Importance of single molecular determinants in the fidelity of expanded genetic codes

TyrRS.

Fersht et al. identified two residues of the Bacillus stearothermophilus (Bst) TyrRS, Tyr 34, and Asp 176, as hydrogen bonding partners for the hydroxyl -OH group of substrate tyrosine (Fig. 1) (9, 34, 35). Their observations were based on crystallographic evidence (36), and their observations can be applied to the highly homologous Methanocaldococcus jannaschii (Mj) TyrRS (37). In the majority of Mj TyrRSs evolved to charge NAAs, these two residues (Mj numbering Tyr32 and Asp158) have been mutated to more hydrophobic residues as a requirement for achieving high fidelity incorporation of the NAA versus native tyrosine (Table S1).

Whereas most Mj TyrRS variants have changed Tyr32 to a smaller, more hydrophobic residue, the TyrRS evolved to charge 3-aminoTyrosine (3-aminoTyr) featured Gln, Glu, Lys, or Arg at this position (38). This can be rationalized from the overall structural similarity between 3-amino Tyr and Tyr and their relative ability to engage in salt bridging interactions. There are just three reports of an evolved TyrRS mutant preserving either of the Mj wild-type (WT) TyrRS residues Tyr32 or Asp158. The two TyrRSs that were reported to charge tRNA_CUA with GlcNAc-Ser or GalNAc-Thr retained Tyr32 or Asp158, respectively (39, 40). However, both reports have been corrected, (31, 32) and it seems these TyrRSs were likely mischarging tyrosine in place of the NAA (41). Similarly, the Mj TyrRS reported to incorporate 3-OH Tyr retained Asp158 (33). The latter report did not include SDS-PAGE analysis as is customary, but a mass spectrum was reported. Two masses were observed in the spectrum at 18,432.3 and 18,448.5 Da. It was previously established that a mass of ∼18,432 Da corresponds to incorporation of tyrosine into their common test protein myoglobin (42). These authors suggested that the signal at 18432.3 was due to fragmentation of the 3-hydroxy tyrosine side chain through, “….loss of oxygen, or oxygen and a proton.” However, this is not consistent with reported fragmentation patterns for the parent molecule (43, 44). We contend that it is more likely that the 18432.3 Da signal is due to competing incorporation of tyrosine, which is consistent with other mass spectra but also the importance of this particular TyrRS’s Asp158 binding the Tyr -OH group (vide infra).

The importance of Tyr32 as a “gatekeeper” residue of TyrRS fidelity, especially in NAA systems, was investigated by reverting three evolved TyrRSs to code for Tyr at this position. The three TyrRSs chosen corresponded to the enzymes reported for p-iodoPhe (45), p-acetylPhe (28), and p-boronoPhe (46). The evolved enzymes were reported with either a Leu or Ser in place of Tyr at position 32. Each revertant and cognate tRNA_CUA were expressed along with a UAG mutant for T4 lysozyme in the presence (+) and absence (-) of each NAA. The results suggest high fidelity tRNA_CUA charging for the original enzymes with p-iodoPhe and p-boronoPhe, and marginal fidelity for charging with p-acetylPhe (Fig. 2). The results also show that the three TyrRS revertants would appear to charge tRNA_CUA in the absence of NAA. The presence of full-length protein in the absence of NAA (Fig. 2) is consistent with WT enzyme activity where tRNA_CUA is charged with a canonical amino acid, presumably tyrosine. It was also observed that the revertants displayed similar fidelity to the AARS reported to have high fidelity for GlcNAc-Ser. Again, this is consistent with the fact that the TyrRS variant reported to have high fidelity for GlcNAc-Ser possessed the same Tyr32 residue found in the WT enzyme (39).

Söll has suggested that an exogenous AARS produced in Escherichia coli requires as little as 1% of the activity of the WT enzyme to complement an auxotrophic phenotype (15). Our results suggest that the activity of the GlcNAc-SerRS and the revertants are well above this threshold. Other insight into why the GlcNAc-SerRS and revertant TyrRSs would charge tRNA_CUA with a canonical amino acid comes from Fersht et al.’s reports on the effects of mutating each of a Bst TyrRS’s molecular determinants. A Tyr34Phe mutation (Bst TyrRS numbering; equivalent to Mj Tyr32) resulted in a 15-fold decrease in the specificity of Tyr versus Phe (9). Attempts to mutate Asp176 (the equivalent of Mj Asp158) resulted in inactive enzymes, suggesting the latter residue is likely the more important molecular determinant of fidelity (34).

Crystallographic evidence suggests that the mutations to a TyrRS, such as those found in the GlcNAc-Ser TyrRS (total mutations: Glu107Pro, Asp158Cys, Ile159 Tyr, and Leu162Arg) would be expected to reconstitute some of the hydrogen bonding network lost by the Asp158Cys mutation (again, Tyr32 was retained). The marginal fidelity that the p-acetylPheRS demonstrated in the absence of p-acetylPhe (Fig. 2) may be similarly rationalized from this enzyme’s Ile159Cys and Leu162Arg mutations forming such a secondary hydrogen bonding network for the substrate Tyr -OH group. Previously, direct characterization of these TyrRSs was performed using isothermal calorimetry (41). K_d values of 330–522 μM were reported for the p-acetylRS with p-acetylPhe; the same experiment with the p-acetylRS and Tyr was below the limit of detection of the ITC method. Conversely, a K_d of 662–2659 μM was reported for the GlcNAc-Ser TyrRS with Tyr, but attempts to measure the binding affinity of this TyrRS mutant with GlcNAc-Ser were also below the limit of detection. Such results, along with Fersht’s analysis and available crystallographic evidence, can be used to rationalize why one amino acid (i.e., p-boronoPhe) but not others (i.e., GlcNAc-Ser or 3-OH Tyr) would be incorporated by high fidelity enzymes.

TrpRSs: Analysis of TrpRS Sequence and Structure.

TrpRS sequences from E. coli, Bacillus subtilis (Bs), and Bst are highly homologous and these enzymes are known to cross react with the others’ tRNA^Trp because of shared identity elements (47, 48). These bacterial TrpRSs also do not require mutations to charge tRNA^Trp with a variety of Trp analogs including: 4-nitroTrp, 4-flouroTrp, 5-fluoroTrp, 5-hydroxyTrp (5-OH Trp), and other azaTrp isosteres (13, 47, 49–52). The reported high fidelity incorporation of 5-OH Trp by a mutated Bs TrpRS thus caught our attention (25). In the report, the TrpRS was subjected to a single mutagenesis (V144P). The mutant enzyme’s amino acid binding pocket retained residue Asp133. This residue is homologous to residue Asp132 of Bst TrpRS and Asp 176 of the Bst TyrRS, and like in the TyrRSs, structurally forms a hydrogen bond with the original substrate; and with this TrpRS, the bond is to the indole N1 nitrogen (3, 16, 53) We suggest that because this residue serves as the primary molecular determinant of amino acid specificity for this enzyme, that any NAA-incorporating system based on this, or similar TrpRS should target this residue to achieve high fidelity incorporation over the canonical substrate.

In the original report V144 was targeted for saturation mutagenesis to create a library of 19 variants. Because the structure of the Bs TrpRS was unavailable, this residue was chosen based on a Bst TrpRS structure. A screen of 19 total variants identified a V144P variant as having high fidelity for 5-OH Trp. This result is unprecedented; all other evolved TyrRSs exhibiting high fidelity (10) have been selected from TyrRS libraries featuring randomization of 4–8 active site residues^†. Furthermore, their target of incorporation, 5-OH Trp, is a well-documented substrate for the WT Bs TrpRSs (54). The burden on the V144P mutant is therefore twofold: it must ablate the enzyme’s inherent affinity for both Trp and 5-OH Trp and yet, switch the enzyme’s fidelity solely to 5-OH Trp. This is a different starting point from which each evolved Mj TyrRS mutant is selected; for example, the NAAs listed in Table S1 are not appreciable substrates for the Mj WT TyrRS.

There are reports from Tirrell and Yokoyama of AARSs requiring only one mutation to incorporate Tyr, Phe and Trp analogues with high fidelity (12, 55, 56). Yokoyama and coworkers demonstrated preferential incorporation of 3-iodo-L-tyrosine into a protein. Their SDS-PAGE analysis (figure 2 in ref. 12), however, shows that full-length protein is produced in the absence of 3-iodo-L-tyrosine. This result suggests their AARS exhibits only moderate fidelity in the absence of NAA. This is explainable because they only mutated one of the Bst TyrRS’s molecular determinants (Y34, equal to Y32 in Mj). Tirrell and coworkers (55, 56) reported incorporation of several Phe and Trp analogs with high fidelity. This required the use of auxotrophic (Phe) strains of E. coli, presumably to keep the pool of competing endogenous amino acids at a minimum. Furthermore, the kinetic parameters of their single-point mutant AARSs suggest that incorporation of Trp and/or Phe would occur in media with normal levels of these canonical amino acids. Whereas one AARS mutation may be sufficient to achieve incorporation of a NAA, either auxotrophic or depleted media conditions appear to be necessary for achieving high fidelity incorporation of an NAA, and neither of these conditions were reported for the incorporation of 5-OH Trp.

The Bs V144P TrpRS mutant crystallized in the presence of both Trp and 5-OH Trp. Crystals diffracted to 2.8 Å (Table S2), but electron density corresponding to either amino acid was not observed. A structural alignment of our V144P TrpRS structure with a Bst TrpRS structure with bound Trp shows that the alpha helix containing Asp133, again the major determinant for Trp specificity, is not appreciably displaced by the mutation at residue 144 (Fig. 3). Although the mutation does occur in the vicinity (∼6 Å) of the 5-position of Trp, the effect on the secondary structure appears rather modest. The most dramatic structural changes appear to occur where Asp146 binds ribose of ATP (57).

To rationalize the high fidelity of their V144P variant, Schultz and coworkers modeled the active site of the Bst TrpRS V144P variant with 5-OH Trp (25). They found that 5-OH Trp preferred to reorient its indole ring ∼180° so the indole nitrogen participates in a hydrogen bond with Ser7 and the 5-OH interacts with Asp132. We have conducted modelling experiments and found that both the Bst WT and Bs V144P TrpRSs prefer Trp and 5-OH Trp in the same orientation found in Schimmel, Crane, and Carter’s crystallographic work of homologous enzymes (Fig. 4) (17, 57, 58). Both models placed the indole nitrogen in close proximity to the key Asp residue, consistent with the residue’s role in determining fidelity. The docking results only became more variable when the critical Asp residue was modified with an Asp → Ala mutation (Fig. S1).

Bs WT and V144P TrpRS Complementation of an Auxotrophic Phenotype.

The E. coli strain trpS4040 is a Trp auxotroph that has been used to assess the activity of exogenous TrpRSs (15, 59). Briefly, the strain has a debilitated TrpRS gene that makes it difficult to culture in the absence of Trp. Growth on media enriched with Trp elevates the intracellular concentration to the point where tRNA^Trp become sufficiently charged. Alternatively, expressing an AARS gene that is able to efficiently charge E. coli tRNA^Trp complements the auxotrophic phenotype under Trp-starved conditions. Our results show complementation after 48 h on either 0.05 μM Trp or 5 μM 5-OH Trp plates in the presence of the Bs WT TrpRS, but not with the V144P mutant (Fig. 5). Even in liquid media supplemented at 50 μM 5-OH Trp, the Bs WT TrpRS allowed cultures to readily grow, whereas the cultures expressing the V144P mutant were still no different from the control experiments (Fig. S2).

Isothermal Calorimetry Suggests Lack of Fidelity of 5-OH TrpRS in Vitro.

We used isothermal calorimetry (ITC) to directly characterize the fidelity of the Bs TrpRSs (Fig. 6 and Fig. S3). ITC can be used to measure binding constants of TrpRSs, and when two substrates are directly compared, a greater relative binding affinity (lower dissociation constant) gives some insight into the enzyme’s activity. By ITC, the Bs WT TrpRS exhibited characteristics similar to other bacterial (60) and human TrpRSs (61). The WT enzyme favors Trp (K_d = 149 μM) over 5-OH Trp (K_d = 1,079 μM) as would be expected as a result of evolutionary optimization. The V144P mutant showed decreased binding to Trp (K_d≥2,000 μM) relative to WT and, due to experimental limitations in fitting, we place a lower bound on the K_d at greater than 2,500 μM for the mutant with 5-OH Trp. The observation that the V144P TrpRS consistently exhibited greater binding affinity for Trp vs. 5-OH Trp conflicts with the premise of the prior study. It was reported that the Bs TrpRS V144P variant failed to charge Bs tRNA^Trp in the absence of 5-OH Trp (see ref. 25 and Fig. 4A, lane 2). However, the previous experiments were performed in 293T cells, so contextual issues within the mammalian expression system cannot be ruled out. But given the importance of intracellular expression levels of the orthogonal pair on the efficiency of incorporation, (62–64) it is unclear how the V144P mutant, an enzyme with a fraction of the activity of the WT enzyme, achieved efficient charging of Bs tRNA^Trp with 5-OH Trp (to the complete exclusion of Trp) in a more challenging mammalian expression system.

Söll and coworkers reported the aminoacylation kinetics for Bst TrpRS single and double mutants (15), including a V141Q mutant. This one mutation, in a similar region as the Bs V144P mutation, caused reductions of ∼300 and 1,000-fold in catalytic efficiency for Trp and ATP activation, respectively. Schimmel and coworkers, however, were able to completely disrupt a human TrpRS’s affinity for Trp by introducing targeted mutations to the enzyme’s determinants of amino acid specificity (61). Consistent with the ideas exposed here, such targeted mutations would be a requirement to break the enzyme’s capacity to bind its canonical substrate and any closely related analog; and with these bacterial TrpRSs, this involves a critical Asp interaction to the indole nitrogen of Trp.

Schultz and coworkers reported the utility of incorporating a single 5-OH Trp by performing fluorescence and electrochemical crosslinking experiments on their purified 5-OH Trp labeled protein (25). However, because their targeted recombinant protein contained a single Trp residue (25), the same single substitution experiment was already permissive using standard mammalian (6965) and bacterial hosts; a bacterial host would have been easier to culture, have produced significantly more labeled protein and was already well-precedented to permit such substitutions due to the inherent promiscuity of E. coli and Bs TrpRSs for Trp analogs, including 5-OH Trp (14, 51, 54, 66).