authors suggest, but it is also consistent with net transport of uncharged CH3NH2 as our results imply.

    All the observations noted so far are reconciled by the recruitment of NH 4+, reduction of its pKa and conductance of NH3 as the primary mechanism. However, observations of ammonium-dependent currents through ammonia channels are not and demand further examination. A two-electrode voltage clamp was used to vary the transmembrane potential in oocytes transiently expressing the tomato paralog LeAMT1. Inward currents increased with voltage and with external ammonium ions from ∼3 μM Am upward. The currents show a Hill coefficient of 1, implying that there is no cooperative effect of one conducted molecule of Am on another and suggesting a single binding site (26), which we presume could be Am1. This electrogenic behavior suggests that the channel transports ammonium ions (NH 4+, or NH3 plus a H+) rather than uncharged NH3. The measured currents are the same from pH 5.5 to 8.5 (26), consistent with conductance of NH 4+ ions rather than with NH3 and H+, which might have showed some pH dependence. Although we have shown that AmtB conducts predominantly NH3 (assayed in proteoliposomes), we have not shown that it does not conduct any occasional nonstoichiometric NH 4+/H+ ions that could give rise to these currents. A stoichiometric measure of conductance of each species NH3/NH 4+ will be required to establish whether this takes place at some low level. We do not see any hydrogen-bonded pathway that could act as a conductor for H+, nor any change in conformation of AmtB determined in the presence of Am/MA. Given with these observations, one possibility is therefore that an occasional NH 4+ ion can reach Am2, stabilized by ring currents of the rich aromatic environment at the constriction, and pass through the two-histidine region, possibly using the acid/base properties of the imidazole nitrogens to assist in proton transfer. This would require a transient conformational change that could be induced. The structure can clearly guide experimental measures of the ratio of conductances. Alternatively, the currents could be carried by another ion or another Am dependent pathway in the oocytes.

    The Mechanisms of Rh Factors

    In light of the AmtB structure, the sequence of Rh factors can now be mapped onto the three dimensional structure of AmtB to address questions of Rh function. Variabilities in the Rh factors versus Amt/MEPs are seen in lengthened N- and C-terminal domains and in Rh-specific internal changes. The Rh proteins complement mutants in which all three of the yeast MEP proteins are inactivated (mep1Δ, mep2Δ, and mep3Δ), showing that the theRh proteins can provide similar function in the transport of Am to support the growth of yeast (13, 28, 31).

    The rate of MA uptake in RhAG was saturable and well fitted by a Michaelis Menten equation with a Km = 1.6 mM for the ionic species. NH 4+ competed with MA (27). However, Km for RhaG is 300-fold higher than for AtB and the Am1 site differs in lacking some of the π-cation stabilizing rings. Trp148 is Leu or Val in Rh factors; Phe/Tyr103 is Ile in Rh; Phe107 is conserved. This is consistent with low affinity for cations.

    Based on several lines of evidence, Soupene et al. speculate that certain Rh factors, prominent in mammals and found alongside Amt in, for example, green algae, may function physiologically as channels for CO2 (32, 33). The stoichiometry of Rh factor associations in erythrocytes and their impact on the maternal immune response to fetal Rh are not known. Thus, the structure allows docking of models of the various Rh factors against each other to define this heterotrimeric assembly.

    Biological Relevance of This Mechanism

    Conductance of uncharged NH3, versus NH 4+ ion, can solve several biological problems. First, because K+ channels conduct the very similarly sized NH 4+ ion, there is the reverse possibility that an NH 4+ ion channel could “leak” potassium and hence leak membrane potential in eukaryotes. Amt/MEP proteins are not permeable to any other ions (26). Transport of only uncharged NH3 and not NH 4+ assures this selectivity against all ions that would require replacement for their hydration shell while in the narrow portion of the channel. The energetic cost of removing even a single water of hydration from an ion is prohibitive. Second, NH 4+ or any other ion is progressively energetically unstable as it approaches the center of the hydrophobic bilayer, whereas NH3 is much less so because it is electrically neutral. Potassium channels for example, have solved the problem by providing 16 carbonyl oxygens, 8 around each K+ ion position on the way into the channel (34). Each oxygen offers a partial charge of 0.4 electrons to stabilize each K+ ion. The KcsA channel also provides a water-filled cavity in the most energetically costly position at the center of the bilayer (35). The narrow hydrophobic channel of AmtB solves this energetic problem, as it also selects against the ionic form of NH 4+ or any organic molecule larger in cross section than a single NH3. Third, passage of uncharged NH3 versus NH 4+ would not leak proton motive force in conduction. Thus, neither energy nor any counter ion would be needed to accumulate ammonia.

    Note added in proof: In a recent publication (37), it is shown that Am conductance by RhBG in oocytes is electroneutral, in contrast to the currents reported by the same group for another homolog (26), and can explain observations of electric currents in oocytes by an indirect mechanism. If general to the Amt/MEP/Rh family (Fig. 1), this would eliminate the only data that we discuss as potentially inconsistent with the mechanism we deduce.

    Supporting Online Material

    Materials and Methods

    References and Notes

    References and Notes

    View Abstract

    Stay Connected to Science

    Navigate This Article