Natural amino acids and ncAAs were purchased from Chem Impex Inc (Wood Dale, IL, United States). The ncAA SeCbzK (10) (Wang Z.U. et al., 2012) and CbzKOH (11) (Li et al., 2012; Wang, 2012) (Scheme 1) were synthesized by the reported methods. PCR was performed using the KOD hot start polymerase kit (Merck). Oligonucleotide synthesis and DNA sequencing were done by Genomics Inc. (Taipei, Taiwan). The primer sequences are listed in the Supplementary Materials. Construction of the pET-pylT-sfGFP-TAG2 and the pET-pylT-sfGFP-TAG27 plasmids followed general cloning protocols. The sfGFP-TAG2 gene was amplified by PCR from pET-pylT-sfGFP (Guo et al., 2014) with the primers sfGFP-TAG2-NdeI-F and sfGFP-SacI-R2. The sfGFP-TAG27 was generated by overlap extension PCR with two fragments (fragment 1: primer sfGFP-NdeI-F1 and sfGFP-TAG27-R; fragment 2: sfGFP-TAG27-F and sfGFP-SacI-R2). The overlapped sfGFP gene products and pET-pylT plasmid were then double-digested with restriction enzyme NdeI and SacI, and then T4 DNA ligase was used for DNA ligation.

In the construction of the plasmid pCDF-N-PylRS variants, the PylRS-R61K/H63Y/S193R (N-PylRS) gene was generated by overlap extension PCR with three fragments using the template pCDF-PylRS: fragment 1: PylRS-NcoI-F and PylRS-R61K/H63Y-R; fragment 2: PylRS-R61K/H63Y-F and PylRS-S193R-R; fragment 3: PylRS-S193R-F and PylRS-BamHI-R. In preparing pCDF-PylRS-R61K, pCDF-PylRS-H63Y, and pCDF-PylRS-S193R, the same procedures were performed with the following primers to produce PylRS-R61K, PylRS- H63Y, and PylRS-S193R genes with pCDF-PylRS template and following the overlap extension PCR; PylRS-R61K gene: fragment 1, PylRS-NcoI-F and PylRS-R61K-R; fragment 2, PylRS-R61K-F and PylRS-BamHI-R; PylRS-H63Y gene: fragment 1, PylRS-NcoI-F and PylRS-H63Y-R; fragment 2, PylRS-H63Y-F and PylRS-BamHI-R; PylRS-S193R gene: fragment 1, PylRS-NcoI-F and PylRS-S193R-R; fragment 2, PylRS-S193R-F and PylRS-BamHI-R. In preparing pCDF-PylRS-ND, the PylRS-ND gene fragment was generated by PCR with two primers, PylRS-NcoI-F and PylRS-P149-BamHI-R, and pCDF-PylRS was used as a template. In the preparation of pCDF-PylRS-D1, the spacer sequence (TGAAAAAAGCGATG), including the TGA stop codon and the ATG start codon, was installed between positions P149 and A150 of the PylRS gene. The PylRS-D1 gene fragment was generated by overlap extension PCR with two fragments using pCDF-PylRS as a template. Fragment 1 was generated by the primers PylRS-NcoI-F and PylRS-P149SSS-R1; fragment 2 was generated by the primers PylRS-P149SSS-F2 and PylRS-BamHI-R. Different lengths of linker were installed in PylRS variants between sposition 149 and 150. In preparing pCDF-PylRS-L1, pCDF-PylRS-L2, and pCDF-PylRS-L3 (Figure 1B), the same procedures were performed by the following primers to produce PylRS-L1, PylRS-L2, and PylRS-L3 genes with pCDF-PylRS template and following overlap extension PCR: PylRS-L1 gene, PylRS-NcoI-F, and Linker-1XG4S-R1, Linker-1XG4S-F2, and PylRS-BamHI-R; PylRS-L2 gene: fragment 1, PylRS-NcoI-F and Linker-2XG4S-R1; fragment 2, Linker-2XG4S-F2 and Linker-2XG4S-R2; fragment 3, Linker-2XG4S-F3 and PylRS-BamHI-R; PylRS-L3 gene: fragments 1 and 3 were from the PylRS-L2 gene; fragment 2, Linker-3XG4S-F2 and Linker-3XG4S-R2.

The pCDF-N-PylRS plasmids with the inserted linker, pCDF-N-PylRS-D1, pCDF-N-PylRS-L1, pCDF-N-PylRS-L2, and pCDF-N-PylRS-L3 (Figure 1B), were then constructed using the procedures mentioned above. pCDF-ZRS was derived from pEVOL-mKRS1-pylT (Wang Z.U. et al., 2012). The N-PylRS mutations were transplanted to ZRS by the methods mentioned earlier to construct pCDF-N-ZRS. The linkers were inserted into ZRS to generate pCDF-N-ZRS-D1, pCDF-ZRS-L1, pCDF-ZRS-L2, and pCDF-ZRS-L3 plasmids. All the ligated products were transformed to E. coli DH5α, and the colonies were selected for DNA sequencing, respectively.

To produce ncAA-encoded sfGFP proteins, the pET-pylT-sfGFP-TAG2 or pET-pylT-sfGFP-TAG27 plasmid was co-transformed with different pCDF-PylRS variants into E. coli BL21 (DE3) individually. After an hour of recovery, the bacteria were spread on a plate containing ampicillin (Amp) (100 μg/ml) and streptomycin (Sm) (100 μg/ml). A single colony was chosen from the plate and cultured in 1 ml LB medium overnight. The cultured bacteria were then transferred to 50 ml fresh LB medium and incubated at 37°C until the OD₅₉₅ reached 0.6–0.8. Protein expression was induced with the supplement of 1 mM IPTG and ncAA (except for sfGFP-3 and sfGFP-4 protein production, where the medium was changed to GMML medium supplemented with 2 mM ncAA) and incubated at 37°C for 12 h. The bacteria were then harvested and resuspended in lysis buffer [1X phosphate-buffered saline (PBS), pH 7.4] and sonicated. After centrifugation (60 min, 20,000 × g, 4°C), the supernatant was collected and incubated with 0.5 ml Ni²⁺-NTA resin (Roche) for protein purification. A total of 5 ml lysis buffer and 2.5 ml washing buffer (1X PBS, 5 mM imidazole, pH 7.4) were used to remove proteins bound non-specifically to the resin. The target protein was eluted from the resin with 2.5 ml elution buffer (1X PBS, 200 mM imidazole, pH 7.4). The buffer of the eluted fractions was changed to 1X PBS with Amicon Ultra-15 Centrifugal Filter Units (MWCO 10 kDa). Purified sfGFP was analyzed by 12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) with instant blue staining.

Whole cells were collected and lysed at 100°C with SDS loading dye for 15 min and then subjected to 12% SDS-PAGE analysis. The gels were stained with InstantBlue^TM Stain to visualize the target proteins with the expected molecular weight of around 28 kDa. The suppression efficiency of the amber codon in the sfGFP proteins with a C-terminal His tag was observed by western blot with an anti-6X His tag antibody. Western blots were performed using a Trans-Blot Turbo System (Bio-Rad) and an RTA transfer kit. Anti-His (SignalChem, H99-61M-100) and horseradish peroxidase (HRP)-conjugated secondary antibody (Cell Signaling Technology, 7076P2) were used for immunoblotting. After SDS-PAGE analysis, the gel was immersed in the transfer buffer and then blotted with a polyvinylidene fluoride (PVDF) membrane (25 V/1.3 A, 10 min). After the transfer process was finished, the PVDF membrane was washed thrice with phosphate-buffered saline with Tween^® (PBST) buffer for 5 min. Next, the membrane was blocked with 5% skimmed milk for 1 h at room temperature. Then, the membrane was washed thrice with PBST buffer for 5 min (washing step). The primary antibody (1:1,000 dilution) was added and incubated with the membrane for 1 h at room temperature following the washing step. Subsequently, the HRP-conjugated secondary antibody (1:5,000 dilution) was added and incubated with the membrane for 1 h at room temperature. The membrane was then treated by a washing step. Finally, the WesternBright ECL HRP substrate (Advansta, K-12045-D50) was mixed and spread onto the membrane to visualize the band signals using the ChemiDoc Imaging Systems (Bio-Rad) in bioluminescence detection mode.

The pure protein was diluted with 50% acetonitrile and 1% formic acid. An aliquot corresponding to 1 pmol of the pure protein was injected into an ESI source (Waters LockSpray Exact Mass Ionization Source) with a syringe pump (Harvard Apparatus, MA, United States) and a flow rate of 5 μl/min was held throughout the analysis. The mass of the intact proteins was determined using Waters Synapt G2 HDMS mass spectrometer (Waters, Milford, MA, United States). The acquired spectra were deconvoluted to single-charge state using the MaxEnt1 algorithm of the MassLynx 4.1 software (Waters).

Details of in-gel digestion are as follows: after the staining procedure, the gel bands were excised and cut into small pieces. The gel pieces were sequentially washed with 25 mM NH₄HCO₃, 40% methanol solution, and 100% acetonitrile before being treated with DTT and then iodoacetamide. Washing of the gel pieces with 25 mM NH₄HCO₃ and 50% acetonitrile and then drying in a vacuum centrifuge provided the materials for trypsin digestion. A solution of 65 to 100 ng of sequencing-grade modified trypsin (Promega) in 25 mM NH₄HCO₃ and 10% acetonitrile (25–30 μl) was added and incubated with the gel pieces for 12–16 hrs at 37°C. The reaction was stopped by adding 1–2 μl of 5% formic acid. Details of the MS and data analysis are as follows: the digested samples (0.5 μl) were carefully mixed with the matrix solution (0.5 μl of 5 mg/ml DHB in 0.1% TFA/30% acetonitrile) and 0.5 μl of the mixture was deposited onto a MTP 600/384 AnchorChip (Bruker Daltonics). All mass spectrometry experiments were done using a Bruker Autoflex III MALDI TOF/TOF mass spectrometer (Bremen, Germany) equipped with a 200-Hz SmartBean Laser in positive ion mode with delayed extraction in the reflectron mode. Data acquisition was done manually with FlexControl 3.4, and data processing was performed with Flex-Analysis 3.4 (both Bruker Daltonik). Protein database searches through Mascot, using combined PMF and tandem mass spectrometry MS/MS datasets, were performed via Biotools 3.2 (Bruker).

To understand the substrate range of PylRS variants, screening of a 359 ncAAs library was performed. The plasmid pET-pylT-sfGFP-TAG2 or pET-pylT-sfGFP-TAG27 and pCDF-PylRS variants were co-transformed into E. coli BL21 (DE3) individually. The bacteria were spread on a plate supplemented with Amp (100 μg/ml) and Sm (100 μg/ml). The plate was incubated at 37°C overnight. Ten colonies were then inoculated and cultured in 3 ml LB medium at 37°C overnight before 500 μl of each cultured bacteria was transferred to 25 ml of fresh LB medium and incubated at 37°C until the OD₅₉₅ reached 0.6–0.8. The cells were harvested and washed twice with M9 salts and suspended in M9 medium (M9 salts, 1% glycerol, 2 mM MgSO₄, and 0.1 mM CaCl₂) containing 1 mM IPTG. Aliquots (50 μl) of the suspended cells were loaded into a 384-well plate containing a different ncAAs (1 mM) in 359 wells (Supplementary Table S2). Cells were incubated in a plate reader (BioTek) at 37°C for 12 h, with continuous monitoring of the fluorescence intensity (excitation 535 nm and emission 595 nm) as well as OD₅₉₅. Twelve wells were used as controls to measure the background signals (six wells without ncAAs and IPTG; six wells without ncAAs but containing IPTG). The fluorescence intensity of sfGFP was divided by the OD₅₉₅ following the subtraction of the control signals (containing IPTG but no ncAAs) to generate the relative fluorescence intensity.

To explore novel substrate ranges, this work aimed to generate novel PylRS variants without active site mutations for studying remote effects in altering the interaction between the tRNA and the PylRS NTD or CTD. In a previous study, the evolved PylRS, HarRS, had R61K, H63Y, S193R, N203T, L305H, L309W, N346D, C348S, L367M, Y384F, K429M, K431M, D433G, and G444E mutations. These mutations of HarRS have been found to enhance the activity and the selectivity in charging homoarginine (Mukai et al., 2015). To understand how the mutations on NTD affect the suppression efficiency of PylRS, the first three mutation sites in HarRS are transplanted to PylRS to generate N-PylRS (R61K/H63Y/S193R) (Figure 1). The R62K and H63Y mutations are in the NTD and at the interface between the PylRS-NTD and the tRNA T loop region (Figure 2). The S193R mutation is in the CTD and is located at the interface with the tRNA D-loop region (Figure 3) based on an overlapped model of the DhPylRS CTD/tRNA^Pyl co-crystal and the MmPylRS CTD crystal structures. In phage-assisted non-continuous evolution (PANCE) approach for evolving chPylRS, some mutants were found to be active in charging BocK with two separated genes by inserted TGA stop codon and following ATG start codon between NTD and CTD (Suzuki et al., 2017). The chPylRS was generated by fusing MbPylRS NTD (1–149 residues) with the MmPylRS C-terminus (185–454 residues). In this work, wild-type MmPylRS (wt-PylRS) was used to translate the inserted mutation between P149 and A150 to form the PylRS-D1 construct (Figure 1B) in generating MmPylRS NTD (1–149 residues) and MmPylRS CTD (150–454 residues) proteins. The truncated MmPylRS NTD (1–149 residues), namely, PylRS-ND, was also generated for comparison.

To probe the crosstalk between the PylRS NTD and the CTD in charging amino acid substrates, three flexible loops of different lengths were inserted between these domains in PylRS-D1. The linkers were the hexapeptide SGGGGS (PylRS-L1), the tridecapeptide S(GGGGS)₂ (PylRS-L2), and the non-adecapeptide S(GGGGS)₃ (PylRS-L3) (Figure 1B). In addition to these three mutants, we also compared wt-PylRS and PylRS-D1; these five PylRS variants were subjected to test substrate range with the supplement of co-transformed E. coli carrying MmPylRS/tRNA^Pyl gene cassettes and reporter gene sfGFP-TAG2 or sfGFP-TAG27, respectively. The fluorescence intensities of the sfGFP-UAG2 and sfGFP-UAG27 gene products indicated the read-through of the amber codon in response to the ncAA. Generally, the sfGFP-UAG2 suppression test has 4.7 to 1.6 times higher signal than sfGFP-UAG27 suppression test in charging ncAA 1 and 2 (Scheme 1), but the reverse results are seen with ncAA 3–5. After producing the sfGFP-UAG2 and sfGFP-UAG27 gene products, 359 ncAAs (see Supplementary Table S1 for structures) were tested as substrates. The screening results (Supplementary Figures S15–S24) show various intensities of signals in charging BocK (1) and AlloK (2) (Scheme 1), which are considered as good substrates for wt-PylRS recognition (Yanagisawa et al., 2008). In sfGFP-UAG2 gene production, PylRS-D1 preserved 34% of activity in charging AlloK (2) compared to wt-PylRS. The PylRS variants with linkers generate better activities (Figure 4A). All three PylRS variants (PylRS-L1, PylRS-L2, and PylRS-L3) were rescued by inserted linkers and showed enhanced activity at 120–230% compared to the activity of wt-PylRS. In sfGFP-UAG27 gene production (Figure 4B), however, the activity of all five PylRS variants maintained a similar pattern with less charging of ncAA 1 and 2. Small signals in charging 3MeH (5) (Scheme 1) were found in wt-PylRS, PylRS-L1, PylRS-L2, and PylRS-3 in this amber suppression test.

The effects of PylRS-R61K, PylRS-H63Y, and PylRS-S193R variants were determined and found to improve the suppression efficiency against ncAA 1–2 in sfGFP-UAG2 production compared to wt-PylRS, whereas no fluorescence signals were observed in PylRS-ND (Supplementary Figure S47). Combining the three beneficial mutations, N-PylRS harboring R61K/H63Y/S193R mutations was evaluated in its substrate range by sfGFP-UAG2 and sfGFP-UAG27 gene production (Supplementary Figures S25, S26). Thus, we decided to investigate N-PylRS and its four variants, N-PylRS-D1 and N-PylRS-L1–L3 (Supplementary Figures S27–S34). Noticeably, N-PylRS showed nearly 5.6 and 4.1 times higher fluorescent signals in charging BocK (1) and AlloK (2) than wt-PylRS in the sfGFP-UAG2 gene suppression yield (Figure 4A). In the sfGFP-UAG27 gene suppression study, N-PylRS was found to recognize ncAA 1–5. While AlloK (2) is still the best substrate, 3MeH (5) has a higher signal than BocK (1) (Figure 4B). In addition, N-PylRS was capable of incorporating S-benzyl cysteine analogs MbzlC (3) and MeObzlC (4) in low suppression efficiencies in sfGFP-UAG27 production. In contrast to PylRS-D1, the substrate specificity profiles of N-PylRS-D1 have revealed abolished fluorescence intensities in charging ncAA 1–5 in sfGFP-UAG2 and sfGFP-UAG27 suppression tests. The N-PylRS-L1–L3 variants showed reinstalled signals in sfGFP-UAG2 suppression with similar activities in charging BocK (1) and AlloK (2). The sfGFP-UAG27 gene suppression test in N-PylRS-L1–L3 had a gradual decrease in activity along with increasing linker length in charging ncAA 1–5.

ZRS was evolved from MmPylRS with Y306M/L309A/C348T/T364K/Y384F mutations at the active site (Wang et al., 2010). To validate the influence on the activity of ZRS by the R61K/H63Y/S193R mutations, as well as the introduction of the linker, these variations in the wt-PylRS study were transplanted and tested on ZRS, which was initially found to incorporate CbzK (6) and CbzK analogs (7, 10–11) (Scheme 1) (Wang Z.U. et al., 2012). Thus, ZRS variants, N-ZRS, ZRS-D1, ZRS-L1, ZRS-L2, and ZRS-L3, were constructed for evaluating their effects with active site mutation. The screening results of the substrate range in sfGFP-UAG2 and sfGFP-UAG27 suppression for the ZRS and the five variants are shown in Supplementary Figures S35–S46. NcAA 3–11 (Scheme 1) had a positive response and are illustrated in Figure 5.

CbzK (6) was efficiently incorporated into sfGFP in response to the amber codon as reported for ZRS. The CbzK analogs, ClCbzK (7), SeCbzK (10), and CbzKOH (11) (Scheme 1), also showed similar intensity in the sfGFP-UAG2 suppression study. Two D-form CbzK analogs, DCbzK (8) and DClCbzK (9), were positive but had lower signals (Figure 5A). sfGFP-UAG27 suppression in ncAAs library screening of ZRS showed 12.8 times lower signal in charging CbzK compared to the sfGFP-UAG2 suppression study but was only 1.6 times lower in charging SeCbzK. The substrate range of N-ZRS remained the same as that of ZRS; nevertheless, the substrate specificity profiles of N-ZRS in the sfGFP-UAG2 suppression study showed a higher fluorescence intensity as compared to ZRS, especially in charging D-ncAA 8 and 9. Unexpectedly, S-benzyl cysteine analogs ncAA 3 and 4 were also incorporated into sfGFP with low efficiencies by N-ZRS in the sfGFP-UAG27 suppression study. The ZRS-D1 screening results showed a significant decrease in activity in the sfGFP-UAG2 and sfGFP-UAG27 suppression studies compared to ZRS and other ZRS variants. The ZRS-D1 activity with CbzKOH (11), however, was the best in the sfGFP-UAG2 screening and had diminished activity in the sfGFP-UAG27 screening (Figure 5). The addition of a linker did not raise the suppression efficiencies of ZRS albeit a slight increase in the fluorescence intensities of ZRS-L2 can be observed with ncAA 6 and 10 in response to sfGFP-UAG2 and sfGFP-UAG27 (Figure 5).

Some amber screening results show reverse or different intensity patterns, and western blotting analysis was used to confirm sfGFP protein yield to support the sfGFP fluorescence intensity results. N-ZRS and ZRS-D1 were chosen for analyzing sfGFP-UAG27 and sfGFP-UAG2 protein productions (Figure 6A). An anti-His tag antibody was used to detect the C-terminal-His tag present in full-length sfGFP, which indicates the amount of read-through at the amber stop codon. Western blotting analysis of N-ZRS/sfGFP-UAG27 indicates strong amber read-through for ncAA 6–8 and 10 and weaker response for ncAA 9 and 11 in SDS-PAGE and anti-His channel. Clearly, sfGFP-11 showed an additional band at lower molecular weight, near the 25-kDa protein marker in SDS-PAGE, which was also detected by western blotting (Figure 6A). These results are in partial agreement with the sfGFP fluorescent screening. They match for ncAA 3, 4, 9, and 10, not for ncAA 6–8 and 11 (Figure 5B). A ZRS-D1/sfGFP-UAG2 analysis (Figure 6A), however, matches well with sfGFP fluorescent screening results (Figure 5A), confirming that ZRS-D1 charges CbzKOH to generate acylated tRNA with high activity. The purified sfGFP-11^∗ protein produced by the ZRS-D1⋅tRNA^Pyl pair generated two additional major mass peaks because of ester bond cleavage and Cbz group deprotection: 27,773 Da (without the N-terminal Met) and 27,639 Da (without the N-terminal Met and Cbz group) (Table 1 and Supplementary Figure S7). Purified sfGFP-6, sfGFP-7, and sfGFP-11, which were produced by N-ZRS⋅tRNA^Pyl pair, were analyzed by electrospray ionization-mass spectrometry (ESI-MS) (Table 1 and Figure 6B, Supplementary Figures S4–S6). The experimental mass of sfGFP-6 (28,086 and 27,955 Da) and sfGFP-7 (28,120, and 27,988 Da) matched well to the calculated molecular weight of sfGFP-6 (28,085 and 27,955 Da) and sfGFP-7 (28,120 and 27,989 Da). A matrix-assisted laser desorption/ionization-time of flight-tandem mass spectrometry (MALDI-TOF-MS/MS) analysis of peptide ²⁷XSV³⁰R fragments which encoded ncAA 6 and 7 further provides the correct molecular mass and sufficient b and y fragments to indicate ncAA purity at position 27 of the sfGFP (Table 1 and Supplementary Figures S10, S11). In characterizing the mass of sfGFP-11, the replacement of the amino group with a hydroxyl group in the main chain results in additional mass peaks. Three mass peaks, at 27,955, 27,820, and 25,036 Da, were observed in the sfGFP-11 mass spectrum. The calculated mass of full-length and truncated sfGFP-11 is 27,955 (-Met) and 25,320 Da, respectively. The full-length sfGFP-11 mass agrees with the calculated mass, and the peak at 27,820 Da matches the expected mass of sfGFP-11 with Cbz group deprotection. The mass of sfGFP truncated at 28 position, 25,036 Da, does not fully match the calculated mass 25,057 Da and instead shows a loss of 21 Da, which indicates CbzKOH deletion at 27 position of the truncated sfGFP-11 (Table 1 and Figure 6B, Supplementary Figure S6).

The purified sfGFP-3 and sfGFP-4 proteins with ncAA 3 and 4 encoded at 27 position that were produced by the N-ZRS⋅tRNA^Pyl pair in E. coli were characterized, and they matched the calculated mass (Table 1 and Supplementary Figures S8, S9). The mass spectra of sfGFP-3 and sfGFP-4 show a molecular peak indicating tryptophan incorporation, 27,878 Da. A MALDI-TOF-MS/MS analysis of sfGFP-3 and sfGFP-4 also confirmed the presence of ncAA 3 and 4 at position 27 of sfGFP. To confirm the ncAAs screening results mentioned in the N-PylRS⋅tRNA^Pyl pair study, sfGFP-3, sfGFP-4, and sfGFP-5 were expressed and purified, respectively, and then subjected to ESI-MS analysis (Table 1 and Supplementary Figures S1–S3). The calculated molecular weight of sfGFP-3, sfGFP-4, and sfGFP-5, generated by N-PylRS, matched the observed mass in the full-length sfGFP (Figure 7). The incorporation of ncAA 3, 4, and 5 into sfGFP at position 27 was shown by MALDI-TOF-MS/MS (Figure 8 and Supplementary Figures S12–S14). This multiple evidence indicates that MbzlC (3), MeObzlC (4), and 3MeH (5) are incorporated into proteins by non-active, mutated N-PylRS⋅tRNA^Pyl pair site-specifically.