The ternary complex of bacterial elongation factor Tu (EF-Tu),

Transcription

1 Tuning the affinity of aminoacyl- to elongation factor Tu for optimal decoding Jared M. Schrader, Stephen J. Chapman, and Olke C. Uhlenbeck 1 Department of Molecular Biosciences, Northwestern University, Evanston, IL Contributed by Olke C. Uhlenbeck, February 8, 2011 (sent for review December 22, 2010) To better understand why aminoacyl-s (aa-s) have evolved to bind bacterial elongation factor Tu (EF-Tu) with uniform affinities, mutant s with differing affinities for EF-Tu were assayed for decoding on Escherichia coli ribosomes. At saturating EF-Tu concentrations, weaker-binding aa-s decode their cognate codons similarly to wild-type s. However, tighter-binding aa-s show reduced rates of peptide bond formation due to slow release from EF-Tu GDP. Thus, the affinities of aa-s for EF-Tu are constrained to be uniform by their need to bind tightly enough to form the ternary complex but weakly enough to release from EF-Tu during decoding. Consistent with available crystal structures, the identity of the esterified amino acid and three base pairs in the T stem of combine to define the affinity of each aa- for EF-Tu, both off and on the ribosome. The ternary complex of bacterial elongation factor Tu (EF-Tu), GTP and aminoacyl- (aa-) binds to the ribosome and participates in a multistep decoding pathway in which GTP is hydrolyzed, EF-Tu GDP is released, and the aa- enters the ribosomal A site (1 6). Although all elongator aa-s bind EF-Tu GTP with similar affinities (7 9), studies with misacylated s reveal that the protein shows substantial specificity for both the esterified amino acid and the body (10 12). The nearly uniform EF-Tu binding affinity observed for s acylated with their correct (cognate) amino acid occurs because the sequence of each has evolved to compensate for the variable thermodynamic contribution of the esterified amino acid. Thus, weak-binding esterified amino acids such as glycine and alanine have corresponding s that bind the protein tightly, while tight-binding amino acids such as tyrosine or glutamine have corresponding s that bind poorly. The crystal structure of Thermus aquaticus EF-Tu GTP bound to Saccharomyces cerevisiae Phe- Phe (13) reveals that the protein primarily forms extensive interactions with the helical phosphodiester backbone of the acceptor and Tstems of Phe. Recent protein (14) and (15, 16) mutagenesis experiments indicate that much of the specificity is the result of interactions made between three amino acids of EF-Tu and three adjacent base pairs in the T stem (16). Additional mutagenesis experiments indicate that the thermodynamic contribution of each of the three base pairs is independent of the others, making it possible to adjust the affinity of aa-s to EF-Tu in a predictable manner. Although a detailed structural and thermodynamic understanding of how EF-Tu achieves uniform binding with different aa-s is beginning to emerge, the underlying selective pressures that lead to uniform binding are less clear. While aa-s must bind EF-Tu tightly enough to participate in translation, the high intracellular concentration of EF-Tu (17) ensures that they do not significantly compete with one another for the protein. It therefore seems unlikely that a minimum threshold binding affinity would provide sufficient selective pressure to ensure the observed uniform binding properties. Additional selective pressure could potentially come at the step in translation where aa-s are released from EF-Tu GDP prior to peptide bond formation. This could limit the affinity to a maximal threshold, thereby selecting for uniform binding. The recent crystal structure of a ribosome-bound ternary complex trapped just before release of the aa- from EF-Tu (18) reveals that although the ribosome forms multiple contacts with both the and EF-Tu, the overall arrangement of the three domains of EF-Tu and its interface with aa- is quite similar to that of the free ternary complex. This suggests that the thermodynamic interplay between the esterified amino acid and the T-stem sequence may also occur during decoding and provide an additional selective pressure for uniform binding. This paper tests this possibility by examining the decoding properties of aa-s that have been altered to have either an increased or decreased affinity for EF-Tu by changing the T-stem sequence or introducing a different esterified amino acid. Results Derivatives of Escherichia coli Val GAC with Differing Affinities to E. coli EF-Tu. The E. coli Val GAC species that decodes GUC/U codons was chosen because its body has an intermediate affinity for EF-Tu (10), making it possible to both increase and decrease its binding affinity by mutating the T stem. In addition, this is one of the few in E. coli that lacks posttranscriptional modifications in its anticodon hairpin, thereby maximizing the likelihood that it will decode effectively as an unmodified. Based upon the binding affinities of individual base-pair substitutions in S. cerevisiae Phe to T. thermophilus EF-Tu (16), T-stem mutations of Val GAC (Fig. 1A) were designed to bind tighter (T1 to T3), weaker (W1 to W3), or similar (Ψ) to the wild-type (WT) sequence. After aminoacylation with [ 3 H]-valine, a ribonuclease protection assay was used to determine the dissociation rates of WT and the seven mutants of Val from E. coli EF-Tu GTP at 4 C (Fig. 1B, Table 1) (19, 20). Dissociation rates for several s were also determined at 20 C, and k off values were ninefold to 15-fold faster than at 4 C (Table S1). As previously observed with other s,(12, 21, 22) the unmodified WT Val GAC bound EF-Tu very similarly to native, fully modified Val GAC consistent with the absence of modifications in the EF-Tu binding site. The seven mutations showed the desired broad range of binding affinity to E. coli EF-Tu that is similar to the range in affinities observed for different bodies with T. thermophilus EF-Tu (10, 16). T1 to T3 bound from 0.5 to 1.0 kcal mol tighter, W1 to W3 bound from 1.0 to 2.0 kcal mol weaker, and Ψ bound nearly the same as the WT Val. These measured values correlate closely with the values predicted by adding the ΔΔG values obtained from data measuring the binding of single base-pair substitutions in S. cerevisiae Phe to T. thermophilus EF-Tu (Fig. 1C). This not only supports the notion that the thermodynamic contributions of the three T-stem base pairs are independent but indicates that both E. coli and T. thermophilus EF-Tu interact with aa-s in a thermodynamically similar manner. Author contributions: J.M.S. and O.C.U. designed research; J.M.S. and S.J.C. performed research; S.J.C. contributed new reagents/analytic tools; J.M.S. analyzed data; and J.M.S. and O.C.U. wrote the paper. The authors declare no conflict of interest (such as defined by PNAS policy). 1 To whom correspondence should be addressed. o-uhlenbeck@northwestern.edu. This article contains supporting information online at doi: /pnas /-/dcsupplemental. BIOCHEMISTRY PNAS Early Edition 1of6

2 Fig. 2. Steps of EF-Tu catalyzed selection of aa- by the ribosome This simplified scheme, adapted from ref. 28, indicates the steps that were assayed to evaluate the activities of the mutants of Val- Val on E. coli ribosomes. Fig. 1. Engineering Val- Val mutants with altered EF-Tu binding affinities. (A) E. coli Val GAC with mutated T-stem base pairs in box. T-stem sequences of WT and seven mutant transcripts are on the right. (B) Individual rates of dissociation (k off ) for WT ( s 1 ), W1 ( s 1 ), T1 ( s 1 ), and native ( s 1 ) Val- Val from E. coli EF-Tu at 0 C in RB buffer using the ribonuclease protection assay. Means and standard errors from multiple determinations are in Table 1. (C) Comparison of the free energies of binding (ΔG ) between Val- Val T-stem mutants to E. Coli EF-Tu with measured or calculated free energies of the same mutants in S. cerevisiae Phe- Phe binding to T. thermophilus EF-Tu. The line corresponds to a least square fit with a slope of 1.1. Table 1. k off and k pep values of Val- Val mutants Rate of aa- dissociation from EF-Tu 10 4 k off ðs 1 Þ Free energy of aa- Binding to EF-Tu ΔG (kcal mol) Rate of peptide bond formation k pep ðs 1 Þ Native 5.6 (±) (±) (±) 0.4 WT 13 (±) (±) (±) 0.7 T1 2.0 (±) (±) (±) 0.02 T2 3.4 (±) (±) (±) 0.04 T3 5.3 (±) (±) (±) 0.2 Ψ 22 (±) (±) (±) 0.8 W1 110 (±) (±) (±) 2.3 W2 130 (±) (±) (±) 2.5 W3* 950 (±) (±) (±) 0.1 *k off value extrapolated from lower NH 4 Cl concentrations. Decoding Properties of Val T-stem Mutants. Three quantitative assays were used to evaluate the ability of ternary complexes made with native Val and the unmodified WT, T1, and W1 s to decode the GUC codon on E. coli ribosomes containing fmet- fmet in the P site (Fig. 2, Table 2). The experiments were performed at 20 C in a buffer containing 10 mm MgCl 2 to ensure that the unmodified s were fully folded (23 25). The first assay measures the equilibrium binding constant of the ternary complex to ribosomes by using the active site mutant of EF-Tu(H84A) to prevent GTP hydrolysis (26, 27). The resulting equilibrium constant reflects the initial binding, codon recognition, and GTPase activation steps of the decoding pathway (28). As shown in Fig. 3A, all of the Val derivatives show very similar K d values. Thus, neither the modifications nor the T-stem mutations significantly alter the initial interaction of the ternary complex with the ribosome. The second assay uses ternary complexes made with γ-[ 32 P]- GTP to measure the rate of GTP hydrolysis as a function of encoded ribosome concentration. This rate reflects a conformational change in EF-Tu that occurs before bond cleavage (28 30). As shown in Fig. 3B, the rate of GTP hydrolysis at 2.5 μm ribosomes differed by about twofold between the four Val- Val derivatives tested. When the rates of GTP hydrolysis were plotted as a function of ribosome concentration (Fig. 3C), values of K 1 2 and k GTPmax (Table 2) also differ by about twofold. Thus, either increasing (T1) or decreasing (W1) the affinity of Val- Val for EF-Tu by about 1.0 kcal mol had little effect on the ability of the ternary complex to catalyze GTP hydrolysis on the ribosome. The modifications also had little effect on k GTP. The final assay measures k pep, the rate of ribosome-catalyzed formation of fmetval- Val. This final step in decoding encompasses the release of both inorganic phosphate and Val- Val from EF-Tu GDP, the accommodation of the Val- Val into the A site and the formation of the peptide bond. It remains uncertain which one of these substeps limits the value of k pep for wild-type aa-s, and it may depend upon the assay conditions and the aa- that are used (28, 31, 32). Here k pep is measured with ternary complexes containing 3 -[ 32 P]- labeled Val- Val using a P1 nuclease assay (33). The ternary complexes containing native Val, the WT, and W1 derivatives all show similar k pep values and extents of reaction (Fig. 3D, Table 2). Strikingly, the ternary complex containing tight-binding T1 Val is fully active but shows a 14-fold slower k pep value than WT. This slow k pep can potentially be explained by its slow release from EF-Tu GDP (Fig. 2). To further test this, k pep values for the remaining five Val derivatives were determined (Table 1). The values of k pep for both the weaker-binding W2 mutation and the very weak-binding W3 mutation are similar to WT, although higher concentrations of EF-Tu are required in the latter case to achieve high levels of dipeptide. The Ψ mutation, which binds similarly to WT, also shows a similar k pep value. In contrast, T2 and T3, which both bind to EF-Tu more tightly, are fully active but show significantly slower k pep values. Taken together, the data indicate that while removing the modifications or weakening the affinity to EF-Tu has little effect on the rate of peptide bond formation, tightening the interaction with EF-Tu clearly reduces it. Rescuing the Slow k pep with a Sequence-Specific EF-Tu Mutation. If the reduced k pep values observed with the three hyperstabilized Val derivatives are the result of slow release from EF-Tu GDP, it should be possible to reverse this effect with an EF-Tu mutation that destabilizes binding. To test this, we made use of a contact between EF-Tu and that stabilizes the Table 2. Performance of Val- Val mutants in decoding A/T site affinity K d (nm) Maximal rate of GTP hydrolysis k GTPmax (s 1 ) Half maximal [70S] for GTP hydrolysis K 1 2 (μm) Native 1.1 (±) (±) (±) 1.9 WT 0.8 (±) (±) (±) 1.4 T1 0.5 (±) (±) (±) 1.4 W1 2.1 (±) (±) (±) 1.2 Standard errors for A/T site K d and k pep are calculated from at least three independent Standard errors for k GTPmax and K 1 2 are fit to the Michaelis Menton equation with at least six k GTPapparent determinations per curve. 2of6 Schrader et al.

3 Fig. 3. Activities of Val- Val mutants in decoding. Individual determinations of the performance of WT, W1, T1, and native Val- Val on encoded ribosomes at 20 C in RB using three assays. Means and standard errors of multiple determinations are in Table 2. (A) Binding affinities of H84A (GTPase deficient) EF-Tu ternary complexes to the ribosomal A/T site using <0.23 nm ternary complex. Lines are best fit simple binding curves with K d ¼ 0.8 nm, 1.5 nm, 0.7 nm, and 1.4 nm, respectively. (B) Rates of γ-[ 32 P]- GTP hydrolysis (k GTP ) at 200 nm ternary complex and 2.5 μm ribosomes. Lines are best fit single exponential rates with k GTP ¼ 10 s 1, 16 s 1, 7.5 s 1, 11 s 1, respectively. (C) Ribosome saturation curve fit to the Michaelis Mentenequation for K 1 2 and k GTPmax (see Table 2). (D) Rates of peptide bond formation (k pep ) measured at 25 nm ternary complex and 1 μm ribosomes. Lines are single exponential fits with k pep ¼ 5.7 s 1, 8.1 s 1, 0.34 s 1, and 3.8 s 1, respectively. interaction in a sequence-specific manner (15). In the ternary complex structure of E. coli EF-Tu with E. coli Phe- Phe (PDB ID code 1OB2), E378 makes a hydrogen bond with the amino group of G63 (Fig. 4A). Mutation of the homologous E390 in T. thermophilus EF-Tu to an alanine reduced the binding affinity of the protein to s containing a G at either position 51 or 63 but not to s lacking a G (15). A similar hydrogen bond is observed between E390 of T. thermophilus EF-Tu and G63 of E. coli Thr- Thr in the structure of the ternary complex bound to T. thermophilus ribosomes in the presence of kirromycin (Fig. 4B) (18). Because this structure depicts EF-Tu after GTP Fig. 4. The EF-Tu interface maintains similar interactions on the ribosome. (A) Cocrystal structure of E. coli EF-Tu bound to S. cerevisiae Phe- Phe with kirromycin (PDB) highlighting the interactions made between E378, Q329, and T338 10B2 (E390, Q341, T350 in T. thermophilus) with base pair and the 2 OH of bases 64 and 65. (B) Cocrystal structure of T. thermophilus EF-Tu bound to E. coli Thr- Thr stalled on the ribosome with kirromycin post GTP hydrolysis (PDB ID code 2WRN) (18). hydrolysis, the hydrogen bond could affect the rate of release of aa- prior to accommodation. A more recent structure of the ribosome-bound ternary complex before GTP hydrolysis is very similar (34). To confirm that the E378A mutation of E. coli EF-Tu shows the expected specificity for binding to the Val derivatives, k off values were determined for ternary complexes of this mutant protein with WT, Ψ, and the three tight Val derivatives (Table 3). The k off for the WT Val- Val was not significantly affected by the E378A mutation, consistent with the fact that it contains a U51-A63 base pair. However, the Ψ, T1, and T2 derivatives, which contain G51-C63, dissociate fivefold to sixfold faster from EF-Tu(E378A) than they do from WT EF-Tu (Table 1), consistent with the presence of a stabilizing hydrogen bond with G51. In contrast, the tight-binding derivative T3, which contains an A51-C63 pair, is not affected by the E378A mutation. While it is unclear how the A51-C63 mismatch stabilizes EF-Tu binding, it appears to do so without the use of E378 and therefore is a valuable control. k pep values for WT, Ψ, T1, T2, and T3 versions of Val- Val were then determined using ternary complexes containing EF-Tu (E378A) (Table 3). The value of k pep for the WT with the mutant EF-Tu (4.4 s 1 ) was identical to that observed using WT EF-Tu (4.5 s 1 ). Because WT Val- Val has a U51-A63 base pair and is not stabilized by the E378 residue, this establishes that EF-Tu(E378A) is fully active in decoding. However, the k pep values for the tightly binding T1 and T2 s with EF- Tu(E378A) are 2.7 s 1 and 3.3 s 1, respectively. Because these values are substantially faster than the 0.32 s 1 and 0.55 s 1 observed with WT EF-Tu, it is clear that the EF-Tu mutation that weakens binding to the hyperstabilized T1 and T2 Val derivatives also reverts their abnormally slow k pep values. This experiment shows that the hydrogen bond between E378 and G63 in the ribosome-bound ternary complex contributes to its stability and thereby influences the rate of release of an aa- into the A site. Two critical control experiments are the k pep measurements of the Ψ and T3 s with EF-Tu(E378A). Despite the fact that Ψ contains a G51-C63 and thus binds more weakly to the E378A protein, k pep for Ψ with EF-Tu(E378A) (3.7 s 1 ) is similar to WT EF-Tu (4.5 s 1 ). This experiment shows that the E378A mutation only stimulates k pep when the is hyperstabilized for EF-Tu binding. The second control shows that the slow k pep observed for the T3 Val with WT EF-Tu (1.4 s 1 ) remains slow with EF-Tu(E378A) (0.6 s 1 ). Because the A51-C63 pair present in T3 is not stabilized by E378, its k pep is unaffected by the E378A mutation. This control shows that when the ternary complex is hyperstabilized in a different way, the slow k pep no longer requires the stabilizing contact using E378. Taken together, the experiments with EF-Tu(E378A) confirm that the slow k pep values observed for the hyperstabilized Val derivatives are caused by the slow release from EF-Tu. Table 3. k off and k pep values of Val- Val mutants with EF-Tu (E378A) Rate of aa- dissociation from EF-Tu 10 4 k off (s 1 ) Free energy of aa- binding to EF-Tu ΔG (kcal mol) Rate of peptide bond formation k pep (s 1 ) WT 18 (±) (±) (±) 1.5 T1 10 (±) (±) (±) 0.4 T2 19 (±) (±) (±) 0.3 T3 5.2 (±) (±) (±) 0.3 Ψ 140 (±) (±) (±) 0.6 BIOCHEMISTRY Schrader et al. PNAS Early Edition 3of6

4 Contribution of the Esterified Amino Acid to k pep. While the above experiments establish that the sequence of the T stem can influence the rate of release from EF-Tu GDP during decoding, it is not clear whether the identity of the esterified amino acid can influence this step in a manner similar to what is observed in the formation of the ternary complex. In the ribosome-bound ternary complex (18), the position of the esterified amino acid is different than in the free ternary complex, so it is uncertain whether the thermodynamic effect of the amino acid would be similar. One way to detect an effect of the amino acid on k pep is to introduce the hyperstabilizing T1 sequence into other E. coli s. Because the specificity of EF-Tu for is primarily defined by the T-stem sequence (16), the T1 chimeras are expected to have a similar tight affinity to EF-Tu GTP, and any difference in k off should primarily reflect differences in the esterified amino acid. This T1 upgrade mutation was introduced into the bodies of E. coli Gly CCC, Ala GGC, Phe GAA, and Tyr GUA, and the resulting chimeras were acylated with their corresponding cognate amino acids. k off values for the chimeras were compared with the WT aa-s in Table 4. In general, the T1 mutation stabilized the binding of each aa- to EF-Tu GTP in the expected manner. Because Gly CCC and Ala GGC are tight-binding sequences (10), their T1 chimeras only bound EF-Tu twofold and fourfold tighter than their WT counterparts. The T1 chimera of Phe GAA, an intermediate-binding, binds 56-fold tighter than the WT Phe. For the T1 chimera of the very weakbinding Tyr GAC, dissociation from EF-Tu GTP was too slow to obtain a reliable k off, even when the ionic strength was increased to 3 M NH 4 Cl or the temperature raised to 20 C. An estimated k off for the T1 Tyr- Tyr, calculated based on its sequence (16), was 62-fold tighter than WT Tyr. k pep values for the eight s are also shown in Table 4. The four unmodified WTaa-s had k pep values that were reasonably similar to WT Val (4.5 s 1 ). The slightly slower rates for Tyr GUA (1.4s 1 ) and Phe GAA (2.3 s 1 ) probably reflect the absence of modifications in the anticodon hairpin, which is known to weaken ribosome binding (35). The slightly slower rate for Gly CCC (1.4 s 1 ) probably reflects the lower pka of glycine (36). The k pep values for the four T1 derivatives were all slower than the corresponding WT s and correlated closely with their k off rates. Thus, k pep values for the T1 derivatives of Gly- Gly CCC and Ala-Ala GCC were only twofold to threefold slower than the WT s. However, k pep of the T1 mutant of Phe- Phe GAA was 13-fold slower than that of WT Phe- Phe. Finally, the T1 mutation of Tyr- Tyr shows a very slow k pep of s 1, which is 57-fold slower than measured for WT Tyr- Tyr. The 104-fold range in k pep for the T1 mutations correlates remarkably well with the greater than Table 4. k off and k pep values of WT and T1 mutations Rate of aa- dissociation from EF-Tu 10 4 k off (s -1 ) Free energy of aa- binding to EF-Tu ΔG (kcal mol) Rate of peptide bond formation k pep (s 1 ) Ala- Ala WT 21 (±) (±) (±) 0.6 Ala- Ala T1 4.8 (±) (±) (±) 0.1 Gly- Gly WT 29 (±) (±) (±) 0.6 Gly- Gly T1 17 (±) (±) (±) 0.3 Phe- Phe WT 35 (±) (±) (±) 1.1 Phe- Phe T1 0.6* (±) (±) (±) 0.11 Tyr- Tyr WT 9.9 (±) (±) (±) 0.5 Tyr- Tyr T (±) *Beyond limit of assay, estimated value. k off value extrapolated from higher NH 4 Cl concentrations. 100-fold difference in their k off values, supporting the view that the compensatory relationship between the esterified amino acid and body observed for the free ternary complex occurs in an identical way during decoding. Although the above data strongly suggests that the identity of the esterified amino acid can contribute to the rate of release from EF-Tu GDP during decoding, experiments with misacylated-s are needed to show this directly. We made use of the observation that Val can be misacylated by high concentrations of yeast-phers, and the resulting Phe- Val binds EF-Tu twofold to threefold tighter than Val- Val (10). Table 5 shows that the k pep values of WT and T1 are 4.3- and 2.2-fold slower when phenylalanylated than when valylated, which agrees reasonably well with the twofold to threefold difference reported in k off (10). In contrast, the k pep of the phenylalanylated W1 (5.1 s 1 ) is the same as when it is valylated (5.4 s 1 ) because the relatively large (1.1 kcal mol) destabilizing effect of the W1 mutation more than offsets the relatively small (0.5 kcal mol) effect of introducing the phenylalanine, so the aa- is not hyperstabilized. In order to show that misacylation can also lead to a faster k pep, the acceptor stem of the T1 derivative of Tyr was mutated to contain a G3-U70 base pair so it could be alanylated by AlaRS(37). As shown in Table 5, k pep for this Ala Tyr is 61-fold faster than when esterified with the tightbinding tyrosine. These experiments with misacylated s confirm that the esterified amino acid contributes to the rate of release from EF-Tu GDP during decoding in a compensatory manner that is very similar to that observed for the formation of ternary complex. Discussion Several conclusions of this paper can be conveniently summarized by plotting k off, the rate of dissociation of aa- from EF-Tu GTP, versus k pep; the rate of peptide bond formation, for the many different aa-s studied (Fig. 5). Because the values differ over such a large range, both rates are plotted on a logarithmic scale. All five of the WT s aminoacylated with their cognate amino acids cluster in a narrow region of k off, consistent with previous studies showing that E. coli aa-s bind EF-Tu with similar affinity(19). The slightly greater variance in k pep for these s is primarily due to their lack of posttranscriptional modifications in the anticodon loop that influence their activity (27, 32). When aa-s were modified such that k off was slower than WT, the early decoding steps were not affected, but their k pep values were reduced. These slow k pep values are best explained by the slow release of these hyperstabilized aa-s from EF-Tu GDP becoming rate-limiting for peptide bond formation. Experiments using the E378A mutation of EF-Tu to revert the slow k pep values in the expected sequenceselective manner strongly support this conclusion. In contrast, when the T stem of Val- Val was mutated such that k off was faster than the WT sequence (variants W1 to W3), all of the steps in decoding were unaffected, provided that sufficiently high EF-Tu concentrations were present to permit the ternary complex to form. Unlike for the tight s, k pep values for the weak aa-s were not correlated with the value of k off, indicating that EF-Tu release does not limit k pep for any of them. Because the native aa-s all have k pep values similar to the weak s, it appears that their sequences have evolved to Table 5. k pep of misacylated s Rate of peptide bond formation k pep (s 1 ) Phe- Val WT 1.5 (±) 0.7 Phe- Val T (±) 0.04 Phe- Val W1 5.1 (±) 0.8 Ala- Tyr T1 1.4 (±) 0.4 4of6 Schrader et al.

5 Fig. 5. Comparison of k off and k pep for all aa-s tested. Rates of dissociation from E. coli EF-Tu for: WT aa-s, Val- Val mutants, T1 mutants (s Ala, Gly, Phe, and Tyr), and misacylated s are compared to rates of peptide bond formation of the same aa-s. The solid line is a linear fit to all tight-binding mutants with k off < s 1, while the dotted line is a linear fit to WT and weaker-binding mutants. release from EF-Tu at a rate that is either faster or similar to the rate-limiting step. An interesting feature of Fig. 5 is that k pep and k off are proportional for all the tight-binding aa- mutations tested. While the k off data in Fig. 5 were obtained at 4 C to permit the weaker aa-s to be measured, k off values for a selected number of tight s were determined at the same conditions used for k pep (Table S1). These experiments show that k off and k pep remain proportional for the different aa-s, but k pep is consistently about 100-fold faster than k off. While both rates reflect dissociation of the tight aa-s from the protein, the structures of the complexes are very different. In the ribosome-bound complex, EF-Tu is in the GDP form and the aa- is in a distorted, presumably high-energy conformation, which are both expected to destabilize the protein interface (18, 38). However, the proportional effects on k off and k pep for both a substitution of the esterified amino acid and several T-stem mutations suggest that the entire interface between EF-Tu and aa- is thermodynamically similar both on and off the ribosome. Additional protein and aa- mutations will be needed to confirm this point. It is striking that the proportionality between k off and k pep is maintained over a wide range of values with several species whose affinities were increased with different T-stem sequences and esterified amino acids. This indicates that, despite the differences in the two structures, the specificity of EF-Tu for the different esterified amino acids and bodies previously established for the free ternary complex also occurs at the release step on the ribosome. This explains why aa-s have evolved to have a uniform affinity for EF-Tu. Two different steps in the decoding pathway combine to constrain the affinity to a narrow range. Each aa- must initially bind EF-Tu tightly enough to efficiently form the ternary complex but also must bind weakly enough to be released from the protein after GTP hydrolysis during decoding. As a result, the WT aa-s have evolved to be in the general region of Fig. 5 where the affinity is as tight as possible without slowing k pep. A mutation of any that either tightens or weakens its affinity to EF-Tu will impact translational performance and presumably be selected against. Since nearly all the EF-Tu residues that contact aa- are conserved among bacteria, it is likely that the conclusions made here for E. coli are generally applicable. Indeed, the thermodynamic effects of T-stem mutations determined here using E. coli EF-Tu are virtually identical to those previously determined for T. thermophilus (10, 15, 16). However, as discussed elsewhere, the sequences of individual species vary considerably among bacteria because multiple T-stem sequences can have a similar EF-Tu affinity (16, 39, 40). By introducing the T-stem sequence of a tight into a weak, the resulting chimera may bind EF-Tu so tightly that k pep becomes very slow. An extreme example of this was the T1 mutation of Tyr- Tyr that gave a k pep of sec 1. This chimera may be useful to more accurately place the EF-Tu release step in the decoding pathway. In addition, because a high-resolution structure is available that depicts the ternary complex just prior to release (18), detailed structure-function studies are now possible. By stabilizing binding to EF-Tu, the T1 chimeras may also be helpful for improving the efficiency of incorporating certain nonnatural amino acids into protein (41, 42). Finally, the decoding properties of the T1 chimera of Tyr- Tyr are interesting to compare with the highly specialized system that has evolved to introduce selenocysteine into proteins. Like EF-Tu, its distant orthologue SelB has evolved to selectively bind its cognate Sec- Sec in a manner that is specific for both the esterified amino acid and the body (43). Like the T1 Tyr mutation, SelB binds Sec- Sec extremely tightly. It will be interesting to see whether k pep for selenocysteine incorporation is also unusually slow. Materials and Methods Materials. Tight-coupled 70S ribosomes from E. coli MRE600 cells were prepared as described (44) and purified pellets were resuspended in buffer RB (50 mm HEPES [ph 7.0], 30 mm KCl, 70 mm NH 4 Cl, 10 mm MgCl 2,and 1 mm DTT). Ribosomes were flash frozen and stored at 80 C and activated as previously described (45). mrna derivatives of the T4 gp32 mrna fragment 5 -GGCAAGGAGGUAAAAAUGXXXGCACGU-3, where XXX indicates the A site codon, were purchased from Dharmacon. E. coli EF-Tu, EF-Tu(H84A) and EF-Tu(E378A) containing an N-terminal histidine tag and TEV protease sequence were over expressed in BL21-Gold (DE3) cells (Agilent) and purified by Ni-NTA chromatography as in (46). The concentration was determined using Pierce A660 protein assay using BSA as a standard. Purified E. coli fmet was purchased from Sigma Aldrich. Transcribed s were prepared as described (16) and stored in TE buffer (10 mm Tris-HCl [ph 7.6], 0.1 mm EDTA) at concentrations greater than 100 μm. Native E. coli Val was purified by hybridization using biotinylated DNA oligonucleotides designed in (47). A 1 ml reaction of 4 mg ml unfractionated E. coli (Roche) was heated to 65 C in 10 mm EDTA for 10 min and 20X SSC buffer was added to a final concentration of 1X SSC (150 mm NaCl, 15 mm NaCitrate ph 7.0). The appropriate 3 biotinylated DNA was added at a concentration of 17 μm and annealed at 65 C for 15 min and at room temperature for 20 min. 500 μl streptavidin resin (Sigma Aldrich) in 1X SSC buffer was added and incubated at room temperature for 20 min to capture the probe. The resin was washed in a spin column three times with 1 ml 0.1X SSC, and the was eluted from the resin by the addition of 250 μl water at 65 C for 10 min. The purified was then EtOH precipitated and additionally purified by 10% denaturing PAGE and stored in a TE buffer. Aminoacylation of s with [ 3 H]-amino acids (Amersham) was performed as in (16). 3 -[ 32 P] labeling of and aminoacylation was performed as previously described(33) with typical aminoacylation yields of >70%. Misacylated Val GAC and Tyr GUA (G3-U70) derivatives were prepared as described (10, 11). Methods. The equilibrium binding affinity of ternary complexes to programmed E. coli ribosomes was determined essentially as described (27). Ternary complexes were formed with 150 nm EF-Tu(H84A) and <0.7 nm [3-32 P] aa- and added to a series of ribosome concentrations programmed with excess mrna and fmet to give final concentrations of 50 nm EF-Tu(H84A), <0.23 nm aa- and from 600 nm to 0.5 um encoded ribosomes. After incubation for 2 min at 20 C, samples were filtered using a dual filter system and filters were quantified using a PhosphorImager (Storm 820). After subtraction of the counts on the lower filter due to 32 P deacylated, the data were fit to a simple binding isotherm using Kaleidagraph. Sample data is shown in Fig. 2A, and the data in Table 2 reports at least three separate measurements using different preparations of programmed ribosomes, fmet- fmet and ternary complex. BIOCHEMISTRY Schrader et al. PNAS Early Edition 5of6

6 The rate of [ 3 H]-aa- dissociation from EF-Tu (k off ) was determined using an RNase protection assay as described (16) with the following exceptions: E. coli EF-Tu GTP was activated for only 20 min (instead of 3 h for T. thermophilus EF-Tu), and reactions were carried out in RB buffer. The fraction of [ 3 H]-aa- bound by EF-Tu over time was fit to a single exponential using Kaleidagraph. mutations with k off values greater than 0.02 s 1 or less than s 1 were measured at multiple lower or higher NH 4 Cl concentrations, and their k off was extrapolated to 70 mm NH 4 Cl present in RB (12). All measurements were performed in triplicate. k off values were converted to ΔG values using k on ¼ M 1 s 1 (19), (21). Apparent rates of GTP hydrolysis were measured as in (27) (30). Ternary complex was formed in RB buffer by incubating 2 μm EF-Tu GDP, 0.6 μm, 50 μm [γ- 32 P]-GTP, 3 mm phophoenol pyruvate, 12 U ml pyruvate kinase, 125 μm Valine, 4 mm ATP, 10 U ml yeast pyrophosphatase, and 0.25 μm ValRS for 1 h at 37 C and then purified through two P30 spin columns (Biorad). Ternary complex was mixed with equivalent volume of 1.0 to 8.0 μm programmed ribosomes in a rapid mixing device (Kintek), quenched in 40% formic acid at the desired time points, and separated using TLC (30). After subtraction of the excess EF-Tu GTP present, the time course of the fraction of GTP hydrolyzed was fit to single exponential equation to obtain k GTP. The concentration dependence was fit to the Michaelis Menton equation to give k GTPmax, the rate at saturation. Because each point on the k GTPmax curve represents k GTP hydrolysis determined using various batches of ribosomes, fmet, and ternary complex, the errors are significant. However, errors are much less when a single set of components were used to obtain each k GTP. The rate of peptide bond formation was performed essentially as described (27). Programmed ribosomes were prepared by combining 2 μm ribosomes, 6 μm mrna and 4 μm fmet- fmet and mixed in a rapid mixing device (Kintek) with equivalent volume of 50 nm 32 P-labeled ternary complex (1 μm EF-Tu GTP and 50 nm [3-32 P] aa-). Reactions were quenched at set time points with 5 mm NaOAc ph mm EDTA, digested with P1 nuclease, and separated on PEI cellulose TLC plates as in (33). After correcting for the fraction of fmet which was not charged or formylated, the fraction of dipeptide formed vs. time was fit to a single exponential by Kaleidagraph. All measurements were performed in triplicate. 1. Agirrezabala X, Frank J (2009) Elongation in translation as a dynamic interaction among the ribosome,, and elongation factors EF-G and EF-Tu. Q Rev Biophys 42: Marshall RA, Aitken CE, Dorywalska M, Puglisi JD (2008) Translation at the singlemolecule level. Annu Rev Biochem 77: Rodnina MV, Wintermeyer W (2001) Fidelity of aminoacyl- selection on the ribosome: Kinetic and structural mechanisms. Annu Rev Biochem 70: Krab IM, Parmeggiani A (1998) EF-Tu, a GTPase odyssey. Biochim Biophys Acta 1443: Zaher HS, Green R (2009) Fidelity at the molecular level: Lessons from protein synthesis. Cell 136: Schmeing TM, Ramakrishnan V (2009) What recent ribosome structures have revealed about the mechanism of translation. Nature 461: Louie A, Ribeiro NS, Reid BR, Jurnak F (1984) Relative affinities of all Escherichia coli aminoacyl-s for elongation factor Tu-GTP. J Biol Chem 259: Asahara H, Uhlenbeck OC (2005) Predicting the binding affinities of misacylated s for Thermus thermophilus EF-Tu.GTP. Biochemistry 44: Ott G, et al. (1990) Ternary complexes of Escherichia coli aminoacyl-s with the elongation factor Tu and GTP: Thermodynamic and structural studies. Biochim Biophys Acta 1050: Asahara H, Uhlenbeck OC (2002) The specificity of Thermus thermophilus EF-Tu. Proc Natl Acad Sci USA 99: Dale T, Sanderson LE, Uhlenbeck OC (2004) The affinity of elongation factor Tu for an aminoacyl- is modulated by the esterified amino acid. Biochemistry 43: LaRiviere FJ, Wolfson AD, Uhlenbeck OC (2001) Uniform binding of aminoacyl-s to elongation factor Tu by thermodynamic compensation. Science 294: Nissen P, et al. (1995) Crystal structure of the ternary complex of Phe-Phe, EF-Tu, and a GTP analog. Science 270: Sanderson LE, Uhlenbeck OC (2007) Exploring the specificity of bacterial elongation factor Tu for different s. Biochemistry 46: Sanderson LE, Uhlenbeck OC (2007) The base pair of confers specificity for binding by EF-Tu. RNA 13: Schrader JM, Chapman SJ, Uhlenbeck OC (2009) Understanding the sequence specificity of binding to elongation factor Tu using mutagenesis. J Mol Biol 386: Bosch L, Kraal B, Van der Meide PH, Duisterwinkel FJ, Van Noort JM (1983) The elongation factor EF-Tu and its two encoding genes. Prog Nucleic Acid Res Mol Biol 30: Schmeing TM, et al. (2009) The crystal structure of the ribosome bound to EF-Tu and aminoacyl-. Science 326: Louie A, Jurnak F (1985) Kinetic studies of Escherichia coli elongation factor Tu-guanosine 5 -triphosphate-aminoacyl- complexes. Biochemistry 24: Vorstenbosch EL, Potapov AP, de Graaf JM, Kraal B (2000) The effect of mutations in EF-Tu on its affinity for as measured by two novel and independent methods of general applicability. J Biochem Biophys Methods 42: Nazarenko IA, Harrington KM, Uhlenbeck OC (1994) Many of the conserved nucleotides of (Phe) are not essential for ternary complex formation and peptide elongation. EMBO J 13: Liu JC, Liu M, Horowitz J (1998) Recognition of the universally conserved 3 -CCA end of by elongation factor EF-Tu. RNA 4: Perret V, et al. (1990) Conformation in solution of yeast Asp transcripts deprived of modified nucleotides. Biochimie 72: Vermeulen A, McCallum SA, Pardi A (2005) Comparison of the global structure and dynamics of native and unmodified val. Biochemistry 44: Sampson JR, Uhlenbeck OC (1988) Biochemical and physical characterization of an unmodified yeast phenylalanine transfer RNA transcribed in vitro. Proc Natl Acad Sci USA 85: Daviter T, Wieden HJ, Rodnina MV (2003) Essential role of histidine 84 in elongation factor Tu for the chemical step of GTP hydrolysis on the ribosome. J Mol Biol 332: Ledoux S, Uhlenbeck OC (2008) Different aa-s are selected uniformly on the ribosome. Mol Cell 31: Pape T, Wintermeyer W, Rodnina MV (1998) Complete kinetic mechanism of elongation factor Tu-dependent binding of aminoacyl- to the A site of the E. coli ribosome. EMBO J 17: Rodnina MV, Fricke R, Kuhn L, Wintermeyer W (1995) Codon-dependent conformational change of elongation factor Tu preceding GTP hydrolysis on the ribosome. EMBO J 14: Cochella L, Green R (2005) An active role for in decoding beyond codon: anticodon pairing. Science 308: Blanchard SC, Gonzalez RL, Kim HD, Chu S, Puglisi JD (2004) selection and kinetic proofreading in translation. Nat Struct Mol Biol 11: Pavlov MY, et al. (2009) Slow peptide bond formation by proline and other N-alkylamino acids in translation. Proc Natl Acad Sci USA 106: Ledoux S, Uhlenbeck OC (2008) [3-32P]-labeling with nucleotidyltransferase for assaying aminoacylation and peptide bond formation. Methods 44: Voorhees RM, Schmeing TM, Kelley AC, Ramakrishnan V (2010) The mechanism for activation of GTP hydrolysis on the ribosome. Science 330: Ashraf SS, et al. (1999) Single atom modification (O S) of confers ribosome binding. RNA 5: Johansson M, et al. (2010) ph-sensitivity of the ribosomal peptidyl transfer reaction dependent on the identity of the A-site aminoacyl-. Proc Natl Acad Sci USA 108: Hou YM, Schimmel P (1988) A simple structural feature is a major determinant of the identity of a transfer RNA. Nature 333: Dell VA, Miller DL, Johnson AE (1990) Effects of nucleotide- and aurodox-induced changes in elongation factor Tu conformation upon its interactions with aminoacyl transfer RNA. A fluorescence study. Biochemistry 29: Saks ME, Conery JS (2007) Anticodon-dependent conservation of bacterial gene sequences. RNA 13: Schrader J, Saks M, Uhlenbeck OC (2011) The specific interaction between aminoacyls and elongation factor Tu. Ribosomes: Structure, Function & Dynamics, eds M Rodnina, R Green, and W Wintermeyer (Springer, NewYork), in press. 41. Guo J, Melancon CE, III, Lee HS, Groff D, Schultz PG (2009) Evolution of amber suppressor s for efficient bacterial production of proteins containing nonnatural amino acids. Angew Chem Int Ed Engl 48: Nakata H, Ohtsuki T, Abe R, Hohsaka T, Sisido M (2006) Binding efficiency of elongation factor Tu to s charged with nonnatural fluorescent amino acids. Anal Biochem 348: Paleskava A, Konevega AL, Rodnina MV (2010) Thermodynamic and kinetic framework of selenocysteyl-sec recognition by elongation factor SelB. J Biol Chem 285: Powers T, Noller HF (1991) A functional pseudoknot in 16S ribosomal RNA. EMBO J 10: Fahlman RP, Dale T, Uhlenbeck OC (2004) Uniform binding of aminoacylated transfer RNAs to the ribosomal A and P sites. Mol Cell 16: Sanderson LE, Uhlenbeck OC (2007) Directed mutagenesis identifies amino acid residues involved in elongation factor Tu binding to yeast Phe-(Phe). J Mol Biol 368: Miyauchi K, Ohara T, Suzuki T (2007) Automated parallel isolation of multiple species of non-coding RNAs by the reciprocal circulating chromatography method. Nucleic Acids Res 35:e24. 6of6 Schrader et al.