Histidine 66 in Escherichia coli Elongation Factor Tu Selectively Stabilizes Aminoacyl-tRNAs*

Background: The cognate-esterified amino acid is critical for optimal delivery of aminoacyl-tRNAs to the ribosome by EF-Tu. Results: Mutation of His-66 in EF-Tu alters the specific binding of many, but not all aminoacyl-tRNAs, but does not affect decoding. Conclusion: His-66 is critical for the specificity of EF-Tu for the esterified amino acid. Significance: Selective mutation of His-66 could improve the incorporation efficiency of unnatural amino acids into proteins. The universally conserved His-66 of elongation factor Tu (EF-Tu) stacks on the side chain of the esterified Phe of Phe-tRNAPhe. The affinities of eight aminoacyl-tRNAs were differentially destabilized by the introduction of the H66A mutation into Escherichia coli EF-Tu, whereas Ala-tRNAAla and Gly-tRNAGly were unaffected. The H66F and H66W proteins each show a different pattern of binding of 10 different aminoacyl-tRNAs, clearly showing that this position is critical in establishing the specificity of EF-Tu for different esterified amino acids. However, the H66A mutation does not greatly affect the ability of the ternary complex to bind ribosomes, hydrolyze GTP, or form dipeptide, suggesting that this residue does not directly participate in ribosomal decoding. Selective mutation of His-66 may improve the ability of certain unnatural amino acids to be incorporated by the ribosome.

Bacterial elongation factor Tu (EF-Tu) 2 in its activated, GTPbound form binds all elongator aminoacyl transfer RNAs (aa-tRNAs) to form the ternary complexes that are substrates for the ribosome. The presence of the esterified amino acid is required for the aa-tRNAs to bind tightly to the protein (1)(2)(3)(4), and experiments with misacylated tRNAs have established that the thermodynamic contribution of the esterified amino acid depends significantly on the identity of its side chain (5)(6)(7)(8). A hierarchy of thermodynamic contributions for the different amino acids has been defined ranging up to as much as 2.8 kcal/mol between the "weak" amino acids such as glycine or aspartate and the "tight" amino acids such as tyrosine or glutamine (6,8). Because tRNA sequences have evolved to thermodynamically compensate for the variable contributions of the esterified amino acids, correctly acylated aa-tRNAs bind to EF-Tu with similar affinities (8 -10).
The differing thermodynamic contributions of the esterified amino acids can at least partially be explained by the interac-tions made between the side chain of the amino acids and a cleft or pocket formed between domains 1 and 2 that is large enough to fit all of the amino acids. In the cocrystal structure of the yeast Phe-tRNA Phe and Escherichia coli EF-Tu⅐GMPPNP ( Fig.  1), the side chain of the esterified phenylalanine stacks on His-66 at the top of the binding pocket, whereas its amino group forms hydrogen bonds with the carboxyl oxygen of Asn-273 and the backbone nitrogen from Phe-261 (11). In a structure of Cys-tRNA Cys bound to the very similar Thermus aquaticus EF-Tu, the position of the cysteine side chain is quite similar to that of phenylalanine with the ␤ carbons superimposing and the SH group in the plane of the phenylalanine ring (12). Two additional ribosome-bound ternary complex structures in preand post-GTP hydrolysis states are available that contain Trp-tRNA Trp and Thr-tRNA Thr , respectively (13,14). Although Trp-tRNA Trp before GTP hydrolysis maintains a very similar environment around the esterified amino acid as the free ternary complexes, the post-GTP hydrolysis structure of Thr-tRNA Thr shows subtle alterations of the pocket residues possibly induced by GTP hydrolysis. It is not known how other esterified amino acids fit into the amino acid binding pocket, although its asymmetric environment is expected to stabilize and/or sterically position each amino acid in a unique manner. In addition, two more distal acidic residues, Glu-215 and Asp-216, are primarily responsible for the overall negative charge of the pocket and help to account for the relatively weak binding of esterified aspartate and glutamate (6,8,15).
One goal of this paper is to evaluate the contribution of His-66 in E. coli EF-Tu to the specificity of the protein for binding different esterified amino acids. It is known that the H66A and H66L mutations of E. coli EF-Tu destabilize the binding to Phe-tRNA Phe (16,17), suggesting that the imidazole side chain can stabilize the esterified Phe. Here we test how the H66A mutation affects the binding of tRNAs esterified with other amino acids and how other mutations of His-66 can affect the specificity for aa-tRNAs.
A second goal of this paper is to use presteady state kinetics to evaluate the ability of the H66A mutant of EF-Tu to support tRNA decoding on E. coli ribosomes. Although the position of His-66 does not change substantially during initial binding, substantial structural rearrangements occur in nearby regions of both the tRNA and EF-Tu upon GTP hydrolysis (13,14), suggesting that the H66A mutation may affect the kinetics of decoding. Although earlier experiments indicated that the H66A mutation had little effect on polyphenylalanine synthesis (16), this steady state assay could potentially obscure a kinetic effect in one of the sub-steps in decoding.

EXPERIMENTAL PROCEDURES
tRNA Labeling and Aminoacylation Reactions-All purified native tRNAs were purchased from Sigma, except tRNA Gly and tRNA Ala , which were purchased from Subriden. The tRNAs were 3Ј-32 P-labeled and aminoacylated for ribosome experiments as described in Ledoux and Uhlenbeck (18). All [ 3 H]amino acids containing aminoacylation reactions for EF-Tu binding experiments were performed as described in Dale et al. (6).
EF-Tu Mutagenesis-The plasmid containing the E. coli tufb gene with a cleavable N-terminal His 6 linker was provided by Rachel Green (Johns Hopkins University School of Medicine). Mutations were introduced via the QuikChange XL site-directed mutagenesis kit (Stratagene). The sequence of each EF-Tu variant was confirmed by DNA sequencing.
Protein and Ribosome Purification-Tight-coupled 70 S ribosomes from E. coli MRE600 cells were purified as described in Powers and Noller (19). EF-Tu from E. coli was overexpressed and purified as described in Sanderson and Uhlenbeck (20) with the following exceptions; the EF-Tu plasmids were transformed into BLR(DE3) cells and grown to an A 600 of 0.5 followed by an incubation with 1 mM isopropyl 1-thio-␤-D-galactopyranoside for another 2 h to an A 600 of 1.2. In buffer A, 4-(2-aminoethyl)benzenesulfonyl fluoride (ABSF) was replaced with a protease inhibitor mixture (Roche Applied Science complete mini). The cell lysate was cleared first by centrifugation at 10,000 ϫ g for 30 min then at 185,000 ϫ g for 90 min. Fractions containing EF-Tu were pooled and dialyzed against buffer A to remove imidazole before incubating with His 6 -tobacco etch virus protease for 3 h at 24°C. The reaction was then run over a nickel-nitrilotriacetic acid column to remove uncleaved EF-Tu and tobacco etch virus protease, and the eluate was stored in buffer A with 50% glycerol.
EF-Tu Assays-The dissociation rates (k off ) of [ 3 H]aa-tRNAs from EF-Tu⅐GTP were determined as described in Sanderson and Uhlenbeck (20) at 0°C in buffer A containing 50 mM Hepes (pH 7.0), 20 mM MgCl 2 , 50 mM NH 4 Cl, 5 mM DTT, 20 M GTP, 3 mM phosphoenolpyruvate, and 50 g/ml pyruvate kinase. The association rate constants for aa-tRNAs to EF-Tu⅐GTP were determined as described in Schrader et al. (21) at 0°C in buffer A. The equilibrium dissociation constant (K D ) was measured in buffer A as described in Sanderson and Uhlenbeck (20). The apparent rate of GTP hydrolysis was performed in buffer B (50 mM Hepes (pH 7.0), 30 mM KCl, 70 mM NH 4 Cl, 10 mM MgCl 2 , and 1 mM DTT) with 1 M ribosomes as described in Schrader et al. (22). The rate of dipeptide formation (k pep ) was performed in buffer B as described in Schrader et al. (22). The ternary complex binding assay was performed in buffer B as described in Ledoux and Uhlenbeck (23).

RESULTS
The H66A Mutation Only Destabilizes Binding to Certain aa-tRNAs-As shown in Fig. 1, the imidazole ring of His-66 is stacked between the esterified phenylalanine and the side chain of Val-79 and is positioned 4 -6 Å from three potential hydrogen bond partners Gln-97, Asn-273, and Glu-215. As a result, the H66A mutation would be expected to enlarge the amino acid binding pocket and thereby reduce the affinity for esterified amino acids with aromatic or polar side chains without affecting the affinity for smaller esterified amino acids which are unlikely to be stabilized by the imidazole. To test this prediction, a ribonuclease protection assay was used to determine the dissociation rates of 10 [ 3 H]aa-tRNAs from wild-type E. coli EF-Tu and the H66A protein ( Table 1). As had been reported previously for both Thermus thermophilus and E. coli EF-Tu (8,9), the dissociation rates of different aa-tRNAs from wild-type E. coli EF-Tu are quite similar, with k off values ranging from 0.08 Ϯ 0.02 min Ϫ1 for Tyr-tRNA Tyr to 0.21 Ϯ 0.06 min Ϫ1 for Lys-tRNA Lys . In contrast, the H66A protein shows a broader range of k off values, ranging from 0.16 Ϯ 0.02 min Ϫ1 for Gly-tRNA Gly to 1.5 Ϯ 0.38 min for Trp-tRNA Trp . The values of k off for Gly-tRNA Gly and Ala-tRNA Ala are quite similar for the wildtype and mutant proteins, indicating that the removal of the imidazole ring does not globally distort the amino acid binding pocket. However, most of the other aa-tRNAs, especially those esterified with aromatic amino acids, show significantly faster k off values from the H66A protein, indicating that His-66 par-  ticipates in stabilizing many of these aa-tRNAs. The destabilization observed for Phe-tRNA Phe is similar to the value measured previously using a different assay (16).
To assure that neither the identity of the esterified amino acid nor the H66A mutation affected the rate of association of EF-Tu with aa-tRNA, k on values for Gly-tRNA Gly and Tyr-tRNA Tyr binding to the wild-type and H66A proteins were determined by measuring the rate of formation of ternary complex at varying EF-Tu concentrations and using the slope of the resulting linear plot to obtain k on (21,24). As summarized in Table 2, k on values for both aa-tRNAs were similar with wildtype EF-Tu and for Gly-tRNA Gly with H66A EF-Tu. Because this assay does not easily yield an accurate k on value for the weak complex of Tyr-tRNA Tyr with H66A EF-Tu, k on values were verified indirectly by measuring the fraction of ternary complex formed at equilibrium as a function of EF-Tu concentration to give K D and then by calculating k on ϭ k off /K D . Although these calculated k on values were all about 2-fold lower, all four complexes gave similar k on values despite a more than 55-fold variation in k off (Table 2). Taken together, it appears that neither the identity of the esterified amino acid nor the H66A mutation affects k on .
A constant value of k on and the values of k off determined in Table 1 permit calculation of the ⌬G°of formation for each ternary complex. These data, depicted graphically in Fig. 2, clearly show that whereas the binding of aa-tRNAs to wild-type EF-Tu is fairly uniform, the H66A mutant protein destabilizes all of the aa-tRNAs by differing amounts with the exception of Gly-tRNA Gly and Ala-tRNA Ala . This clearly establishes that His-66 contributes substantially to the specificity of EF-Tu for different esterified amino acids.
His-66 Substitutions Alter Specificity for aa-tRNAs-In an attempt to obtain a mutation of EF-Tu, which bound certain aminoacyl-tRNAs tighter than wild-type, His-66 was mutated to Trp, Phe, Tyr, and Arg. Although these bulkier amino acids may not position themselves in the same orientation as histidine, the size of the pocket is large enough to accommodate them, and they may be better than histidine at stabilizing hydrophobic amino acids. Although all bacterial and most archaeal EF-Tus have a histidine at the orthologus position, it is interesting that about 15% of archaea have a Phe or, rarely, a Tyr in the corresponding site in their somewhat different amino acid binding pockets. When the dissociation rates of Gly-tRNA Gly from each of the four His-66 mutations were determined, the H66W and H66F proteins had k off values very similar to wild-type and the H66A protein (Tables 1 and 3). However, the values of k off for H66Y and H66R were 3-and 6-fold faster than wild type (data not shown). Because the esterified glycine would be more than 5 Å away from the residue 66, Gly-tRNA Gly would not be expected to be affected by His-66 mutations. Thus, the faster off rates for the H66Y and H66R proteins indicates that the structure of EF-Tu was compromised outside of the amino acid binding pocket. Because of this, the H66Y and H66R proteins were not studied in further detail.
k off values for the 10 different aa-tRNAs from the H66W and H66F mutants are presented in Table 3, and the corresponding ⌬G°values are shown in Fig. 2. It is clear the affinities of the H66W and H66F proteins for the different aa-tRNAs are different from each other as well as from the wild-type and the H66A proteins. For example, Phe-tRNA Phe and Tyr-tRNA Tyr bind much better to the H66W and H66F proteins than to the H66A protein, consistent with a productive stacking interaction between the aromatic esterified amino acid and the aromatic Trp-66 and Phe-66 residues. However, it is interesting that His-66, present in the wild-type protein, is even better than Trp-66 or Phe-66 at binding the esterified Phe and Tyr even though it would not be expected to stack any better. It is possible that the partial protonation of His-66 (25) may aid its positioning and thereby improve its ability to stabilize aromatic amino acids.
Additional intriguing examples of ⌬G o differences between the four proteins include the fact that the ⌬G o of Glu-tRNA Glu is reduced to a similar extent in all three mutant EF-Tus, whereas the ⌬G o of Gln-tRNA Gln is only reduced in the H66A mutation and even slightly stabilized in the H67F mutation. Ala-tRNA Ala and Gly-tRNA Gly bind to all four proteins with a similar ⌬G o . In general, Fig. 2 makes it clear that the identity of residue 66 is an important feature in establishing the specificity of aa-tRNAs to EF-Tu.
Decoding Properties of H66A Mutant-Three assays were used to evaluate the role of His-66 in ribosomal decoding. The apparent binding affinity (K D ) of ternary complex to ribosomal entry site, the rate of GTP hydrolysis (k GTP ), and the rate of dipeptide bond formation (k pep ) were measured for wild-type

constants for wild-type and H66A EF-Tu
Experiments were performed in buffer A at 0°C. S.E. of k on and K D were calculated from four independent determinations. k off /K D ϭ k on was calculated using experimental k off from Table 1 and K D . ND, not determined.

Direct determination
Gly-tRNA Gly 6.8 Ϯ 2 5.9 Ϯ 0.9 Tyr-tRNA Tyr 7.6 Ϯ 2 N D and H66A EF-Tu using both Ala-tRNA Ala and Phe-tRNA Phe on their respective cognate codons. These two aa-tRNAs were chosen because the esterified phenylalanine interacts with His-66, whereas the esterified alanine does not. To measure K D , the H84A mutation was introduced into the wild-type and H66A EF-Tu proteins, thereby blocking GTP hydrolysis. Kinetic experiments have established that decoding by the H84A mutant is quite similar to the wild-type protein up to the point of GTP hydrolysis, including similar rates of initial binding and GTPase activation (26). Thus, the value of K D reflects both the reversible binding of the ternary complex and the subsequent conformational change in both the ribosome and EF-Tu associated with GTPase activation. Control experiments measuring the k off values from both the H84A and H84A/H66A versions of EF-Tu show that the H84A mutation does not significantly alter the binding affinity of aa-tRNAs to EF-Tu (data not shown). A synergistic effect between the two mutations was not expected as the two histidines are more than 14 Å apart. The K D values for Ala-tRNA Ala and Phe-tRNA Phe ternary complexes binding to the ribosomal entry site are presented in Table 4. Both H84A ternary complexes bind with K D values similar to those previously measured (23,26). When H66A/H84A EF-Tu is used, the K D values are only slightly different, indicating that the H66A mutation does not significantly alter the early events of initial selection on the ribosome. The final step of initial selection, GTP hydrolysis, was then investigated for H66A. Apparent rates of GTP hydrolysis (k GTP ) were measured at 1 M ribosomes, a concentration close to K m maximizing any potential effect of the mutation. The rates of GTP hydrolysis were similar for Ala-tRNA Ala with wild-type and H66A EF-Tu, suggesting that His-66 does not significantly affect the final step of initial selection, k GTP . Phe-tRNA Phe did not form a stable enough complex with the H66A protein to allow purification of the ternary complex from excess [ 32 P]GTP, thereby prohibiting the measurement of k GTP .

Wild-type EF-Tu H66A EF-Tu
To investigate whether His-66 alters the rate of peptide bond formation (k pep ), Ala-tRNA Ala and Phe-tRNA Phe were both assayed with wild-type and H66A EF-Tu on E. coli ribosomes ( Table 4). As previously observed, both aa-tRNAs undergo k pep with wild-type EF-Tu with equivalent rates (23). When assayed with H66A EF-Tu, no impact for Ala-tRNA Ala was observed on k pep as expected by the lack of interaction between the residues. This suggests that the loss of the stacking energy between the aminoacyl phenylalanine and His-66 does not dramatically affect the rates of accommodation or peptide bond formation on the ribosome. Interestingly, despite a reduced affinity of H66A EF-Tu for Phe-tRNA Phe off the ribosome, only a slight impact was observed in k pep similar to what had been previously observed in polyphenylalanine synthesis (16).

DISCUSSION
Because His-66 is the residue that contributes the most surface area to the amino acid binding pocket of EF-Tu, it is not surprising that His-66 is critical in determining the specificity of the protein for different esterified amino acids. The mutation of His-66 to alanine in E. coli EF-Tu causes larger amino acids to be destabilized by as much as 18-fold, whereas the smaller amino acids such as glycine or alanine are not affected at all. Although mutation of Glu-215 and Asp-216 in the back of the pocket have also been found to alter the specificity for certain esterified amino acids (6,15), the effects are considerably smaller. Understanding the physical mechanism of stabilization by the His-66 residue is not straightforward. Although the structure of the ternary complex suggests the presence of stabilizing stacking interactions between the imidazole and aromatic side chains, our data suggest several non-aromatic amino acids also appear stabilized by His-66. However, an understanding of the amino acid binding specificity also requires consideration of the structures of the unbound forms of EF-Tu⅐GTP and aa-tRNA. The x-ray structure of the free EF-Tu⅐GTP indicates that the position of His-66 with respect to the neighboring side chains is identical to the ternary complex. Although no high resolution structure of a tRNA with an ester-  ified amino acid is available, there is evidence that aromatic esterified amino acids stack upon the 3Ј terminal adenosine of tRNA in a very different position from where it is in the ternary complex (27,28). Other amino acids would be expected to interact with the terminal A to differing degrees. Because this interaction must be disrupted to form the ternary complex, this complicates a simple interpretation of the contribution of His-66 to amino acid binding specificity simply in terms of the ternary complex structure. Natural mutations of His-66 have not been found in bacteria presumably because if this occurred, the relative affinities of many aa-tRNAs would be altered, thereby compromising their equal access to the ribosome. The only way to reestablish uniform binding of aa-tRNAs in such an EF-Tu mutation would be to simultaneously mutate the T-stems of multiple tRNA genes. Because this would not occur easily, His-66 is universally conserved in bacteria. In archaeal EF-1␣, the orthologus histidine is present in a majority of organisms, although many species contain a phenylalanine. In the case of eukaryotic EF-1␣, this position is conserved as Leu. Although it is unclear whether thermodynamic compensation between the amino acid and tRNA body even occurs with archaeal and eukaryotic EF-1␣, if it does, the specific hierarchies of esterified amino acid and tRNA affinity are likely to be different.
Despite structural evidence that His-66 makes interactions with the esterified amino acid before and after GTP hydrolysis on the ribosome (13,14), the mutation of H66A does not appear to significantly impact the early or late events of aa-tRNA selection. For Ala-tRNA Ala and Phe-tRNA Phe we found that H66A did not significantly affect either ribosome entry site binding or k pep . This suggests that the mutation of H66A does not alter the global structure of the protein. Additionally, H66A with Ala-tRNA Ala did not impact the rate of GTP hydrolysis, suggesting that His-66 plays a role in discriminating the esterified amino acid side chains in aa-tRNA binding but does not affect the catalytic functions of EF-Tu on the ribosome. These results confirm previous evidence showing that H66A mutation does not significantly affect polyphenylalanine synthesis (16).
The experiments presented here provide some insight in efforts to adapt EF-Tu to optimally incorporate unnatural amino acids (Uaa) into proteins. Like the 20 natural amino acids, each esterified Uaa is expected to fit into the amino acid binding pocket in EF-Tu in a unique manner and thereby have an associated characteristic contribution to the ⌬G°of binding of the Uaa-tRNA to the protein. If the resulting ⌬G°is too weak, the ternary complex containing the Uaa will not form efficiently, and incorporation into protein will be reduced. If the ⌬G°is too tight, a reduced rate of release from EF-Tu⅐GDP on the ribosome may reduce the rate of incorporation of the Uaa into protein (22). One way to adjust the ⌬G°to the optimal value is to modify the T-stem sequence of the chosen tRNA in a way that compensates for the altered affinity of the Uaa (29). However, some Uaas that are very large or have unusual structures may weaken binding to wild-type EF-Tu by so much that no stable complex forms even when the tightest possible T-stem is used. In these cases, it will be necessary to mutate the amino acid pocket to accommodate the Uaa. However, because this will alter the specificity for the natural amino acids, such a modified EF-Tu must be used in the presence of the wild-type EF-Tu.
An initial partially successful example of such an "orthogonal" EF-Tu system was the discovery that the E215A or D216A mutations of E. coli EF-Tu improved incorporation of DL-2-anthraquinonylalanine, L-2-pyrenelyalanine, and L-1-pyrenylalanine into protein compared with the wild-type EF-Tu (30). However, the overall incorporation efficiency of these Uaas remained low, possibly because the activity of EF-Tu was compromised by the mutations.
A second example of an orthogonal EF-Tu system is an elegant selection of an E. coli EF-Tu variant that can function with phosphoserine, a Uaa that does not work well with the wildtype protein (31). In this case six residues in the amino acid binding pocket, including His-66, were changed to create an EF-Sec that promoted efficient incorporation of a single phosphoserine residue into several proteins. However, this system was not efficient at introducing multiple phosphoserines into proteins, again perhaps because the intrinsic activity of EF-Sec was not very high.
Our experience with various mutations of His-66 indicate that whereas it is possible to enlarge the amino acid binding pocket without compromising the function of EF-Tu, the identity of the introduced amino acid can be critical to the activity of the protein. One convenient way to test whether the function of a mutant EF-Tu is altered is to test it using Gly-tRNA Gly or Ala-tRNA Ala , which are minimally sensitive to mutations in the pocket. If these aa-tRNAs are fully active, the mutations do not disrupt the structure of the binding site.