Construction of the natural/ in vitro transcribed (IVT)-hybrid tRNA set
E. coli has 37 tRNAs that decode 61 codons into 20 proteinogenic amino acids. Among these, tRNASer, tRNALeu, and tRNATyr are longer (85–93 bases) than the other tRNAs (74–77 bases) (Fig. 2A). This characteristic allows us to prepare a tRNASer and tRNALeu deficient natural tRNA set without having to isolate each individual tRNA.
Initially, we analyzed the E. coli tRNA extract using denaturing PAGE containing urea (Fig. 2A, lane 1). The mobility of band I was consistent with that of IVT tRNASer, tRNALeu, and tRNATyr (Fig. 2A, lanes 3–5). Thus, we inferred that this band represents the natural tRNASer, tRNALeu, and tRNATyr. We harvested the tRNAs in band II (Fig. 2A, lane 2) and defined this tRNA mixture as the ∆SLY natural tRNA set.
Subsequently, we prepared IVT tRNATyr, chimeric tRNASerGAG, and chimeric tRNALeuGGA using T7 RNA polymerase. We then combined these with the ∆SLY natural tRNA set to assemble the natural/IVT-hybrid tRNA set for the Ser/Leu-swapped genetic code, which we termed the hybrid-SL tRNA set (Fig. 2B). Similarly, we assembled the natural/IVT-hybrid tRNA set for the standard genetic code (hybrid-Std tRNA set) using the IVT tRNASer and tRNALeu instead of chimeric tRNAs. We also assembled a tRNA set with 21 IVT tRNAs for the Ser/Leu-swapped genetic code, termed the IVT-SL tRNA set. Using one of these three tRNA sets, we prepared reconstituted cell-free translation systems for further study (Figure S2).
Synthesis of peptides in a cell-free translation system with the hybrid-Std tRNA set
To verify the accurate decoding of codons in the cell-free translation system using the hybrid-Std tRNA set, we first assessed the synthesis of model peptides (Met-Tyr-Tyr-Tyr-Xaa-Asp-Asp-Arg-Asp, where Xaa represents 17 types of amino acids; Fig. 3A). The matrix assisted laser desorption/ionization-time of flight mass spectrometer (MALDI-TOF-MS) analysis demonstrated that all peptides were synthesized in the cell-free translation system (Fig. 3B top and Figure S3A; calculated and observed masses are listed in Tables S1 and S2). Tricine SDS-PAGE analysis followed by autoradiography ([14C]-Asp labeling) showed comparable amounts of peptides synthesized across all samples (Fig. 3C and Figure S3B). These results suggest that the natural tRNA species corresponding to 17 natural amino acids in the ΔSLY natural tRNA set accurately decode Xaa codon. Tyr insertion was unexpectedly observed for peptides with Glu, Phe, Trp, and Lys. This anomaly might be due to translational slippage on repeated Tyr codons preceding the Xaa codon. Various translational slippage, such as − 1, − 4, and + 2 frameshift at the NNA-AAC-AAG sequence in an E. coli cell-free translation system20, as well as + 2 frameshift at the GUG-UG sequence and + 6 ribosome hopping at the GUG-UGA-GUU sequence in E. coli21, have been reported. In addition, the depletion of an aminoacyl-tRNA that corresponds to an A-site codon can increase the − 1 frameshifting in an E. coli cell-free translation system22. This suggests that the unexpected insertion of Tyr observed in the mass spectra might be due to the low concentration of aminoacyl-tRNAs of Glu, Phe, Trp, and Lys (vide infra).
Accuracy of Ser/Leu-swapping in a cell-free translation system with the hybrid-SL tRNA set
The presence of natural tRNASer and tRNALeu in the IVT-SL tRNA set could compromise accurate translation in the Ser/Leu-swapped cell-free translation system, leading to the synthesis of proteins that contain both Ser and Leu at the same positions. While the PAGE analysis indicated no detectable amount of natural tRNASer and tRNALeu in the ∆SLY natural tRNA set (Fig. 2A, lane 2), we examined the accuracy of Ser/Leu-swapping by translating model peptides (Met-Tyr-Tyr-Tyr-Xaa-Asp-Asp-Arg-Asp, where Xaa represents amino acid at UCU or CUU codons; Fig. 3A).
Using a cell-free translation system with the hybrid-SL tRNA set, we synthesized the model peptides with Ser or Leu at the Xaa position. MALDI-TOF-MS analysis confirmed that Ser and Leu were accurately assigned to their respective swapped codons in the Ser/Leu-swapped genetic code without mutual contamination (Fig. 3B middle and Table S1). Tricine SDS-PAGE analysis further supported this finding as translated products formed a single intense band for each mRNA (Fig. 3C).
To assess the importance of excluding natural tRNASer and tRNALeu from the E. coli tRNA extract, we additionally tested a mix of intact E. coli tRNA extract with chimeric tRNASerGAG and tRNALeuGGA. The mixture of peptides with Ser and Leu were detected in both MALDI-TOF-MS (Fig. 3B bottom and Table S1) analysis and tricine SDS-PAGE (Fig. 3C). These results indicated that the excluding the natural tRNASer and tRNALeu from E. coli tRNA extract is crucial for the accurate Ser/Leu-swapping in the cell-free translation system.
Optimization of the hybrid tRNA set for translating a protein
We next attempted to synthesize a model protein in the cell-free translation system using the hybrid-Std tRNA set. We employed standard sfGFP gene encoding superfolder GFP (sfGFP) according to the standard genetic code and assessed the sfGFP production by a spectrofluorometer (Figure S5A, Std-22C). Unexpectedly, the fluorescence intensity of sfGFP synthesized using hybrid-Std tRNA set was only 4% compared to synthesis using E. coli tRNA extract (Figure S6).
Considering the possibility that the concentration of certain natural tRNA(s) may have been reduced during the PAGE purification process, we attempted to identify these tRNA(s) by adding IVT tRNAs. We divided 18 IVT tRNAs into four groups (Groups A–D) and assessed whether the sfGFP fluorescence intensity improved upon adding each group of tRNAs. Notably, only Group A tRNAs (comprising tRNAPhe, tRNALys, tRNATrp, tRNACys, and tRNAAla) significantly enhanced sfGFP production (Figure S6). Then, we added each tRNA from Group A to the cell-free translation system. We found that only the addition of tRNAPhe increased the production of sfGFP to approximately 78% of that from the natural tRNA extract (Fig. 4). The denaturing PAGE analysis revealed a smeared band corresponding to IVT tRNAPhe (Fig. 2A, lane 6), suggesting partial removal of tRNAPhe during the PAGE purification. Therefore, we decided to include IVT tRNAPhe in the hybrid tRNA sets for subsequent experiments.
Ser/Leu-swapped synthesis of a model protein using the hybrid-SL tRNA set
To study the synthesis of the model protein sfGFP in the Ser/Leu-swapped cell-free translation system, we prepared a Ser/Leu-swapped sfGFP gene encoded according to the Ser/Leu-swapped genetic code (Figure S5A, SL-22C). Native PAGE analysis showed that the amount of active sfGFP in the cell-free translation system with the hybrid-SL tRNA set was comparable to that with the E. coli tRNA extract (Fig. 5, lane 4 vs. 1). Furthermore, the amount of active sfGFP was significantly improved to 3.5 times compared to that with the IVT-SL tRNA set (Fig. 5, lane 4 vs. 5). This result clearly indicates that the cell-free translation system with the hybrid-SL tRNA set is useful for in vitro protein production.
We also studied the orthogonality of the Ser/Leu-swapped genetic code and the standard genetic code by translating the protein using mismatched tRNA set/gene combinations: i.e., the Ser/Leu-swapped sfGFP gene with E. coli tRNA extract and the standard sfGFP gene with the hybrid-SL tRNA set. Indeed, no active sfGFP was detected in either cell-free translation system (Fig. 5, lanes 2 and 3). This is because the translation products synthesized by incorrect combinations contained 30 mutations (10 mutations of Ser to Leu and 20 mutations of Leu to Ser) (Figure S5B); hence, 13% of the residues in the sfGFP (comprising 239 residues) were mutated. Autoradiography of the same gels also showed that proteins with different mobilities from sfGFP were synthesized in both cell-free translation systems with mismatched tRNA set/gene combinations (Fig. 5, lanes 6–10).
Comparison of hybrid-SL tRNA and IVT-SL tRNA sets for translating genes encoded by 22 codons or 47 sense codons
Using the IVT-SL tRNA set, only 35 of the 61 sense codons (Figure S2) are available to encode a protein, as the tRNA set consists of 21 IVT tRNAs. In contrast, 53 of the 61 sense codons are available to encode a protein when using the hybrid-SL tRNA set, as the tRNAs for 16 standard amino acids (excluding Ser, Leu, Tyr, and Phe) were isolated from the E. coli tRNA extract (Figure S2).
In the aforementioned study, we used a Ser/Leu-swapped sfGFP gene containing 22 sense codons (Figure S5A, SL-22C). To verify codon availability, we prepared a Ser/Leu-swapped sfGFP gene containing 47 sense codons (excluding rare codons) (Figure S5A, SL-47C). The synthesis of inactive proteins in a cell-free translation system using natural tRNA extract from this gene demonstrated the orthogonality of the Ser/Leu-swapped genetic code with 53 sense codons against the standard genetic code (Figure S8).
In the cell-free translation system with the hybrid-SL tRNA set, a similar amount of active sfGFP was synthesized from both genes (Fig. 6, lanes 1, 2, 5, and 6), suggesting that at least 47 sense codons are available to encode a protein. In contrast, while sfGFP was synthesized from the Ser/Leu-swapped sfGFP gene containing 22 sense codons (Fig. 6, lanes 4 and 8), none was produced from the gene containing 47 sense codons (Fig. 6, lanes 3 and 7) in the cell-free translation system using the IVT-SL tRNA set. This is attributed to the IVT-SL tRNA set’s lack of tRNAs to decode 14 codons, representing 26% of the coding sequence, in the SL-47C_sfGFP gene.
Recent studies have reported that the choice of the nonsynonymous codons affects the accuracy of an amino acid incorporation23 and a protein folding through the control of the translation rate24, 25 for certain proteins. Hence, the cell-free translation system with the hybrid-SL tRNA set could be invaluable for designing a gene.