The gene of Thr-free super-folder GFP (sfGFP) was ordered from Integrated DNA Technologies (Coralville, IA, United States). The sequence is in the Supplementary Material. The primers for generating sfGFP variants with a single-threonine residue are listed in Supplementary Table S1. The mutations were made by the Q5 Site-Directed Mutagenesis Kit from New England Biolabs (Ipswich, MA, United States) following the manufacturer’s protocol. The DNA sequences of these variants were confirmed by DNA sequencing (Eurofins Genomics, Louisville, KY, United States). The gene of sfGFP or its variants was transformed into Escherichia coli BL21 DE3 cells (New England Biolabs, Ipswich, MA, United States) for expression. Detailed procedures for vector construction are in the Supplementary Material.

The strain harboring the plasmid to express wild-type (WT) sfGFP or its variants was inoculated into 2 ml minimal medium (M9 medium and 0.4% glucose) with 100 μg/ml streptomycin and incubated at 37°C overnight, individually. Ten microliters of each overnight culture was added into 190 μl fresh minimal medium with 100 μg/ml streptomycin to OD_600nm ∼ 0.15. The 200 μl mixture was pipetted into a well in a 96-well plate with supplementary 0.1 mM IPTG to induce protein expression. The fluorescence intensity (excitation wavelength 485/20 nm; emission wavelength 528/20 nm) and cell growth (OD_600nm) were monitored by a microplate reader with continuous shaking at 37°C for 6 h. Means and standard deviations were calculated by three replicates for initial screening and five replicates for measuring mistranslation rates. Normalized fluorescence was the fluorescence reading divided by the cell culture density (OD_600nm) or by the protein yield quantified by ELISA. The mistranslation rate equals 1 minus the rate of non-mistranslation, which is the normalized fluorescence of cells expressing threonine-tRNA synthetase (ThrRS) variants divided by that of cells expressing WT-ThrRS.

For easy purification, a His₆-tag fused to the WT sfGFP or its variants by PCR used the NEBuilder HiFi DNA Assembly Master Mix Kit. The expression and purification procedures followed the previous protocol (Venkat et al., 2017b). Detailed procedures were in the Supplementary Material. The site-specifically acetylated ThrRS variant was generated by our established genetic incorporation system (Venkat et al., 2017a). Protein concentrations were measured by the Bradford Protein Assay (Bio-Rad, Hercules, CA, United States). Purified proteins were fractionated on a 12% SDS-PAGE gel and visualized by the Bio-Safe Coomassie stain (Bio-Rad). To quantify GFP yields, ELISA was performed with anti-His₆ tag (Abcam, Cambridge, MA, United States).

The LC-MS/MS analyses were performed by Yale University Keck Proteomics Facility following previous protocols (Venkat et al., 2019). The purified ThrRS was digested in gel by trypsin and analyzed by LC-MS/MS on an LTQ Orbitrap XL equipped with a nanoACQUITY UPLC system. The Mascot search algorithm was used to search for the acetyllysine modifications. CD spectrometry was performed by previous protocols (Chen et al., 2019). The CD spectra were recorded on a J-1500 CD spectrometer. Purified proteins were diluted to a concentration of 0.1 mg/ml in 5 mM Tris-HCl pH 7.8, 0.1 M KCl, and scanned from 190 to 250 nm with a 20 nm/min speed. Scanning was performed three times for each sample, and the average was plotted.

ThrRS uses a zinc ion to discriminate against the valine (without hydroxyl group) at the activation step and utilizes the N-terminal editing domain to hydrolyze mischarged tRNA^Thr with serine (with a smaller size) (Sankaranarayanan et al., 2000; Dock-Bregeon et al., 2004). Because ThrRS solves such a unique double-discrimination problem with isosteric amino acids, it has attracted much attention for studying its mistranslation (Dock-Bregeon et al., 2000). However, there is still no facile method to quantify mistranslation of threonine codons in living cells. In this study, we engineered the sfGFP for this purpose. As mentioned, ThrRS can mischarge tRNA^Thr with serine. Besides serine, alanyl-tRNA synthetase (AlaRS) can also mischarge tRNA^Thr with alanine (Sun et al., 2016). The sfGFP has 18 threonine residues in total. All these threonine codons could be mistranslated as serine or alanine. Thus, we actually observe the overall effect of mistranslation of all these threonine codons on GFP fluorescence. Mistranslation at some sites could increase fluorescence, while that at others could impair fluorescence. So we may underestimate or overestimate the actual mistranslation. To eliminate possible interference from mistranslation at other threonine codons and focus on one specific threonine codon, we aimed to generate an sfGFP variant containing only one threonine residue which is essential for its fluorescence (Figure 1).

First, we mutated all the threonine codons in the gene of sfGFP to serine codons because serine is the most common mistranslated amino acid for threonine codons. As expected, this Thr-free sfGFP variant (TF-sfGFP) had no fluorescence (Figure 2). The protein yield of TF-sfGFP was similar to that of WT-sfGFP with the same expression condition (Supplementary Figure S1), indicating that substitution of threonine residues with serine at all 18 sites does not affect GFP expression significantly. We also performed CD spectrometry analyses with both TF-sfGFP and WT-sfGFP to see whether such substitution can affect GFP folding. The result showed that there was no significant difference between them (Supplementary Figure S2). In the next step, we put the threonine codon back to its original position in the gene of TF-sfGFP individually to generate 18 single-threonine sfGFP variants in total. We measured fluorescence intensities for these sfGFP variants (Figure 2). Among these variants, the TF-sfGFP T203 variant restored ∼20% fluorescence with one threonine residue alone. We also noted that protein yields of all these variants were similar to those of WT-sfGFP with the same expression condition (Supplementary Figure S1).

T203 plays a critical role in forming the structure similar to a proton pump for GFP and is well conserved among GFP-like proteins (Ong et al., 2011), which explained the recovery of sfGFP fluorescence. Besides T203, T65 is well known to increase the fluorescence of GFP (Heim et al., 1995). But the TF-sfGFP T65 variant in this study had no significant fluorescence, probably because it is a good enhancer, but not essential for producing fluorescence. Actually, the original amino acid at this site in WT-GFP is serine. Besides serine, AlaRS can also mischarge tRNA^Thr with alanine (Sun et al., 2016). To eliminate the possibility that the substitution of threonine with alanine at position 203 can also generate fluorescence, we generated the TF-sfGFP A203 variant. This variant had no fluorescence (Figure 2). Thus, the TF-sfGFP T203 variant can be used as the reporter for quantifying mistranslation of the threonine codon, because (i) mistranslation of the threonine codon to serine or alanine will eliminate the fluorescence of sfGFP; (ii) there is no interference of mistranslation of other threonine codons in the sfGFP; and (iii) the fluorescence intensity depends on the ratio of mistranslated sfGFP and total amounts of sfGFP.

In the previous quantitative MS study on ThrRS-mediated mistranslation, the ThrRS editing-deficient variant ThrRS C182A was used to evaluate the mistranslation rate at threonine codons (Mohler et al., 2017a). To compare our method with the MS approach, we expressed the TF-sfGFP T203 variant in the strain containing ThrRS C182A. The strain containing WT-ThrRS was used as the control (Table 1). Both the protein yield and cell culture density were decreased in the strain containing ThrRS C182A, which is consistent with the previous study (Ling and Söll, 2010). To address potential effects of E. coli cell autofluorescence on GFP fluorescence readings, which could be caused by different cell densities and stress conditions (Mihalcescu et al., 2015; Surre et al., 2018), we used the corresponding stains harboring the same vector but without the reporter gene as controls. Fluorescence readings were subtracted with corresponding backgrounds. Moreover, we compared two approaches to normalize fluorescence intensities, either by cell culture densities from OD_600nm (Supplementary Figure S3) or by protein yields from ELISA. Both normalization approaches had similar results and showed that the mistranslation rate caused by the ThrRS C182A variant is ∼3%, which is consistent with previous quantitative MS studies (Mohler et al., 2017a). As cell culture densities were monitored simultaneously with fluorescence reading, we used this more facile approach for later experiments.

To compare TF-sfGFP T203 and WT-sfGFP in evaluating mistranslation, we tested the effect of the ThrRS C182A variant on their fluorescence, individually. We also applied a strategy similar to that used in previous mistranslation studies (Lee et al., 2006) and added serine in growth media to force threonine-to-serine mistranslation (Figure 3). With the increase of serine concentrations, both reporters had increased mistranslation rates as expected. WT-sfGFP gave higher mistranslation rates than the single-threonine sfGFP reporter at all conditions, probably because threonine-to-serine substitution at other threonine sites further decreased its fluorescence. In this case, we could overestimate mistranslation rates by using WT-sfGFP as the reporter. Furthermore, the single-threonine sfGFP reporter was less sensitive to serine concentrations than WT-sfGFP, which is another advantage of this reporter.

In our recent study on acetylation of ThrRS, we found that acetylation of K169 in E. coli ThrRS can generate Ser-mischarged tRNA^Thr in vitro (Chen et al., 2019). However, the effect of acetylation of ThrRS on mistranslation of threonine codons in living cells is unknown. So we applied the single-threonine sfGFP reporter to evaluate the impact of acetylation of ThrRS on threonine mistranslation in vivo. The site-specifically acetylated E. coli ThrRS at K169 (ThrRS-169AcK) was generated by the genetic code expansion strategy in E. coli cells and confirmed by LC-MS/MS (Supplementary Figure S4). The TF-sfGFP T203 variant was co-expressed in the strain containing ThrRS 169AcK. Results showed that the mistranslation rate caused by ThrRS acetylation was 4.13 ± 0.06%. K169 is located at the opening of the editing site of ThrRS (Supplementary Figure S5). We proposed that the ThrRS-169AcK variant has an impaired activity to hydrolyze mischarged Ser-tRNA^Thr due to steric hindrance from the additional acetyl group. The crystallography study on this acetylated ThrRS variant is ongoing.