Inhibitory complex of the transmembrane ammonia channel, AmtB, and the cytosolic regulatory protein, GlnK, at 1.96 Å

Trimer-to-Trimer Interaction.

The complex crystallized as the physiological association of the trimer of AmtB, with the trimer of GlnK, related by crystallographic threefold symmetry (Fig. 1). Each AmtB monomer consists of a central hydrophobic channel surrounded by a right-handed helical bundle of 11 transmembrane α-helices (7, 8). Within the bundle, there are two five-helix motifs that are similar in structure but of opposite polarity with respect to the membrane. The first five helices begin on the outside of the cell membrane and end on the cytoplasmic side, whereas the next five begin on the cytoplasmic side and return to the outside. This structural organization reflects an early gene duplication event that lead to the duplication of these motifs and results in a quasi-twofold symmetry in the membrane plane and perpendicular to the threefold axis of the trimer. This duplication thus generates quasisymmetric vestibules on opposite ends of the channel (Fig. 2A). However, the vestibules have diverged significantly in response to their different functions, which now include binding of GlnK on the cytoplasmic surface. The entry and exit from water-filled vestibules to the central hydrophobic channel are both narrow hydrophobic portals. The external portal must undergo dynamic opening to “select” for passage of the neutral NH₃; however, the inside portal is more open and open enough to accept the arginine side chain R41 of GlnK.

The central region of GlnK consists of a four-stranded β-sheet that forms the interfaces between monomers of the GlnK trimer and supports two α-helices that form the outer surface of the trimer (15). The T- (residues 35–56) and B-loops (residues 81–89) extend from GlnK toward AmtB. On the opposite side, a β-strand, turn and 3₁₀ helix extend to the C terminus to comprise the C-loop. The initial portion of the T-loop, the entire B-loop, and the C-loop from an adjacent monomer form the nucleotide-binding site. The basis of the interaction with AmtB is an intimate association of the T-loop of GlnK with the cytoplasmic vestibule of AmtB. The remainder of the interface between the two trimers is highly hydrated, with many ordered water molecules, as is the vestibule of AmtB itself (Fig. 3Left Inset).

R47 of GlnK Blocks the Channel of AmtB.

The extended T-loops of the GlnK monomers allow R47 of each GlnK to insert deeply into the cytoplasmic vestibule of successive monomers of the AmtB channel. (Fig. 3). The cleft in the cytosolic vestibule is preformed in uncomplexed AmtB. The buried guanidinium cation makes five donor hydrogen bonds inside the channel, completely pairing all five potential hydrogen bond donors with acceptors. The NεH and Nη2H1 of R47 are both hydrogen bond donors to the lone pair of electrons of the hydroxyl of S263 in the cytoplasmic vestibule, which in turn donates its hydrogen to the carbonyl of L259. The Nη2H2 of R47 also donates a hydrogen bond to a water molecule within the channel, termed Am6. The R47 Nη1H1 donates a hydrogen bond to the carbonyl of AmtB C312 and Nη1H2 to a water in the cytosolic vestibule, which is hydrogen-bonded to D313, a counter ion to the guanidinium of R47. The carbonyl of C312 is also a hydrogen bond acceptor from the channel water Am6. Van der Waals contacts between R47 Cγ and L259 of AmtB and between R47 Cβ and V299 of AmtB may also serve to stabilize this conformation of R47. The effective coordination of R47 completely occludes any NH₃ or water from passage through the cytosolic portal.

Although roles for P_II proteins as regulators of ammonia channels have not been established outside of bacteria, the conservation of these key residues of AmtB that coordinate R47 of GlnK in prokaryotes and yeast suggests there may be a common mechanism for blocking the channel in yeast. S263 and D313 are strictly conserved in bacterial and fungal Amts [see supporting information (SI) Fig. 7]. β-branched residues are conserved at the hydrophobic contact positions 259 and 299 in prokaryotic ammonia channels. Although plants have P_II transduction proteins, they have different roles (16). Correspondingly, these AmtB residues are not conserved in plants. This, therefore, suggests that conservation at these sites is a diagnostic for possible P_II regulation of ammonia channels.

T-Loop Interaction with AmtB.

The total interaction surface area between the two protein trimers is 5,389 Ų, of which the T-loop (residues 36–55 of GlnK) accounts for ≈77%. Of the total, 47% consists of polar interactions involving 27 hydrogen bonds and 18 salt bridges. Side chains account for 87% of the interaction. The total, polar, and nonpolar surface areas per monomer are within the normal distribution for physiological heterooligomeric complexes. Because there are three such interfaces, the buried surface area is three times larger and so accounts for higher avidity (17). The T-loop sequence of GlnK and GlnB from E. coli is highly homologous throughout those residues involved in the interaction with AmtB, which suggests common types of interactions.

Regulation by Uridylylation.

Uridylylation of Y51 on GlnK abrogates binding to AmtB. We mutated Y51 to phenylalanine (Y51F) to ensure homogeneity. The aromatic ring of F51 was stacked on the aliphatic chain of K194 and has additional hydrophobic contacts with A192. Modeling the hydroxyl of Y51 onto F51 shows that in the native complex, it would be hydrogen-bonded to the backbone NH of AmtB F193. Therefore, uridylylation would sterically block the T-loop region from interacting with AmtB (Fig. 4).

Although there is currently no evidence of posttranslational modification of GlnK Y46 in E. coli, this tyrosine is also in the T-loop, which is highly mobile in uncomplexed trimeric GlnK and would be similarly accessible for modification. In complex with AmtB, the side chain of Y46 is fixed in a pocket, with its OH hydrogen-bonded to AmtB, suggesting that any posttranslational modification of it would also inhibit AmtB binding. In view of the flexibility of the uncomplexed T-loop, the urydyl transferase GlnD is presumably quite sequence-specific, targeting only Y51.

In cyanobacteria, S49, also on the T-loop, is substituted for A49 of E. coli GlnK and becomes the site for posttranslational regulation, in this case by phosphorylation (18, 19). A49 is positioned in a cleft on AmtB, suggesting that this posttranslational modification in cyanobacteria would also sterically inhibit binding of the T-loop to AmtB.

The Role of Nucleotide Binding to GlnK.

ATP enhances the binding of GlnK to AmtB (12, 20). However, our complex incubated in 2 mM ATP shows it is ADP rather than ATP-bound. ATP hydrolysis may have been the result of a trace ATPase contaminant. However, the proximity of the ADP terminal phosphate to side chains of three arginines, a lysine, and a highly coordinated buried water in GlnK suggests this site might itself serve to catalyze ATP hydrolysis.

The nucleotide-binding site lies in the interface between two monomers of GlnK, as is seen in the ATP-bound structure of GlnK alone (ref. 15; see also Fig. 5). However, unlike the ATP-bound structures, the β-phosphate end of the ADP is buried due to interactions with the main-chain N-Hs of residues 38 and 39 at the base of the T-loop. These residues are part of a tight turn that may be induced by ADP to position the T-loop for the interaction with AmtB. The β-phosphate of ADP is very close to the position of the γ-phosphate of ATP in the ATP-GlnK structure in the absence of AmtB. It is yet undefined whether binding of GlnK to AmtB is influenced by hydrolysis of ATP or by the ADP concentration. The T-loop and ADP-binding regions of GlnK are highly conserved across known P_II proteins, suggesting a common ATP/ADP-dependent mechanism (see SI Fig. 8).

Regulation by 2-KG.

The AmtB/GlnK complex reveals two pockets in contact with the base of the T-loop in GlnK and a third in the interface between the T-loop and the vestibule, that present as plausible 2-KG-binding sites (Fig. 6). To test these sites, we ran flexible docking using DOCK6 (21). The two sites on GlnK (2KG1 and 2KG2) achieved docking poses that had complementary shape and charge. Binding of 2-KG at either of these locations could alter conformation at the base of the T-loop and so affect the ability of GlnK to bind to AmtB. The 2KG2 site, previously proposed to be a possible 2-KG site based on the GlnK structure alone (15), is in close proximity to the ADP site. Thus a counter ion such as Mg²⁺ may be necessary to augment binding of 2-KG to ADP at this location. This would be consistent with the inhibition of AmtB/GlnK complexation by 2-KG in the presence of ATP and Mg²⁺ and weaker inhibition if Mg²⁺ is absent (12). The interstitial site (2KG3) had compatible shape but suboptimal electrostatic potential. The nature of the site suggests that another small-molecule effector could potentially modulate AmtB/GlnK association by binding at this allosteric site. For example, glutamate has been shown to affect GlnB interaction with uridylyltransferase (20). It is possible that 2KG3 could act as an allosteric binding site for another metabolite, such as glutamate.

Bound-State Conformational Changes.

The loops and the C-terminal residues on the cytosolic face of AmtB are rearranged and ordered with respect to previous AmtB structures (refs. 7 and 8; Fig. 2B, region i). Portions of the C-terminal region of AmtB interact with the T-loop of GlnK and may account for increased order in the C terminus as compared with the prior AmtB structures [Protein Data Bank (PDB) identifiers 1U7G and 1XQF]. The C-terminal conformation is now helical and similar to the C terminus conformation of Amt1 from Archaeoglobus fulgidus (ref. 22; Fig. 2B, region ii), suggesting that the primary change in our structure is due to the different interactions with the C terminus. This also suggests that the P_II regulator, GlnB-1 in A. fulgidus, probably interacts with Amt1 in a similar manner. The conformations of loops from residues 301–316 and 181–195 in our structure of AmtB were already different from in the other Amt structures, suggesting they are somewhat flexible and easily reordered in the complex.

Comparison of GlnK in this complex to previous GlnK and other P_II proteins revealed very high structural similarity throughout, except in the T-loop region (residues 36–55). The T-loop has various different conformations in previous structures of GlnK alone (refs. 15 and 23–26; see also Fig. 2B, region iii). In the three known P_II protein structures with an occupied nucleotide-binding site (PDB ID codes 2GNK, 1V3S, and 1V9O; refs. 15 and 25), this T-loop region is disordered. Thus, before complexation, the T-loop is loosely structured, and it becomes ordered in the physiological complex, as is seen here.

Periplasmic Ammonium Ion Recruitment.

The periplasmic vestibule of AmtB contains many carbonyl oxygens and is highly hydrated, with ≈30 ordered water sites. Three aromatic side chains form a favorable π-cation interaction with methyl ammonium ions and may aid in the recruitment of ammonium ions right at the entrance to the hydrophobic channel (7, 8). The AmtB/GlnK complex bound two quaternary ammonium-containing Tris molecules [2-amino-2-(hydroxymethyl) propane-1,3 diol] in the vestibule, of which one occupies the π-cation ammonium ion-stabilizing site. The two quaternary ammonium cations are coordinated by polar and van der Waals interactions in pockets of the vestibule. Otherwise, the structure of the periplasmic face is virtually identical to the prior AmtB structure. The sequence of AmtB used here has the sequence of the K12 strain throughout.

Two hydrophobic residues form the narrow portal that selects a nonpolar NH₃ substrate, F215 and F107 (ref. 7; see also Fig. 3 Right Inset). Within the channel, the H168 and H318 appear to be the critical, but weak C-H hydrogen bond donors facilitate the passage of the ammonia molecules through an otherwise hydrophobic central channel. The hydrogen bonding between N-H to N: of the two coplanar histidines is unambiguous. The hydroxyl of T273 donates a hydrogen bond to the carbonyl of L269 and accepts a hydrogen bond from the protonated Nε of H168. The unprotonated Nδ1 of H168 accepts a hydrogen bond from the protonated Nδ1 of H318. In our structure of AmtB (7), three ammonia molecules were defined based on electron density and hydrogen bonding (Am2–4). Two of these sites were confirmed to be on the hydrophobic ammonia pathway by Xe substitution in the structure of Amt1 (22). In the structure of the AmtB/GlnK complex, these ammonia molecules are not present, but two waters seen in the native Amt1 are conserved. The positions of these molecules are likely to be on the ammonia conduction pathway and are more probably water in our complex. They are labeled as Am5 and Am6 to reflect this.

The cytoplasmic portal is not comprised of aromatic side chains and is more open than on the periplasmic side. The pore conformation in the complex with GlnK is very close to that seen in the uncomplexed Amt1 structure (22), suggesting that the principal adaptation on complexation is focused on the organization of the T-loop. The mechanism of inhibition of AmtB ammonia conduction is the blockade of the cytosolic pore by R47 of GlnK and not consistent with the proposal of AmtB adopting a “closed” conformation on GlnK binding (8). The pockets on GlnK and in the interface suggest the prospect of other allosteric factors that may serve to further tune the critical balance of nitrogen supply and demand.