Histidine 114 Is Critical for ATP Hydrolysis by the Universally Conserved ATPase YchF*

Background: YchF is a universally conserved hydrophobic amino acid-substituted ATPase whose catalytic mechanism is not understood. Results: Biochemical, rapid kinetics, mutagenesis, and molecular dynamics data identify His-114 as critical for ATP hydrolysis. Conclusion: The His-114-containing flexible loop exists in nucleotide-dependent conformations and is involved in catalysis. Significance: These results expand the current understanding of enzymatic strategies employed by nucleotide hydrolases. GTPases perform a wide range of functions, ranging from protein synthesis to cell signaling. Of all known GTPases, only eight are conserved across all three domains of life. YchF is one of these eight universally conserved GTPases; however, its cellular function and enzymatic properties are poorly understood. YchF differs from the classical GTPases in that it has a higher affinity for ATP than for GTP and is a functional ATPase. As a hydrophobic amino acid-substituted ATPase, YchF does not possess the canonical catalytic Gln required for nucleotide hydrolysis. To elucidate the catalytic mechanism of ATP hydrolysis by YchF, we have taken a two-pronged approach combining classical biochemical and in silico techniques. The use of molecular dynamics simulations allowed us to complement our biochemical findings with information about the structural dynamics of YchF. We have thereby identified the highly conserved His-114 as critical for the ATPase activity of YchF from Escherichia coli. His-114 is located in a flexible loop of the G-domain, which undergoes nucleotide-dependent conformational changes. The use of a catalytic His is also observed in the hydrophobic amino acid-substituted GTPase RbgA and is an identifier of the translational GTPase family.

tions, including protein synthesis in Escherichia coli, iron utilization in Brucella melitensis and Vibrio vulnificus, protein degradation and virulence in Streptococcus pneumoniae and V. vulnificus, the oxidative stress response and infection defense response in rice, and centrosome regulation through interaction of its human homolog with the breast and ovarian cancer-specific tumor suppressor BRCA1 (9, 11, 19 -28). Due to its initial characterization as a GTPase, only one x-ray crystallography structure of YchF bound to an adenine nucleotide is available, that of the human homolog OLA1 in complex with AMPPCP (10). Unfortunately, these x-ray crystallography structures do not provide structural information for critical regions of the protein, including both switch regions and a flexible loop located in the G-domain. Furthermore, mutational analyses have revealed that two residues, Asn-13 and Lys-78 (E. colinumbering),playimportantrolesinthepotassiumdependence of ATP hydrolysis, but amino acid residues essential for ATP hydrolysis have yet to be identified (12). Thus, the current structural and biochemical information on YchF is insufficient to devise a reliable mechanism for ATP hydrolysis by YchF.

Experimental Procedures
All chemicals were obtained from VWR International, Sigma-Aldrich, or Invitrogen unless indicated otherwise. DH5␣ cells were purchased from New England Biolabs, and BL21(DE3) competent cells were purchased from Novagen. Restriction enzymes were obtained from Fermentas, radiochemicals were from PerkinElmer Life Sciences, and nucleotides and fluorescent nucleotide analogs were purchased from Invitrogen. All buffers were filtered through Whatman 0.45-m nitrocellulose membranes.
Sequence Alignments-All YchF protein sequences were obtained from the UniProt and Entrez Gene databases. Multiple-sequence alignments (data not shown) were performed using ClustalW2 provided by the European Bioinformatics Institute (EBI).
Protein Expression and Purification-Overexpression and protein purifications were carried out as described previously (11).
CD Spectroscopy-CD spectroscopy was performed using a Jasco J-815 CD spectrometer with a cuvette of 1-mm path length. WT YchF and variants were dialyzed overnight in 50 mM potassium phosphate buffer (pH 7.5) to reduce the sodium ion concentration to Ͻ15 mM. The dialyzed samples were diluted with phosphate buffer to a final concentration of 1.5 M YchF and then equilibrated to 20°C before measuring. All samples were subjected to 10 successive scans (320 to 190 nm) with a digital integration time of 1 s, a bandwidth of 1 nm, and a scanning speed of 50 nm/min. The obtained spectra were averaged and corrected for the phosphate buffer spectrum.
Fluorescence Measurements-Fluorescence measurements were obtained using a QuantaMaster fluorescence spectrophotometer (Photon Technology International, London, Ontario, Canada) with a 3 ϫ 3 mm quartz cuvette (Starna Cells, Inc., Atascadero, CA). YchF from E. coli (E. coli YchF) contains a single Trp residue located in its C-terminal domain whose fluorescence decreases upon nucleotide binding (11). YchF (1 M) in Buffer A (50 mM Tris-Cl (pH 7.5), 70 mM NH 4 Cl, 30 mM KCl, and 7 mM MgCl 2 ) was titrated with the respective adenine nucleotides and excited at 280 nm with a slit width of 1 nm. Fluorescence emission was monitored from 295 to 400 nm at a 1-nm step size with an emission slit width of 5 nm. Following the addition of nucleotide, the solution was equilibrated for 1 min prior to excitation. Equilibration was not performed when ATP was added due to the intrinsic ATPase activity of YchF. The fluorescence of Buffer A with nucleotides was subtracted from all nucleotide-binding experiments, and fluorescence intensities were corrected for dilution. The equilibrium dissociation constant (K D ) was determined by plotting the change in fluorescence at 337 nm as a function of the nucleotide concen-FIGURE 1. YchF is a member of the HAS-GTPase family. Shown is an alignment of the G3 motif of Ras (gray), YqeH (black), MnmE (yellow), HflX (purple), EHD2 (green), FeoB (pink), Era (red), YsxC (brown), and YlqF (orange) with YchF (cyan). Also shown is GDP⅐AlF 3 and the attacking water molecule (red sphere in line with AlF 3 ) from Ras⅐RasGap (PDB ID 1WQ1). Alignment was based on the structure of the conserved G1 motif (P-loop). Gln cat , catalytic Gln. tration. Each data set was fit with a hyperbolic function (Equation 1) using TableCurve software (Jandel Scientific Software) and Prism (GraphPad Software).
Here, F represents the fluorescence at 337 nm, [nt] is the nucleotide concentration, F 0 is the initial fluorescence, and ⌬F max is the amplitude of the signal change. Final K D values and their S.D. values were calculated from at least three independent experiments. ATP Hydrolysis Assays-The multiple-turnover ATP hydrolysis (ATPase) activities of WT YchF and YchF variants were measured at 20°C in Buffer A (unless indicated otherwise) by following the liberation of 32 P i from [␥-32 P]ATP (200 dpm/ pmol). To convert any ADP present, the ATP-containing solutions were charged by incubation with 0.25 g/l pyruvate kinase and 3 mM phosphoenolpyruvate for 15 min at room temperature and then for 15 min at 37°C.
The intrinsic ATPase activities of WT YchF and variants were determined as described (11) using reactions containing 5 M YchF and 125 M [␥-32 P]ATP. Thin-layer chromatography was used to separate 32 P i from [␥-32 P]ATP, followed by visualization using a Typhoon Trio scanner (GE Healthcare). The relative amount of 32 P i formed was determined using ImageJ (29). Ribosome-stimulated ATPase activity was determined under these conditions in the presence of 6 M ribosomes (70S) purified from E. coli MRE600 cells (11).
The pH-dependent ATPase assays were performed similarly to the intrinsic ATPase assays with Buffer A used at pH 7-9. Experiments at pH Ͻ7 were performed in MES-containing buffer. Results were corrected for the background hydrolysis of ATP at 20°C, and ATPase assays were performed at least in triplicate to determine S.D. values. Michaelis-Menten analyses of the ribosome-dependent ATPase activities of YchF(H114A) and YchF(H114R) were performed similarly to the method described previously (11) at 37°C using purified 70S ribosomes and the respective YchF variant (1 M).
Pre-steady-State Kinetics-All rapid kinetics measurements were performed using a KinTek SF-2004 stopped-flow apparatus at 20°C. N-ethylanthraniloyl (mant)-nucleotides were excited via FRET from the single Trp ( ex ϭ 280 nm) present in E. coli YchF and measured after passing through LG-400-F cutoff filters (Newport Corp., Irvine, CA).
For dissociation experiments, 20 M mant-nucleotide and 2 M YchF were preincubated in Buffer A for 30 min at 37°C. Pyruvate kinase (0.25 g/l final concentration) and phosphoenolpyruvate (3 mM final concentration) were added to the mixture when mant-ATP was used. To observe dissociation, 25 l of YchF⅐mant-ATP/mant-ADP (1 M after mixing) was rapidly mixed with 25 l of ATP/ADP (100 M after mixing). Due to the excess of unlabeled nucleotide present, only the dissociation of the mant-labeled nucleotide contributed to the observed fluorescence change, and rebinding of mant-nucleotide was negligible. Each time course was fit with a one-or two-exponential function (Equations 2 and 3, respectively), where F is the mant fluorescence at time t, F ∞ is the final fluorescence, and A is the respective amplitude of the observed fluorescence change. The characteristic apparent rate constants (k app , k app1 , and k app2 ) correspond to the respective nucleotide dissociation constants (k off , k off1 , and k off2 ). Calculations were performed using TableCurve and Prism.
The corresponding apparent rate constants for association were k app1 , k app2 , and k app3 . The bimolecular association rate constant (k 1 ) was determined from the slope of the linear concentration dependence of k app1 .
Molecular Dynamics (MD) Simulations-The initial model for E. coli YchF was obtained by constructing a homology model using the SWISS-MODEL server (30) and the crystal structure of Haemophilus influenzae YchF as a template (Protein Data Bank (PDB) ID 1JAL) (31). The conformation of Switch I (residues 28 -41, which are missing in the 1JAL structure) was modeled based on the structure of Switch I in Thermus thermophilus YchF bound to GDP (PDB ID 2DBY). To model the adenine nucleotide bound to E. coli YchF, we used the conformation of AMPPCP bound to the human homolog of YchF, OLA1 (PDB ID 2OHF). Transformation of the AMPPCP to ATP and ADP was done manually. To position the magnesium ion associated with the bound nucleotide, we aligned the nucleotide with crystal structures of EF-Tu bound to either GDP or GMPPNP (PDB ID 1EFC and 1EFT, respectively). Hydrogen atoms were added to all models using psfgen in the NAMD software package, and His side chains were protonated at the ⑀-nitrogen only (32). Initial models were minimized and then placed in a water box extending at least 10 Å from the protein in all directions. Water molecules present in the template crystal structures were included in this box, and all other waters were added at random using the Solvate package in NAMD (32). Relaxation of the solvated system was achieved by minimizing the positions of water molecules, followed by minimization of the protein/ligand atoms in two iterative rounds (10,000 steps each). Sodium ions were then added in random positions by the Autoionize package in VMD to neutralize the total charge of the system, followed by a final minimization of all components of the system. YchF⅐ATP⅐Mg 2ϩ and YchF⅐ADP⅐Mg 2ϩ had total charges of Ϫ17 and Ϫ16; therefore, was considered to be minimized when no change in energy was observed for at least 1000 steps (33). Minimizations, subsequent equilibrations, and equilibrium MD simulations were performed with periodic boundary conditions using the NAMD software package (32). Minimized models were initially equilibrated at 300 and 350 K for 150 ps at constant pressure (1 atm). Production-phase simulations were started using velocities from the 300 K equilibration and coordinates from the 350 K equilibration. Simulations were performed over 50 ns at 300 K with a step size of 0.5 fs using the CHARMM22 parameters for proteins and CHARMM27 parameters for nucleic acids as implemented in the NAMD package (32,34). MD simulations were performed in an NPT ensemble in which the pressure was maintained at 1 atm with a Nosé-Hoover Langevin piston, and the temperature was controlled using Langevin dynamics. All simulations were performed in the NAMD software package, and visualization was carried out in VMD (32,33).
Snapshots of each MD simulation were saved every 0.5 ps, and trajectories were fitted with the software Carma to remove any rotations of the protein complex or translation of the center of mass (35). The root mean square deviation (r.m.s.d.) and root mean square fluctuation (r.m.s.f.) calculations were performed using scripts written in-house and invoked with the VMD software package (33). To assess whether the nucleotide present had an effect on the conformation of the His-114-containing loop, we measured the distance between the ␣-carbons of Ser-16 in the P-loop (as a reference point) and His-114. The distribution of distances during the different simulations was analyzed using a histogram plot for each simulation with bin sizes of 0.4 Å. Distance measurements were taken every 0.5 ps during the equilibrium MD simulations.

Results
pH Dependence of ATPase Activity-The mechanistic details of how ATP is hydrolyzed by the HAS-ATPase YchF are not understood. To identify amino acid residues that are involved in the catalytic mechanism, we determined the pH dependence of the intrinsic ATPase activity of E. coli YchF (Fig. 2). Such an analysis can provide information regarding the number of ionizable groups involved and their respective pK a values. The latter can indicate the identity of the involved ionizable group. Our data reveal a YchF activity profile consistent with a single ionizable group (Fig. 2B). The intrinsic ATPase activity was maximal at pH Ն7.5 (0.20 Ϯ 0.02 min Ϫ1 ) and negligible at pH Ͻ6. From the corresponding log plot of the ATPase activities as a function of pH, a pK a of ϳ6.7 for the single ionizable group could be obtained (Fig. 2B). A pK a of 6.7 suggests the participation of a His residue (pK a of 6.04) during catalysis (36).
The two common mechanisms of catalysis in GTPases involve aligning a water molecule for nucleophilic attack of the ␥-phosphate (Gln-61 in Ras and His-85 in EF-Tu) and the stabilization of a developing negative charge in the transition state (Arg-178 in G i ␣ 1 and Arg-174 in G t ␣) (5,37,38). Of the four histidines in YchF (positions 102, 114, 145, and 308), only positions 102 and 114 are located within the G-domain and therefore close enough to the ␥-phosphate of the bound nucleotide to participate in hydrolysis (Fig. 3). Because the catalytic amino acid in YchF is likely to be conserved, we investigated the conservation of each amino acid in E. coli YchF to assist in the search for a catalytic residue. To this end, we aligned 116 bacterial YchF sequences to determine the conservation of the histidines present in E. coli YchF. Histidines at positions 102 and 114 are found in 76 and 98% of the bacterial YchF sequences used in the alignment. Position 102 shows significant variability, containing Asn (8%), Met (1%), Gln (13%), or Tyr (2%) instead of His. On the other hand, His-114 is 98% conserved, and only Lys (1%) or Arg (1%) is also found at this position. However, His-114 and His-102 are Ͼ10 Å away from a position that would be compatible with a catalytic role such as aligning a nucleophilic water or stabilizing a developing negative charge in the transition state.   JULY 24, 2015 • VOLUME 290 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 18653
MD Simulations of YchF-Structural information obtained by x-ray crystallography provides only limited insight into the dynamic features of a molecular system. We therefore used MD simulations to assess the potential of His-102 and His-114 to participate in the catalysis of ATP hydrolysis by YchF. To this end, we constructed three homology models of E. coli YchF, in complex with ATP or ADP or in the apo state, and performed 50-ns MD simulations (see "Experimental Procedures"). The stability of these models during simulations was scored by measuring the r.m.s.d. of the backbone atoms with respect to the initial structure throughout the whole simulation (Fig. 4A). After an initial increase in r.m.s.d. of ϳ2-3 Å, depending on the particular simulation, the models appeared to stabilize after 5 ns. The flexibility of each amino acid in the different YchF complexes was determined by calculating the r.m.s.f. of each ␣-carbon over the final 40 ns of each simulation. The r.m.s.f. profiles vary only marginally among complexes, indicating that, on the whole, the dynamics of each structure is similar (Fig. 4B). Interestingly, in all three simulations, His-102 showed very small r.m.s.f. values. In fact, the side chain of His-102 is pointing away from the nucleotide-binding pocket due to a stable hydrogen bond between N ␦1 of His-102 and H ⑀1 /H ⑀2 of Tyr-204 (the hydrogen bond flips between H ⑀1 and H ⑀2 around 22 ns during the YchF⅐ATP⅐Mg 2ϩ simulation) (Fig. 5). This interaction is also maintained when YchF is bound to ADP and in its apo state (data not shown).
A closer look at His-114 reveals that it is located in an extended loop structure (residues 105-121) within the G-domain. This loop displays greater r.m.s.f. values during simulations in the order ATP Ͼ ADP Ͼ apo (Fig. 4B). The observed high level of flexibility is consistent with the fact that electron density for this loop is observed in only one of the five experimentally determined structures of YchF/human OLA1 available in the PDB (IDs 2DBY, 1JAL, 1NI3, 2DWQ, and 2OHF) (9,10). This prompted us to reference this loop as the "flexible" loop. Interestingly, a visual comparison of loop conformations before and after 50 ns of simulation revealed that the flexible loop is found in a more open conformation in the YchF⅐ADP⅐Mg 2ϩ complex and in a more closed conformation (closer to the nucleotide-binding pocket) in the YchF⅐ATP⅐Mg 2ϩ complex or apo state (Fig. 6A). To quantify the movement of the flexible loop during simulation, the distances between the ␣-carbons of His-114 and Ser-16 were measured and plotted over the simulation time (0 -50 ns) (Fig. 6B). Ser-16 is a conserved amino acid in the P-loop responsible for coordinating the magnesium ion and showed essentially no fluctuations in its position during the 50-ns simulations. A similar plot was constructed for each of the remaining histidines (His-102, His-145, and His-308). None of these measurements revealed any nucleotide-dependent changes in dynamics over the course of the simulations (Fig. 7).
To evaluate the distribution of different conformations explored by the flexible loop, the distances plotted in Fig. 6B were binned and plotted as a histogram, counting the number of times the C ␣ -C ␣ distance fell into a particular distance bin (Fig. 8). Analysis of the three 50-ns MD trajectories (YchF⅐ATP⅐Mg 2ϩ , YchF⅐ADP⅐Mg 2ϩ , and apo-YchF) revealed distinct distribution profiles, indicating that the presence and identity of the bound nucleotide influence the preference for different conformations of the flexible loop. In the YchF⅐ATP⅐Mg 2ϩ MD simulation, the flexible loop favors at least three different C ␣16 -C ␣114 distances centered near 14, 17, or 23 Å. In comparison, the YchF⅐ADP⅐Mg 2ϩ trajectory favors distances of ϳ24 and 27 Å, which appear to be sampled by the YchF⅐ATP⅐Mg 2ϩ simulation as well. Notably, the C ␣16 -C ␣114 distance rarely drops below 20 Å in the YchF⅐ADP⅐Mg 2ϩ simulation, whereas YchF⅐ATP⅐Mg 2ϩ mainly samples distances Ͻ20 Å during the 50-ns simulation. In contrast, the apo state MD simulation of YchF exhibits a broad distribution with a maximum around 17 Å and a long tail extending toward 30 Å. The broad distribution of   C ␣16 -C ␣114 distances observed for the apo state simulation is consistent with the rapid fluctuations observed in the time-dependent analysis (Fig. 6B).
These observations clearly support a nucleotide-sensitive behavior of the flexible loop, with primarily open conformations of the loop in the ADP-bound form of YchF, which convert into primarily closed conformations in the ATP-bound and apo states of YchF. Although the flexible loop exists primarily in closed conformations in the ATP-bound and apo states, it is still capable of sampling long C ␣16 -C ␣114 distances characteristic of the open conformation sampled by the YchF⅐ADP⅐Mg 2ϩ simulation (Fig. 8). This suggests a potential for interconversion between the conformations, enabling nucleotide-dependent molecular switching in YchF.
Nucleotide-binding Properties of YchF Variants-Our MD simulations in conjunction with the reported biochemical data suggest that His-114 is most likely involved in the catalysis of ATP hydrolysis by YchF and that it is located in a highly mobile, nucleotide-sensing structural element. To confirm a role for His-114 in catalysis, we constructed four single-amino acid substitution variants of YchF containing Ala, Phe, Arg, or Lys (YchF(H114A), YchF(H114F), YchF(H114R), and YchF(H114K)) in place of His-114. First, we confirmed that introducing these substitutions did not alter the overall structure of the protein using CD spectroscopy (Fig. 9). We then assessed adenine nucleotide binding in each variant using equilibrium fluorescence titrations to exclude any effect of these substitutions on the nucleotidebinding properties. The TGS domain of YchF contains a single Trp residue, which was previously used to determine the equilibrium binding constants (K D ) for the interaction of YchF and ADP/ATP (11). Upon monitoring the Trp fluorescence between 295 and 400 nm, a decrease in fluorescence could be observed following the addition of ADP or ATP (Fig. 10, A and  C). The fluorescence decrease at 337 nm was plotted as a function of nucleotide concentration and fitted with a hyperbolic equation to obtain a K D (Fig. 10, B and D). All YchF variants tested were capable of binding adenine nucleotides with an affinity comparable with WT YchF (Table 2). Interestingly, although WT YchF and YchF(H114K) did not show a signal change upon the addition   of ATP (11), all other variants did. The K D values for all tested mutants are summarized in Table 2 and are in the low micromolar range, similar to values obtained for ADP.
Kinetics of Nucleotide Binding-On the basis of our observation in the MD simulations that nucleotide binding triggered different conformations within YchF and in particular the His-114-containing flexible loop, we performed fluorescence-based rapid kinetics experiments using fluorescent analogs (mant) of the respective nucleotides. Dissociation of the mant-nucleotide resulted in a decrease in mant fluorescence, whereas the respective association experiments yielded fluorescence time courses with increasing fluorescence (Fig. 11).
The dissociation of mant-ADP from YchF was fit with a oneexponential function, yielding a k off of 30 Ϯ 2 s Ϫ1 . Association time courses were also best fit with a one-exponential function, and the value for the association rate constant (k on ) was 3.3 Ϯ 0.5 M Ϫ1 s Ϫ1 (Fig. 11A). The two determined rate constants can be used to calculate the value of the respective equilibrium binding constant (K D ϭ 9 Ϯ 1 M), which is comparable with values (4 Ϯ 2 M) reported previously by Becker et al. (11) and is consistent with the values obtained from the amplitude changes of nucleotide association (Fig. 12). Compared with the catalytically inactive variant YchF(H114A), both the rate constants for mant-ADP association and dissociation were slightly faster (2-fold), yielding an equilibrium binding constant for the nucleotide that is similar to the WT enzyme (Table 3).
In contrast, the dissociation of mant-ATP from YchF was best fit with a two-exponential function, yielding two rate constants (k off1 ϭ 5 Ϯ 2 s Ϫ1 and k off2 ϭ 0.33 Ϯ 0.05 s Ϫ1 ) (Fig. 11A), whereas the association of mant-ATP and YchF was best fit with a three-exponential function. In the latter case, only one of the apparent rate constants exhibited the linear concentration dependence expected for a bimolecular reaction (Fig. 11A). The two remaining rates observed were not concentration-dependent, suggesting two conformational changes as a result of nucleotide binding. These data are consistent with a three-step linear binding mechanism in which binding is followed by two conformational changes. The determined association and dissociation rate constants are summarized in Table 3. To determine which observed rate constant (k off1 or k off2 ) is associated with mant-ATP dissociation as opposed to a conformational change, we compared the determined values with those obtained from the k on plot (y axis intercept). The k off value determined from the k on plot (3.4 Ϯ 0.5 s Ϫ1 ) suggests that the faster step (5 Ϯ 2 s Ϫ1 ) represents dissociation of the nucleotide and that the 10-fold slower step (0.33 Ϯ 0.05 s Ϫ1 ) is associated with a conformational change. Although the pre-steady-state analysis revealed a three-step binding process, the corresponding nucleotide dissociation experiments showed two-phase kinetics. This suggests that one of the dissociation steps is either extremely fast and cannot be observed using the current system or that the rate constants for two of the steps are of the same magnitude and cannot be discriminated. Therefore, we calculated the missing rate constant (k off3 ) based on the K D  value obtained from the amplitude plot (Fig. 12) using the equation K D ϭ (k off1 ϫ k off2 ϫ k off3 )/(k on1 ϫ k on2 ϫ k on3 ). The resulting value (k off3 ϭ 0.24 s Ϫ1 ) is comparable with the value for k off2 and therefore supports the interpretation that dissociation precedes two conformational changes that cannot be kinetically resolved.
Interestingly, the observed dissociation kinetics of mant-ATP from YchF(H114A) varied from that of WT YchF (see above) in that it was best fit with a one-exponential equation, yielding a single rate constant for nucleotide dissociation that was ϳ3-fold faster than the fastest step observed for WT YchF dissociation (Table 3). A similar reduction in complexity of the binding mechanism is also reflected in the observed association kinetics. Time courses obtained from mant-ATP binding to YchF(H114A) were best fit with a two-exponential function instead of a three-exponential function as in the case of WT YchF. Consistent with a two-step binding mechanism, only one of the observed apparent rate constants was concentration-dependent and yielded a 2-fold higher k on than compared with WT YchF (Fig. 11B). Similar to WT YchF, the observed association kinetics for mant-ATP suggest a two-step binding mechanism, whereas the dissociation seems to follow a one-step mechanism for YchF(H114A). Using the equation K D ϭ (k off1 ϫ k off2 )/(k on1 ϫ k on2 ) and the K D value obtained from the amplitude plot, the missing rate constant (k off2 ) can be calculated as 0.02 s Ϫ1 . Given the very slow conversion to the second state and the rapid dissociation, the majority of the nucleotide-bound complex will exist in the initial binding state. Given the small overall fluorescence change associated with the conformational change, this might prevent detection due to a too small signal change.
Intrinsic Nucleotide Hydrolysis Activity-To confirm biochemically that His-114 is required for ATP hydrolysis by YchF, we determined the ability of variants YchF(H114A), YchF(H114F), YchF(H114R), and YchF(H114K) to hydrolyze ATP. Similar to the experiments shown in Fig. 2A, the rate of ATP hydrolysis was determined by thin-layer chromatography (Table 4). Although WT YchF and all variants were able to bind adenine nucleotides with similar affinities, the ATP hydrolysis rates were strongly affected. No YchF-dependent ATP hydrolysis could be detected for YchF(H114A), YchF(H114F), YchF(H114R), and YchF(H114K). This strongly suggests that His-114 is indeed important for efficient catalysis of ATP hydrolysis by YchF.
To investigate a putative role of the other conserved His located in the G-domain, we also constructed a variant (YchF(H102Q)) in which it is replaced with a side chain that is able to maintain the hydrogen bond observed in the MD simulations (see above). The observed intrinsic rate of ATP hydrolysis was unaffected by this substitution compared with WT YchF (Table 4), supporting our assumption that His-102 does not participate in catalysis.
Based on the fact that YchF is a member of the class of HAS-NTPases (Table 1), we wanted to determine whether reverting the Leu directly following the G3 motif, a hallmark for this class of hydrolytic enzymes, to the canonical Gln found in Ras would improve catalysis of ATP hydrolysis. We therefore constructed a variant of YchF with the native Leu ( Fig. 1) replaced with Gln at position 76 (YchF(L76Q)) and determined the intrinsic ATPase activity, which was slightly lower than for WT YchF (Table 4).
Ribosome-stimulated ATPase Activity-To score the role of His-114 against a biologically significant function of YchF, we determined the rate of ATP hydrolysis for two of the His-114 variants (YchF(H114A) and YchF(H114R)) and YchF(H102Q) in the presence of 6 M purified 70S ribosomes, which have been shown previously to stimulate the ATPase activity of WT YchF by 10-fold (11). Under these conditions, the rate of ATP hydrolysis by the YchF(H102Q) variant was identical to that observed for WT YchF (k ATPase 70S ϭ 1.24 Ϯ 0.04 and 1.24 Ϯ 0.03 min Ϫ1 , respectively), reflecting a rate enhancement of 10-fold compared with the intrinsic activity (Table 4). Interestingly and consistent with a role for His-114 in catalysis, the rate of ATP hydrolysis for the two tested variants was 7-8-fold slower (YchF(H114A), k ATP 70S ϭ 0.18 Ϯ 0.03 min Ϫ1 ; and YchF(H114R), 0.16 Ϯ 0.02 min Ϫ1 ). To further investigate if this reduction is due to an effect on the catalytic step or, for example, a change in affinity of these variants for the 70S ribosome, we determined the Michaelis-Menten parameters (k cat and K m ) ( Table 5) for this reaction as described (11). These parameters strongly support a role of His-114 during catalysis of ATP hydrolysis by YchF, as the substitution of His-114 with Ala and Arg reduced k cat by 5-and 10-fold, respectively.

Discussion
Although YchF hydrolyzes ATP more efficiently than GTP, it was originally classified as a GTPase and might have retained certain functional aspects associated with GTPases. Furthermore, it belongs to the enzymatically poorly understood class of     HAS-ATPases. Understanding their enzymatic properties and molecular mechanism is of fundamental importance. YchF is of particular interest because this highly conserved ATPase is found in all domains of life and shares Ͼ45% sequence identity with its human homolog, which is implicated in breast cancer (28). His-114 Is Important for ATP Hydrolysis-To address the question regarding the molecular mechanism of nucleotide hydrolysis in the highly conserved HAS-ATPase YchF, we conducted a combined biochemical and in silico study. The pH dependence of the ATPase activity of YchF revealed that catalysis depends on a single ionizable group, with a pK a consistent with a catalytic His residue. Although YchF from E. coli contains four histidines, only two of them are located in the same domain as the nucleotide-binding pocket. Interestingly, both residues are highly conserved. MD simulations of an E. coli YchF homology model revealed that one of these histidines (His-114) is located in a flexible loop that visits different conformations depending on whether ATP, ADP, or no nucleotide is bound. In contrast, His-102 is engaged in a highly stable hydrogen bond with Tyr-204, locking it in an orientation that renders it inaccessible for catalysis. Furthermore, its movement is restricted due to a vast number of surrounding hydrophobic residues, excluding His-102 as a candidate for participation in catalysis of ATP hydrolysis. To biochemically test the hypothesis that His-114 has a role in catalysis, we determined the ATPase activity of several YchF variants bearing substitutions at position 114. Replacing His-114 with Ala, Phe, Arg, and Lys completely abolished the intrinsic ATPase activity, whereas replacing His-102 with Gln had no effect. Interestingly, in the 116 bacterial species examined, two species showed a substitution of His-114 with either Lys or Arg (Wigglesworthia glossinidia and S. pneumoniae, respectively). However, the flexible loop is poorly conserved in W. glossinidia YchF and contains a four-amino acid insert in S. pneumoniae YchF, suggesting that this loop may have a different conformation or function than that in E. coli YchF. Together with the fact that substitution of His-114 does not affect nucleotide binding, our findings strongly support a catalytic role of His-114 during ATP hydrolysis.
YchF Undergoes Nucleotide-dependent Conformational Changes-GTPases typically function as molecular switches, existing in at least two nucleotide-dependent conformations. YchF has evolved as a functional ATPase, and it is unclear if it has retained the ability to switch between two states, as no such behavior has been reported previously. Our pre-steady-state measurements suggest that YchF indeed behaves as a molecular switch. Differential analysis of the kinetic mechanism of mant-ADP and mant-ATP binding revealed that ADP binds to YchF in a single-step binding process, but the binding of ATP involves two observed conformational changes. This difference in binding mechanisms likely reflects different conformations of ATP-and ADP-bound YchF.
Different conformations of YchF⅐ADP and YchF⅐ATP are also supported by the differential behavior of the flexible loop in our MD simulations of YchF⅐ATP⅐Mg 2ϩ and YchF⅐ADP⅐Mg 2ϩ . In the YchF⅐ATP⅐Mg 2ϩ complex, the flexible loop is in an overall closed conformation wrapped over the bound nucleotide, with an average C ␣16 -C ␣114 distance of ϳ15 Å and the ability to sample at least two sub-states (Fig. 13). On the other hand, the conformations closest to the nucleotide-binding pocket are essentially not accessed by the loop in the YchF⅐ADP⅐Mg 2ϩ simulation, resulting in an open conformation (average C ␣16 -C ␣114 distance of 25 Å) (Fig. 13). In the apo state, the flexible loop is able to sample all states (Figs. 6B and 13), consistent with a model that would limit the conformational space of the flexible loop in response to the binding of a particular nucleotide.
Although the mechanism of ADP binding was essentially unaffected by substitution of His-114 with Ala, a marked difference was observed when mant-ATP interacted with WT YchF or YchF(H114A). In the YchF(H114A) variant, only a single conformational change was observed, as opposed to two for the WT enzyme. These results indicate that upon binding ATP, WT YchF undergoes at least two conformational changes, one of which relies on His-114. Because His-114 is located in the flexible loop and is essential for catalyzing ATP hydrolysis, one conformational change observed during ATP binding is likely a movement of the flexible loop that causes ATPase activation (Fig. 13) by bringing the catalytic His into the proximity of the bound ATP. Using steered MD simulations, it is possible to place His-114 within 7 Å of the ␥-phosphate into a position similar to the catalytic Glu-282 in MnmE (data not shown). This conformation of the loop does not require unfavorable ⌽ and ⌿ angles of the peptide backbone, further supporting the ability of the flexible loop to reach a catalytically active conformation.
Catalytic Mechanism of ATP Hydrolysis-The catalytic mechanisms of only a few HAS-GTPases have been proposed to date, including those for MnmE, YqeH, RbgA, and FeoB (13,15,39,40). Of these, FeoB does not appear to use a catalytic side chain to align an attacking water molecule; MnmE and YqeH provide a catalytic residue from helix ␣2; and RbgA provides a catalytic residue from a flexible linker that is located in a position analogous to helix ␣2. Compared with most HAS-GTPases, helix ␣2 in YchF is shifted away from the nucleotidebinding pocket, and helix ␣4 is located adjacent to the nucleotide-binding pocket instead. This is similar to RbgA, in which helix ␣2 is also displaced away from the nucleotide-binding pocket. YchF and RbgA may therefore share similar solutions with respect to catalyzing ATP hydrolysis. In fact, RbgA has also been shown to possess a catalytic His residue in a flexible loop region (40). We suspect that upon ATPase activation, the flexible loop of YchF undergoes a conformational change to place the catalytic His residue in the active site at a position analogous to that of either the catalytic Gln in Ras-like GTPases or the catalytic Glu in MnmE. This raises the question of the role that the conserved His-114 plays during catalysis in YchF. The fact that the ATPase activity drops at pH Ͻ6.5, where His will likely be protonated, suggests that His-114 is not involved in the sta-bilization of the evolving negative charge on the ␥-phosphate during hydrolysis. It rather supports a role of His-114 either in positioning and/or activating the hydrolytic water during the nucleophilic attack. This would be consistent with the role and the pH dependence for GTPase activation observed in EF-Tu, where experimental conditions prevented analysis below a pH of 6.5 (41).
Furthermore, a role in positioning the catalytic water is consistent with our observation that the ribosome stimulates the ATPase activity of YchF(H114R) to a 10-fold lower rate (k cat ) than WT YchF. It suggests that besides the catalytic His-114, additional catalytic residues are provided either in trans directly by the ribosome or in cis by YchF upon interaction with the 70S ribosome, which, in addition to the contribution of His-114, could stabilize the charge developing on the ␥-phosphate during hydrolysis. We therefore suspect that the interaction with the ribosome will stabilize the ATPase-activated conformation of the flexible loop in YchF, locking His-114 in a catalytically active position and in turn providing the observed additional 10-fold rate enhancement of ATP hydrolysis in the presence of ribosomes.
In summary, we have for the first time identified a single amino acid (His-114) that is critical for ATP hydrolysis by E. coli YchF, located in a highly flexible loop able to adopt nucleotide-dependent conformations. The highly conserved His-114 is important for sensing the presence of the ␥-phosphate in the bound nucleotide and contributes critically to catalysis of ATP hydrolysis by YchF.