The C-terminal Helix of Pseudomonas aeruginosa Elongation Factor Ts Tunes EF-Tu Dynamics to Modulate Nucleotide Exchange*

Little is known about the conservation of critical kinetic parameters and the mechanistic strategies of elongation factor (EF) Ts-catalyzed nucleotide exchange in EF-Tu in bacteria and particularly in clinically relevant pathogens. EF-Tu from the clinically relevant pathogen Pseudomonas aeruginosa shares over 84% sequence identity with the corresponding elongation factor from Escherichia coli. Interestingly, the functionally closely linked EF-Ts only shares 55% sequence identity. To identify any differences in the nucleotide binding properties, as well as in the EF-Ts-mediated nucleotide exchange reaction, we performed a comparative rapid kinetics and mutagenesis analysis of the nucleotide exchange mechanism for both the E. coli and P. aeruginosa systems, identifying helix 13 of EF-Ts as a previously unnoticed regulatory element in the nucleotide exchange mechanism with species-specific elements. Our findings support the base side-first entry of the nucleotide into the binding pocket of the EF-Tu·EF-Ts binary complex, followed by displacement of helix 13 and rapid binding of the phosphate side of the nucleotide, ultimately leading to the release of EF-Ts.

the release of the bound aa-tRNA followed by accommodation of the aa-tRNA into the 50S ribosomal A site. The intrinsic rate of GDP dissociation from E. coli EF-Tu is extremely slow (0.002 s Ϫ1 (9)), requiring the guanine nucleotide exchange factor EF-Ts (10) to facilitate the rapid conversion of EF-Tu⅐GDP into its active GTP-bound state and to maintain rates of protein synthesis (ϳ12 amino acids/s (11)) observed in vivo. Despite the critical cellular role of EF-Ts, its sequence is highly divergent among different bacterial species. Poor conservation of the primary sequence within bacterial EF-Ts sequences may give rise to differences in the enzymatic properties, raising the question of how and if the enzymatic properties are maintained among different bacterial EF-Tu⅐EF-Ts pairs. This is particularly interesting because the structure of the eukaryotic exchange factor is completely unrelated to its bacterial counterpart, whereas the respective structure for eEF1A remains similar to EF-Tu (12).
Based on the experimental approach used in previous studies to dissect the kinetic details of the interaction between EF-Tu, EF-Ts, and the functionally relevant nucleotides GDP and GTP (13)(14)(15) (Fig. 1), we performed a comparative mutational and rapid-kinetics analysis of the enzymatic properties of EF-Ts and EF-Tu from E. coli and the clinically relevant pathogen Pseudomonas aeruginosa. These two gram-negative bacterial species share 84 and 55% sequence identity between their respective EF-Tu and EF-Ts, enabling us to investigate the conservation of the overall enzymatic strategy for nucleotide exchange and the respective kinetic parameters among closely related species. In particular, it is interesting that EF-Ts from P. aeruginosa contains an extended C-terminal helix that interacts with a low conservation region on EF-Tu and that may play a functional role during nucleotide exchange.

Results
Nucleotide Binding Properties of P. aeruginosa EF-Tu (k 1 , k Ϫ1 , k 5 , and k Ϫ5 )-Prior to a detailed investigation of the kinetic mechanism of P. aeruginosa EF-Ts mediated nucleotide exchange reaction, we first conducted a comparative analysis of the nucleotide binding properties of P. aeruginosa EF-Tu and E. coli EF-Tu. Rate constants for guanine nucleotide association and dissociation were determined using the stopped flow technique by measuring FRET between a single tryptophan in P. aeruginosa EF-Tu (Trp-40) or E. coli EF-Tu (Trp-184) and fluorescent methylanthraniloyl (mant)-labeled guanine nucleotides (13)(14)(15). Nucleotide dissociation rate constants were obtained in a chase experiment by mixing EF-Tu⅐mant-GTP/ GDP with excess unlabeled guanine nucleotide. Consistent with a first-order reaction, dissociation of guanine nucleotides from EF-Tu occurs with an observed single-exponential decay (Fig. 2, A and B). Rate constants (k Ϫ5 and k Ϫ1 for GTP and GDP) were directly obtained by fitting the observed time courses with a single-exponential function (Equation 1 under "Experimental Procedures") and are summarized in Table 1. Despite the high degree of sequence conservation between P. aeruginosa and E. coli EF-Tu, dissociation of GDP and GTP from P. aeruginosa EF-Tu (k Ϫ1 ϭ 0.0007 Ϯ 0.0001 sϪ 1 , k Ϫ5 ϭ 0.007 Ϯ 0.001 s Ϫ1 ) is slightly slower than from E. coli EF-Tu (k Ϫ1 ϭ 0.0018 Ϯ 0.0001 s Ϫ1 , k Ϫ5 ϭ 0.013 Ϯ 0.001 s Ϫ1 ).
To determine the rate constants of GTP and GDP association, guanine nucleotide association was measured using a constant concentration of nucleotide-free EF-Tu and increasing concentrations of mant-GTP or mant-GDP (Fig. 2, C and D). The respective apparent rate constants (k app ) at different nucleotide concentrations were determined from the obtained fluorescence time courses by fitting with a single-exponential func-tion (Equation 1). The linear dependence of k app on the concentration of mant-GTP/mant-GDP was used to determine the respective association rate constants (k 5 and k 1 ) from the slope of the concentration dependence. The values for P. aeruginosa and E. coli EF-Tu are summarized in Table 1 and indicate a 3-fold faster rate of mant-GTP association (k 5 ) for P. aeruginosa EF-Tu but no difference in the mant-GDP association (k 1 ) rate constants. These rate constants reveal 5-and 2.8-fold higher affinities (K d ) of mant-GTP and mant-GDP for P. aeruginosa EF-Tu when compared with E. coli EF-Tu (summarized in Table 1).
EF-Ts-mediated Nucleotide Dissociation-To maintain in vivo rates of protein synthesis, EF-Ts is required to enhance the low intrinsic rate of guanine nucleotide dissociation from EF-Tu. This becomes even more relevant for P. aeruginosa EF-Tu with its tighter binding to nucleotides compared with E. coli EF-Tu. The interaction of EF-Tu⅐GTP or EF-Tu⅐GDP with EF-Ts can be described as two consecutive steps: the formation of the EF-Tu⅐nucleotide⅐EF-Ts ternary complex followed by the release of the bound guanine nucleotide (Fig. 1).   To study the EF-Ts-catalyzed dissociation of guanine nucleotides from EF-Tu, experiments were carried out using mant-GTP/mant-GDP. Upon mixing of the respective preformed EF-Tu⅐mant-nucleotide complexes with EF-Ts, a single exponential fluorescence decrease reflecting the mant-GTP/mant-GDP dissociation was observed. At low concentrations of EF-Ts, k app increases linearly with the EF-Ts concentration, consistent with the formation of the EF-Tu⅐mant-GTP/mant-GDP⅐EF-Ts ternary complex. At high concentrations of EF-Ts, the rate of the reaction reaches saturation and equals the rate constant of nucleotide dissociation, i.e. k Ϫ7 (GTP) or k Ϫ4 (GDP) (Fig. 3). To assess the interchangeability of EF-Ts from either E. coli or P. aeruginosa, the experiment was also performed with EF-Tu⅐mant-GTP/mant-GDP from each species in the presence of EF-Ts from E. coli or P. aeruginosa . (summarized in  Tables 2 and 3). Consistent with previous reports, E. coli EF-Ts stimulates the dissociation of mant-GTP and mant-GDP from its EF-Tu more than 9,000-fold (k Ϫ7 /k Ϫ5 ) and more than 130,000-fold (k Ϫ4 /k Ϫ1 ), respectively. Similar to the E. coli system, mant-GTP and mant-GDP dissociation from P. aeruginosa EF-Tu is efficiently catalyzed by its own EF-Ts, more than 5000-fold (k Ϫ7 /k Ϫ5 ) and 50,000-fold (k Ϫ4 /k Ϫ1 ), respectively. However, dissociation of mant-GTP and mant-GDP from P. aeruginosa EF-Tu is accelerated only 700-fold (k Ϫ7 /k Ϫ5 ) and 14,000-fold (k Ϫ4 /k Ϫ1 ) by the E. coli exchange factor. In contrast to this, the P. aeruginosa exchange factor is able to stimulate the dissociation of mant-GTP and mant-GDP from E. coli EF-Tu . more than 21,000-fold (k Ϫ7 /k Ϫ5 ) and 270,000-fold (k Ϫ4 /k Ϫ1 ), respectively.
It is surprising that P. aeruginosa EF-Ts is able to catalyze the dissociation of guanine nucleotides from E. coli EF-Tu . more efficiently than the corresponding E. coli exchange factor. This is particularly astonishing because the interaction surface between EF-Tu and EF-Ts is highly conserved between the two species (Fig. 4A). The only exceptions to this are variations found within the C-terminal module (helix 13) of EF-Ts, pointing at a putative role for this helix during EF-Ts stimulated nucleotide dissociation from the EF-Tu⅐nucleotide⅐EF-Ts complex. The difference in sequence is subtle and includes a twoamino acid extension of helix 13 (Fig. 4B). The increased helix length and variation in its sequence might provide additional interactions between domain I of EF-Tu and helix 13 of P. aeruginosa EF-Ts that are not possible with the shorter E. coli helix.
The Role of Helix 13 during Nucleotide Exchange-To dissect the role of helix 13 for the nucleotide exchange mechanism in EF-Tu, a C-terminal truncation variant, EF-Ts (⌬283), of P. aeruginosa EF-Ts (Fig. 4B) was constructed, and the kinetic

Nucleotide Exchange in EF-Tu⅐EF-Ts from P. aeruginosa
parameters describing the interactions between EF-Tu, the EF-Ts variants, and guanine nucleotides were determined using the approach described above (summarized in Table 3). Con-sistent with a role of the C-terminal extension of helix 13 during EF-Ts-mediated nucleotide dissociation, the C-terminal truncation variant of P. aeruginosa EF-Ts (EF-Ts(⌬283)) stimulated GTP and GDP dissociation from P. aeruginosa EF-Tu 44-fold (k Ϫ7 /k Ϫ7 ) and 13-fold (k Ϫ4 /k Ϫ4 ) less efficiently than wild type EF-Ts, respectively ( Fig. 5 and Tables 2 and 3). A similar effect (47-and 16-fold for GTP and GDP, respectively) was also observed for EF-Ts(⌬283)-mediated nucleotide dissociation from E. coli EF-Tu. Similar to the P. aeruginosa system, the effect is strongest for GTP dissociation, suggesting that this is a common feature of the C-terminal extension present in P. aeruginosa EF-Ts. The conserved relative effect is particularly interesting because this EF-Ts is able to facilitate 6 -10fold faster GTP (k Ϫ7 /k Ϫ7 ) and GDP (k Ϫ4 /k Ϫ4 ) dissociation in E. coli EF-Tu then in its P. aeruginosa counterpart. Overall the P. aeruginosa EF-Ts-stimulated nucleotide dissociation from E. coli EF-Tu was twice as fast (k Ϫ7 /k Ϫ7 and k Ϫ4 /k Ϫ4 , respectively) as when stimulated by its own EF-Ts (Table 2). Thus, the C-terminal extension of helix 13 in EF-Ts from P. aeruginosa affects nucleotide dissociation from both its own EF-Tu and that of the E. coli to a similar extent, independent of the nucleotides present.
To further investigate the role of the C-terminal extension for the nucleotide exchange mechanism, we replaced the last four residues in E. coli EF-Ts (SKQS) with the last six residues (VAATKQ) from P. aeruginosa, constructing a chimera of EF-Ts (EF-Ts(Chimera); Fig. 4B). Interestingly, the dissociation rate constants describing EF-Ts(Chimera)-catalyzed mant-GTP/mant-GDP from E. coli EF-Tu are k Ϫ7 ϭ 120 Ϯ 20 s Ϫ1 and k Ϫ4 ϭ 380 Ϯ 50 s Ϫ1 (Table 2) and are similar to wild type E. coli EF-Ts. However, the EF-Ts(Chimera) reduces the association rate constants for mant-GTP/mant-GDP binding (k 7 and k 4 , respectively) to the EF-Tu . ⅐EF-Ts complex 25-and 10-fold ( Table 2). This effect is comparable with what we determined for k 7 and k 4 in the presence of P. aeruginosa EF-Ts (15-and 5-fold reduction, respectively), suggesting a role of the C-terminal helix in modulating the nucleotide association kinetics. Ultimately, given the unchanged dissociation rate constant, replacing the C-terminal residues in E. coli EF-Ts with those from P. aeruginosa lowers the nucleotide affinity for mant-GTP/mant-GDP (K 7 and K 4 ) to the EF-Tu . ⅐EF-Ts complex 24and 16-fold, making it similar to the affinity for the E. coli EF-Tu P. aeruginosa EF-Ts complex ( Table 2). This is consistent with the fact that the EF-Ts(Chimera) is able to catalyze dissociation of mant-GTP/mant-GDP from P. aeruginosa EF-Tu more like the EF-Ts from P. aeruginosa . (Table 3), despite the fact that 98% of its sequence is identical to E. coli EF-Ts.
When comparing the interactions of the helix 13 residues in the structure of the EF-Tu⅐EF-Ts complex from E. coli (16) with TABLE 3 Rate constants for the kinetic mechanism of nucleotide exchange in P. aeruginosa EF-Tu The steps are as defined for Fig. 1.
Step We therefore substituted Gln-283 in P. aeruginosa EF-Ts (which is the corresponding residue to E. coli Met-278) with Met (EF-Ts (Q283M); Fig. 4B). This substitution resulted in only slightly increased nucleotide dissociation rate constants for P. aeruginosa EF-Tu ( Fig. 4B and Table 2 and 3). Interestingly, the nucleotide association was 3-fold (k 4 ) faster for mant-GDP and 32-fold faster (k 7 ) for mant-GTP (Table 2). Therefore, the substitution of Gln-283 mainly affects the association step and is most pronounced for GTP. The Role of Helix 13 for Ternary Complex Stability-To determine whether helix 13 has a mechanistic role in the dissociation of EF-Ts from the EF-Tu⅐GTP/GDP⅐EF-Ts complex, a GTP/GDP titration against a fixed concentration of purified EF-Tu⅐EF-Ts complexes was performed similar to previous studies (13)(14)(15). The observed change in the intrinsic tryptophan fluorescence of EF-Tu reflects the dissociation of EF-Ts from the transiently formed EF-Tu⅐GTP/GDP⅐EF-Ts complex. The rates of EF-Ts dissociation were determined from time courses using a single-exponential fit and plotted as a function of increasing guanine nucleotide concentration ( Fig. 6). At saturating concentrations of guanine nucleotide, where rebinding of EF-Ts is negligible and EF-Ts dissociation is not limited by guanine nucleotide binding, the rate constants for EF-Ts dissociating from the EF-Tu⅐GTP/GDP⅐EF-Ts ternary complex (k Ϫ6 and k Ϫ3 ) can be determined ( Fig. 1 and Table 2). From the slope of the concentration dependence, the corresponding association rate constants (k 6 and k 3 ) can be calculated (9). The location of the tryptophan residue within P. aeruginosa EF-Tu did not allow for a detectable change in the tryptophan signal upon GDP-induced dissociation of the Tu⅐Ts complex. Therefore, k Ϫ3 , k 3 , and k 4 could not be determined for the respective P. aeruginosa EF-Tu-containing complexes.

EF-Ts (E. coli) EF-Ts (P. aeruginosa) EF-Ts (Q283M) EF-Ts (⌬283) EF-Ts (chimera)
Using E. coli EF-Tu, we were able to compare the rate constants for dissociation of P. aeruginosa and E. coli EF-Ts from ternary EF-Tu⅐nucleotide⅐EF-Ts complexes. For both EF-Tu⅐GTP⅐EF-Ts and EF-Tu⅐GDP⅐EF-Ts complexes, the rates of EF-Ts dissociation (k Ϫ6 and k Ϫ3 ) were 2-fold slower than E. coli EF-Ts. A similar 2-3-fold (k 3 and k 6 ) effect was observed for the association rate constants for the binding of P. aeruginosa EF-Ts to both E. coli EF-Tu⅐GDP and EF-Tu⅐GTP (Table  2). Interestingly, the removal of the C-terminal extension in P. aeruginosa EF-Ts, EF-Ts(⌬283), increases the rate constants for dissociation from the GTP and GDP-complexes 19-and 13-fold (k Ϫ6 and k Ϫ3 ), respectively. When compared with wild type E. coli EF-Ts, the rate constants are 10-and 6-fold faster. These effects on the dissociation rate constants are stronger than for the association rate constants, where a 4-fold increase (k 6 ) in association rate constant for the P. aeruginosa EF-Ts binding to E. coli EF-Tu⅐GTP suggests a role for helix 13 in the interaction with the GTP conformation of EF-Tu. Surprisingly, grafting the C-terminal six amino acid residues from P. aeruginosa on EF-Ts from E. coli has no effect on either the association or dissociation rate constants to the GTP (k 6 and k Ϫ6 ) or GDP (k 3 and k Ϫ3 ) conformation of E. coli EF-Tu (Table 2). However, substitution of Gln-283 with Met reduces the rate constant of P. aeruginosa EF-Ts dissociation from the EF-Tu⅐GTP⅐EF-Ts complex 5-fold (k Ϫ6 ) and increases the rate constant for binding 2-fold (k 6 ), which results in a 10-fold increased affinity (K 6 ) of EF-Ts(Q283M) to the E. coli EF-Tu⅐GTP complex. A similar 5-fold (k Ϫ3 ) decrease in the dissociation rate constant is also observed for the GDP containing ternary complex, which is accompanied by a small 1.5-fold (k 3 ) reduction in the association rate constant. Based on this, the affinity of EF-Ts(Q283M) for the GDP-bound E. coli EF-Tu is affected in the similar direction by this single amino acid substitution in helix 13 but is less pronounced than for the GTPcontaining complex and only 3-fold (K 3 ) higher than wild type P. aeruginosa EF-Ts.
The Role of Helix 13 for EF-Tu⅐EF-Ts Binary Complex Stability-To investigate how the effects of altered C-terminal sequences in EF-Ts relate to the formation of the nucleotidefree binary EF-Tu⅐EF-Ts complex, we used the same approach as above monitoring the change of the intrinsic tryptophan fluorescence of E. coli EF-Tu (Fig. 7). The obtained values for k 2 and k Ϫ2 are summarized in Table 2. In agreement with previously published data (9), k 2 for the EF-Tu and EF-Ts complex from E. coli is 19 Ϯ 5 ϫ 10 6 M Ϫ1 s Ϫ1 . On the other hand, the association rate constant for E. coli EF-Tu and EF-Ts from P. aeruginosa is slightly reduced (3.8-fold). The dissociation rate constant (k Ϫ2 ) for the EF-Tu⅐EF-Ts complexes can be calculated from the remaining rate constants in the kinetic mechanism (9,15). Interestingly, k Ϫ2 for the E. coli EF-Tu⅐EF-Ts and the complex containing E. coli EF-Tu and EF-Ts from P. aeruginosa were 0.05 Ϯ 0.04 s Ϫ1 and 0.0006 Ϯ 0.0002 s Ϫ1 , respectively (Table 2). This results in an affinity (K 2 ) for the mixed E. coli/ P. aeruginosa complex of 0.12 nM, which is 25-fold higher than for the E. coli EF-Tu⅐EF-Ts complex (K 2 ϭ 3 nM). Consistent with a role of the C-terminal extension in increasing the affinity of EF-Ts for EF-Tu, removing the last six amino acids from P. aeruginosa EF-Tu (K 2 ϭ 200 nM) reduces the affinity 1600fold. This is mainly due to a 1000-fold increased dissociation rate constant (k Ϫ2 ) without a change in the corresponding association rate constant (Table 2). This trend is continued in the EF-Ts variant containing the last six amino acids of P. aeruginosa EF-Ts. Although the effect is less pronounced, the affinity of the EF-Ts(Chimera) to EF-Tu from E. coli is still 6-fold higher (K 2 ) than wild type E. coli EF-Ts, mainly because of a 6-fold (k Ϫ2 ) decreased dissociation rate constant. Interestingly, the identity of the amino acid in position 283 of P. aeruginosa EF-Ts is not responsible for this effect, because this substitution variant has only a 2-fold lower (K 2 ) affinity than wild type EF-Ts from P. aeruginosa.
Molecular Dynamics Simulations Reveal Destabilization of Helix F in EF-Tu by EF-Ts-Based on our kinetic analysis above, it is evident that the C-terminal module of EF-Ts (helix 13) strongly modulates the interaction between EF-Tu and guanine nucleotides. To obtain a structural understanding of this effect, all-atoms models of E. coli EF-Tu⅐EF-Ts and EF-Tu⅐EF-Ts (Chimera) were constructed and simulated using molecular dynamics similar to our previous studies (15,17,18). After constructing a model of the EF-Tu⅐EF-Ts(Chimera) complex, the C terminus of EF-Ts(Chimera) was subjected to simulated annealing to investigate which regions of EF-Tu it may interact with. Based on 20 independent annealing simulations, the Nucleotide Exchange in EF-Tu⅐EF-Ts from P. aeruginosa OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44

JOURNAL OF BIOLOGICAL CHEMISTRY 23141
C-terminal module of EF-Ts(Chimera) likely contacts EF-Tu either at the switch I region or helix F (near the base-side interactions of the nucleotide-binding pocket; Fig. 4A). Evidence suggests, however, that the switch I region of EF-Tu is disordered when bound to EF-Ts (16). Thus, differences in the structural dynamics between E. coli EF-Tu⅐EF-Ts and EF-Tu⅐EF-Ts(Chimera) complexes that can account for the observed differences in nucleotide interaction in vitro likely originate in the helix F region (Fig. 4A).
The MD simulations of these models were stable after 5 ns based on backbone root mean square deviation and root mean square fluctuation (RMSF) calculations (Fig. 8, A and B), and the overall dynamics of the complexes are largely similar (Fig.  8B). Throughout both simulations the C-terminal helix of EF-Ts remains in contact with the G-domain of EF-Tu. Interestingly, the amino acids following the C-terminal helix of EF-Ts show different dynamics in the E. coli EF-Tu⅐EF-Ts and EF-Tu⅐EF-Ts(Chimera) simulations. In the E. coli EF-Tu⅐EF-Ts complex, the three C-terminal amino acids (280 -282) pack against the G-domain of EF-Tu (Fig. 9, A and C). In the EF-Tu⅐EF-Ts(Chimera) simulation, however, the five C-terminal amino acids (280 -284) exhibit higher mobility and do not pack efficiently on EF-Tu (Fig. 9, B and D). This difference in the C termini of E. coli EF-Ts and EF-Ts(Chimera) is correlated with the stability of a nearby hydrogen-bonding network (Lys-37/Glu-185/Asp-181) in EF-Tu (Fig. 9, C and D).
In E. coli EF-Tu⅐EF-Ts, the C-terminal amino acids of EF-Ts form hydrophobic interactions with the side chain of Lys-37, which appears to stabilize the hydrogen bond network formed by Lys-37, Glu-185, and Asp-181 (Fig. 9). Because the C-terminal amino acids in EF-Ts(Chimera) are mobile, they do not form the same hydrophobic interactions, and the resulting increase in mobility of the Lys-37 side chain destabilizes the Lys-37/Glu-185/Asp-181 hydrogen bonding network (Fig. 9). In quantitative terms, the hydrogen bond between Glu-185 and Asp-181 in helix F of EF-Tu is stable throughout the entire simulation of the E. coli EF-Tu⅐EF-Ts complex but only for 75% of the simulation of EF-Tu⅐EF-Ts(Chimera) (Fig. 10A).
To investigate how the differential stabilities of this hydrogen-bonding network affect the structural dynamics of the complex, the ⌽ and ⌿ angles for each amino acid in EF-Tu were plotted as a "Ramachandran histogram" (17). This analysis revealed highly flexible amino acids as those that occupied mul-tiple conformations. Interestingly, the loop connecting helix E and helix F (Fig. 4A) in EF-Tu (Gly-180, Asp-181, and Ala-183) was found to be more flexible when bound to EF-Ts(Chimera) compared with E. coli EF-Ts (Fig. 10B). Because the N terminus of helix E interacts with the guanine base of the bound nucleotide, backbone destabilization of the loop between helix E and F may ultimately destabilize bound nucleotides. Taken together, our in vitro and in silico results suggest that the C terminus of EF-Ts(Chimera) destabilizes the loop between helixes E and F in EF-Tu, which compromises interactions between bound nucleotides and helix E, thus increasing the rates of nucleotide dissociation.  coli EF-Tu⅐EF-Ts and the E. coli complex containing EF-Ts(Chimera). The root mean square deviation was computed for the nonhydrogen backbone atoms of EF-Tu⅐EF-Ts (blue) and EF-Tu⅐EF-Ts(Chimera) (red) models during 20-ns MD simulations. B, the overall dynamics of EF-Tu⅐EF-Ts and EF-Tu⅐EF-Ts(Chimera) are similar in MD simulations. The RMSF of each C␣ atom was measured every 2 ps from 10 to 20 ns of MD simulation. The RMSF is plotted as a function of amino acid for EF-Tu⅐EF-Ts (blue) and EF-Tu⅐EF-Ts(Chimera) (red). The shaded regions indicate the switch I region of EF-Tu and helices 10 and 11 of EF-Ts from left to right, respectively. The colored bar indicates EF-Tu (blue) and EF-Ts (yellow) as they are shown in Fig. 9.

Discussion
Guanine Nucleotide Exchange in E. coli and P. aeruginosa EF-Tu-Kinetic parameters describing the interaction of the essential elongation factor Tu with guanine nucleotides are available for a number of different model systems, such as yeast and E. coli, revealing large functionally relevant differences in affinities (9,19,20). However, little is known about the conservation of these fundamental and functionally critical parameters in bacteria, particularly in clinically relevant pathogens such as P. aeruginosa. The rapid kinetics analysis performed here, comparing the nucleotide binding properties of EF-Tu from E. coli and P. aeruginosa, revealed that EF-Tu from P. aeruginosa binds GDP with subnanomolar and GTP with low nanomolar affinity. Compared with the well established parameters for E. coli (9,15), the affinity is 3-5-fold higher for GDP and GTP, respectively. This effect is mainly due to a decreased rate constant for dissociation of the nucleotide from the factor, suggesting either that the nucleotide-bound state has lower energy than the corresponding E. coli complex or that the transition state energy is higher. Either way, this will have implications for the mechanistic strategies employed by the nucleotide release factor. Interestingly, the EF-Ts stimulated release of GTP or GDP is also 3-6-fold slower than in E. coli. Nevertheless, the strategy to equalize the dissociation rates for GTP and GDP (9) is maintained, as well as the 10-fold difference in dis-sociation rate of the nucleotide from the respective EF-Tu⅐nucleotide complex (9).
The high degree of sequence conservation between the two EF-Tu (84%, data not shown) raises questions regarding the structural and sequence origins of the observed differences in the nucleotide binding properties and how the P. aeruginosa EF-Ts is able to overcome them. Mapping the sequence differences between the two species onto the structure of the EF-Tu⅐EF-Ts complex from E. coli reveals that in addition to the conserved interactions between domains 1 and 3 of EF-Tu and EF-Ts, helix 13 of EF-Ts contains an extension that maps to a region on the surface of EF-Tu corresponding to helix F and E. Interestingly, this region contains the highest number of nonconserved residues between EF-Tu from E. coli and P. aeruginosa (Fig. 4A). A functional role of this interaction for finetuning the nucleotide binding properties in the EF-Tu⅐EF-Ts pair is consistent with our observation that removal of this region of the helix has a strong effect (16 to 44-fold decrease) not only on the EF-Ts stimulated nucleotide dissociation from P. aeruginosa EF-Tu but also from E. coli. This is in line with previous work that found that removal of the whole helix 13 affects nucleotide release (21). However, this study removed the helix completely, whereas in our work, only the C-terminal six amino acids were removed, pointing to a mechanistic role of this particular region in helix 13. Furthermore, the fact that the  OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44 JOURNAL OF BIOLOGICAL CHEMISTRY 23143 effect occurs in E. coli EF-Tu, as well as P. aeruginosa indicates that this mechanistic strategy is likely common in bacteria. This transferability is further supported by the effect of transplanting these six residues from P. aeruginosa EF-Ts onto E. coli EF-Ts. Notably, this chimera has the ability to increase the rate of EF-Ts stimulated nucleotide release using the E. coli exchange factor scaffold to the level observed for the P. aeruginosa factor, a surprising result given the low overall sequence conservation of the factors (55%, data not shown).

Nucleotide Exchange in EF-Tu⅐EF-Ts from P. aeruginosa
A more detailed mechanistic understanding of the role of the C-terminal segment of helix 13 during stimulated nucleotide release can be derived from the kinetic parameters describing the interaction in the heterologous system containing EF-Tu from E. coli and EF-Ts from P. aeruginosa. Here the P. aeruginosa EF-Ts performs as a "super" nucleotide release factor because it is able to catalyze nucleotide release more efficiently than the E. coli exchange factor, lowering the nucleotide affinity an additional 10-fold (K 4 ) for GDP and over 30-fold (K 7 ) for GTP. This additional destabilization of nucleotide binding is mainly achieved by destabilizing binding of the nucleotide to the EF-Tu⅐EF-Ts binary complex accompanied by a slight increase in the rate of nucleotide dissociation (Table 2). Reduction in the nucleotide association rate constant is more pronounced for GTP than for GDP. This effect can be attributed to the last six amino acids in helix 13 of EF-Ts, because the EF-Ts chimera contains just these residues from P. aeruginosa in an otherwise E. coli sequence background. This six-residue substitution essentially turns E. coli EF-Ts into a P. aeruginosa EF-Ts from a kinetics point of view. Interestingly, this effect does not have a nucleotide preference and does not involve nucleotide dissociation. Together with the structural dynamics data obtained from our MD simulations of the EF-Tu⅐EF-Ts(Chimera) binary complex, this suggests a model where the C-terminal end of helix 13 of EF-Ts interacts with helices E and F in EF-Tu, destabilizing the structure and increasing the flexibility of helix E. Because the N-terminal end of helix E is involved in nucleotide binding through hydrogen bonding of Ser-173 with the carbonyl oxygen of the nucleobase, this should ultimately lead to decreasing the nucleotide-binding energy. This is supported by our kinetic data showing 10-and 20-fold (k 7 ) reductions in the nucleotide association (GDP and GTP) rate when comparing EF-Ts (Chimera) with wild type EF-Ts. A functionally relevant interaction network between helix 13 in the region of helix F and E is further supported by the observation that removal of the C-terminal extension selectively increases EF-Ts dissociation from EF-Tu. This suggests that the incoming nucleotide, when bound to the base-side of the binding pocket on EF-Tu, will stabilize the structural dynamics of the F-E helix network. In turn, this will induce dissociation of helix 13 from this region, which then is followed by disruption of the interaction that the N-terminal end of helix 13 makes with helix A and the phosphate side of the nucleotide-binding pocket (Fig. 4A).
Our observations lead to a model that supports the base sidefirst entry of the nucleotide into the binding pocket of the EF-Tu⅐EF-Ts binary complex, which will be followed by displacement of helix 13 and rapid binding of the phosphate side of the nucleotide, ultimately leading to the release of EF-Ts (Fig.  11). This model is consistent with SMD simulations inducing nucleotide dissociation via extraction of the phosphate side that shows the junction between helix F and helix E as highly flexible (17). It also provides a strategy for efficient discrimination between different nucleotides (adenine versus guanine), because these base side interactions must occur prior to complete accommodation of the nucleotide and the subsequent stabilization of binding by interactions with the nucleotide phosphates, which are energetically different for the di-and triphosphate forms. This ultimately allows the EF-Tu⅐EF-Ts complex to effectively screen the identity of the incoming nucleotide. Our findings demonstrate that the overall strategy for nucleotide exchange is likely conserved among different bacterial species and that alterations in the individual sequences and enzymatic properties of EF-Tu and EF-Ts seem to have evolved to compensate and to maintain the overall catalytic strategy of equalizing the dissociation rates (k Ϫ4 and k Ϫ7 ) for the nucleotide, maintaining a rate that is almost 10 times faster than the protein synthesis rate. Furthermore, the role of helix 13 identified here might explain how the corresponding structurally unrelated eukaryotic nucleotide exchange factor has evolved as helix 13 and eEF1B␣ approach the nucleotidebinding pocket from the same side (12).
In summary, we report for the first time that the nucleotide binding properties of EF-Tu vary even between closely related bacterial species, that nucleotide binding occurs via the base side-first entry into the nucleotide-binding pocket, and that fine-tuning of the initial base side entry kinetics via modulation of the structural dynamics of the respective side of the EF-Tu nucleotide-binding pocket by helix 13 of EF-Ts is the strategy used by bacterial EF-Ts to adjust for nucleotide binding affinity differences.

Experimental Procedures
Molecular Biology-All restriction enzymes used were purchased from Fermentas, and primers were obtained from Integrated DNA Technologies. For the recombinant expression of E. coli EF-Tu and EF-Ts, the previously reported constructs pKECAHIS (22), pCA24Ntsf (23), and pHK1Ts (13) were used.
Cloning of P. aeruginosa tufA and tsf-The EF-Tu encoding tufA gene was PCR-amplified from P. aeruginosa genomic DNA using Phusion polymerase (Finnzymes) utilizing the primers 5Ј-AGA GGA TCC CTG TCG TGG CTA AAG GA-3Ј and 5Ј-ATG GCA GGA GCT CCG ATT ACT CGA-3Ј, introducing BamHI and SacI restriction sites (underlined). In a similar fashion the tsf gene coding for EF-Ts was amplified from P. aeruginosa genomic DNA, using the primers 5Ј-TTC CAT ATG GCA GAA ATT ACT GCA GC-3Ј and 5Ј-CAG TCG AAT TCG TCT TTG TTA CTG-3Ј, with engineered NdeI and EcoRI restriction sites, respectively. The resulting PCR products were ligated into SmaI-digested pUC19 (New England Bio-Labs) using T4 DNA ligase (Invitrogen). The ligation mixture was transformed into subcloning efficiency E. coli DH5␣ cells (New England BioLabs). The resulting pUC19 constructs were digested with restriction enzymes corresponding to the recognition sequence introduced via the respective PCR primers. The resulting insert fragments were then ligated (T4 DNA ligase) into similarly digested pET28a to yield pET28tufPa and pET28tsfPa. Sequence and orientation was confirmed by DNA sequencing (Macrogen DNA Sequencing Services). Protein Expression and Purification-EF-Tu and EF-Ts variants derived from P. aeruginosa were expressed in E. coli BL21-DE3 (Novagen) containing the respective plasmid. The cells were grown at 37°C in LB medium supplemented with 50 g/ml kanamycin to the mid-log phase (A 600 ϭ 0.6) and induced with 1 mM isopropyl thio-␤-galactoside (BioBasic). The cells were grown for an additional 3 h at 37°C, harvested by centrifugation (5000 ϫ g for 10 min at 4°C), and stored at Ϫ80°C.

Mutagenesis-The
EF-Tu from E. coli was expressed in LB supplemented with 100 g/ml ampicillin as described (24), and EF-Ts from E. coli (using plasmid pCA24Ntsf obtained from the ASKA collection (23)) was expressed in LB supplemented with 30 g/ml of chloramphenicol. EF-Ts variants derived from E. coli were grown at 37°C in LB medium supplemented with 100 g/ml ampicillin. The cells containing the respective plasmid were grown to the mid-log phase (A 600 ϭ 0.6) and induced with 1 mM isopropyl thio-␤-galactoside. Following induction the temperature was reduced to 25°C for 2 h and then to 16°C overnight. The cells were harvested by centrifugation as described above.
Similar purification procedures were followed for EF-Tu from E. coli and P. aeruginosa, as well as EF-Ts variants from P. aeruginosa and EF-Ts from E. coli (23). The cell pellets were resuspended in 7 ml of buffer A (50 mM Tris-Cl pH 8.0 (4°C), 60 mM NH 4 Cl, 7 mM MgCl 2 , 300 mM KCl, 7 mM ␤-mercaptoethanol, 1 mM PMSF, 10 mM imidazole, 15% glycerol, and 50 M GDP) per gram of cells and lysed with 1 mg/ml lysozyme. Cell debris was removed through centrifugation at 3000 ϫ g for 30 min followed by 30,000 ϫ g for 45 min using a JA-16 rotor (Beckman). The respective protein was purified from the cleared lysate (S30 extract) using affinity chromatography (7 ml of Ni 2ϩ -Sepharose resin (GE Healthcare) equilibrated with buffer A). The resin was washed three times with 50 ml of buffer A and four times with 50 ml of buffer B (buffer A with 20 mM imidazole). The protein was eluted in 10 washes of 7 ml of buffer C (buffer A with 250 mM imidazole). The obtained His 6tagged protein was then concentrated via ultrafiltration (Vivaspin 20, 30,000 MWCO for EF-Tu and Vivaspin 20, 10,000 MWCO for EF-Ts (Sartorius)) and further purified by size exclusion chromatography (XK26/100 column; Superdex 75 prep grade (GE Healthcare)) equilibrated in buffer E (Buffer A without GDP). Fractions were analyzed by SDS-PAGE, and those containing protein that was Ͼ90% pure were pooled, concentrated via ultrafiltration (see above), and stored at Ϫ80°C. The final protein concentration was determined spectrophotometrically at 280 nm using a molar extinction coefficient of 32,900 M Ϫ1 cm Ϫ1 for EF-Tu (calculated using ProtParam (25)) and confirmed using Bradford Protein Assay (Bio-Rad).
EF-Ts variants from E. coli were purified using the following method. The cell pellets were resuspended in 7 ml of buffer E (25 mM Tris-Cl, pH 7.5 (4°C), 50 mM NH 4 Cl, 10 mM MgCl 2 , 0.1% PMSF, and 50 M GDP) per gram of cells and lysed with 1 mg/ml of lysozyme. Cell debris was removed via centrifugation as above to obtain a cleared lysate (S30 extract). The respective protein was purified from the cleared lysate using affinity chromatography (5 ml of chitin beads (New England BioLabs)) equilibrated with buffer E. The resin was washed three times with 50 ml of buffer E and five times with 50 ml of buffer F (20 mM Tris-Cl, pH 8.0 (4°C), 50 mM NH 4 Cl, 0.1 mM EDTA and 50 M GDP). Approximately 20 ml of buffer G (20 mM Tris-Cl, pH 8.0 (4°C), 50 mM NH 4 Cl, 0.1 mM EDTA, 60 mM DTT, and 400 mM KCl) was added to induce release of EF-Ts caused by the cleavage of the intein fusion followed by incubation for ϳ30 h at room temperature. The EF-Ts protein was eluted with 8 washes of 4.5 ml of buffer H (20 mM Tris-Cl, pH 8.0 (4°C), 50 mM NH 4 Cl, 0.1 mM EDTA, 400 mM KCl, and 50 M GDP). Elutions containing EF-Ts were pooled and concentrated via ultrafiltration (Vivaspin 20, 10,000 MWCO (Sartorius)) and subsequently subjected to size exclusion chromatography (XK26/ 100 column; Superdex 75 prep grade (GE Healthcare)) equilibrated in buffer A. The fractions were analyzed by SDS-PAGE, and those containing protein with a purity of Ͼ90% were pooled, concentrated via ultrafiltration (see above), and stored at Ϫ80°C for further use. The final EF-Ts protein concentration was determined using Bradford protein assay (Bio-Rad).
Preparation of EF-Tu⅐mant-GDP and EF-Tu⅐mant-GTP-EF-Tu⅐GDP was incubated with a 10-fold molar excess of mant-GDP or mant-GTP in buffer A for 30 min at 37°C. When mant-GTP was used, the incubations were carried out in the presence of 3 mM phosphoenolpyruvate and 0.1 mg/ml pyruvate kinase (Roche Diagnostics) to convert any nucleotide di-phosphate present into the nucleotide triphosphate form (26).
Preparation of Nucleotide-free EF-Tu and EF-Tu⅐EF-Ts-To promote the dissociation of GDP, which is tightly bound to EF-Tu and co-elutes with the factor during purification, EF-Tu⅐GDP was incubated in buffer I (25 mM Tris-Cl, pH 7.5 (20°C), 50 mM NH 4 Cl, 10 mM EDTA) for 30 min at 37°C. GDP and EF-Tu were separated using size exclusion chromatography (Acorn 10/300 GL column; Superdex 75 (GE Healthcare)) equilibrated in buffer J (25 mM Tris-Cl pH 7.5 (4°C), 50 mM NH 4 Cl). To prepare the respective EF-Tu⅐EF-Ts complexes, EF-Tu⅐GDP was incubated with equimolar amounts of EF-Ts in buffer I and subsequently treated as described above.
Rapid Kinetics-Fluorescence stopped flow measurements were performed using a stopped flow apparatus (KinTek SF-2004), as described in Ref. 27. Tryptophan fluorescence was excited at 280 nm and measured after passing through LG-305-F cut-off filters (NewPort). Mant-nucleotides were excited via FRET from the single tryptophan ( ex ϭ 280 nm) present in EF-Tu and measured after passing through LG-400-F cut-off filters (NewPort).
The apparent rate for the bimolecular association of mantnucleotides to nucleotide-free EF-Tu was determined by rapidly mixing 25 l of nucleotide free E. coli EF-Tu (0.3 M after mixing) or 25 l of nucleotide free P. aeruginosa EF-Tu (0.5 M after mixing) with 25 l of varying concentrations of mantnucleotides (ranging from 1 to 10 M after mixing) to ensure pseudo-first order conditions. The experiments were carried out at 20°C in buffer A. Fluorescence time courses were evaluated by fitting with a single exponential equation (Equation 1), where F(t) is the fluorescence at time t, F ∞ is the fluorescence signal at equilibrium, and k is the apparent rate constant of association (k app ) measured independently for each guanine nucleotide concentration. The apparent rate constants were then plotted as a function of nucleotide concentration.
The apparent rate for the bimolecular association of nucleotides to nucleotide free EF-Tu⅐EF-Ts was determined by rapidly mixing 25 l of nucleotide free EF-Tu⅐EF-Ts (final concentration, 0. To determine the association rate constants (k 2 ) for the EF-Tu⅐EF-Ts interaction, increasing concentrations of EF-Ts (final concentration, 1-5 M) were mixed with nucleotide-free EF-Tu (final concentration, 1 M), and the resulting decrease in tryptophan fluorescence was observed as a function of time. The resulting concentration dependence of the observed association rates yielded the respective k 2 .
Equilibrium binding constants (K d ) were calculated from the respective rate constants (k and k Ϫ ). The corresponding error ( Kd ) was calculated using Equation 2.
The calculations were performed using TableCurve (Jandel Scientific) and Prism (GraphPad Software).
Sequence Alignment-Protein primary sequences for EF-Ts and EF-Tu were obtained from the ExPASy website (Swiss-Prot database) (28), and the generated multiple sequence alignments (CLUSTALW (29)) were analyzed using GeneDoc (30).
Construction of EF-Tu⅐EF-Ts Models in Silico-A model of the E. coli EF-Tu⅐EF-Ts complex was constructed using the x-ray structure of this complex solved by Leberman and coworkers (16). This model lacked atomic coordinates for the N terminus (amino acids 1-8) and switch I regions of EF-Tu, which were presumably disordered in the crystal. A model for the N terminus of EF-Tu was obtained from the crystal structure of EF-Tu⅐GDPNP (7). The only conformation of switch I of EF-Tu available in the Protein Data Bank that is compatible with the presence of the C-terminal module of EF-Ts is found in a model of EF-Tu bound to the ribosome (31). The C terminus of EF-Ts(Chimera) was modeled by extending the E. coli EF-Ts model using Swiss-Pdb Viewer software (32). This model was solvated in a box of water molecules that extended 10 Å from the protein in each direction, and the aqueous system was minimized by relaxing the coordinates of protein, water, protein, and water in sequential energy minimization calculations for 10,000 steps each. A final all-atom minimization was performed until no change in total energy was measured over 1000 steps. To alleviate close contacts with the switch I region and generate a disordered conformation, the switch I and the N terminus were subjected to simulated annealing. During simulated annealing only the water molecules, the N terminus and switch I region of EF-Tu, as well as the C-terminal extension of EF-Ts(Chimera) (when present), were allowed to move. The melting phase involved heating the system from 100 to 1000 K by increasing the temperature 100 K every 5 ps. The system was maintained at 1000 K for 150 or 300 ps prior to cooling to 300 K by decreasing the temperature 10 K every 5 ps. Each cooling phase was performed 10 times with different random velocities to yield a total of 20 different cooling trajectories. After reaching 300 K, each trajectory was extended for an additional 150 ps at 300 K. Simulated annealing was performed with constant volume to maintain close proximity of water molecules and the protein elements. Following simulated annealing, the protein complex was solvated in a new box of water molecules, and the aqueous system was again minimized as before (see above). The electrostatic charge of the model was then neutralized by addition of 24 Na ϩ ions using the autoionize module in VMD (33).
Each model was then equilibrated in two independent simulations: at 300 and 350 K for 150 ps at 1 atm pressure using an NPT ensemble wherein the number of atoms, pressure, and temperature were kept constant. During equilibration the temperature was maintained using Langevin dynamics, and pressure was controlled by a Nosé-Hoover Langevin piston. At the start of each production phase simulation, the coordinates from the 350 K equilibration were cooled to 300 K by supplying the atomic velocities from the final step in the 300 K equilibration. Production phase simulations were carried out for 20 ns on each EF-Tu⅐EF-Ts model in an NPT ensemble; temperatures were maintained at 300 K by rescaling atomic velocities every 100 steps, and the Nosé-Hoover Langevin piston was used to maintain 1 atm pressure. All molecular dynamics simulations were carried out using the NAMD (34) software with a conservative step size of 0.5 fs/step and the CHARMM27 forcefield (35,36).
Analysis of Molecular Dynamics Simulations-Snapshots of each MD simulation were saved every 2 ps, and trajectories were fitted by the software Carma (37) to remove water molecules, as well as any rotations of the protein complex or translation of the center of mass. All metrics in the simulation were measured in VMD (33) using scripts written in-house (15,17). Further processing of MD data and subsequent plotting were performed using the software R (38). Quantitative analysis of hydrogen bonds was performed as previously described (15,17).