Structural and biochemical analysis of Escherichia coli ObgE, a central regulator of bacterial persistence

The Obg protein family belongs to the TRAFAC (translation factor) class of P-loop GTPases and is conserved from bacteria to eukaryotes. Essential roles in many different cellular processes have been suggested for the Obg protein from Escherichia coli (ObgE), and we recently showed that it is a central regulator of bacterial persistence. Here, we report the first crystal structure of ObgE at 1.85-Å resolution in the GDP-bound state, showing the characteristic N-terminal domain and a central G domain that are common to all Obg proteins. ObgE also contains an intrinsically disordered C-terminal domain, and we show here that this domain specifically contributed to GTP binding, whereas it did not influence GDP binding or GTP hydrolysis. Biophysical analysis, using small angle X-ray scattering and multi-angle light scattering experiments, revealed that ObgE is a monomer in solution, regardless of the bound nucleotide. In contrast to recent suggestions, our biochemical analyses further indicate that ObgE is neither activated by K+ ions nor by homodimerization. However, the ObgE GTPase activity was stimulated upon binding to the ribosome, confirming the ribosome-dependent GTPase activity of the Obg family. Combined, our data represent an important step toward further unraveling the detailed molecular mechanism of ObgE, which might pave the way to further studies into how this GTPase regulates bacterial physiology, including persistence.

Guanine nucleotide-binding proteins (GNBPs, G proteins, or GTPases) are ubiquitously found in all living organisms where they are known to play essential roles in a myriad of cellular processes, including protein synthesis and translocation, membrane trafficking, development, signal transduction, and cell cycle control (1)(2)(3). G proteins share a common structural module, the G domain that acts as a molecular switch by cycling between a GDP-bound "off" state and a GTP-bound "on" state. Both states mainly differ in the conformation of two regions (Switch I and II), allowing reversible interaction and activation of downstream effector proteins (4,5).
In the intensively studied small GTPases of the Ras family, nucleotide exchange is very slow and exchange of GDP for GTP is stimulated by guanine nucleotide exchange factors (6,7). Moreover, the intrinsically slow GTP hydrolysis is accelerated through the interaction with GTPase activating proteins (GAPs) 3 (6,8). The latter proteins stabilize and/or complement the active site of the GTPase by providing one or two catalytic residues in trans that counter negative charge development at the phosphate groups of GTP or orient the nucleophilic water molecule during hydrolysis (4,6,8). More recently a number of other activation mechanisms of G proteins have been described. One class of G proteins is activated through homoor heterodimerization of their G domains, hence reciprocally activating each other (9 -11). This class of G proteins is known as the G proteins activated by nucleotide-dependent dimerization (12). Another expanding family of G proteins recruits specific cations (such as K ϩ ) to stabilize the GTPase transition state (13)(14)(15)(16). Some G proteins use a combination of the two above strategies, as is the case for MnmE, a G protein involved in tRNA modification (17).
The Obg family belongs to the TRAFAC (for translation factor) class of P-loop GTPases and consists of high molecular weight proteins like Obg/CgtA, YchF/YyaF, Drg/Rbg, Nog1, and Ygr210 (3). Among them, Obg is a highly versatile GTPase that is conserved in bacteria, as well as in many eukaryotes (18). Obg displays a three-domain arrangement with an N-terminal glycine-rich domain typical for Obg (Obg domain), a central G domain and a C-terminal domain that is highly variable in length and sequence among Obg proteins from different species (19,20). Obg homologs are widely present in eukaryotic organelles, including chloroplasts and mitochondria (21)(22)(23). However, in contrast to bacterial Obg, the rice homolog OsYchF and the human homolog hOLA1 (human Obg-like ATPase 1) were found to bind and hydrolyze ATP more efficiently than GTP (24,25). Obg displays a very low GTP hydrolysis rate in combination with a high guanine nucleotide exchange rate, seemingly abrogating the need for guanine nucleotide exchange factors (3,26). No "classical" Obg-specific GAP proteins have so far been described, although the 50S ribosomal subunit has recently been found to increase the GTPase activity of Obg (27). Furthermore, it has been proposed, based on sequence analysis, that Obg might belong to the class of G proteins that is specifically activated by K ϩ ions (15). Obg was originally discovered in Bacillus subtilis, where it was shown to have a critical role in sporulation (28). Moreover, Obg expression was found to be essential for viability in nearly all bacterial species (3,26), and various functions including ribosome assembly and maturation, cell cycle control, DNA replication, stress response, sporulation and morphological development have been proposed (29 -34). The role of Obg in the cellular response to environmental stress is particularly interesting (27,33,(35)(36)(37). Recently, we showed that, in Escherichia coli as well as in Pseudomonas aeruginosa, Obg plays a central role in the regulation of bacterial persistence in response to nutrient starvation (38). Although the underlying mechanisms need further elucidation, Obg-mediated persistence depends on the presence of the alarmone (p)ppGpp and proceeds through induction of the expression of the HokB toxin.
Here, we report the first crystal structure of Obg from E. coli (ObgE) bound to a GDP molecule at 1.85-Å resolution. SAXS and MALS experiments reveal that ObgE behaves as a monomer in solution. A detailed biochemical analysis confirms that ObgE has a modest affinity for nucleotides and ppGpp and fast guanine nucleotide dissociation rates, combined with a low GTP hydrolysis rate. Interestingly, we show that the C-terminal intrinsically disordered domain plays a role in GTP binding, whereas deletion of this region has a negligible effect on GDP and ppGpp binding and GTP turnover. Finally, kinetic experiments show that ObgE is neither activated by K ϩ ions, nor by homodimerization, in contrast to what was previously suggested (15,20). However, we do observe a weak but significant stimulation of the ObgE GTPase activity by the (70S) ribosome. Together, these data provide new insights into the structure and function of ObgE and will be instrumental to drive further studies aiming to unravel the role of Obg in bacterial persistence.

Results and discussion
Overall structure of GDP-bound ObgE E. coli Obg (ObgE) is a 390-amino acid protein that is predicted to consist of a conserved N-terminal glycine-rich Obg domain (amino acids 1-157) followed by a central Ras-like G domain (amino acids 158 -340), in analogy to the previously reported crystal structures of Obg from B. subtilis and Thermus thermophilus (19,20). In most Obg proteins, these two domains are followed by a third domain, which is much less conserved. Unlike the proteins from B. subtilis and T. thermophilus, in ObgE this domain consists of a stretch of amino acids (341-390) that are composed for 60% of charged amino acids and are predicted to be intrinsically disordered (Fig. 1a) (39,40). So far, the function of the latter domain remains largely enigmatic.
To allow a straightforward interpretation of biochemical and biological data on E. coli ObgE we set out to solve its crystal structure. However, all attempts to crystallize the full-length protein (ObgE_FL) have failed so far, probably due to the presence of the intrinsically disordered C terminus. Therefore, we decided to generate a C-terminal deletion construct (ObgE_340) lacking the last 50 amino acids, by replacing the codon for amino acid Glu-341 by a stop codon. This protein was crystallized successfully in the presence of Mg 2ϩ ions and GDP in space group C222 1 with one molecule in the asymmetric unit, allowing to solve its crystal structure at a resolution of 1.85 Å by molecular replacement. We were able to confidently build 334 of the 340 amino acids in the electron density map. Residues 133-136, corresponding to a loop region in the Obg domain and C-terminal residues 339 -340, were not built due to weak electron density. The crystal structure clearly shows density for a molecule of GDP and a Mg 2ϩ ion in the G domain. The final model has a crystallographic R factor of 0.198 and a free R factor of 0.239 (Table 1).
The crystal structure of ObgE_340 shows the typical arrangement of an N-terminal "Obg domain" followed by a Ras-like G domain (Fig. 1b). Similar to B. subtilis and T. thermophilus Obg, two regions can be discerned in the Obg domain: a region with six left-handed type II helices connected on one side by long loops, and an eight-stranded ␤-barrel containing an ␣-helix between the second and third strand (19,20). This ␤-barrel makes extensive contacts with the G domain. The largest differences in this domain comparing ObgE to the proteins from B. subtilis and T. thermophilus are present in two of the three long loops connecting each pair of type II helices.
The G domain of ObgE displays a Ras-like fold, consisting of a six-stranded ␤-sheet and five ␣-helices. Superposition of the G domain of ObgE on the corresponding domains of T. thermophilus and B. subtilis Obg (supplemental Fig. S1) shows that the biggest differences in conformation are localized in the Switch I and II regions. In the GDP-bound structure of ObgE, both Switch loops could be fully traced, although they are in an "open" conformation where they do not interact with the nucleotide. However, interactions are made between both Switch regions and the N-terminal Obg domain. The Switch II region is folded in two helices, classically termed ␣ 1 Ј and ␣ 2 Ј. The conformation of the Switch loops of GDP-bound ObgE most closely resembles the conformation of these loops in the nucleotide-free protomer of the B. subtilis Obg crystal structure, whereas especially Switch II adopted a different conformation in the T. thermophilus crystal structure, mainly due to partial unwinding of helix ␣ 1 Ј. In the ppGpp-bound protomer of B. subtilis Obg the Switch regions could not be fully traced (19,20).
Crystal structures of Obg available in the PDB are either bound to ppGpp (B. subtilis Obg) or show Obg in the apo state (B. subtilis and T. thermophilus Obg). Here we report the first high resolution structure of ObgE bound to GDP. GDP is anchored tightly to the G domain of ObgE_340 via interactions with residues from the P-loop and the G4 and G5 motif (Fig. 1,  c and d). The ␣and ␤-phosphates are bound by the P-loop, with the ␣-phosphate interacting with Thr-174 (side chain and main chain) and the ␤-phosphate with Asn-169 (main chain), Gly-171 (main chain), Lys-172 (side chain and main chain), and Ser-173 (side chain and main chain). Specificity for the guanine nucleotide is provided by the G4 motif via interactions with Asn-283 (with the N7 of guanine) and Asp-286 (interactions with the N1 and N6 of guanine), and by the G5 motif via interactions of the guanine O6 carbonyl with Ser-314, Ala-315, and Ala-316 (main chain).

ObgE is a monomer in solution
In the crystal structure, the G domain of ObgE_340 interacts with the G domain of the crystal symmetry neighbor. This interface buries a surface area of 836 Å 2 and is mainly formed by interactions between residues of the P-loop and the first ␣-helix (6 residues), Switch I (10 residues), and Switch II (7 residues), where Switch I of one protomer interacts with Switch II of the adjacent protomer and vice versa ( Fig. 2a and supplemental Fig.  S2). This orientation places one arginine residue (Arg-177) of the first ␣-helix as well as one lysine residue (Lys-183) of Switch I into the GDP/GTP binding pocket of the neighboring protomer (Fig. 2b). The orientation of the amino group of Lys-183 is reminiscent of the position of the guanido group of the catalytic "arginine finger" (Arg-789) in the Ras-RasGAP complex (Fig. 2c, see below) (41). This observation raises the intriguing question whether ObgE could form a dimer in solution, where two adjacent G domains could reciprocally enhance the GTPase activity of each other. Interestingly, a functional dimerization was previously also suggested for the Obg protein from T. thermophilus, where the C-terminal domain of one protomer was found to interact with the G domain of the adjacent protomer (20).
To determine the oligomeric state of ObgE_340 and ObgE_FL in solution, first small angle X-ray scattering experiments coupled to size exclusion chromatography (SEC-SAXS) were performed using the proteins in a nucleotide-free form (Fig. 3). The SEC-SAXS profile of ObgE_340 shows a symmetric peak with a constant radius of gyration (R g ) across the elution peak (supplemental Fig. S3a). An average R g (from Guinier analysis) and maximal intramolecular distance (D max ) of 30.9 and 108 Å, respectively, is obtained that is very close to the R g and D max values calculated from the crystal structure of an ObgE_340 monomer (R g ϭ 29 Å, D max ϭ 110 Å). Furthermore, the molecular mass obtained from the Porod volume is 41 kDa, whereas the molecular mass derived from the Porod invariant (Q R ) is 36 kDa (42) (supplemental Table S1). These molecular mass estimations match very well with the theoretical monomer molecular mass of 39.1 kDa calculated from the protein sequence, indicating that ObgE_340 mainly adopts a monomeric form in solution. Subsequently, SAXS was also used to determine the molecular mass of ObgE_FL ( Fig. 3 and supplemental Fig. S4a). The SAXS data of ObgE_FL in a nucleotidefree state yield R g (from Guinier analysis) and D max values of 37.0 and 160 Å, respectively, whereas the molecular mass estimation from the Porod volume gives a value of 60 kDa (supplemental Table S1). The latter value is higher than the expected value for an ObgE_FL monomer (45.4 kDa). The overestimation of the molecular mass using the Porod volume and the large values of R g and D max are, however, in agreement with the presence of an intrinsically disordered C-terminal domain that occupies a large volume in solution (see further). Correspondingly, calculation of the molecular mass using the Porod invariant (Q R ) gives a molecular mass of 44.8 kDa, which is much closer to the theoretical molecular mass of an ObgE_FL monomer. Together these data show that ObgE in its nucleotide-free state behaves as a monomer in solution.
To further test whether dimerization could potentially occur upon nucleotide binding, we also collected SAXS data of ObgE_340 and ObgE_FL in the presence of an excess of GDP or the nucleotide triphosphate analogue Gpp(NH)p (supplemental Figs Table S1). Finally, the molecular mass of ObgE_FL (apo), ObgE_FL in the presence of Gpp(NH)p, and ObgE_FL in the presence of the transition state analogue GDP-AlF x was further validated using SEC-MALS. SEC-MALS gave a monodisperse peak with apparent molecular mass in the range of 49 -51 kDa (supplemental Fig. S7), again close to the value expected for a monomer (45.4 kDa). Together, we can conclude from these experiments that ObgE behaves mainly as a monomer in solution, regardless of concentration and nucleotide state. This is a strong indication that the dimeric form observed in the crystal structure, generated from crystallographic symmetry operations, is a crystallographic artifact. However, at this point we cannot completely exclude that very transient dimer interactions do occur (see below).

Interdomain orientation and influence of nucleotides on the conformation of ObgE
Although the individual Obg and G domains of the Obg crystal structures from E. coli (this study), B. subtilis (19), and T. thermophilus (20) superimpose very well (see above), the relative orientation of these domains in the crystal structures differs. Indeed, it had been noted before that upon superimposing the G domains of B. subtilis (nucleotide-free or ppGpp bound) and T. thermophilus (nucleotide-free) Obg, the respective Obg domains are rotated by about 180°around the G domain axis (20). Moreover, in a low resolution (5.5 Å) EM structure of E. coli Obg, bound to the 50S ribosomal subunit, an orientation of the Obg domain intermediate to that observed in B. subtilis Obg and T. thermophilus Obg was observed (27). A recent molecular dynamics study suggested that the relative orientation of the Obg domain, vis à vis the G domain changes depending on the nucleotide state of the protein, with the largest difference occurring in the GDP-bound state compared with the apo-and GTP-bound states (43). However, in the current high resolution structure of ObgE_340 bound to GDP, the orientation of the Obg domain with respect to the G domain is very similar to the B. subtilis Obg structure, either in the apo state or bound to ppGpp (supplemental Fig. S1).
To further investigate the influence of nucleotides on the conformation of E. coli ObgE_340 and ObgE_FL in solution we again turned to SAXS. To test whether our crystal structure

PDB code 5M04
a Values for the highest-resolution shell are given in between brackets. CC 1/2 values were used as a guide for selecting the highest usable resolution shell (67). b A subset of 5% of the reflections was used for calculating R free .
corresponds to a main conformation in solution, we first compared the theoretical scatter curve of the ObgE_340 crystal structure, after flexible modeling of the N-terminal purification tag, with the experimental scatter curve (supplemental Fig. S8). Because both curves overlay very well, we can conclude that the crystal structure indeed represents a relevant conformation in solution. Next, we compared the scattering curves of both ObgE_340 and ObgE_FL bound to GDP or Gpp(NH)p to the corresponding curve of the nucleotide-free protein (supplemental Figs. S5 and S6). This superposition does not reveal any significant nucleotide-dependent changes in the scattering curves, although small differences are observable when comparing the pair-distance distribution functions and dimensionless Kratky plots of ObgE_FL in different nucleotide states (supplemental Figs. S5 and S6). These data seem to indicate that no large-scale nucleotide-induced conformational changes are taking place in E. coli ObgE. However, alternatively, it remains possible that such changes only take place in the presence of certain partner proteins, such as the ribosome (27), or that the suggested rotation of the Obg domain vis à vis the G-domain does not lead to an observable difference in the low-resolution spherically averaged SAXS profiles. High-resolution crystal structures of ObgE in different nucleotide-bound states could shed further light on this issue.

Conformation and role of the C-terminal domain
The last 50 C-terminal amino acids of ObgE_FL are mainly composed of charged amino acids (about 60%) and are predicted to be intrinsically disordered (39,40). The flexible disordered nature of this peptide region is also in agreement with our SAXS data for ObgE_FL (see above), which show that (i) the calculated Porod volume is at least 15% higher than would be expected for a compact monomeric protein, and (ii) the normalized Kratky plot has a broader peak and converges at higher angle than the corresponding plot for ObgE_340 (Fig. 3c). Correspondingly, we did not succeed in crystallizing the full-length ObgE protein. Consequently, so far no information regarding the structure and function of this C-terminal domain is available. To get an idea of the conformational space adopted by the C-terminal domain in solution, we used the experimental SAXS data of the full-length ObgE, in combination with our partial crystal structure and the amino acid sequence, to perform ensemble modeling. The conformation of the last 50 residues was probed using all-atom modeling, followed by model validation using the Molprobity score and further refinement using iterative normal mode analysis (NMA). Model selection based on the experimental SAXS data indicated that the latter could be accounted for by an ensemble of 5 models as depicted in Fig.  4. However, we would like to stress that the presented ensemble only gives a representative subset of possible orientation of the C-terminal domain in solution that is consistent with the experimental scattering data.
To investigate the function of this intrinsically disordered C-terminal domain we determined its contribution to nucleotide binding and hydrolysis by comparing the full-length protein with the C-terminal deletion construct ObgE_340. First, we determined its contribution to nucleotide binding affinity and kinetics using stopped flow fluorescence experiments (using mant-labeled nucleotides). In agreement with previous reports (35), we find that ObgE_FL displays a relatively low affinity for GDP (K D ϭ 0.18 Ϯ 0.01 M) and GTP (K D ϭ 0.37 Ϯ 0.16 M) due to fast nucleotide binding and dissociation (see Fig. 5 for all data). Deletion of the C-terminal domain (ObgE_340) leads to a small decrease in affinity for GDP (K D ϭ 0.46 Ϯ 0.02 M), whereas the affinity for GTP is decreased more than 10-fold (K D ϭ 4.21 Ϯ 0.46 M). This decrease in affinity is nearly entirely due to an increased GTP dissociation rate (k off -(GTP, ObgE_FL) ϭ 0.44 Ϯ 0.19; k off (GTP, ObgE_340) ϭ 3.67 Ϯ 0.31; Fig. 5).
Subsequently, considering that ppGpp is required for ObgEmediated persistence (38), we also determined the affinity of ObgE_FL and ObgE_340 for ppGpp using ITC experiments ( Fig. 6 and supplemental Fig. S9). ObgE_FL binds ppGpp with a K D ϭ 0.66 Ϯ 0.03 M. Deletion of the C terminus leads to a very small decrease in affinity for ppGpp (K D ϭ 0.78 Ϯ 0.07 M), in agreement with the observed small effect of the C-terminal domain on GDP binding.
Finally, we also assessed the influence of the C-terminal domain on GTP turnover. To this end we measured Michaelis-Menten kinetics. ObgE_FL has a k cat ϭ 0.064 Ϯ 0.001 min Ϫ1 and a K m ϭ 10.40 Ϯ 0.70 M, compared with a k cat ϭ 0.046 Ϯ 0.001 min Ϫ1 and a K m ϭ 11.07 Ϯ 0.92 M for ObgE_340 (Table  2 and supplemental Fig. S10). This indicates that although the C terminus is involved in GTP binding, it is not substantially involved in GTP turnover.
In conclusion, we show here that the intrinsically disordered C-terminal domain of E. coli ObgE specifically contributes to GTP binding with very little effect on GDP or ppGpp binding. Such a discriminatory effect of the C-terminal domain on GTP versus GDP binding seems to indicate that this domain transiently folds back on the G domain in the GTP-bound state. Such a conformational change of the C-terminal domain is, however, not reflected in large changes in our SAXS profiles (see above), and further research is required to reveal the underlying mechanisms.

Testing potential GTPase-activating mechanisms of ObgE
ObgE, and Obg proteins in general, show a very low intrinsic GTPase activity (k cat (ObgE) ϭ 0.064 min Ϫ1 , Table 2). Classically, the rate of GTP hydrolysis of small Ras-like GTPases is    ). b, the R g distribution of the ensemble (red curve) compared with the R g distribution of a random pool of models (gray filled area) generated using an ensemble optimization method coupled to all-atom modeling, model validation, and NMA refinement. c, ribbon representation of 5 ensemble models, selected by the genetic algorithm, which give an average theoretical curve that fits the experimental ObgE_FL SAXS profile as shown in a. enhanced by dedicated GAPs (6,8). However, so far no such GAPs for Obg have been reported, although it was recently described that the GTPase activity of Obg is increased upon binding to the ribosome (27,43). Moreover, recently a number of alternative mechanisms to increase the GTPase activity of G proteins have been reported, including binding of specific ions and reciprocal complementation of active sites by homodimerization (9 -16).
With respect to a potential mechanism of GTPase activation though homodimerization, we were particularly triggered by the observation of a potential homodimer in the ObgE_340 crystal structure, formed through crystal symmetry (see above). This dimeric organization would place both an arginine (Arg-177) and a lysine residue (Lys-183) coming from, respectively, the first ␣-helix and the Switch I region of one subunit into the GTP binding pocket of an adjacent subunit (Fig. 2b). Despite our observation that ObgE_340 is mainly monomeric in solution (see above), it can a priori not be excluded that a transient dimeric complex could be formed, where Lys-183 and Arg-177 would function as a catalytic lysine or arginine "finger" by electrostatically stabilizing the GTPase transition state in the adjacent subunit. Such a catalytic role of Lys-183 and/or Arg-177 would be reminiscent of the catalytic arginine finger (Arg-789) used by the RasGAP protein for activation of the Ras GTPase activity (41). Even more intriguingly in this respect we find that superposition of the G domain of ObgE_340 onto Ras in the Ras-RasGAP complex (41) places the amino group of Lys-183 from the adjacent ObgE subunit nearly perfectly on the guanidino group of the catalytic arginine finger of the RasGAP protein (Fig. 2c). To test whether Lys-183 or Arg-177 contribute to GTP hydrolysis, we mutated both residues to alanine in ObgE_FL, and compared the steady state kinetic parameters to the wild type protein ( Table 2 and supplemental Fig. S10). These data show that the K183A mutation lowers k cat by a factor of 3, whereas the K m value is nearly unaffected (k cat ϭ 0.020 Ϯ 0.000 min Ϫ1 ; K m ϭ 10.27 Ϯ 0.75 M). On the other hand, the R177A mutant has unaffected k cat and K m values (k cat ϭ 0.060 Ϯ 0.002 min Ϫ1 ; K m ϭ 6.63 Ϯ 0.69 M). We thus conclude that, although the K183A mutation has a somewhat lowered k cat value, the observed effects are too small to account for a genuine catalytic finger.
Although many small GTPases use an arginine finger from a GAP protein to stabilize the transition state of GTP hydrolysis, other G proteins including MnmE, YqeH, FeoB, RbgA, and EngA bind a K ϩ ion in the active site to neutralize the negative charge of the transition state (13, 14, 44 -46). Based on sequence analysis and the presence of an asparagine residue in the so-called K-loop of the Switch I region, Obg proteins were recently proposed to be potential members of these so-called potassium-selective cation-dependent GTPases (15). However, so far this hypothesis has not been experimentally validated. To investigate the effect of K ϩ ions on the kinetic constants of ObgE, the Michaelis-Menten kinetic parameters of ObgE_FL were determined in the presence of 150 mM KCl and compared with those in the presence of 150 mM NaCl (Table 2 and supplemental Fig. S10). A small (less than 2-fold) decrease in k cat was observed in the presence of KCl compared with NaCl.  Therefore, we can rule out a mechanism whereby K ϩ acts as a GTPase activating factor. Because we did not find any activation via dimerization or K ϩ binding, we also reinvestigated the reported effect of the ribosome. Indeed, it was recently shown that a 1:1 complex of ObgE and the 50S ribosomal particle stimulates the ObgE GTPase activity about 120-fold (27). In another study it was shown that the GTPase activity of Vibrio cholerae Obg (or CgtA) was also stimulated up to 5-fold by catalytic amounts of 50S or 70S ribosomes (43). Hereto we measured the ObgE_FL activity upon addition of increasing amounts of ribosome. However, at the highest ribosome concentrations we found a significant GTPase activity in the ribosome preparation, most likely due to GTPases that bind and co-purify with the ribosome. After subtracting this background activity we found a gradual increase in the GTPase activity of ObgE with increasing (catalytic) amounts of ribosome (Fig. 7). Addition of 0.1 M 70S ribosome to 0.5 M ObgE increases the initial rate (at 100 M GTP) about 10-fold. We can thus conclude that although ObgE does not seem to be activated by homodimerization or K ϩ ions as previously suggested, it is activated upon binding to the ribosome.

Conclusion
ObgE is an essential G protein implicated in ribosome maturation, cell cycle control, DNA replication, and bacterial persistence. The present study provides new insights into the structure and function of ObgE. We determined the first crystal structure of ObgE bound to GDP at 1.85-Å resolution and we unequivocally show, using SAXS and MALS, that ObgE behaves as monomer in solution. Biochemical and kinetic analysis unravels a role of the intrinsically disordered C-terminal domain in binding of GTP. Because this domain is less conserved among Obg family members, it remains to be determined whether such a role for this domain is general. Finally, it has been suggested lately that the GTPase activity of Obg proteins might be accelerated by an alternative mechanism, such as homodimerization or binding of K ϩ ions. We show that neither of these hypotheses are valid, but rather confirm that the GTPase activity is stimulated in the presence of ribosomes. These findings thus contribute to our understanding of the detailed molecular mechanism of Obg proteins, which in turn might lead to novel insights into the role of ObgE in cellular physiology, including bacterial persistence.

Protein expression and purification
The open reading frame (ORF) coding for full-length ObgE (ObgE_FL) was amplified by PCR from genomic E. coli DNA as described before (38), and subsequently cloned within the NdeI and HindIII restriction sites of a pET28 vector containing an N-terminal His 6 tag. In addition, another ObgE_FL construct was cloned with a C-terminal strep tag in a pET22b vector within the NdeI and XhoI restriction sites. An ObgE construct, lacking the last 50 amino acids (ObgE_340), was generated by replacing the codon for Glu-341 by a stop codon in pET28-ObgE_FL using the QuikChange site-directed mutagenesis method (Stratagene).
All protein constructs were expressed in either E. coli BL21(DE3) or Rosetta (DE3) pLysS cells. Cells were grown in TB medium at 37°C and induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside when an A 600 of 0.7 was reached. After induction for 6 h at 25°C, cells were harvested by centrifugation and resuspended in buffer A (20 mM Hepes/NaOH, pH 7.5, 300 mM NaCl, 5 mM MgCl 2 , 5 mM ␤-mercaptoethanol, 5% glycerol) containing protease inhibitors (4-(2-aminoethyl)benzenesulfonyl fluoride and leupeptin, 0.1 mg/ml and 1 g/ml final concentrations, respectively). After cell disruption using a cell disruptor system (Constant Systems) and clearance of the lysate by centrifugation at 30,000 ϫ g, the supernatant with His 6tagged proteins was applied to a Ni 2ϩ -Sepharose HP column (GE Healthcare), equilibrated with buffer A. After extensive washing with buffer A containing increasing amounts of imidazole (20 and 30 mM), the protein was eluted with buffer A containing 250 mM imidazole. In the case of the C-terminal strep-tagged protein, the supernatant of the cell lysate was loaded on a Strep-tactin column (IBA Lifesciences), washed extensively with buffer A, and eluted with buffer A containing 2.5 mM desthiobiotin. Preparation of nucleotide-free ObgE was achieved by incubation with calf intestine alkaline phosphatase (Roche Diagnostics) followed by dialysis against buffer A. The nucleotide load of the protein was monitored via reversed phase chromatography on a C 18 column (Jupiter, 25 cm ϫ 4.6 mm) coupled to an Alliance HPLC (Waters) system as described previously (47). The residual alkaline phosphatase was removed by Ni 2ϩ -Sepharose HP (for His 6 -tagged ObgE) or a HiTrap Q HP anion exchange (for strep-tagged ObgE) column. Size-exclusion chromatography (Superdex 75, 16/60) in buffer containing 20 mM Hepes/NaOH, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , and 2 mM DTT was used as the final purification step. Fractions containing purified ObgE were concentrated, flash frozen in liquid nitrogen, and stored at Ϫ80°C. All subsequent experiments were started from these nucleotide-free Obg protein batches.

Purification of 70S ribosome
15 g (wet weight) of E. coli MRE600 cells grown in LB medium were resuspended in a buffer solution containing 20 mM Tris/HCl, pH 7.5, 10 mM MgCl 2 , 100 mM NH 4 Cl, and 6 mM ␤-mercaptoethanol (buffer B). Cells were lysed by two passes through a cell disruptor system (Constant Systems). The lysate was centrifuged twice in a Beckman JA-20 rotor at 17,000 rpm for 30 min. Subsequently the supernatant was loaded onto two 35-ml sucrose cushions (1.1 M sucrose, 20 mM Tris/HCl, pH 7.5, 500 mM NH 4 Cl, 10 mM MgCl 2 , and 0.5 mM EDTA) in 75-ml ultracentrifuge polycarbonate tubes, and centrifuged for 20 h at 35,000 rpm at 4°C in a Ti-45 Beckman rotor. The translucent ribosome pellet was gently washed with buffer B. The pellet was resuspended in 5 ml of the same buffer and the volume was adjusted to 100 ml with buffer B containing 0.5 M NH 4 Cl. An ultracentrifugation step for 6 h at 25,000 rpm was repeated and the pellet was washed with buffer B again. The purified ribosomes were resuspended in 1.6 ml of buffer B containing 50 mM Tris/HCl, pH 7.5. The ribosome concentration was calculated using a ⑀ 260 nm of 4.35 ϫ 10 7 M Ϫ1 cm Ϫ1 . Finally, the ribosomes were flash frozen in liquid nitrogen and stored at Ϫ80°C.

Crystallization, data collection, phasing, and refinement
ObgE_340 was loaded with 1 mM GDP and crystals were obtained by mixing protein (at a concentration of 8 mg/ml) with an equal volume of crystallization buffer in a hanging drop vapor diffusion set-up. The crystallization solution contained 16% (w/v) polyethylene glycol (PEG) 6000, 15% (v/v) 2-propanol, and 100 mM sodium citrate/HCl, pH 5.6. Crystals were flash frozen in liquid nitrogen using crystallization buffer supplemented with 15% glycerol as cryoprotectant, and data were collected at 100 K at the PROXIMA II beamline of the Soleil Synchrotron (Paris, France). Data indexing, integration, and scaling were done using the XDS suite (48). Data quality was assessed in phenix.xtriage (49). Crystals belong to the space group C222 1 with unit cell dimensions a ϭ 64.6 Å, b ϭ 83.0 Å, c ϭ 177.4 Å and ␣ ϭ ␤ ϭ ␥ ϭ 90° (Table 1).
Phases were obtained by molecular replacement using PHASER (50,51) from the CCP4 software package (52) and using the structure of B. subtilis Obg (PDB code 1LNZ) as search model. ARP/wARP was used for automated model building (53). Model building was finalized by manual building cycles in COOT (54), alternated with refinement in Refmac (55). TLS refinement was implemented in the refinement protocol, using six individual TLS groups determined by the TLSMD server (56,57). The obtained model was validated with the Molprobity server (58). All figures were prepared in PyMOL. Data collection and refinement statistics are summarized in Table 1.

SAXS measurement and modeling
SAXS data for ObgE_FL and ObgE_340 (C-terminal streptag and N terminally His 6 -tagged constructs, respectively) were collected either at the SWING beamline of the Soleil Synchrotron (Paris, France) or using an in house Rigaku BioSAXS-2000 instrument. Measurements of ObgE_FL and ObgE_340 in a nucleotide-free state and in the presence of Gpp(NH)p and ObgE_FL in the presence of GDP were performed using an in-line HPLC-SEC setup at the SWING beamline (59). The scattering intensities were recorded after injection of 70 l of 8 -15 mg/ml of protein on an Agilent Bio-SEC 3 column preequilibrated with 20 mM Hepes/NaOH, pH 7.5, 300 mM NaCl, 250 mM imidazole, 5 mM MgCl 2 , and 2 mM DTT (SEC buffer) using a flow rate of 0.2 ml/min. For measurements of ObgE_340 in the presence of GDP, 1 ml of concentrated protein sample (10 mg/ml) was injected on a S75 10/300 column pre-equilibrated with SEC buffer and connected to an AKTA purifier (GE Healthcare) using a flow rate of 1.0 ml/min, and subsequently the elution fractions of the protein peak were collected and concentrated to 2.0, 4.0, and 8.0 mg/ml. The SAXS data were immediately afterward collected in batch mode on a Rigaku BioSAXS-2000 instrument. For data collection of nucleotidebound proteins, the protein was preincubated with 1 mM of the nucleotide, and 400 M of the corresponding nucleotide was added to the running buffer of the size-exclusion chromatography.
The radial averaging and buffer subtraction of the resulting data frames were performed using FOXTROT for the data collected at the synchrotron and using Rigaku SAXSLab for the data collected on the Rigaku BioSAXS-2000 instrument. A final scattering curve used for processing of the data collected in house, was obtained by merging the lower q value data of the 4.0 mg/ml curve with the higher q value data of the 8.0 mg/ml curve. In case of data collection at the synchrotron, DATASW was used for calculation of the invariants (60). The averaged data, corresponding to a peak of interest, was further processed using ATSAS (61) and SCATTER (42) software packages. The molecular mass of the scattering particle was derived using the Q R method (42) and Porod volume (62). Further interpretation of the SAXS data involved representation using the dimensionless Kratky plot and calculation of the pair-distance distribution function using the ATSAS program GNOM (63), whereas calculation of the theoretical scattering profile of the X-ray crystal structure was done using CRYSOL (64).
Prior to flexibility assessment, two short internal regions within the ObgE_340 structure (residues 133-136 and residues 339 -340), missing from the crystal structure, were modeled using the "model missing loop" tool of MODELLER within CHIMERA (65). A similar procedure was performed for ObgE_FL before modeling the flexible C-terminal domain including the C-terminal purification tag. For flexibility assessment of ObgE_340, the crystal structure with modeled loops and the amino acid sequence of the 340 residues of the protein along with the 20 residues of N-terminal purification tag were used. On the other hand, for ObgE_FL, the amino acid sequence of all 390 residues together with the C-terminal puri-fication tag was used. In both cases, a random pool of structures was generated using the ensemble optimization method suite (66), which gives the combination of an all-atom model, corresponding to the crystal structure, and the C-␣ traces of the missing flexible regions. Next, a computational pipeline FULCHER 4 was used to convert the random pool of combined all-atom/C-␣ trace models to all-atom models, with subsequent model validation using the Molprobity clash score and iterative NMA refinement of the top 50% scoring models to probe the conformational dynamics of the molecule. Finally, the genetic algorithm GAJOE was used to obtain an ensemble of models that best describes the experimental SAXS data. Selected ensembles included refined models that had r.m.s. deviation Ͻ0.4 Å compared with the initial crystal structure.

MALS analysis
Multi-angle light scattering experiments coupled to SEC (SEC-MALS) were performed using a Dawn Heleos (Wyatt Technology) detector (using 9 angles) connected to an Agilent Bio-SEC 3 or Shodex KW-800 column attached to an HPLC system (Waters). 10 -20 l of ObgE_FL (with N-terminal His 6 tag) at 4 -8 mg/ml was injected on the column at a flow rate of 0.2 ml/min in 20 mM Hepes/NaOH, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , and 2 mM DTT. In case of the nucleotide-bound sample, the buffer was supplemented with 400 M Gpp(NH)p or 1 mM GDP-AlF x . The molar mass was calculated with the ASTRA 5.3.4.20 software.

Steady state kinetic measurements
GTP hydrolysis rates of ObgE were measured by following the production of GDP in function of time on HPLC (Waters). All measurements were performed at 25°C in 20 mM Hepes/ NaOH, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , and 2 mM DTT. For multi-turnover analysis, 0.5 M ObgE was incubated with different GTP concentrations ranging from 2.5 until 100 M. At different time points 50-l aliquots were taken and the reaction was stopped by heating for 5 min at 100°C. Hydrolysis of GTP was monitored by separation of the nucleotides (GTP and GDP) using a C18 column (Jupiter, 25 cm ϫ 4.6 mm) attached to an HPLC Alliance system (Waters). The nucleotides were eluted using a buffer solution containing 100 mM KH 2 PO 4 , pH 6.4, 10 mM tetrabutyl ammonium bromide, and 7.5% acetonitrile, as mobile phase. Nucleotides were detected at 254 nm and peak areas were converted to concentration using a standard curve derived from known GDP concentrations. Initial rates were obtained as the slope of the [GDP] versus time plot and fitted to the Michaelis-Menten equation. The influence of the 70S ribosome on ObgE GTPase activity was determined by preincubating increasing amounts of the 70S ribosome (0.01, 0.02, 0.05, and 0.1 M) with 0.5 M ObgE at 25°C for 30 min. GTPase activity was determined at 100 M GTP. Because a low GTPase activity co-purified with the ribosome, identical time traces were recorded in the absence of ObgE at each ribosome concentration. These time traces were subtracted from the measurements in the presence of ObgE before calculating initial rates.

Fluorescence stopped flow kinetics
Nucleotide binding kinetics were determined via stopped flow fluorescence (SX18.MV; Applied Photophysics). ObgE in a concentration range of 2-16 M was rapidly mixed with 0.2 M 2Ј,3Ј N-methylanthraniloyl-labeled nucleotides (mant-GDP/ mant-GTP; Jena Bioscience), providing conditions for pseudofirst order binding kinetics. Mant nucleotides were excited at 360 nm and the change in fluorescence was monitored through a 405-nm cut-off filter. For each protein concentration, the data of at least 5 time traces were averaged and fitted to a single exponential function, yielding the observed rate constant k obs . The association and dissociation rate constants (k on and k off ) were obtained from the slope and intercept of the plot of k obs versus the protein concentration. Alternatively, k off was obtained by mixing 200 M of unlabeled nucleotide with a mixture of 0.4 M protein and 1.5 M mant-nucleotide. The resulting time traces were fitted on a single exponential, yielding k off . The K D values were calculated from the ratio of k off and k on . The experiments were performed at 25°C in 20 mM Hepes/NaOH, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , and 2 mM DTT.

Isothermal titration calorimetry
Binding affinity of ObgE for guanosine-3Ј,5Ј-bisdiphosphate (ppGpp; TriLink Biotechnologies) was determined by isothermal titration calorimetry (ITC), using the MicroCal iTC200 system (GE Healthcare) with a reference power of 10 microcal/s. The thermodynamic parameters upon titration of ppGpp (750 M) to nucleotide-free ObgE (75 M) were measured at 25°C in a buffer consisting of 20 mM Hepes/NaOH, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , and 2 mM DTT. For each experiment a preliminary 0.4-l injection (not included in data analysis) was followed by 20 injections of each 2 l with duration of 4 s at a stirring speed of 400 rpm and intervals of 180 s. The initial delay was 120 s. All experiments were performed in triplicate. Origin version 7.0 software was used for data integration and fitting to a single binding site model using the standard Marquardt non-linear regression method as provided in the Microcal Origin routines. The equilibrium binding dissociation constant (K D ) as well as the binding stoichiometry (n) reported in the main text are given as the mean Ϯ S.D. of the three independent measurements.
Author contributions-S. G., R. K. S., M. F., N. V., J. M., and W. V. conceived and designed the experiments. S. G. solved the crystal structure. S. G., J. M., and W. V. performed and analyzed kinetic data. R. K. S. and A. V. S. performed and analyzed SAXS measurements. S. G., R. K. S., and W. V. wrote the paper. All authors reviewed the manuscript.