In conditions of nitrogen limitation, ammonium uptake is

The crystal structure of the Escherichia coli AmtB–GlnK complex reveals how GlnK regulates the ammonia channel Matthew J. Conroy*, Anne Durand†, Domenico Lupo‡, Xiao-Dan Li‡, Per A. Bullough*, Fritz K. Winkler‡§, and Mike Merrick† *Department of Molecular Biology and Biotechnology, University of Sheffield, Firth Court, Western Bank, Sheffield S10 2TN, United Kingdom; †Department of Molecular Microbiology, John Innes Centre, Norwich NR4 7UH, United Kingdom; and ‡Biomolecular Research, Paul Scherrer Institut, CH-5232 Villigen, Switzerland

Amt proteins are ubiquitous channels for the conduction of ammonia in archaea, eubacteria, fungi, and plants. In Escherichia coli, previous studies have indicated that binding of the PII signal transduction protein GlnK to the ammonia channel AmtB regulates the channel thereby controlling ammonium influx in response to the intracellular nitrogen status. Here, we describe the crystal structure of the complex between AmtB and GlnK at a resolution of 2.5 Å. This structure of PII in a complex with one of its targets reveals physiologically relevant conformations of both AmtB and GlnK. GlnK interacts with AmtB almost exclusively via a long surface loop containing Y51 (T-loop), the tip of which inserts deeply into the cytoplasmic pore exit, blocking ammonia conduction. Y51 of GlnK is also buried in the pore exit, explaining why uridylylation of this residue prevents complex formation. regulation 兩 x-ray structure 兩 PII protein

I

n conditions of nitrogen limitation, ammonium uptake is facilitated by a family of integral membrane proteins known as the ammonium transporter (Amt) family (1). They are almost ubiquitous among archaea, eubacteria, fungi, and plants, whereas in animals they are represented by the closely related Rhesus family. The Escherichia coli AmtB protein has become the paradigm for studies on the biology of these proteins. AmtB is a stable homotrimer in the cytoplasmic membrane and retains this structure when purified and reconstituted in two-dimensional crystals (2, 3). Mature E. coli AmtB has 11 transmembrane helices with an Nout and Cin topology (4–7) that appears to characterize all members of the Amt family. The x-ray crystal structures of E. coli AmtB (5, 6) and the related Archaeoglobus fulgidus Amt-1 (8) are very similar. Each subunit of the trimer has a pseudo-twofold symmetry relating helices M1 to M5 and M6 to M10, with M11 across the lipid-accessible face of each monomer. In A. fulgidus, the cytosolic C-terminal region (CTR) of Amt-1 (8) has a defined structure, but in the E. coli structures (5, 6) this region is disordered. The Amt structures suggest that substrate conductance occurs through a narrow, mainly hydrophobic pore located at the center of each monomer and containing two highly conserved histidines (H168 and H318) each of which is essential for conductance of the ammonium analogue methylammonium (9). A potential high-affinity ammonium binding site was identified at the periplasmic entrance to the pore, and these structural features, together with complementary data, suggest that ammonium is the substrate recognized but that ammonia is the translocated species (5, 6, 10, 11). [In this article, we use the term ammonium to refer to both the protonated (NH4⫹) and unprotonated (NH3) forms, and we use the term ammonia to refer specifically to NH3.] Therefore, as previously proposed (12), Amt proteins have been described as ammonia channels (5, 6). www.pnas.org兾cgi兾doi兾10.1073兾pnas.0610348104

In almost all eubacteria and archaea, the amtB gene is invariably linked to glnK, suggesting a functional relationship for AmtB and GlnK (4). GlnK belongs to the PII family of cytosolic signal transduction proteins that act as sensors of cellular nitrogen status and regulate the activities of many different proteins, including enzymes and transcription factors, by protein–protein interaction (13–15). We therefore have suggested that AmtB activity might be regulated by GlnK (16). Indeed, in vivo, GlnK binds rapidly and reversibly to AmtB in response to an extracellular ammonium shock (ⱖ50 ␮M extracellular ammonium) (17), and this interaction has also been shown in a number of other eubacterial species (16, 18–20). Methylammonium uptake by AmtB is also inhibited by a GlnK variant that forms an irreversible complex, supporting the concept that GlnK regulates ammonium flux through AmtB (17). Like AmtB, GlnK is also a homotrimer; its three subunits form a compact barrel in which each monomer folds into a fourstranded ␤-sheet packed against two helices (21). Three notable loops, designated the B-, C-, and T-loops, are present in the structure. A lateral cleft between each subunit is formed between the T- and B-loops of one subunit and the C-loop of another. E. coli expresses two PII proteins, GlnK and GlnB, whose activities are regulated covalently and noncovalently. A conserved residue Y51 within the T-loop is covalently modified by uridylylation in cells that are subjected to nitrogen starvation and this modification is reversed in nitrogen sufficiency (22–24). Only deuridylylated GlnK binds to AmtB (16, 17, 25). PII proteins synergistically bind ATP and 2-oxoglutarate (2-OG), a small-molecule signal of cellular nitrogen status (14, 15, 22) and a metabolite of the Krebs cycle. The crystal structures of E. coli GlnB–ATP and GlnK–ATP co-complexes revealed ATP bound in the lateral clefts between the subunits (21, 26). Very little is known about the mechanism whereby PII proteins interact with their targets, although a number of studies have implicated the T-loop in such interactions (22, 27–29). In E. coli, the failure of uridylylated GlnK to interact with AmtB also implicates the T-loop (16, 17). Conversely the CTR of AmtB is important for GlnK binding (16, 30). In vitro biochemical analysis revealed a direct and stoichiometric interaction between the two Author contributions: M.J.C. and A.D. contributed equally to this work; X.-D.L., P.A.B., F.K.W., and M.M. designed research; M.J.C., A.D., and D.L. performed research; M.J.C., D.L., X.-D.L., and F.K.W. analyzed data; and M.J.C., A.D., P.A.B., F.K.W., and M.M. wrote the paper. The authors declare no conflict of interest. Abbreviations: 2-OG, 2-oxoglutarate; CTR, C-terminal region; PDB, Protein Data Bank. Data deposition: The atomic coordinates and structure factors have been deposited in the Protein Data Bank, www.pdb.org (PDB ID code 2NUU). §To

whom correspondence should be addressed. E-mail: [email protected]

This article contains supporting information online at www.pnas.org/cgi/content/full/ 0610348104/DC1. © 2007 by The National Academy of Sciences of the USA

PNAS 兩 January 23, 2007 兩 vol. 104 兩 no. 4 兩 1213–1218

BIOPHYSICS

Communicated by Stephen C. Harrison, Harvard Medical School, Boston, MA, November 28, 2006 (received for review November 10, 2006)

Table 1. Data collection and refinement statistics Space group Cell dimensions (a, b, and c), Å Resolution, Å*, and wavelength, Å Completeness, %* Rsym, %* I/␴* Reflections, total/unique Rcryst, %* Rfree, %* rmsd bond lengths, Å rmsd bond angles, ° No. of atoms/average B-factors, Å2 Protein PDP Water

P212121 99.69, 107.87, 280.16 50–2.5 (2.6–2.5) and 0.97972 99.8 (99.9) 9.9 (53.6) 13.3 (3.0) 438,641/105,009 17.1 (26.0) 24.9 (32.0) 0.009 1.425 23,314/29.9 162/33.5 550/31.5

Rsym ⫽ 兺hkl 兺i兩I(hkl;i) ⫺ 具I(hkl)典兩/兺hkl 兺i I(hkl;i); Rcryst, Rfree ⫽ 兺hkl兩Fobs(hkl) ⫺ Fcalc(hkl)兩/兺hklFobs(hkl), with summation over working set and test set reflections, respectively. *Values in parentheses are for the outermost-resolution shell.

Fig. 1. Overview of the AmtB–GlnK complex. The surface of AmtB is shown, and GlnK is shown in cartoon representation. Each subunit of AmtB and GlnK is colored independently. (A) View from the cytoplasm along the threefold axis perpendicular to the membrane. (B) View in the membrane plane. All figures were made using PyMOL (40) .

proteins (25). Furthermore, in vitro dissociation of GlnK from AmtB does not require uridylylation of GlnK, and association/ dissociation of the complex is sensitive to 2-OG in the presence of ATP (25). To date, there have been no structural data for any AmtB– GlnK complex or for any PII protein complexed to one of its targets. Here, we describe the crystal structure of the E. coli AmtB–GlnK complex, which provides direct visualization of the effect of GlnK in blocking ammonia conductance by AmtB. Results Overall Structure of the Complex. In vitro reconstitution of the AmtB–GlnK complex is only achievable in specific conditions of effector composition and concentration (25). Consequently, we chose to purify the complex directly from cells grown in physiologically relevant conditions as described in Methods. Crystallization and structure determination are described in Methods, and crystallographic statistical data are given in Table 1. Two copies of an AmtB–GlnK complex make up the asymmetric unit, stacked such that the periplasmic faces of AmtB interact. All six copies of each protein in the asymmetric unit have very similar structures, with the few small differences being attributable to crystal contacts. AmtB and GlnK form a threefold symmetric complex in which the GlnK trimer binds to the cytoplasmic face of AmtB with a stoichiometry of AmtB3:GlnK3 (Fig. 1 A and B), confirming the stoichiometry previously determined analytically (25). The overall shape of the complex is that of a truncated cone with a height of 85 Å and a width of 80 Å at the periplasmic face and 50 Å at the base (Fig. 1 A and B). Each GlnK subunit inserts an ordered T-loop into the cytoplasmic pore exit of an adjacent AmtB subunit (Fig. 1 A and B). GlnK does not pack tightly against the cytoplasmic surface of AmtB but displays a rather open interface. This is the first structure for which we can be confident that the observed conformations of the cytoplasmic loops of AmtB 1214 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0610348104

and the T-loops of GlnK, both prone to disorder, are physiologically relevant because they occur in a naturally formed complex. We discuss below the features that we observe in the structures of the individual proteins and then consider how these features facilitate complex formation. Structure of AmtB in the Complex. The structure of AmtB in this

complex differs significantly from the previously published structures (5, 6) on its cytoplasmic face, whereas the periplasmic face and transmembrane part are essentially unchanged. The side chains of the residues lining the periplasmic vestibule and those forming the ammonia conducting channel, including H168 and H318, are coincident between all structures. When compared with the previous structure with ordered cytoplasmic loops [Protein Data Bank (PDB) entry 1U7G (5)], three of them, M3–M4, M5–M6, and M9–M10, exhibit highly altered conformations (Fig. 2A). A remarkable salt-bridge interaction network formed by the highly conserved residues R37, E121, R185, and D309 and not present in PDB entry 1U7G links four of the five loops. For the AmtB structures with disordered loops (6), the crystals were grown at pH 4.5 and at such a low pH this stabilizing network would be strongly perturbed. Different conformations at the cytosolic end of M10 (residues 310–316) have been previously noted in the free AmtB structures and speculatively linked to different states of the cytoplasmic pore exit (6). One conformation, observed in two E. coli AmtB structures (PDB entries 1U7G and 1XQF), has a narrow pore constriction on the cytoplasmic side of H318, the other conformation, observed in A. fulgidus Amt-1 (PDB entry 2B2F), E. coli AmtB (PDB entry 1XQE) and also in the AmtB–GlnK complex, has a ␲-type N-terminal turn that generates a wider and more polar cytoplasmic opening (see figure 1b of ref. 11). In this conformation, the aromatic ring of F315 at the cytosolic end of M10 forms a small hydrophobic cluster with residues at the cytosolic end of M9, influencing the orientation of the M9–M10 loop, which is fully ordered. Importantly, and in contrast to previous structures of E. coli AmtB, the CTR (residues 383–406) is ordered (Fig. 2 A) and adopts a structure very similar to that predicted by homology modeling using the CTR of A. fulgidus Amt-1 (PDB entry 2B2F) as a template (30). It contains two short helices separated by a tight turn centered on G393. There are a large number of contacts and specific interactions of the cytoplasmic loops with each other and with the CTR that appear to cooperatively stabilize the fully ordered state of the cytoplasmic face of AmtB. Conroy et al.

BIOPHYSICS

Fig. 2. Features of cytoplasmic face of AmtB in the complex. (A) Comparison of cytoplasmic loop conformations observed in the complex with the corresponding loops in PDB entry 1U7G (5) [M3–M4 (red), M5–M6 (green), and M9 –M10 (magenta) in bright and pale colors, respectively]. For clarity, the end points of the loops shown as C␣ traces have been centered on the helical cylinders. CTR is shown in blue (not ordered in PDB entry 1U7G). (B) Interaction of one CTR with the cytoplasmic loops of two adjacent AmtB subunits, colored pink and blue with the CTR of the pink subunit shown in yellow. Residues involved in hydrogen bonding interactions are shown as sticks. The arrow points at the position of G393.

The CTR forms extensive interactions with the cytoplasmic surface of the cognate AmtB chain; namely, the M1–M2, M3– M4, and M5–M6 loops (Fig. 2B). Furthermore, it contacts the neighboring AmtB molecule in the trimer; namely, helix M1 and the M5–M6 and M7–M8 loops. In this way, the last six residues of the CTR form part of the wall of the cytoplasmic pore exit in the neighboring subunit (Fig. 5A). This interaction mode is tightly linked to the cytoplasmic loop conformations, and furthermore helix M6 is one turn shorter than in PDB entry 1U7G (5) to accommodate the C-terminal residues from the neighboring subunit. We consider it possible that the ordered state of the cytoplasmic AmtB face can also exist in the absence of GlnK, although binding of the latter will certainly stabilize it. Structure of GlnK and Its Nucleotide Binding Site in the Complex. The overall fold of GlnK is very similar to that of the uncomplexed protein (21). However, the T-loop exhibits a conformation that has not been seen previously in PII proteins (21, 31–33). It forms a short two-stranded antiparallel ␤-sheet (residues E44–Y46 and A49–Y51), the strands of which are separated by a ␤-turn. The T-loop extends 28 Å from the core of the protein (Fig. 3A), a Conroy et al.

Fig. 3. Features of GlnK in the complex. (A) The GlnK trimer showing the T-loops extending from the main body of the protein. R47 and Y51 are shown as sticks. One molecule of ADP in the nearest lateral cleft (between the green and red subunits) is shown in stick representation. Residues at the base of the T-loop and of the B-loop from the red subunit together with residues of the C-loop from the green subunit contribute to this binding site. (B) Comparison of T-loop conformations observed in the complex (green) with the corresponding loops in E. coli GlnK (PDB entry 1GNK) (21) (pink) and E. coli GlnB (PDB entry 2PII) (26) (yellow). R47 and Y51 are indicated. Superimpositions were carried out by using backbone atoms of residues 1–35 and 58 –103.

completely different conformation from that previously reported for E. coli PII proteins (Fig. 3B). The tip of the T-loop is formed by R47 which, together with the adjacent G48, is totally conserved. Within the lateral cleft between adjacent GlnK subunits, we see clear electron density corresponding to the base, sugar, and two phosphate groups of a bound nucleotide (Fig. 4). Thus, we observe an ADP-bound state of E. coli PII, with a stoichiometry of 3ADP:GlnK3. Previous studies have shown ATP can bind to this site in both E. coli GlnK (21) and GlnB (26). However, given the close contacts with neighboring amino acid main-chain and side-chain atoms all around the diphosphate moiety as verified through omit maps, we rule out that we are observing ATP with a disordered ␥-phosphate. The ADP molecule is almost completely buried in the lateral cleft, and hydrogen bonding interactions with the base and sugar moieties of ADP (Fig. 4) are the same as in GlnK–ATP and GlnB–ATP. Superposition of our structure on that of GlnK–ATP (21) shows significant differences in the conformation and relative position of the two sugar-phosphate moieties. In contrast, superposition with GlnB complexed with ATP in two different crystal forms (26) reveals a strikingly close fit for the nucleotides apart from the extra PNAS 兩 January 23, 2007 兩 vol. 104 兩 no. 4 兩 1215

Fig. 4. Stereo representation of the ADP binding site of GlnK. The difference electron density, contoured at 3␴, was obtained after 20 cycles of restrained refinement with the ADP molecule omitted. Residues of the two GlnK subunits forming the binding site and ADP are shown in gray, cyan, and magenta ball-and-stick models, respectively. Hydrogen bonding interactions including one water molecule (pink sphere) are shown with dashed lines, and selected residues are labeled.

␥-phosphates. In one case, the ATP ␥-phosphate is oriented toward the guanidinium moiety of R103, and in the other it is oriented toward the guanidinium moiety of R101; these differences being accommodated by local structural adjustments and in one case by the additional displacement of the four C-terminal GlnB residues. Thus, the mobility of the R101 and R103 side chains and the very C-terminal chain segment (D108–L112) of one subunit, together with that of the B-loop of the other subunit at the interface, appear to accommodate various nucleotide binding modes of these PII proteins. An important difference between the ADP complex observed here and the E. coli GlnK–ATP (21) and E. coli GlnB–ATP (26) complexes is that in the latter cases the T-loop is disordered beyond residue G37. However, in our structure the T-loop is fully ordered, and we observe the two hydrogen bonds of the main-chain NH groups of residues R38 and Q39 with ␣-phosphate oxygens (Fig. 4). These bonds appear important to fix a 90° kink in the GlnK chain at R38 (see arrow Fig. 5A). The R38 side chain reaches over the ␣-phosphate and makes a salt-bridge interaction with the Cterminal carboxylate of the adjacent subunit. However, this contact appears somewhat labile because the corresponding density is weak in some of the six crystallographically independent copies. The Interaction of GlnK with AmtB. Many of the features of the E. coli AmtB and GlnK structures described above are likely to contribute directly or indirectly to the stabilization of the complex. The interface between the cytoplasmic face of AmtB and GlnK is a remarkably open one with a relatively small surface area buried between the two proteins. The docking interaction in the complex is formed almost exclusively by the T-loops of GlnK acting as extended pins plugging into the cytoplasmic pore exits of AmtB (Fig. 1B), delineated by cytosolic loops M5–M6, M7–M8, and M9–M10 of one subunit and the CTR of the adjacent subunit (Fig. 5A). The corresponding buried surface areas are 1,079 Å2 with the cytosolic loops and 439 Å2 with the CTR. The importance of the CTR for shaping the interaction surface with the T-loop and thus GlnK is strongly indicated by our mutational analysis of the CTR showing that very minor changes abolish the interaction with GlnK (30). For example, mutation of R384 or Y404 in the CTR, which make hydrogen bonds to residues in loop M5–M6 (Fig. 2B), prevent GlnK binding (30). There is one other, albeit rather weak, contact involving F11 of GlnK and R391 of AmtB and some neighboring side chains with partly ill defined density. Overall, GlnK is not bound rigidly to the face of AmtB, as also indicated by its elevated B-factors. 1216 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0610348104

Fig. 5. Interactions of the GlnK T-loop with AmtB and consequent pore blockage. (A) Two AmtB subunits are shown in surface representation, and two GlnK molecules with ADP at their interface are in cartoon representation. The green T-loop (arrow marks kink at R38) inserts into the vestibule of the pink AmtB subunit. The surfaces of its cytoplasmic loops interacting with the T-loop are emphasized in red (M7–M8) and salmon (M5–M6), respectively, and that of the C-terminal residues of the neighboring subunit are emphasized in dark blue. (B) Hydrogen bonding interactions of the T-loop (magenta) inserted into an AmtB subunit (blue) with part of the cytoplasmic vestibule being formed by the neighboring CTR (green). Interacting residues are shown as sticks. Helices are labeled in the gray circles. H318 in the channel is included for orientation. (C) Insertion of R47 into the cytoplasmic end of the ammonia translocating channel. The channel is shown in cross-section with R47 of GlnK shown in spacefill. The rest of the T-loop is shown as a cartoon (red). H168 and H318 in the channel are shown for orientation.

Whereas previous studies have implicated the T-loop in interactions of PII proteins with their targets (27–29), here we describe such an interaction in molecular detail. At the tip of the T-loop, the charged guanidinium moiety of R47 forms hydrogen bonds to the carbonyl of AmtB-C312 and hydroxyl of AmtB-S263 as well as a salt-bridge interaction with AmtB-D313 (Fig. 5B). The guanidinium group of R47 dramatically constricts the exit to the ammonia conducting pore, giving it a minimum radius of 0.6 Å and effectively blocking any possible movement of ammonia through the channel (Fig. 5C). Hence, we conclude that when GlnK is sequestered to AmtB in response to an increase in extracellular ammonium levels (16, 17) it sterically blocks the pore and prevents ammonia flux into the cell. In the complex, the side chain of Y51 is packed against the AmtB M5–M6 loop that forms one wall of the cytoplasmic pore exit. The OH group of Y51 is within hydrogen bonding distance (2.8 Å) of F193 in the M5–M6 loop, and the hydrophobic ring packs against the aliphatic portion of the side chain of K194 (Fig. 5B). Discussion This work provides structural insight into the role of GlnK binding to AmtB while also describing the structure of a PII protein bound to one of its targets. The extension of the GlnK T-loops deep into the cytoplasmic exit pores of each of the AmtB subunits clearly indicates that the mode of action of GlnK is to block sterically the conductance of substrate through the AmtB pore. In a previous publication, a complex between A. fulgidus Amt-1 and GlnB-1 was modeled but necessarily could not show Conroy et al.

Conroy et al.

reflect the intracellular levels of ADP induced by the ammonium shock before purification of the complex. However, in vitro, ATP can promote complex formation (25), and we cannot exclude the possibility that the complex actually binds ATP that has been hydrolyzed to ADP during purification. The conformation of the GlnK T-loop is clearly critical for complex formation, and this is likely to be significantly influenced by binding of nucleotide. It is notable that in our structure, the ␣-phosphate of ADP makes contact with the main chain of R38 and Q39 and fixes a marked kink in the GlnK chain at the base of the T-loop (Fig. 5A). However, because PII proteins can clearly accommodate a variety of nucleotide binding modes, it is not possible to predict the likely conformation if ATP replaced ADP in the present structure. We have also shown that in vitro association/dissociation of the complex is sensitive to the concentration of 2-OG in the presence of Mg-ATP (25). It has also been suggested that 2-OG could bind in the lateral cleft in the vicinity of the phosphate group of the nucleotide (37). This could be accommodated either by the presence of Mg2⫹ serving to balance their respective negative charges (21, 25) or by displacement of the C-terminal GlnK residues and substitution of the C-terminal carboxylate by a carboxylate of 2-OG. These considerations suggest that in vivo complex formation might not exclusively depend on the uridylylation state of GlnK but may be also influenced by the intracellular pools of ATP/ADP and 2-OG. This structure of an AmtB–GlnK complex offers significant insights into the interaction mode, the likely consequences of the interaction for ammonium conduction, and the effect of GlnK uridylylation on complex formation. The structure emphasizes that the primary function of GlnK is almost certainly regulation of ammonium flux into the cell, and the almost ubiquitous linkage of the glnK and amtB genes suggests that regulation of the ammonia channel may be the ancestral role of PII proteins (38). However, in some systems, complex formation might also serve other functions such as sequestration of cytoplasmic proteins involved in regulation of nitrogen metabolism as suggested by recent studies in Azospirillum brasilense (19) and Bacillus subtilis (39). Methods Complex Preparation and Crystallization. To optimize the AmtB– GlnK interaction, AmtB was His-tagged on the periplasmic face, leaving GlnK and the interacting cytosolic face of AmtB in their native forms. Complex formation was induced by ammonium shock, and purification was carried out by using lauryldimethylamine-N-oxide as solubilizing detergent throughout the purification as described (25). The N-terminal sequence of the recombinant AmtB is APAVAHHHHHHA(3)VADK corresponding to the wild-type sequence from residue 3. The pure complex was concentrated to 40 mg䡠ml⫺1 in a buffer containing 50 mM Tris䡠HCl (pH 8.0), 100 mM NaCl, 10% (vol/vol) glycerol, and 0.05% (wt/vol) lauryldimethylamine-N-oxide. Crystals were grown in sitting drops by mixing equal volumes of protein and precipitant solution [65% (vol/vol) 2-methyl-2,4-pentanediol in 100 mM Tris䡠HCl at pH 8.0] at room temperature. The complex crystallized in space group P212121 with cell constants a ⫽ 99.7 Å, b ⫽ 107.9 Å, and c ⫽ 280.2 Å. Data Collection and Structure Determination. For details of data

collection and structure determination, see supporting information (SI) Methods. We thank J. Thornton for assistance with fermentation and membrane preparation, E. Kunji and J. Rafferty for the very considerable time and advice given in helping us with crystal screening in the early stages of the project, and R. Dixon and A. Javelle for helpful comments on the manuscript. Crystallographic data were collected at the Swiss Light Source (Paul Scherrer Institut, Villigen, Switzerland) with the excellent PNAS 兩 January 23, 2007 兩 vol. 104 兩 no. 4 兩 1217

BIOPHYSICS

any of the molecular details seen here for the E. coli AmtB–GlnK binding (8). Furthermore, in the absence of any structure for A. fulgidus GlnB-1, the authors used E. coli GlnB as a template to predict the structure of GlnB-1, and it is now apparent that the T-loop in the E. coli GlnB structure has a totally different conformation from that now determined in our structure (Fig. 3B). The occurrence of the AmtB–GlnK pair is highly conserved in both eubacteria and archaea, and the interaction of these proteins has already been demonstrated in a number of cases (16, 18–20). Hence, we might expect that the mode of binding is also conserved. Comparison of predicted AmtB and GlnK sequences indicates that R47 at the tip of the GlnK T-loop is completely conserved, as is the adjacent G48, which potentially mediates the turn. Of the residues in AmtB that make contact with R47 of GlnK, S263 and D313 are highly conserved, whereas C312 is not, presumably because the interaction is with the backbone carbonyl rather than with the side chain. The structure of the GlnK T-loop observed in the complex is distinct from those seen in previous T-loop structures, none of which was crystallized in a functional context. In each of the four previous PII structures in which the T-loop was ordered, the ability to resolve the loop is a direct consequence of stabilization of the structure by crystal-packing contacts (21, 31–33). Although two of the previous structures exhibit an extended conformation, in each of those cases this is a consequence of extension of ␤-strands 2 and 3 (32, 33). Neither exhibits the short ␤-strands observed in our structure, which are composed of distinctly different residues. We suggest that the observed T-loop structure is likely to be a consequence of binding to AmtB, but it is also conceivable that its conformation would be inf luenced by nucleotide binding. Although Y51 is not completely protected from solvent in the complex, it is clear that there is not sufficient room within the pore exit to accommodate a bound uridine monophosphate group on this residue, and thus the uridylylated form would be sterically prevented from binding to AmtB. It would also seem unlikely that uridylyltransferase is able to access Y51 to uridylylate it while the complex is intact. Hence, we conclude that uridylylation of Y51 is not the driving force for dissociation of the complex, and this is consistent with our in vitro studies in which dissociation can occur in the absence of uridylylation (25). Another feature of the complex is the presence of ADP bound to each subunit of GlnK. Bound ADP has been reported on two previous occasions: in Thermus thermophilus GlnK where ADP was added during crystallization (33), and in Thermotoga maritima where ADP was found in the crystal structure of a PII protein purified from cells where the growth conditions were not defined (34), the latter case being the only one to date where nucleotide was found to copurify with PII. Consequently for E. coli GlnB and GlnK, the only comparisons that can be made are with bound ATP that was added during crystallization (21, 26). An important difference between the ADP complex observed here and those GlnK–ATP and GlnB–ATP complexes is that in the latter cases, the T-loops are disordered, whereas in our structure they are fully ordered (21, 26). The role of ATP or ADP binding to PII proteins is presently unclear, but it has been proposed that in cyanobacteria and Rhodospirillum rubrum, these ligands could play a role in mediating signaling of the cellular energy status by PII proteins (15, 35). In E. coli, the relationship between ATP levels and cellular nitrogen status has not been extensively studied, and the physiological role, if any, of ATP/ADP binding to PII proteins is unclear. However, ammonium shock of nitrogen-limited cells has been shown to cause a 10-fold drop in intracellular ATP levels within 15 sec (36), which might be expected to cause a concomitant rapid decrease in the ATP/ADP ratio. Hence, our observation that ADP is bound to GlnK in the complex could

help of T. Tomizaki. We thank D. Ollis for sending the coordinates of GlnB–ATP complexes. This work was supported by a grant from the Biotechnology and Biological Sciences Research Council (to P.A.B. and

M.M.). F.K.W. acknowledges support from the Swiss National Science Foundation within the framework of the National Center of Competence in Research in Structural Biology.

1. von Wire´n N, Merrick M (2004) Trends Curr Genet 9:95–120. 2. Conroy MJ, Jamieson SJ, Blakey D, Kaufmann T, Engel A, Fotiadis D, Merrick M, Bullough PA (2004) EMBO Rep 5:1153–1158. 3. Blakey D, Leech A, Thomas GH, Coutts G, Findlay K, Merrick M (2002) Biochem J 364:527–535. 4. Thomas GH, Mullins JG, Merrick M (2000) Mol Microbiol 37:331–344. 5. Khademi S, O’Connell J, III, Remis J, Robles-Colmenares Y, Miercke LJ, Stroud RM (2004) Science 305:1587–1594. 6. Zheng L, Kostrewa D, Berne`che S, Winkler FK, Li, XD (2004) Proc Natl Acad Sci USA 101:17090–17095. 7. Thornton J, Blakey D, Scanlon E, Merrick M (2006) FEMS Microbiol Lett 258:114–120. 8. Andrade SL, Dickmanns A, Ficner R, Einsle O (2005) Proc Natl Acad Sci USA 102:14994–14999. 9. Javelle A, Lupo D, Zheng L, Li X-D, Winkler FK, Merrick MJ (2006) J Biol Chem 281:39492–39498. 10. Javelle A, Thomas G, Marini AM, Kramer R, Merrick M (2005) Biochem J 390:215–222. 11. Winkler FK (2006) Pflu ¨gers Arch 451:701–707. 12. Soupene E, He L, Yan D, Kustu S (1998) Proc Natl Acad Sci USA 95:7030– 7034. 13. Arconde´guy T, Jack R, Merrick M (2001) Microbiol Mol Biol Rev 65:80–105. 14. Ninfa AJ, Jiang P (2005) Curr Opin Microbiol 8:168–173. 15. Forchhammer K (2004) FEMS Microbiol Rev 28:319–333. 16. Coutts G, Thomas G, Blakey D, Merrick M (2002) EMBO J 21:536–545. 17. Javelle A, Severi E, Thornton J, Merrick M (2004) J Biol Chem 279:8530–8538. 18. Detsch C, Stulke J (2003) Microbiology 149:3289–3297. 19. Huergo LF, Souza EM, Araujo MS, Pedrosa FO, Chubatsu LS, Steffens MB, Merrick M (2006) Mol Microbiol 59:326–337. 20. Strosser J, Ludke A, Schaffer S, Kramer R, Burkovski A (2004) Mol Microbiol 54:132–147. 21. Xu Y, Cheah E, Carr PD, van Heeswijk WC, Westerhoff HV, Vasudevan SG, Ollis DL (1998) J Mol Biol 282:149–165.

22. Atkinson M, Ninfa AJ (1999) Mol Microbiol 32:301–313. 23. Son HS, Rhee SG (1987) J Biol Chem 262:8690–8695. 24. van Heeswijk WC, Hoving S, Molenaar D, Stegeman B, Kahn D, Westerhoff HV (1996) Mol Microbiol 21:133–146. 25. Durand A, Merrick M (2006) J Biol Chem 281:29558–29567. 26. Xu Y, Carr PD, Huber T, Vasudevan SG, Ollis DL (2001) Eur J Biochem 268:2028–2037. 27. Jiang P, Zucker P, Atkinson MR, Kamberov ES, Tirasophon W, Chandran P, Schefke BR, Ninfa AJ (1997) J Bacteriol 179:4342–4353. 28. Martinez-Argudo I, Contreras A (2002) J Bacteriol 184:3746–3748. 29. Pioszak AA, Jiang P, Ninfa AJ (2000) Biochemistry 39:13450–13461. 30. Severi E, Javelle A, Merrick M (2006) Mol Membr Biol, 10.1080/ 09687860601129420. 31. Cheah E, Carr PD, Suffolk PM, Vasudevan SG, Dixon NE, Ollis DL (1994) Structure 2:981–990. 32. Xu Y, Carr PD, Clancy P, Garcia-Dominguez M, Forchhammer K, Florencio F, Vasudevan SG, Tandeau de Marsac N, Ollis DL (2003) Acta Crystallogr D 59:2183–2190. 33. Sakai H, Wang H, Takemoto-Hori C, Kaminishi T, Yamaguchi H, Kamewari Y, Terada T, Kuramitsu S, Shirouzu M, Yokoyama S (2005) J Struct Biol 149:99–110. 34. Schwarzenbacher R, von Delft F, Abdubek P, Ambing E, Biorac T, Brinen LS, Canaves JM, Cambell J, Chiu HJ, Dai X, et al. (2004) Proteins 54:810–813. 35. Zhang Y, Pohlmann EL, Ludden PW, Roberts GP (2001) J Bacteriol 183:6159– 6168. 36. Schutt H, Holzer H (1972) Eur J Biochem 26:68–72. 37. Benelli EM, Buck M, Polikarpov I, de Souza EM, Cruz LM, Pedrosa FO (2002) Eur J Biochem 269:3296–3303. 38. Javelle A, Merrick M (2005) Biochem Soc Trans 33:174–176. 39. Heinrich A, Woyda K, Brauburger K, Meiss G, Detsch C, Stulke J, Forchhammer K (2006) J Biol Chem 281:34909–34917. 40. DeLano WL (2002) PyMOL Molecular Graphics System (DeLano Scientific, San Carlos, CA).

1218 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0610348104

Conroy et al.