Effectors of the Stringent Response Target the Active Site ofEscherichia coli Adenylosuccinate Synthetase*

Guanosine 5′-diphosphate 3′-diphosphate (ppGpp), a pleiotropic effector of the stringent response, potently inhibits adenylosuccinate synthetase from Escherichia coli as an allosteric effector and/or as a competitive inhibitor with respect to GTP. Crystals of the synthetase grown in the presence of IMP, hadacidin, NO3 −, and Mg2+, then soaked with ppGpp , reveal electron density at the GTP pocket which is consistent with guanosine 5′-diphosphate 2′:3′-cyclic monophosphate. Unlike ligand complexes of the synthetase involving IMP and GDP, the coordination of Mg2+ in this complex is octahedral with the side chain of Asp13 in the inner sphere of the cation. The cyclic phosphoryl group interacts directly with the side chain of Lys49 and indirectly through bridging water molecules with the side chains of Asn295 and Arg305. The synthetase either directly facilitates the formation of the cyclic nucleotide or scavenges trace amounts of the cyclic nucleotide from solution. Regardless of its mode of generation, the cyclic nucleotide binds far more tightly to the active site than does ppGpp . Conceivably, synthetase activity in vivo during the stringent response may be sensitive to the relative concentrations of several effectors, which together exercise precise control over the de novosynthesis of AMP.

the stringent response stems from a severe shortfall, relative to the demands of ribosomal protein biosynthesis, in one or more of the aminoacylated tRNAs. During the stringent response in E. coli, pools of ATP and GTP decrease, but concentrations of pppGpp and ppGpp rise to millimolar levels. The reduction in the de novo synthesis of purine nucleotides putatively is a consequence of direct inhibition of adenylosuccinate synthetase and IMP dehydrogenase by ppGpp (2)(3)(4).
The catalytic mechanism is a two-step process (3)(4)(5) as follows: the formation of 6-phosphoryl-IMP by nucleophilic attack of the 6-oxo group of IMP on the ␥-phosphoryl group of GTP, followed by a second nucleophilic displacement of the 6-phosphoryl group by aspartate to form adenylosuccinate. ppGpp potently inhibits the synthetase (K m ϳ50 M; 6 -8). The mechanism of inhibition by ppGpp, however, remains unsettled. ppGpp may inhibit the synthetase by binding to an allosteric site, as suggested by its noncompetitive inhibition with respect to GTP (6,7), or by simply binding to the GTP pocket, as suggested by its competitive inhibition with respect to GTP in a different study (8).
Reported here are crystal structures of adenylosuccinate synthetase from E. coli complexed with IMP, NO 3 Ϫ , Mg 2ϩ , hadacidin, and guanosine 5Ј-diphosphate 2Ј:3Ј-cyclic monophosphate (hereafter the ppG2Ј:3Јp complex), and with IMP, NO 3 Ϫ , Mg 2ϩ , hadacidin, and GDP (hereafter the GDP complex). Hadacidin, a fermentation product of Penicillium frequentans (9), is a competitive inhibitor (K i ϳ10 Ϫ6 M) with respect to aspartate (10,11). Even though ppGpp was used as a ligand in crystal soaking buffers, the electron density in the guanine nucleotide pocket is consistent only with ppG2Ј:3Јp. Evidently, ppG2Ј:3Јp binds with much greater affinity to the synthetase than does ppGpp. Crystals either absorb a minor impurity of the cyclic nucleotide from solution or the synthetase itself transforms ppGpp to ppG2Ј:3Јp. Regardless of its mechanism of formation, the ppG2Ј:3Јp complex here represents the first structure of an effector of the stringent response with one of its target enzymes. Results here and from past investigations raise the possibility of several effectors of the stringent response acting in a coordinated way to modulate synthetase activity.

MATERIALS AND METHODS
Purification of Enzyme from E. coli-The synthetase was prepared as described previously from a genetically engineered strain of E. coli (12). The enzyme was at least 95% pure on the basis of SDS-polyacrylamide gel electrophoresis.
Crystallization-Hadacidin was a generous gift of Drs. Fred Rudolph and Bruce Cooper, Department of Biochemistry and Cell Biology, Rice University. All other reagents came from Sigma. Crystals were grown by the method of hanging drops under conditions similar to those employed in previous work (13). For reasons provided below, acetate was excluded from the crystallization droplets, and in preparing crystals for ppGpp soaks, GDP was omitted as well. Droplets (total volume, 12 l) contained equal parts of an enzyme solution (HEPES (30 mM), IMP (4 mM), hadacidin (4 mM), GDP (either 0 or 4 mM), and protein (18 mg/ml) at pH 6.8) and a crystallization buffer (polyethylene glycol 8000 (16% w/v), cacodylic acid/sodium cacodylate (100 mM, pH 6.5), magnesium nitrate (100 mM)). The pH of the droplets was 6.5. Wells contained 500 l of crystallization buffer. Crystals of approximately 0.8 mm in all dimensions, belonging to the space group P3 2 21, grew in about 1 week. The omission of GDP and/or acetate had no significant effect on unit cell parameters; the crystalline complexes here are isomorphous to previously published ligand complexes of the synthetase.
Preparation of ppG2Ј:3Јp Complex-ppGpp (preparation 33A) was prepared as described previously (14). The sample was checked for purity by thin layer chromatography, using polyethyleneimine plates and 1.5 M KH 2 PO 4 , pH 3.4 (14). Synthetase crystals, grown in the absence of guanine nucleotides, were exposed for 12 h to a solution containing polyethylene glycol 8000 (18% w/v), cacodylic acid/sodium cacodylate (100 mM, pH 6.5), magnesium nitrate (100 mM), IMP (2 mM), hadacidin (2 mM), and ppGpp (1 mM). A soaked crystal then was transferred sequentially at 10-min intervals to fresh solutions, which contained the components above along with glycerol at concentrations of 5, 10, 15, 20, and 25% (v/v). Glycerol is a cryoprotectant of adenylosuccinate synthetase crystals. After passing through the glycerol equilibration steps, a single crystal was mounted and flash-frozen to 100 K.
Thin Layer Chromatography-Possible degradation of ppGpp under conditions employed in soaking crystals was monitored by thin layer chromatography (14). Aliquots were taken at intervals of 2, 5, 12, and 36 h from a solution containing Mg(NO 3 ) 2 (100 mM), hadacidin (2 mM), cacodylic acid/sodium cacodylate (100 mM, pH 6.5), and ppGpp (2 mM), spotted on polyethyleneimine plates, and developed in 1.5 M KH 2 PO 4 (pH 3.4). Parallel time courses were run in the absence and presence of adenylosuccinate synthetase at a concentration of 2 mg/ml (approximately a 40 M subunit concentration). GTP and ppGpp were used as standards in thin layer chromatography. The developed plates were inspected for UV-absorbing components. A series of solutions, obtained by the dilution of guanine nucleotide standards of known concentration, were processed by the chromatographic procedure above, in order to determine the detection limit for guanine nucleotides.
Data Collection-Data from single crystals of the GDP complex and the ppG2Ј:3Јp complex were collected on a Siemens area detector at 100 K and were reduced by XENGEN (15). The data sets were at least 97% complete to nominal resolutions (at which the average of I/(I) is 2) of 2.3 and 2.5 Å for the GDP and ppG2Ј:3Јp complexes, respectively (Table I).
Model Refinement-Starting phases were calculated from the GDP complex (13), omitting all ligands and solvent. Refinement procedures are as described previously (13). The ligand models were fit to omit electron density maps, followed by a cycle of refinement using XPLOR (16). Constants of force and geometry for the protein came from Engh and Huber (17), and those for hadacidin from Poland et al. (13). Refinement parameters for ppG2Ј:3Јp were based on those of GTP and the crystal structures of 2Ј:3Ј-monophosphate nucleotides (18 -20). In early rounds of refinement, models were heated to 2000 K and then cooled in steps of 25-300 K. In later rounds of the refinement, the systems were heated to 1000 or 1500 K and cooled in steps of 10 K. After the slow-cool protocol was complete (at 300 K), the models were subjected to 120 steps of conjugate gradient minimization, followed by 20 steps of individual B-parameter refinement. Individual B-parameters were subject to the following restraints: nearest neighbor, main chain atoms, 1.5 Å 2 ; nextto-nearest neighbor, side chain atoms, 2.0 Å 2 ; nearest neighbor, side chain atoms, 2.0 Å 2 ; and next-to-nearest neighbor, side chain atoms, 2.5 Å 2 . Criteria for the addition of water molecules were identical to those of previous studies (5,13). (21) indicates an uncertainty in coordinates of 0.30 Å. The amino acid sequence used in refinement is identical to that reported previously (22,23). Results of data collection and refinement are in Table I. A overview of the ppG2Ј:3Јp complex is in Fig. 1, and a schematic of ppG2Ј:3Јp is in Fig. 2. Gln 10 is the only serious violation of the Ramachandran plot as identified by PROCHECK (24). The conformation of Gln 10 is enforced by its hydrogen bonding environment as described in previous work (22,23). The models have better stereochemistry than is typical for structures based upon data of comparable resolution. Thermal parameters vary in the GDP and ppG2Ј:3Јp complexes from 10 to 64 Å 2 and 10 to 65 Å 2 , respectively. The variation in thermal parameters as a function of residue is comparable to that of other structures of the ligated synthetase (5,13). Thermal parameters for individual ligands average to less than 30 Å 2 in each of the structures. Supperposition of the GDP complex onto the ppG2Ј:3Јp complex results in a root mean square deviation of C␣ coordinates of 0.3 Å, comparable to the estimate of coordinate uncertainty.

Quality of the Refined Models-The method of Luzzati
GDP Complex (Protein Data Base Accession Code 1cib)-Conditions of crystallization for the GDP complex reported here differ from those of an earlier study (13) in the elimination of acetate from the crystallization buffers. Acetate and GDP synergistically inhibit the synthetase, as does nitrate and GDP (25). Although the combination of GDP/nitrate exhibits greater synergism than GDP/acetate, the total concentration of acetate in the previous study was some 40-fold higher relative to the nitrate anion, hence the potential exists for substantial compe- where i runs over multiple observations of the same intensity, and j runs over all crystallographically unique intensities.
b All data are in the resolution ranges indicated. c Includes hydrogens linked to polar atoms. d R-factor ϭ ͚ʈF obs ͉ Ϫ ͉F calc ʈ/͚͉F obs ͉, ͉F obs ͉ Ͼ 0. e R free based upon 10% of the data randomly culled and not used in the refinement.
tition between nitrate and acetate for the putative ␥-phosphoryl pocket of GTP. The GDP complex presented here provides data to higher resolution relative to that of the previous GDP complex (2.3 as opposed to 2.8 Å), and more significantly, the active site in the vicinity of the nitrate anion has stronger electron density and lower thermal parameters. The average distance of the six oxygen atoms nearest to the Mg 2ϩ (Table II) in the present GDP complex has decreased by 0.16 Å relative to the former complex. Most significantly the distance between Mg 2ϩ and OD1 of Asp 13 has fallen from over 3.0 Å in the GDP complex from acetate buffer to 2.7 Å in the GDP complex without acetate. In the acetate-free buffer, the synthetasebound Mg 2ϩ exhibits quasi-octahedral coordination, rather than the square pyramidal coordination of the ligated complex in acetate buffer. In addition His 41 now interacts through a bridging water molecule with Glu 221 , whereas in the acetate buffer no water molecule is evident, and His 41 exhibits an elevated set of thermal parameters. The elimination of acetate, then, results in the tighter coordination of Mg 2ϩ and a more ordered active site in the immediate vicinity of the ␥-phosphoryl pocket. As the ppG2Ј:3Јp complex described below comes from an acetate-free buffer, the GDP complex without acetate reported here serves as a basis of comparison. ppG2Ј:3Јp Complex (Protein Data Base Accession Code 1ch8)-The conformation of the synthetase in the ppG2Ј:3Јp complex is essentially identical to that of the GDP complex ( Fig. 3 and Table II). Interactions of IMP, nitrate, and hadacidin are unchanged from previous structures. Significant differences occur, however, in the ligation of Mg 2ϩ and at the guanine nucleotide pocket. Within the uncertainty of the coordinates, Mg 2ϩ exhibits octahedral coordination, in contrast to the GDP complex, where it exhibits quasi-octahedral coordination. The oxygen atoms that define the equatorial plane of the Mg 2ϩ (one each from 5Ј-␣-and 5Ј-␤-phosphoryl groups of ppG2Ј:3Јp, from NO 3 Ϫ , and from the N-formyl group of hadacidin) average to 2.1 Å, as opposed to 2.3 Å for the corresponding bonds in the GDP complex (Table II). In the GDP complex, OD1 of Asp 13 tightly hydrogen bonds (2.6 Å) to N-1 of IMP and is 2.7 Å from the Mg 2ϩ . In the ppG2Ј:3Јp complex, however, the distance between OD1 of Asp 13 and N-1 of IMP is 3.1 Å,  whereas the distance separating the Mg 2ϩ and OD1 of Asp 13 is 2.5 Å. Evidently, the synthetase coordinates Mg 2ϩ more tightly in the ppG2Ј:3Јp complex than in the GDP complex; the tighter coordination occurring, however, at the expense of a weakened interaction between Asp 13 and N-1 of IMP. The significance of the above to the mechanism of catalysis and inhibition of the synthetase is discussed below. The difference electron density in the guanine nucleotide pocket clearly indicates an analog of GDP (Fig. 4). Although electron density similar to that of the GDP complex is present for the base and 5Ј-pyrophosphoryl moieties, additional electron density extends from both the 2Ј-and 3Ј-hydroxyl groups of the stringent effector. The density can accommodate a single phosphoryl group but not the pyrophosphoryl group expected at the 3Ј position. A difference map, based on observed data from the effector and GDP complexes, and calculated phase angles from the GDP complex, reveals a strong and well defined peak of electron density in the vicinity of the 2Ј-and 3Ј-hydroxyl groups of GDP (Fig. 4). The positive difference density overwhelms any negative difference density due to water molecules bound to the 2Ј-and 3Ј-hydroxyls of GDP. Hence the bound nucleotide cannot be GDP. Furthermore, a 3Ј (or 2Ј)-phosphoryl group cannot occupy the density without causing severe steric conflict with the remaining 2Ј (or 3Ј)-hydroxyl group of the ribose. The only acceptable fit to the electron density was provided by ppG2Ј:3Јp. Thermal parameters for the refined cyclic nucleotide are comparable to those of the surrounding protein, suggesting full occupancy of the ligand at the guanosine pocket.
The synthetase recognizes the 5Ј-pyrophosphoryl group and the guanine base of GDP and ppG2Ј:3Јp in the same manner. Atom O-6 of the cyclic nucleotide interacts with OG of Ser 414 and backbone amide 331. Endocyclic N-1 and exocyclic N-2 of the base hydrogen bond with the side chain of Asp 333 . The 5Ј-␤-phosphoryl group of ppG2Ј:3Јp interacts with backbone amides 15, 16, and 17, the side chain of His 41 , and bound Mg 2ϩ . The 5Ј-␣-phosphoryl group interacts with the Mg 2ϩ , backbone amide 42, and Arg 305 . The ribose moiety of ppG2Ј:3Јp, however, differs significantly in location and conformation relative to that of GDP (Fig. 4). The 5Ј-␣-phosphoryl group of ppG2Ј:3Јp rotates  Table II) are presented with dashed lines .   FIG. 4. Stereoview of electron density associated with ligands in the ppG2:3p complex. A, the electron density, contoured at 6 using a cut-off radius of 0.9 Å, comes from a ligand-excised omit map. ppG2Ј:3Јp and other active site ligands are represented with bold lines. B, difference electron density based on phases from the GDP complex and Fourier coefficients from the ppG2Ј:3Јp and GDP complexes. The contour level is 6 and reveals only residual electron density in the ppG2Ј:3Јp complex that has no counterpart in the GDP complex. The GDP molecule is represented with bold lines and the ppG2Ј:3Јp molecule with fine lines. torsion angle O-4Ј-C-1-N-9 -C-4 ( by convention) is Ϫ118 o (anti), both comparable to those of the GDP complex (␥ ϭ Ϫ66 o and ϭ Ϫ82 o ). The five-member ring (O-2Ј-P-1-O-3Ј-C-3Ј-C-2Ј) is nearly planar, and its angles, bond lengths, and ribose ring pucker (C-2Ј-endo and C-3Ј-exo) are comparable to those of cyclic nucleotides, observed in high resolution crystal structures (19,20).
Thin Layer Chromatography-ppG2Ј:3Јp is putatively an intermediate in the degradation of ppGpp (26). ppGpp is acidand alkali-labile but relatively stable at neutral pH. The stability of ppGpp was monitored by thin layer chromatography in solutions of composition indicated under "Materials and Methods." No detectable hydrolysis product of ppGpp appeared up to 36 h in aliquots without adenylosuccinate synthetase. Nor did a hydrolysis product appear under the same conditions in the presence of the synthetase at a concentration of 2 mg/ml. The minimum concentration of guanine nucleotide detectable by the chromatographic protocol is approximately 80 M. Hence hydrolysis products of ppGpp could be present in the original sample or any of the solutions used to monitor the hydrolysis of ppGpp at a level of up to 4 parts per 100 and still escape detection.

DISCUSSION
Under the conditions of our crystallographic studies ppGpp does not undergo significant hydrolysis. Two possibilities, then, can account for the appearance of ppG2Ј:3Јp in the guanine pocket of the synthetase: either the crystalline synthetase scavenges a minor (and undetectable) impurity of the cyclic nucleotide from our sample of ppGpp or the synthetase itself facilitates the conversion of ppGpp to the cyclic nucleotide. As no build-up of the cyclic nucleotide was observed in the presence of the synthetase, the putative enzyme-mediated process must occur without appreciable turnover. For the latter scenario, ppGpp must bind to the guanine pocket and undergo cyclization, producing an inhibitor that does not readily dissociate from the synthetase.
As a means of better understanding a possible enzyme-mediated mechanism, one can model a transition state complex for the cyclization of ppGpp in the guanine nucleotide pocket (Fig.  5). Mg 2ϩ (total concentration of 100 mM) must be associated with the 3Ј-pyrophosphoryl group of ppGpp under the conditions of crystal soaks. The water molecule, which bridges backbone carbonyls 42 and 417 and hydrogen bonds with the 2Ј-hydroxyl of GDP (Table II), may act as a weak catalytic base and abstract a proton from the 2Ј-OH of ribose (Fig. 5). The activated O-2Ј atom then attacks the 3Ј-␣-phosphoryl group, displacing the 3Ј-␤-phosphoryl group. The side chain of Lys 49 , as well as the Mg 2ϩ associated with the 3Ј-pyrophosphoryl group, would stabilize the development of negative charge on the leaving group and/or the transition state. Lysyl side chains in analogous positions are critical to the catalytic function of adenylate kinase and GTPases in general.
Regardless of its mode of formation, ppG2Ј:3Јp must bind at least 200-fold more tightly than does ppGpp to the active site of the synthetase, assuming the cyclic nucleotide is 4% of the concentration of ppGpp in soaking solutions and that the K i for ppGpp is 50 M. The estimated K i for ppG2Ј:3Јp (ϳ10 Ϫ7 M) compares favorably to the tightest known inhibitors of the synthetase, alanosyl-AICOR (ϳ10 Ϫ7 M (27)) and hydantocidin 5Ј-phosphate (ϳ10 Ϫ8 M (28)). The enhanced binding of ppG2Ј: 3Јp may arise from the following three sources. (i) Cyclization of the 2Ј-and 3Ј-hydroxyls of the ribose reduces conformational freedom of the nucleotide and hence reduces the entropic penalty associated with ligand binding. (ii) The 2Ј:3Ј-phosphoryl group provides additional opportunities for strong hydrogen bonds with the synthetase. (iii) The active site Mg 2ϩ has octahedral coordination, whereas in GDP complexes its coordination is square pyramidal (13) or quasi-octahedral (see above). In fact, an octahedrally coordinated metal cation appears as well in the complex of the tight binding inhibitor hydantocidin 5Ј-phosphate (29). Asp 13 interacts directly with IMP in GDP complexes (see above) and evidently in the absence of an IMP-Asp 13 interaction (as in the hydantocidin 5Ј-phosphate complex (29)) or a weakened IMP-Asp 13 interaction (as here in the ppG2Ј:3Јp complex), Asp 13 favors direct coordination to the Mg 2ϩ . Octahedral coordination of the active site Mg 2ϩ , then, may be a hallmark of tight binding inhibitors of the synthetase. Such inhibitors are of potential significance in chemotherapy and the development of herbicides (27,28).
A cyclic nucleotide at the guanine nucleotide pocket does not explain competitive inhibition between ppGpp and IMP and noncompetitive inhibition between ppGpp and GTP, as observed in two investigations (6,7). Such phenomena could arise from impure preparations of the synthetase (8). Alternatively, ppGpp could bind to a site distinct from the guanine nucleotide pocket. Synthetase dimers, for instance, putatively bind IMP with an affinity 100-fold greater than that of synthetase monomers. Hence, ppGpp will inhibit competitively with respect to IMP and non-competitively with respect to GTP, by binding to a site that blocks dimerization of the synthetase. The existence of such a site is inferred by covalent modification of the synthetase by pyridoxal phosphate (30) and guanosine-5Ј-O-[S-(4bromo-2,3-dioxobutyl)thiophosphate) (31) at a cluster of basic residues, exposed in synthetase monomers, but buried between monomers in the synthetase dimer. The experimental protocol developed here reveals ppGpp interactions at the guanine nucleotide pocket (consistent with all kinetics investigations (6 -8)) but does not probe for a ppGpp-binding site at the buried interface between subunits of the synthetase dimer.
Potent inhibition of adenylosuccinate synthetase in vivo typically impedes growth of the organism. The pur A Ϫ , E. coli cell line grows poorly, even in rich medium. Hydantocidin (which is transformed in plants to the tight binding synthetase inhibitor, hydantocidin 5Ј-phosphate) is a potent herbicide (27). Alanosyl-AICOR and hadacidin are antibiotics (9,28), the latter having only the synthetase as a known target. Hence, if the stringent response is ultimately a coordinated attempt by E. coli to limit its own growth under unfavorable conditions, then the synthetase is an appropriate target. Even though ppGpp is an effective inhibitor of the synthetase at its physiological concentration during the stringent response, other effectors (pppGpp and ppG2Ј:3Јp) may also influence synthetase activity. The transformation of ppGpp into ppG2Ј:3Јp, perhaps at the active site of the synthetase, suggests a parallel and more potent mechanism of inhibition, which may be temporally distinct from the inhibition by ppGpp. pppGpp, putatively the first nucleotide generated during the stringent response, could be an inhibitor of the synthetase, its 5Ј-␥-phosphoryl group being transferred to IMP while the cyclization reaction occurs at the guanine pocket. The result would be an active site complex of 6-phosphoryl-IMP and ppG2Ј:3Јp. The combination of the synthetase with GTP and 6-thio-IMP has resulted in the formation of a stable complex of 6-thiophosphoryl-IMP at the active site (5). Furthermore, the combination of IMP and GTP leads to the analogous 6-phosphoryl-IMP complex (32). Hence, the double chemical transformation of pppGpp at the active site of the synthetase is well within the realm of possibility. Alternatively, if pppGpp is a simple substrate of the synthetase, then in vivo inhibition would depend on the relative concentrations and affinity constants of pppGpp and ppGpp, as well as the concentration of the cyclic effector. Hence the inhibition of the E. coli synthetase by stringent effectors may be far more complex than the competition of ppGpp for the active site.