An Evolved Aminoacyl-tRNA Synthetase with Atypical Polysubstrate Specificity

GFP Expression

GFP expression was carried out as previously described. BL-21 (DE3) E. coli was co-transformed with the appropriate pEVOL plasmid and pET-GFP_Y151X plasmid(1), grown in 2xYT media at 37 °C to saturation with chloramphenicol (40 μg/mL) and ampicillin (100 μg/mL) and diluted in 2YT to an OD₆₀₀ of 0.2. Diluted cultures were grown at 37 °C to an OD₆₀₀ of 0.7 – 1.0, and induced with IPTG (1 mM) and arabinose (0.02%). The culture was immediately aliquoted (100 μL/well) into a 96 well plate containing 10 μL of 10 mM unnatural amino acid per well and background fluorescence was measured on a Spectramax Gemini EM (Ex/Em= 395/509 nm; Molecular Probes). After 14 h at 30 °C fluorescence was measured again to determine the level of GFP expression. Amino acids that afforded a significantly measurable signal were then employed in similar expression experiments in triplicate. Each sample was normalized to cell density using the OD₆₀₀ and fluorescence was measured and compared to wild type expression (Table 2), confirming the UAA incorporation.

Myoglobin Expression

BL-21(DE3) E. coli was co-transformed with the appropriate pEVOL plasmid and pET-Myo_F107X, grown in 2xYT media at 37 °C to saturation with chloramphenicol (40 μg/mL) and ampicillin (100 μg/mL) and diluted in 2YT to an OD₆₀₀ of 0.2. Diluted cultures were grown at 37 °C to an OD₆₀₀ of 0.7, induced with IPTG (1 mM), arabinose (0.2%) and the appropriate unnatural amino acid (1 mM) and grown at 37 °C for 16 h. The cultures were pelletted and lysed using Bug Buster (Novagen) and the protein was purified on Ni-NTA spin columns (Qiagen) according to the manufacturer’s protocol. Myoglobin expression was analyzed on a 4–20% Gly-Tris polyacrylamide gel (see Supporting Information) and the mass of purified protein was determined by LC/MS on an Agilent 1100 Series LC/MSD. The chromatographic peak corresponding to myoglobin (between 6.1 min and 6.5 min) was charge deconvoluted using Agilent LC/MSD ChemStation software (Rev. B.03.02). Deconvolution parameters were set to high M_r= 17,000 and low M_r= 21,000, maximum charge= 50, and minimum peaks in set = 3 –8. Error± 0.02%, as determined from control samples. Liquid chromatography mass spectrometry results for Myo_Y107X expression experiments are listed in the Supporting Information.

Aminoacyl-tRNA synthetase Crystallization

Protein expression and crystallization was carried out as previously described(10–12). The pCNF-RS DNA was amplified from the pBK-CN vector(13) (Fwd Primer: 5′ GCAAGCGCATATGGACGAATTTGAAATGA 3′; Rev Primer: 5′ GTTCGGCTCGAGTAATCTCTTTCTAATTG 3′), inserted into the pET-22b(+) vector using the XhoI and NdeI restriction sites, and transformed into BL21(DE3) cells. A 1L 2X YT (Amp) expression culture was inoculated to OD₆₀₀ 0.2 and grown at 37 °C to an OD₆₀₀ of 0.6 and induced with IPTG (1 mM). The culture was grown at 37 °C for 16 h, pelleted and lysed (50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, 10 mM BME, pH 8.0) by sonication. The sample was then centrifuged (18,000 rpm; 20 min) and the supernatant was incubated with Ni NTA resin (2 mL; Qiagen) for 1 hour at 4 °C. The resin was then washed (100 mL; 50 mM NaH₂PO₄, 300 mM NaCl, 20 mM imidazole, 10 mM BME, pH 8.0) and eluted (5 mL; 50 mM NaH₂PO₄, 300 mM NaCl, 250 mM imidazole, 10 mM BME, pH 8.0), followed by dialysis (3 × 1h rt; 25 mM Tris, 50 mM NaCl, 1 mM EDTA, 10 mM BME, pH 8.0). The aaRS was purified by FPLC on a MonoQ column (Buffer A: 25 mM Tris, 25 mM NaCl, 1 mM EDTA, 10 mM BME, pH 8.5; Buffer B: 25 mM Tris, 1 M NaCl, 1 mM EDTA, 10 mM BME, pH 8.5), concentrated to 2 mL and dialyzed into crystallization buffer (20 mM Tris, 50 mM NaCl, 10 mM BME, pH 8.0). Crystals were grown by the sitting-drop-vapor-diffusion method using a 1:1 mixture of concentrated protein (15 mg/mL, with 2 mM pCNF) and mother liquor (0.2M (NH₄) Tartrate, 20% PEG-3350, pH 6.6) at 20 °C.

Data Collection, Structure Determination and Refinement

Data for the pCNF co-crystals were collected at beamline 5.0.3 of the Advanced Light source (ALS) at a wavelength of 0.9778Å to a maximum Bragg spacing of 2.3Å (Table 1). All data were reduced and scaled using the HKL2000 package(14). The crystal belonged to the monoclinic crystal system with the β angle approaching 90º. A strong peak in the native Patterson map indicated the presence of pseudo-translational symmetry and that the crystal was pseudo-orthorhombic. This was confirmed by the inability of the unreduced data to scale reasonably in the equivalent orthorhombic spacegroup. The presence of non-crystallographic translational symmetry induces considerable deviation from Wilson Statistics. The Wilson Ratio (<I²>/<I>²) for the data set calculated from phenix xtriage is 2.752 and 2.776 for acentric and centric reflections respectively, whereas perfect data should have values of 2.0 and 3.0 respectively. The Wilson distribution, which is based on the assumption of a random distribution of atoms in the cell, is therefore considerably skewed by this translational symmetry and statistics such as the R-factor and procedures such as the calculation of normalized amplitudes which are based on this assumption can be affected. Despite this the structure was easily solved using Phaser(15) using all data between 50 and 2.3Å resolution using the wild type Tyrosyl tRNA synthetase(16) (PDB code 1J1U) as a probe. The molecular replacement solution confirmed the presence of the pseudo-translational symmetry between the two molecules in the asymmetric unit. Electron density produced from an initial refinement with Buster clearly showed the mutated residues between the probe model and the structure as well as electron density for pCNF in both molecules, confirming the validity of the molecular replacement solution. Model building and refinement were performed by iterative building and refinement with Coot and Buster.(17, 18) All other crystallographic manipulations were carried out with the CCP4 program suite.(19) Convergence of the refinement was checked throughout, with weights between geometric and X-ray terms optimized on the basis of the R-free in combination with a criterion for reasonable geometry. The Local Structure Symmetry Restraints (LSSR) procedure was also applied throughout refinement(20), also incorportated based on its effect on the R-free. Validity of the final structure was checked using a combination of the validation tools in Coot (17), including inspection of the Ramachandran plot, real space correlation, rotamer and geometry analysis. Using these refinement and geometrical critia final model converged with R_cryst and R_free of 23.0 % and 30.0 % respectively, with excellent geometry, root mean squared deviation (rmsd) of bonds and angles 0.013Ų and 1.3º respectively, no residues in disallowed regions of the Ramachandran plot (Table 1) and all of the mutations in the structure in a preferred rotamer conformation

Assay Development

To develop a rapid assay to analyze the incorporation of unnatural amino acids by an evolved aaRS, a fluorescence-based GFP reporter was used. A GFP gene with an amber mutation at Tyr151 (GFP_Y151X) was placed under the control of the T7 promoter in a pET101 vector as previously described.(21) Tyrosine 151 is a surface residue that can be substituted with a range of structurally diverse amino acids without affecting protein stability or fluorescence.(21) This reporter affords fluorescence only when the TAG codon is used as a sense codon for the acylated suppressor tRNA_CUA (and not a stop codon). Mutant Methanocaldococcus jannaschii tyrosyl-RS/tRNA_CUA (MjTyrRS/tRNA_CUA) pairs encoded in the efficient pEVOL system(21) were co-transformed into BL-21(DE3) cells with the pET-GFP_Y151X construct to enable rapid screening for substrate permissivity. Aminoacyl-tRNA synthetases specific for pAcF (p-acetylphenylalanine, 7),(22) pAzF (p-azidophenylalanine, 8),(23) pCNF (p-cyanophenylalanine, 1),(13) pIF (p-iodophenylalanine, 2), (24) pBoF (p-borophenylalanine),(25) pPrF (p-propargylphenylalanine, 14),(26) pCMF (p-carboxymethyl-phenylalanine),(27) pBpF (p-benzoylphenylalanine),(28) NapA (napthylalanine),(29) CouA (7-hydroxycoumarin-4-yl ethylglycine),(30) PLA (p-hydroxyL-phenyllactic acid),(31) BipyA (bipyridylalanine),(32) and HQA (8-hydroxyquinolin-3-yl alanine),(33) in addition to the wild type tyrosyl-tRNA synthease, were analyzed.(1, 34) A 96-well format was used to screen 72 unnatural amino acids including para and meta substituted phenylalanine analogues, histidine and alanine derivatives, amine, biaryl, and thiol containing amino acids (see Supporting Information). While fluorescent amino acids were included in the plate, other than CouA (well E7), the wavelengths of excitation and emission were nonoverlapping with GFP, and a background subtraction prior to expression was sufficient to allow for assessment of incorporation. The fluorescence of each well supplemented with 1 mM amino acid was measured after 14 hours of expression in rich media (Figure 1). In order to rapidly assess amino acid incorporation fluorescence readings were taken on the cultures, as opposed to lysed cells, and potential hits were later confirmed by protein isolation and mass spectrometry.

Many of the aaRSs failed to exhibit a significant degree of substrate promiscuity with respect to the UAA in our screen. Not surprisingly, the wild type MjTyrRS showed no detectable incorporation of any unnatural amino acid in the screen. Additionally, the pBoF, pPrF, pCMF, HQA, CouA, PLA, NapA, and BipyA specific aaRSs displayed a very low propensity for incorporation of unnatural amino acids beyond the specific unnatural amino acid for which they were evolved. The pAcF, pIF and pAzF synthetases were able to incorporate several para-substituted phenylalanine analogs included in the screen, but only to a limited extent (see Supporting Information). Most interestingly, pCNF-RS(13) exhibited a substantial degree of polyspecificity as it afforded increased GFP_Y151X fluorescence in the presence of p-chlorophenylalanine (2), p-fluorophenylalanine (5), p-acetylphenylalanine (7), p-alkynylphenylalanine (6), benzylserine (15), and p-phenyl phenylalanine (12) (Figure 1). Screen hits were validated by incorporation of the unnatural amino acid into a his-tagged myoglobin mutant containing an F107TAG amber mutation (Myo_F107X).(21) The protein was purified using a Ni-NTA resin and analyzed by SDS-PAGE. The calculated mass of purified mutant protein was confirmed by liquid chromatography mass spectroscopy (LCMS). In the absence of any unnatural amino acid low levels of phenylalanine incorporation in response to the amber codon could be detected (< 20%); however, in the presence of any of these amino acids (>1mM), only incorporation of the unnatural amino acid was observed within the detection limits of LCMS.(21) The behavior of pCNF-RS in this assay was unique both with respect to incorporation efficiency and substrate diversity relative to the other 13 aminoacyl-tRNA synthetases.

Determination of Substrate Scope for pCNF-RS Polyspecificity

The pCNF-RS appears to be highly tolerant of para-substituted phenylalanine derivatives. Consequently, the pCNF-RS was further screened against additional para-substituted aromatic amino acids (Figure 2). The majority of these amino acids (1–18) were incorporated into GFP_Y151X by the pCN-RS (relative incorporation was assessed in triplicate and normalized to cell density, Table 2); incorporation was further validated by incorporation into Myo_F107X with subsequent mass spectral analysis (Table 2). pCNF-RS substrates include UAAs that have already been genetically encoded in bacteria, e.g., pIF (2), pBrF (p-bromophenylalanine 3), pAcF (7), pAzF (8), pPrF (14), pNO₂F (p-nitrophenylalanine 11), oNO₂F (o-nitrophenylalanine 13), O-methyl Tyr (16), and O-allyl Tyr (17).(2)p CNF-RS substrates also include unnatural amino acids which have not previously been encoded in bacteria, e.g., pClF (4), pFF (5), p-ethynyl phenylalanine (6), p-trifluoromethylacetyl phenylalanine (9), p-phenyl phenylalanine (12), benzylserine (15), and O-tert-butyl tyrosine (18). Based on GFP expression levels, many of the amino acids are good substrates for pCNF-RS, affording >80% protein expression relative to wild type GFP. Interestingly, several amino acids (pIF (2), pBrF (3), pClF (4), p-alkynylF (6), pAzF (8), p-isopropylF (10), and OMeY (16)) afford higher protein yields than did the pCNF substrate (1), the UAA against which the pCNF-RS was originally selected. Only amino acids 19–21 (p-carboxyphenylalanine, p-methyl-phosphonic acid phenylalanine, and p-aminophenylalanine), which have hydrogen bond donating substituents, were not incorporated. It is interesting to note that despite the apparent polysubstrate specificity of the pCNF-RS, 21 (p-aminophenylalanine) was not incorporated while other significantly larger substituents were tolerated, indicating the pCNF-RS has some degree of specificity (21 has been previously shown to be taken up efficiently by E. coli and is stable in the cytoplasm).(35) In total, the pCNF-RS was found to incorporate 18 unnatural amino acids, seven of which had not been previously encoded. A number of the latter UAAs, e.g., p-alkynylphenylalanine (6) and p-trifloroacetylphenylalanine (9), may be useful as orthogonal bioconjugation handles or as warheads for the selective inactivation of enzymes.

X-ray Crystal Structure of pCNF-RS

To better understand the molecular basis of pCNF-RS’s polysubstrate specificity (which in contrast to polyspecific enzymes like p450s, is not a solvent exposed surface cavity), the aaRS was overexpressed and crystallized for x-ray structure analysis (10–12, 36). The crystal structure of the pCNF-RS pCNF complex was solved to 2.3 Å with an R_factor = 0.23 (Table 1 and Supporting Information). The pCNF-RS structure superimposes well onto the WT-TyrRS-Tyr structure (PDB 1J1U) except in the amino acid binding site. The amino acid binding site for the pCNF-RS is relatively hydrophobic containing the following mutations relative to the wild type synthetase: Y32L, L65V, F108W, N109M, D158G, and I159A. A comparison of the WT-TyrRS structure to the pCNF bound pCNF-RS structure (Figure 4A vs. Figure 4B) reveals that mutations create a larger binding pocket which allows the enzyme to accept the nitrile group. In the WT-TyrRS structure, residues Y32 and D158 are involved in hydrogen bonding with the tyrosine hydroxyl group. These residues in the pCNF-RS are mutated to leucine and glycine, respectively, which removes hydrogen bonding potential and likely explains the inability of pCNF-RS to accept tyrosine and other phenylalanine derivatives with hydrogen bond donor substituents in the para position. Additionally, the mutation L65V creates a larger binding pocket and appears to alter the angular position of the bound amino acid. This mutation is unique to the pCNF-RS, as no other previously evolved aminoacyl-tRNA synthetases possess the valine residue at this position. Although the binding pocket is larger in pCNF-RS, key residues (Leu 32, Val65 and Gly158) are still able to stabilize the pCNF (1) through van Der Waals interactions (Figure 4B).

To determine which mutations might be determinants of polyspecificity, the pCNF bound pCNF-RS amino acid binding site structure was overlaid with the substrate-bound pAcF-RS (Y32L, D158G, I159C, L162R; PDB 1ZH6) and NapA-RS (Y32L, D158P, I159A, L162Q, A167V; PDB 1ZH0) structures. The former aaRS accepts relatively few other unnatural amino acids, while the latter displays no polyspecificity (Figure 3). As previously noted, the Y32L mutation increases the amino acid binding site size (especially at the para-position of phenylalanine derivatives); however, this mutation is also found in the pAcF-RS and NapA-RS (among others), which do not exhibit a comparable degree of polyspecificity to pCNF-RS. The unique L65V mutation appears to create unoccupied space at the back of the amino acid binding that is filled by other residues in pAcF-RS. This mutation may enable pCNF-RS to accept a diverse group of para-substituted phenylalanine or tyrosine derivatives (Figure 3). Interestingly, when pCNF is modeled into the pAcF-RS binding pocket, the leucine at residue 65 interferes with the cyano substituent (Figure 4C; the pAcF-RS complex is shown for comparison in Figure 4F). In contrast, pAcF (7) can readily be docked into the active site of the pCNF-RS (Figure 4E). Additionally, mutation of leucine 65 to smaller amino acids has been shown to allow the incorporation of larger unnatural amino acids in other aaRSs. For example, the o-nitrobenzyl tyrosine (ONBY) synthetase contains a L65G mutation, facilitating the incorporation of the sterically bulky bicyclic ONBY amino acid.(37)

While an increased substrate binding site size may be partially responsible for increased polyspecificity, other factors are involved. This is apparent when comparing the pCNF-RS and NapA-RS structures. Although the D158P mutation in the NapA-RS gives it a significantly larger binding pocket than the pCNF-RS, this aaRS exhibits virtually no permissivity. When pCNF is modeled into the amino acid binding site of NapA-RS, proline 158 prevents side chain Van der Waals interactions that might stabilize the bound and amino acid (Figure 4D). Moreover, a number of the other aminoacyl-tRNA synthetases examined with larger binding sites (e.g., that accommodate bipyridyl, 8-hydroxyquinolinyl and benzoylphenyl side chains) do not show significant polyspecificity, suggesting that there is a plasticity unique to pCNF-RS active site.

The uniqueness of the pCNF-RS is also evident when examining backbone perturbations in the mutant aminoacyl-tRNA synthetases. As previously reported, the D158P and I159A mutations terminate the α₈-helix in NapA-RS; a similar backbone alteration is observed in the pAcF-RS (albeit to a lesser extent).(12) Interestingly, this perturbation is not observed in the pCNF-RS, despite having the same D158G mutation as the pAcF-RS and the same I159A mutation as the NapA-RS. In fact, the pCNF-RS aligns much more closely to the wild-type tyrosyl-tRNA synthetase than with the other aminoacyl-tRNA synthetases in this region (Figure 5A). Another distinctive difference is apparent in the pCNF-RS when examining the helical residues 108 and 109. Relative to both the WT-RS and the pAcF-RS, the helix is moderately displaced and is slightly truncated (Figure 5B). This may afford a greater degree of space and plasticity for the accommodation of larger para-substituted amino acids.

Importantly, the pCNF-RS amino acid binding site still highly discriminates against Phe, Tyr and p-aminophenylalanine, while it accepts similar amino acid such as O-methyltyrosine (16), p-chlorophenylalanine (4) and p-fluorophenylalanine (5). Clearly, mutations in a relatively small number of active site residues result in the ability of this aaRS to discriminate subtle changes in substrate structure. This result indicates that the negative selection step in the evolution of pCNF-RS is quite robust in its ability to remove polyspecific mutants that accept endogenous amino acids. Work is currently underway to co-crystallize pCNF-RS and various mutants with additional common and unnatural amino acids in an effort to further decipher this unique enzyme’s specificity.