Structures of the Bacillus subtilis Glutamine Synthetase Dodecamer Reveal Large Intersubunit Catalytic Conformational Changes Linked to a Unique Feedback Inhibition Mechanism*

Background: B. subtilis GS catalyzes the production of glutamine, a key metabolite in nitrogen assimilation. Results: B. subtilis GS forms a dodecamer that undergoes large intersubunit conformational changes during catalysis. Conclusion: B. subtilis GS structures reveal a heretofore unseen active site restructuring that is linked to a novel feedback regulatory mechanism. Significance: The GS specific regulatory/catalytic mechanism may be used to target Gram-positive pathogens. Glutamine synthetase (GS), which catalyzes the production of glutamine, plays essential roles in nitrogen metabolism. There are two main bacterial GS isoenzymes, GSI-α and GSI-β. GSI-α enzymes, which have not been structurally characterized, are uniquely feedback-inhibited by Gln. To gain insight into GSI-α function, we performed biochemical and cellular studies and obtained structures for all GSI-α catalytic and regulatory states. GSI-α forms a massive 600-kDa dodecameric machine. Unlike other characterized GS, the Bacillus subtilis enzyme undergoes dramatic intersubunit conformational alterations during formation of the transition state. Remarkably, these changes are required for active site construction. Feedback inhibition arises from a hydrogen bond network between Gln, the catalytic glutamate, and the GSI-α-specific residue, Arg62, from an adjacent subunit. Notably, Arg62 must be ejected for proper active site reorganization. Consistent with these findings, an R62A mutation abrogates Gln feedback inhibition but does not affect catalysis. Thus, these data reveal a heretofore unseen restructuring of an enzyme active site that is coupled with an isoenzyme-specific regulatory mechanism. This GSI-α-specific regulatory network could be exploited for inhibitor design against Gram-positive pathogens.

Glutamine is a key metabolite in nitrogen assimilation because it is not only incorporated into proteins but also serves as the direct nitrogen donor for the synthesis of nearly 25% of all nitrogen-containing compounds (1). Glutamine is synthesized from ammonium, ATP, and glutamate by the enzyme glutamine synthetase (GS). 2 Underscoring its essential role in nitrogen metabolism and assimilation, GS is found in all organisms and is tightly regulated (2)(3). GS enzymes are all large homooligomeric machines, in which each active site is formed between subunit interfaces (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). There are three different classes of GS enzymes, categorized as GSI, GSII, and GSIII (14). GSI enzymes are specific to prokaryotes and archaea and have thus far all been shown to be dodecamers (3,(7)(8)(9)(10)(11)(12)(13)(14). GSII enzymes are decamers consisting of two stacked pentamer rings and are found in eukaryotes and soil-dwelling bacteria (4,5). GSIII proteins are harbored in certain anaerobic bacteria and cyanobacteria and, like GSI enzymes, are dodecamers but are formed from hexamer-hexamer interactions different from those that form GSI oligomers (6).
Despite their distinct oligomeric arrangements, the active site and residues involved in catalysis are conserved in GS proteins, suggesting similar overall mechanisms. This enzymatic reaction can be divided into two steps, which are facilitated by either Mg 2ϩ or Mn 2ϩ (3,15). In the first step, the ⑀-oxygen of the substrate glutamate attacks the ␥-phosphorus of ATP to form a ␥-glutamyl phosphate intermediate (16 -21). In the second step, ammonia, formed by the deprotonation of an ammonium ion, attacks the reaction intermediate to produce glutamine (Fig. 1A). Because of their central role in metabolism, GSs have emerged as important chemotherapeutic targets. For example, the herbicide glufosinate-ammonium, which inhibits plant GS, has recently become commercially available, and research is under way for the development of compounds that selectively target the Mycobacterium tuberculosis GS (22). The design of pathogen-specific drugs would be greatly facilitated by structural and biochemical studies revealing any unique catalytic or regulatory mechanisms utilized by these enzymes. However, our understanding of bacterial GS isoenzyme structure and function has been limited.
Notably, the GS⅐Gln feedback-inhibited form of B. subtilis GSI-␣ performs an unexpected role in controlling the DNAbinding activity of two global regulatory factors, GlnR and TnrA (29 -31). GlnR and TnrA have similar N-terminal DNA binding domains with putative Mer-like motifs; however, they have distinct C-terminal domains, which are targeted by GS⅐Gln. GS⅐Gln forms a complex with TnrA that prevents it from binding DNA, hence shutting off transcription of genes encoding nitrogen catabolic and scavenging enzymes (29). By contrast, GS⅐Gln acts as a chaperone to stabilize GlnR⅐DNA complexes, which allows it to repress expression of genes such as glnA (encoding GS) (31). Thus, B. subtilis GS serves as an enzyme, a chaperone, and a DNA binding coeffector. To gain insight into the functions of this unique, multitasking GS protein, we obtained structures of all GS catalytic and regulatory states and performed biochemical and in vivo studies.

EXPERIMENTAL PROCEDURES
Protein Purification, Crystallization, and Structure Determination-An artificial gene (codon optimized for E. coli expression) encoding the B. subtilis GS was obtained from Genscript Corp. (Piscataway, NJ) and subcloned into pET15b for protein expression. Gel filtration of GS species indicated a dodecameric oligomer. The His tag was removed for all structural and biochemical studies. Crystals were grown via hanging drop vapor diffusion at room temperature. B. subtilis apo-GS crystals were obtained by mixing the protein (40 mg/ml) at a 1:1 ratio with 40% 4-methyl-2,4-pentanediol and 200 mM MgSO 4 and inverting the drop over the reservoir solution. These crystals contain a dodecamer in the crystallographic asymmetric unit (ASU). To produce GS⅐glutamate⅐AMPPCP crystals, glutamate and AMPPCP were added to GS (at 40 mg/ml) to final concentrations of 5 mM, and the solution was mixed 1:1 with a reservoir of 15% PEG 8000, 0.1 M Hepes, pH 7.5, and 10 mM MgCl 2 . These crystals contain a hexamer in the ASU, and the dodecamer is generated via symmetry.
GS⅐Met-Sox-P⅐ADP was produced by mixing GS with 5 mM MgCl 2 , 5 mM ATP, and 5 mM L-methionine-S-sulfoximine (Met-Sox) for 1 h. This solution was then combined 1:1 with a crystallization reagent of 10% PEG 4000, 0.1 M Hepes, pH 7.5. These crystals contain a dodecamer in the ASU. Two crystal forms of GS⅐Gln were obtained. Form 1 was grown by mixing the GS⅐glutamine complex 1:1 with 40% 4-methyl-2,4-pentanediol and 100 mM MgCl 2 . Form 2 was obtained using 10% PEG 8000, 0.1 M Tris, pH 8.0, 5 mM MgCl 2 as a crystallization solution. Form 1 contains a dodecamer in the ASU, whereas form 2 contains a hexamer (with the dodecamer generated by crystal symmetry). All x-ray intensity data were collected at 100 K at the Advanced Light Source beamline 8.3.1. The space group and cell parameters of each crystal are shown in Tables 1 and 2. Phases were determined first for the apo-GS form by molecular replacement, using the M. tuberculosis GS dodecamer as the search model. Several nonconserved loops were removed from the search model before molecular replacement. Refinement was carried out using the Crystallography and NMR Software pack- Enzyme Active Site Restructuring Linked with Regulation age (CNS) and PHENIX (37,38). The models were all constructed with the modeling program O (39). Enzyme, Kinetic, and in Vivo Assays-The glnA(R62A) mutant was constructed, and overexpression and purification of the mutant R62A GS were performed as described previously (40). Methods for the cultivation of bacteria in the minimal medium of Neidhardt et al. (41) have been described previously (42). All cultures contained 0.5% glucose, whereas the final concentration of all nitrogen sources was 0.2%. ␤-Galactosidase was assayed in crude extracts prepared from cells grown to mid-log growth phase (42). The reported ␤-galactosidase levels were corrected for the endogenous activity present in cells containing the promoterless lacZ fusion vectors integrated into the amyE gene. One unit of ␤-galactosidase activity produced 1 nmol of o-nitrophenol min Ϫ1 . The kinetic constants for glutamate, ATP, and hydroxylamine were measured by assaying the production of ␥-glutamylhydroxamate (32). The ammonium K m values were determined by measuring the production of inorganic phosphate that resulted from the hydrolysis of ATP as described (43). The inhibition of the Mg 2ϩ -dependent biosynthetic reaction by glutamine, AMP, and Met-Sox were determined with glutamate and ATP concentrations of 150 and 18 mM, respectively.
Cooperativity Assays-To assay for GS cooperativity, mixed GS oligomers were produced. The two proteins that were used to produce the mixed dodecamer were the WT, active protein and an inactive GS(E304A/A305G) mutant. The mutant also contained an N-terminal 35 residue extension, which included a His tag, followed by a tobacco etch virus cleavage site and a hydrophilic region. This extension permitted the detection of mixed dodecamer populations by SDS-PAGE. The GS(E304A/ A305G) protein was expressed as for WT and purified via Ni-NTA chromatography. The N-terminal His tag of WT GS was cleaved via thrombin, and the cleaved protein was isolated as the flow-through of a Ni-NTA column. This untagged WT protein was then mixed with the His-tagged GS(E304A/A305G) protein. The mixed sample was dialyzed in buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, and 2 M urea for 12 h. This urea treatment resulted in dodecamer dissolution and monomer formation. The sample was then dialyzed into a buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, and 5 mM MgCl 2 for 16 h with buffer changes at 4 and 8 h to allow dodecamer formation. The sample was subjected to Ni-NTA purification; the imidazole-eluted fractions were isolated and treated with tobacco etch virus protease to remove the His tag from the GS(E304A-A305G) protein. The digested sample was dialyzed into buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, and 5 mM MgCl 2 overnight and applied to a Ni-NTA column to remove any His tag peptide and uncleaved protein. The samples were run on SDS-PAGE to ensure that both proteins were present. The enzymatic activities of the mixed dodecamer were measured by following the production of ␥-glutamylhydroxamate (19). The Mg 2ϩ -dependent forward reaction was assayed in a final volume of 0.5 ml containing 0.05 M imidazole, pH 7.0, 100 mM sodium glutamate, 40 mM hydroxylamine, 100 mM MgCl 2 , and 18 mM ATP. The Mn 2ϩ -dependent forward reaction was measured in a final volume of 0.5 ml containing 0.05 M imidazole, pH 7.0, 100 mM sodium glutamate, 40 mM hydroxylamine, 7.5 mM MnCl 2 , and 7.5 mM ATP. The Mn 2ϩ -dependent reverse (transferase) reaction was assayed in 0.5 ml containing 20 mM sodium arsenate, 50 mM hydroxylamine, 40 mM glutamine, 2 mM ADP, 25 mM Tris, pH 7.5, 0.33 mM MnCl 2 , and 0.4 M K 2 SO 4 . The purified enzymes were pretreated with 0.4 M K 2 SO 4 before addition to the transferase reaction buffer (19). For the biosynthetic reactions, 1 unit of GS activity corresponds to the ATPdependent formation of 1 nmol of ␥-glutamylhydroxamate min Ϫ1 . In the transferase reaction, 1 unit of GS activity is the ADP-dependent formation of 1 nmol of ␥-glutamylhydroxamate min Ϫ1 .
Circular Cichroism (CD) and ITC Studies-CD measurements in the far UV regions (190 -260 nm) were performed with pre-and post-urea-treated WT GS, GS(E304A/A305G) and the mixed complex at 25°C with a 1-mm path length. Each sample solution contained 300 mM NaF and 10 mM Tris, pH 7.5. The temperature of the sample was controlled with a PTC-348 WI Peltier block. ITC experiments were performed using a VP-ITC system with GS loaded in the sample cell and either AMP or AMPPCP loaded in the syringe. The nucleotides were titrated into the sample cell with GS at 25°C, and the resulting isotherms were fitted with the Origin software.

Structure Determinations of B. subtilis GS and Its Complexes-
To deduce the structural mechanism of GSI-␣ catalysis and its regulation, de novo crystals were obtained of the apo-, substrate-bound, glutamine-bound, and transition state complex (Tables 1 and 2). The transition state structure was obtained by reacting GS with ATP and Met-Sox (see "Experimental Procedures"). GS phosphorylates Met-Sox in the presence of ATP to form the transition state analog, Met-Sox-P, which causes irreversible inhibition of the enzyme (44). The B. subtilis apo-GS structure was solved first by molecular replacement using the M. tuberculosis GSI-␤ structure, which is 41% identical to B. subtilis GS (see "Experimental Procedures"). The structure, refined to R work /R free values of 21.7%/26.7% 3.1 Å (Table 1), belong to a test set of 5% randomly selected data.
contains 444 residues/subunit and is composed of 19 ␤-strands and 15 ␣-helices. The subunit can be divided by helix ␣3 into a larger C-terminal domain and a smaller, N-terminal region (Fig. 1A). Two hexameric rings are formed from six GS subunits and stack to form a dodecamer ( Fig. 1, B and C). The hexameric rings are held together primarily by interactions between C-terminal helices ␣14 and ␣15 and ␣14Ј and ␣15Ј (where a prime indicates the neighboring subunit). Each subunit has the classical GS bifunnel structure, whereby key residues that mediate catalysis are located on loops that encapsulate the active site (3,45,46) (Fig. 1A). The active site regions are mainly located in the C-terminal domain, with the N-terminal domain of the adjacent subunit contributing one loop. Previous studies on GSI-␤ enzymes led to the designation of key active site loops (8). Based on homology with GSI-␤ proteins, the corresponding GSI-␣ active site loops within the C-domain are the Glu flap (B. subtilis residues 301-306), the Tyr loop (residues 365-373), the Asn loop (residues 231-242), and the Tyr 179 loop (residues 148 -158). The N-terminal domain provides the Asp 50 Ј loop (also called the "latch"), which contributes a key catalytic aspartic acid. In B. subtilis, this loop consists of residues 52-66, and the catalytic aspartic acid corresponds to Asp 53 Ј (Fig. 1D).
Analysis of the Apo-GS Structure-The Glu flap and latch are arguably the most critical GS active site loops because they contribute residues that are central for catalysis. The aspartic acid from the Asp 50 Ј loop abstracts a proton from ammonium to create ammonia, and the Glu flap harbors the catalytic base (Glu 304 in B. subtilis GS), which abstracts a proton from the transition state to form glutamine and ADP. In the apo-GSI-␣ structure, the Glu flap is almost entirely disordered; only residues 301 and 306 are visible, and they display B-factors (average) Ͼ 100 Å 2 (Table 3). Glu flap flexibility is important because it permits facile substrate entry and product release. The Tyr loop, which contains the tyrosine that is adenylylated in GSI-␤ enzymes, is also disordered. A corresponding tyrosine (Tyr 373 ) is found in the B. subtilis GSI-␣; however, studies have shown that it is not adenylylated, consistent with a different GSI-␣ regulatory mechanism (23). Aside from the Glu flap and Tyr loop, clear density is observed for the other active site regions in the GSI-␣ apo-structure, and with the exception of the Tyr 179 loop, the catalytic regions of the GSI-␣ enzyme are similar in length to GSI-␤ proteins. In GSI-␤ enzymes, the Tyr 179 loop contains 36 residues and extends from the active site into the central dodecamer channel, whereas in the GSI-␣ structure, this region consists of a short, 11-residue loop (Fig. 1, E and F) (3). When the Tyr 179 region is excluded, the C␣ atoms of the B. subtilis GS subunit can be superimposed onto those of S. typhimurium (Protein Data Bank entry 1f52) and M. tuberculosis (Protein Data Bank entry 2bvc) with root mean square

Enzyme Active Site Restructuring Linked with Regulation
(r.m.s.) deviations of 1.3 and 1.5 Å, respectively. The altered structures of the Tyr 179 loop in GSI-␣ compared with GSI-␤ lead to different pore structures for these enzymes. The function(s), if any, of the pores, however, is not known (Fig. 1E).

GSI-␣-Substrate and Transition State Structures Reveal That Large Conformational Changes Are Required for Catalysis-To
gain insight into the enzymatic mechanism employed by GSI-␣ enzymes, we determined the structure of GSI-␣ bound to the substrates glutamate and AMPPCP (which is a non-hydrolyzable ATP analog) to 2.87 Å ( Table 1). The substrate glutamates and AMPPCP molecules bind at each end of the bifunnel. The glutamate substrates are located under the Glu flap. Although better ordered than the apo-structure the Glu flap residues in this structure, as well as the glutamate substrates, display high B-factors (B-factors (average) ϭ 86.7 Å 2 for the glutamate substrates and 85.0 Å 2 for Glu flap residues) ( Table 3). This suggests that the active site may become protected by Glu flap closure only during catalysis. The glutamate substrate peptide carboxyl and amide groups interact with Arg 298 and His 245 , whereas its side chain carboxyl interacts with a Mg 2ϩ ion. The adenine moiety of the AMPPCP is sandwiched between the side chains of Arg 331 and Tyr 201 , whereas hydrogen bonds to the adenine N6, N1, and N7 atoms are provided by the carbonyl oxygen of Ile 328 and the side chains of Ser 249 and Ser 329 . Interestingly, the phosphate groups of AMPPCP point away from the active site and are only contacted weakly by Lys 44 . In fact, density is only observed for the ␣and ␤-phosphates of the AMPPCP molecules. Hence, AMPPCP does not appear to bind GS productively, which is consonant with ITC data showing that GS binds AMPPCP with 100-fold lower affinity than AMP (Fig. 2). Comparisons of the substrate-bound and apo-GSI-␣ structures revealed that not only are the subunits and their corresponding active sites essentially unchanged upon substrate binding, but the dodecameric arrangement remains unaffected as well (superimposition of C␣ atoms ϭ 0.50 -0.60 Å/subunit and 0.60 Å for all C␣ atoms of the dodecamers) (Fig.  3, A-D).
Key active site residues are conserved in all classes of GS, suggesting a similar overall catalytic mechanism. According to this reaction scheme, ATP and glutamate first bind in the active site at each end of the bifunnel. The bound ATP ␥-phosphate is polarized by the metal ions and an arginine residue (corresponding to B. subtilis GSI-␣ Arg 316 ), which facilitates nucleophilic attack by the bound glutamate substrate to yield ␥-glutamyl phosphate (Fig. 3A). The next step in the reaction involves proton abstraction of ammonium by the Asp 50 Ј aspartic acid to create the ammonia. The ammonia then reacts with the ␥-glutamyl phosphate to form the tetrahedral transition state. The final step is dissociation of the phosphate from the intermediate and concomitant abstraction of the proton from the ␥-amino group of the intermediate by the Glu flap glutamate (B. subtilis Glu 304 ).
GSI-␤ structures have defined the residues and specific active site architecture optimal for catalysis (3). At first glance, the minimal shifts in the active site regions of the apo-and substrate-bound B. subtilis GS structures appear to suggest that GSI-␣ catalysis may also involve only moderate active site loop movements. However, close examination of the apo-and substrate-bound GSI-␣ active sites reveal that they lack an optimal  organization of catalytic residues. This is largely due to the deep insertion of the latch, in particular residues Arg 62 Ј and Glu 65 Ј, into the neighboring active site of each subunit. Arg 62 Ј blocks and prevents the correct positioning of the catalytic Asp 53 Ј as well as Ser 56 Ј, which stabilizes the Glu flap. Moreover, the side chain of Glu 65 Ј overlaps the location of the key catalytic arginine, Arg 316 , which assists in phosphoryl transfer. In fact, in the apo-and substrate-bound GSI-␣ structures, the side chain of Arg 316 is located entirely outside the active site. The non-optimal active site organization in the apo-and substrate-bound GSI-␣ structures suggests that either a different catalytic mechanism may be at play in GSI-␣ and GSI-␤ proteins or that significant structural changes must occur during catalysis. To address this question, we generated a GSI-␣⅐transition state complex by reacting the protein with Met-Sox, ATP, and Mg 2ϩ and crystallized and solved the structure to R work /R free values of 20.3%/23.2% to 2.58 Å (see "Experimental Procedures") ( Table 1). Met-Sox-P mimics the GS transition state as its methyl group occupies the ammonium substrate binding site (8). Thus, the Met-Sox-P adduct forms a stable complex with the enzyme that precludes the last step, which is release of the phosphate from the intermediate and concomitant abstraction of the proton from the intermediate ␥-amino group by Glu flap residue Glu 304 . The structure revealed that the expected catalysis had occurred, leading to Met-Sox-P and ADP creation (Fig. 3B).

Enzyme Active Site Restructuring Linked with Regulation
GS enzymes all utilize two Mg 2ϩ ions to aid substrate binding, transfer of the phosphate group, and structural stability. In the GSI-␣⅐transition state structure, a third metal ion is found, which probably functions in neutralizing the negative charge arising from the intermediate. A similarly located metal ion was identified in the M. tuberculosis GSI-␤⅐Met-Sox-P and maize GSII⅐phosphinothricin structures (5,8). Contacts to the adenine and ribose moieties of ADP in the transition state complex are essentially identical to those made to AMPPCP. In the transition state complex, however, the phosphate groups are rotated into the active site near the Met-Sox-P and are contacted by Arg 321 and Arg 331 as well as a Mg 2ϩ ion. The ␤-phosphate lies close to the phosphate transferred to Met-Sox to form Met-Sox-P. There are numerous contacts from GS to the Met-Sox-P. Arg 298 interacts with the carboxyl moiety, and the Met-Sox-P amide nitrogen interacts with the Glu 134 side chain as well as the carbonyl of Gly 241 . The Met-Sox-P phosphate moiety is surrounded by multiple basic residues, including Arg 316 , Arg 321 , Arg 335 , and His 245 (Figs. 3,C-E, and 4). Glu flap residue Glu 304 contacts Asp 53 Ј and is within close distance to the Met-Sox-P methyl group. The Glu 304 -Asp 53 Ј interaction probably helps to stabilize and shield the Met-Sox-P from attack by bulk solvent as well as facilitating proton abstraction of the ammonium to form ammonia. The resultant ammonia is then in an ideal position to attack the ␥-glutamyl phosphate intermediate. Hence, the GSI-␣⅐Met-Sox-P⅐ADP complex mimics the tetrahedral adduct before the generation of glutamine and inorganic phosphate products.
Comparison of the transition state structure with the apoand substrate-bound forms revealed dramatic conformational changes, consonant with the finding that the residues in the transition state structure are optimally positioned for catalysis. The largest structural changes are located on the active site loop regions (r.m.s. deviations of apo-/substrate-bound versus Met-Sox-bound subunits ϭ 1.8 -2.0 Å, whereas the r.m.s. deviations are 0.8 Å when the active site loops are not included). More-over, superimposition of the transition state dodecamer onto the apo-or substrate-bound dodecamers results in r.m.s. deviations of 2.5-3.0 Å, indicating that the catalytically induced structural changes are transmitted between subunits, leading to significant alterations in the overall dodecamer structure (Fig.  3D). Although all of the GSI-␣ active site loops are altered upon transition state formation, residues in the Asp 50 Ј loop undergo the most dramatic structural rearrangements (r.m.s. deviations ϭ 3.7-4.0 Å for superimposition of the Asp 50 Ј loops of the transition state and glutamate-bound structures). As noted, the Asp 50 Ј loop is wedged into the active site in apo-and substrate-bound structures, preventing an optimal configuration of catalytic residues. The restructuring of the active site that takes place during catalysis results in a functional arrangement of catalytic residues (Fig. 3, C and E). These extensive structural changes include the unfolding of one turn of the helix from residue 53 to 59. The structural changes in the Asp 50 Ј loop lead to the expulsion of Glu 65 Ј and Arg 62 Ј more than 10 Å from the active site to the solvent-exposed surface of the dodecamer (Fig. 3E). Concomitantly, the Arg 316 loop moves by ϳ10 Å, resulting in relocation of the Arg 316 side chain into the position vacated by Glu 65 Ј. These alterations create an active site organization that allows GSI-␣ to employ the same general catalytic mechanism used by all characterized GS proteins (3).
Demonstration that B. subtilis GS Is Cooperative Using Mixed Oligomers-The structural changes that take place during GSI-␣ catalysis are found in all the active sites and result in an altered dodecamer structure, suggesting that the enzyme may be cooperative, whereby binding and catalysis at one active site is affected by adjacent subunits. Such cooperativity has been indicated by previous studies revealing active site coupling with E. coli GS (45). However, to directly assess GSI-␣ cooperativity, we developed a mixed dodecamer assay. Previous studies showed that single E304A and A305G mutations abrogated GS activity (40). Thus, in this assay, WT GS subunits were mixed with inactive, double GS(E304A/A305G) mutant subunits in the presence of 2 M urea, allowing subunit exchange. The mixture was then dialyzed into urea free buffer resulting in the formation of mixed oligomers. 2 M urea treatment did not cause denaturation as assessed by circular dichroism but broke up subunit-subunit interactions (see "Experimental Procedures") (Fig. 5). To assay possible effects of urea treatment on GS function, we first tested the activity of WT GS and WT GS that had undergone the identical urea treatment as the mixed oligomer. As shown in Table 4, this procedure had no effect on GS enzyme activity. If GSI-␣ is not cooperative, a mixed WT GS⅐GS(E304A/A305G)-containing dodecamer should display either WT or only slightly reduced enzyme activity. By contrast, if the dodecamer is cooperative, then incorporation of even a few GS(E304A/A305G) subunits into the dodecamer should impair or abrogate activity for a cooperative enzyme. The mixture used in the assay depicted in Fig. 5 contained between a 3and 4-fold excess of the WT protein (lower molecular weight band) compared with the mutant. Thus, this experiment provided a strong test case of cooperativity. Even under these conditions, where there is far less mutant than WT, the activity was abrogated, supporting the hypothesis that the dodecamer is cooperative (Table 4). Note that there is a dodecamer in the crystallographic asymmetric unit, and each active site (formed between subunits) has generated the identical transition state with the same contacts. One active site (between the magenta and salmon subunits) is highlighted, and key residues that contact Met-Sox-P and ADP are shown as sticks (enlarged view to the right). The key catalytic residues Asp 53 and Glu 304 are labeled and highlighted with asterisks. Magnesium ions are colored green.

Molecular Basis for GSI-␣-specific Gln Feedback Inhibition-
GSI-␣ enzymes are distinct from their GSI-␤ counterparts in that they are feedback-inhibited by the reaction product, Gln. Previous GSI-␤-Gln structures showed no interactions between Gln and side chains of the Glu flap, and the bound glutamine and Glu flap were largely disordered in the structure (3). Indeed, this Glu flap flexibility was suggested as being important in facile product release following catalysis. The finding that GSI-␣ enzymes are feedback-inhibited by Gln suggests that differences must exist in GSI-Gln isoenzyme structures. It also indicates that important distinctions must be present between GSI-␣-Glu and GS-␣-Gln structures. Understanding the structural basis for GSI-␣-Gln feedback inhibition is crucial, because this form of the enzyme serves as a key nexus in the Bacilli nitrogen regulation pathway (29 -31). Thus, we next determined structures of B. subtilis GS⅐Gln-Mg 2ϩ complexes. Two structures were obtained using different crystallization conditions (see "Experimental Procedures") ( Table 2). The two structures are identical and reveal that Gln binds within the Glu flap pocket in a similar location as the substrate glutamate (Fig. 6A). However, there are striking differences between the Gln-bound and Glu-bound states. Unlike the Glu-bound state, in the Gln-bound structure, both the side chain and backbone atoms of Glu flap residues are well ordered; the B-factors (average) for Glu flap residues in the Gln-bound structure are 35.9 Å 2 compared with 85.0 Å 2 for the glutamatebound form, and the B-factors (average) for the glutamines are 33.0 Å 2 (Table 3). In fact, although not directly comparable due to their differing resolutions, the B-factors of the Glu flap residues in the Gln-bound structures are on par with those in the transition state structure (which have B-factors (average) ϭ 24.0 Å 2 for Glu flap residues and 20.0 and 20.3 Å 2 for ADP and Met-Sox-P, respectively) ( Table 3).
In the GSI-␣-Gln structures, the Asp 50 Ј loop adopts the same conformation as that observed in the apo-and substrate-bound   GSI-␣ forms (Figs. 6B and 7). Remarkably, however, in the GS-Gln structure, Arg 62 plays a central role in Gln binding and, hence, Gln feedback inhibition. Specifically, the Arg 62 Ј side chain hydrogen-bonds with the key catalytic residue from the Glu flap, Glu 304 (Fig. 6B). The stabilized Glu 304 also makes a tight hydrogen bond to the glutamine N⑀ atom. This intimate hydrogen bond network ties the Glu flap closed, explaining its well ordered state in this structure (Fig. 6B). The vital role played by Arg 62 in Gln feedback inhibition probably explains why, unlike GSI-␤ enzymes, the GSI-␣ Asp 50 Ј loop adopts distinct structures in different states. Asp 50 Ј loop insertion into the active site in GSI-␣ is critical for Gln feedback inhibition, but it can then be readily ejected in the absence of the Glnspecific hydrogen bond network to allow catalysis. In this way, the Asp 50 Ј loop acts as a GSI-␣-specific regulatory trigger. The finding that Glu 304 directly interacts with Gln provides a molecular explanation for previous studies showing that E304D or E304A mutants were defective in Gln feedback inhibition (40). The intricate Arg 62 -Glu 304 -Gln hydrogen bond linkage also explains why GSI-␣, but not GSI-␤, enzymes are feedbackinhibited by Gln because sequence alignments of GS proteins show that Arg 62 is completely conserved in GSI-␣ enzymes but not GSI-␤ enzymes, where this position is typically Gly or Ala (Fig. 6B). Thus, these structures demonstrate that B. subtilis GS uses a residue found only in GSI-␣ enzymes, Arg 62 , along with a conserved catalytic residue (Glu 304 ) to mediate glutamine binding. The structures also reveal the basis for Gln feedback inhi-bition, because the Arg 62 Ј-Glu 304 -Gln hydrogen bond network locks the Glu flap in a closed state, preventing Gln release, substrate binding and catalysis.
The Role of Arg 62 in GS Activity and Regulation; Kinetic, Enzymological, and in Vivo Studies-Our structures indicate that Arg 62 plays a key role in GSI-␣-specific Gln feedback inhibition. The structures also suggest that Arg 62 is not involved in catalysis and, in fact, must be ejected from the active site during transition state formation. To test these structure-based hypotheses, an R62A mutant was constructed. The purified R62A enzyme had only small differences (less than 2-fold) in its steady-state kinetic constants compared with WT, indicating that it does not play a significant role in substrate binding or catalysis, consistent with our structural data ( Table 5). Unlike the E304A substitution in GS, the R62A replacement did not increase the K m for ammonium (Table 5). This result is again consistent with the structural data and argues that Arg 62 is not part of the ammonium substrate binding site. In contrast to E304A, which is highly resistant to Met-Sox inhibition, the R62A enzyme has a sensitivity to Met-Sox that is similar to that of WT GS (Table 6). These data, again, support the idea that Arg 62 is not involved in the GS reaction mechanism. Strikingly, however, and consistent with our structural findings, the R62A mutant was highly resistant to Gln feedback inhibition ( Table  6). The R62A substitution was also found to increase the resistance to AMP inhibition. We have argued previously that there is an interdependence between the glutamate (glutamine) and ATP (AMP) binding sites (40). The activities of the nitrogen transcription factors TnrA and GlnR are regulated by the feedback-inhibited form of GS (29 -31). TnrA-and GlnR-regulated genes were shown previously to be expressed constitutively in mutants encoding feedback-resistant GS enzymes (29 -31). The contribution of the GS Arg 62 residue to the in vivo regulation of TnrA and GlnR was assessed by constructing a B. subtilis mutant containing an R62A substitution in the chromosomal GS gene (glnA) (see "Experimental Procedures") (40). Gene reg-FIGURE 7. The GS⅐glutamine structure active site is similar to that of the apo-and substrate-bound forms. Superimposition of the apo-, glutamate-AMPPCP, and glutamine-bound GSI-␣ structures, focusing on key active site loops. The apo-structure is colored cyan, the glutamate-AMPPCP-bound structure is yellow, and the glutamine-bound structure is green. The active site loops adopt essentially the same conformation. The Glu flap in the apo-state is mostly disordered, whereas backbone atoms of the Glu flap in the substrate-bound state are visible (but not side chains), and in the glutaminebound form, the Glu flap is as well ordered as in the transition state.

TABLE 5 Enzymatic constants for wild-type and mutant GS proteins
The kinetic constants were determined for the Mg 2ϩ -dependent biosynthetic reaction. All assays were performed at least twice. The uncertainty is the S.E. from the nonlinear regression analysis of the data.

TABLE 6 Sensitivities of wild-type and mutant GS proteins to inhibition
The IC 50 is the inhibitor concentration that reduces enzymatic activity by 50%. The Mg 2ϩ -dependent biosynthetic reaction was used to measure the inhibition. All values are the averages of at least two determinations.  (Table 7). These results indicate that the glnA(R62A) mutant has the in vivo phenotypes expected for a strain encoding for feedback-resistant GS enzyme.
In conclusion, we have captured all key catalytic and regulatory states of the B. subtilis GSI-␣ enzyme. Surprisingly, these structures reveal heretofore unseen large structural changes during catalysis. GSI-␣ enzymes are also distinct among GSs in being feedback-inhibited by the product, Gln. Remarkably, our work demonstrates that the catalytic and regulatory features of GSI-␣ are linked. Specifically, Gln feedback inhibition arises from a hydrogen bond network between Gln, the active site Glu 304 , and a GSI-␣-specific residue, Arg 62 Ј, from an adjacent subunit. Notably, Arg 62 Ј is ejected from the active site as part of the GSI-␣ active site reorganization during catalysis. These key Arg 62 roles were supported by biochemical and in vivo studies; mutation of Arg 62 to alanine eliminates Gln feedback inhibition but does not affect catalysis. The extensive conformational changes that take place during catalysis also suggest a cooperative catalytic mechanism, which was supported by a mixed dodecamer enzyme assay. Due to its critical role in nitrogen metabolism, GS is an attractive and validated target for drug design against pathogenic bacteria. The isoenzyme-specific glutamine interaction network uncovered here presents a promising point of intervention in the development of highly specific inhibitors in the treatment of Gram-positive pathogenic bacteria.

TABLE 7
TnrA-and GlnR-dependent gene regulation ␤-Galactosidase activity was determined in extracts of cells harvested during exponential growth in glucose minimal medium containing glutamate plus casamino acids or glutamate as the respective nitrogen sources for excess and limiting nitrogen conditions. Each value is the mean of at least two independent determinations and is reproducible to within 10% (S.E.) of the mean value.