Enzymes for molecular cloning were purchased from New England Biolabs and Takara Bio. Gene fragments were purchased from Integrated DNA Technologies and Twist Bioscience. W.M. Keck Foundation at Yale University provided oligonucleotide synthesis (Supplementary Table S1) and DNA plasmid (Supplementary Table S2) sequencing. Antibiotics and media additives were used at the following final concentrations: ampicillin (amp), 100 μg/mL; spectinomycin (spec), 50 μg/mL; glucose, 1% (w/v); arabinose, 0.1% (w/v); sodium selenite, 10 μM; N _ε-acetyl-l-lysine (AcK), 10 mM; nicotinamide, 20 mM.

Plasmids were transformed into electrocompetent E. coli cells (B-95.ΔA.ΔfabRΔselABC) (Mukai et al., 2015) and plated on Luria Broth (LB) agar containing appropriate antibiotics and incubated overnight at 37°C. Single colonies were grown in 150 μL media at 37°C for 6 h in a 96-well black plate with clear bottoms. Unless otherwise noted, the media containing appropriate antibiotics included glucose, arabinose, sodium selenite, AcK, and nicotinamide was used. After 6 h, 75 μL from each well was transferred to a clean well in the same plate. The original wells were replenished with 75 μL of fresh media, while protein expression was induced in the new wells by adding 75 μL of media containing 2 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). Fluorescent measurements and analysis were performed as described previously (Chung et al., 2022) using a minimum of four biological replicates.

For protein expression, pB_PM17, pB_PM21, or pB_PM22 plasmids were co-transformed with pSecUGA into electrocompetent E. coli cells (B-95.ΔA.ΔfabRΔselABC) (Mukai et al., 2015) before plated on LB agar containing appropriate antibiotics and incubated overnight at 37°C. Single colonies were grown overnight at 37°C in 25 mL LB containing the appropriate antibiotics. Precultures were transferred to 1 L of LB containing antibiotics, arabinose, and sodium selenite and grown at 37°C. At an OD₆₀₀ = 0.6, protein expression was induced with 0.5 mM IPTG, following the addition of nicotinamide, AcK, and extra sodium selenite. Proteins were expressed at 20°C for 16 h before pelleting the cells for purification.

All purification steps were performed under anaerobic conditions (90% N₂, 5% H₂, 5% CO₂) in an anaerobic tent (Coy Laboratories). Each cell pellet was resuspended in 18 mL lysis buffer (50 mM sodium phosphate [pH 8.0], 300 mM NaCl, 30 mM imidazole, 10% (v/v) glycerol, 120 μg/mL lysozyme, 30 μg/mL Dnase, 0.5 mM phenylmethylsulfonyl fluoride [PMSF]) and 2 mL BugBuster^® 10 X Extraction Reagent (EMD Millipore). The lysed cells were incubated at room temperature for 20 min before the lysate was ultracentrifuged at 45,000 rpm (150,000 x g) for 45 min at 4°C. The supernatant was 0.45 μm filtered before being loaded onto a 1 mL Ni-NTA column pre-equilibrated with wash buffer (50 mM sodium phosphate [pH 8.0], 300 mM NaCl, 30 mM imidazole, 10% (v/v) glycerol). The beads were washed with 50 mL of wash buffer and then eluted in 1 mL fractions with elution buffer (50 mM sodium phosphate [pH 8.0], 300 mM NaCl, 250 mM imidazole, 10% (v/v) glycerol). Protein fractions were combined, concentrated, and buffer exchanged into storage buffer (50 mM sodium phosphate [pH 8.0], 300 mM NaCl, 10% (v/v) glycerol) before being stored at −80°C.

LC-MS analysis of dithiothreitol reduced samples were performed on a ThermoFisher Scientific Orbitrap Exploris 240 mass spectrometer, equipped with a heated electrospray ionization source (H-ESI) in positive ion mode with a ThermoFisher Ultimate 3000 RSLCnano HPLC System. On H-ESI source, sheath gas was set to 2 arbitrary units (arb), and auxiliary gas was set to 6 arb. The ion transfer tube was set at 275°C and the vaporizer temp at 200°C. The sample was analyzed on a MAbPac RP, 4 μM, 3.0 mm × 50 mm analytical column (ThermoFisher Scientific), held at 60°C. The protein was eluted at a rate of 500 μL/min for a 10-min gradient, where 0–7 min: 10%–70% acetonitrile + 0.1% formic acid; 7–8.2 min: 95% acetonitrile + 0.1% formic acid, 8.2–10 min: 20% acetonitrile + 0.1% formic acid. MS spectra were acquired using full scans at 15,000 resolution in the orbitrap within a range of 700–2,200 m/z. The maximum injection time was set at auto with a standard AGC target. Ten micro scans were employed, and the RF lens was set to 100%. 15 V of insource CID was applied. Thermo BioPharma Finder 5.1 was used for intact mass deconvolution and peak identification.

Approximately 5 μg of GPx1 protein samples were loaded onto a 4%–20% Mini-PROTEAN TGX stain-free gel (Bio-Rad) in reducing conditions. Samples were transferred to nitrocellulose and acetylation was detected with a primary mouse anti-acetylated lysine mAb (Novus Biologicals) and secondary anti-mouse IgG HRP-linked antibody (Cell Signaling Technology).

Using our established pSecUAG plasmid system for insertion of Sec at UAG codons, we first engineered the anticodon of M. alvus tRNA^Pyl to recode UGA codons (tRNA^Pyl _UCA). Before testing the orthogonality of each tRNA for its respective codon, we confirmed the specificity of MaAcLysRS3-IP for AcK (Seki et al., 2020). E. coli cells expressing only pB_PM08 showed significantly higher fluorescence (p < 0.0001) in the presence of AcK compared to cells without the amino acid (Figure 2A). This suggests that AcK promotes readthrough of the UGA codon in sfGFP with the MaAcLysRS3:tRNA^Pyl _UCA pair.

When using multiple translation systems to suppress more than one stop codon, confirming the codon specificity and orthogonality of engineered suppressor tRNAs is imperative. Therefore, we tested the ability of each system to readthrough UAG or UGA independently. To first investigate the ability of tRNA^Pyl _UCA to specifically suppress UGA, we expressed the plasmid pB_PM08 or pB_PM09 in the presence of AcK. Notably, a significant increase (p < 0.0001) in GFP fluorescence for readthrough of 151UAG relative to 2UGA was observed (Figure 2B), suggesting that tRNA^Pyl _UCA has a propensity to decode UAG codons better than its cognate UGA codon. We also tested this for allo-tRNA^UTu1D _CUA by expressing plasmids pB_PM02 or pB_PM03 together with pSecUAG in the presence of sodium selenite. Interestingly, allo-tRNA^UTu1D _CUA translated both 2UGA and 151UAG, with a significant increase in fluorescence (p < 0.01) observed for the intended UAG codon (Figure 2C). Together, these results show that tRNA^Pyl _UCA and allo-tRNA^UTu1D _CUA can suppress both UGA and UAG codons in isolation, displaying a higher efficiency to recode UAG (Figure 2D). The propensity to recode UAG was also observed when testing the suppression efficiency of each codon individually (pB_PM08 or pB_PM09) and together (pB_PM10) compared to wild-type (pB_PM07) in the presence of AcK and sodium selenite (Supplementary Figure S1A). Thus, these two tRNA variants are not entirely orthogonal.

To test whether codon specificity can be attained, we swapped the anticodons of both tRNAs, resulting in tRNA^Pyl _CUA and allo-tRNA^UTu1D _UCA. This led to the pB series (pB_PM14, pB_PM15, pB_PM16, and pB_PM17) and pSecUGA to decode UAG and UGA, respectively. We again confirmed that the MaAcLysRS3:tRNA^Pyl _CUA pair relies on AcK for efficient readthrough (p < 0.01) (Figure 3A). Importantly, we found that tRNA^Pyl _CUA almost exclusively translates the UAG codon, with minimal decoding of UGA. Expressing only pB_PM15 or pB_PM16 we observed fluorescence when 151UAG is present but not in the presence of 2UGA (p < 0.0001) (Figure 3B). Similarly, allo-tRNA^UTu1D _UCA displayed a higher specificity for UGA than UAG. Expressing pSecUGA with pB_PM02 or pB_PM03, we observed the opposite, fluorescence with 2UGA and not 151UAG (p < 0.0001) (Figure 3C). Furthermore, the suppression efficiencies of each codon individually (pB_PM15 or pB_PM16) and together (pB_PM17) were comparable and roughly 60%–70% of wild-type sfGFP (pB_PM14) in the presence of AcK and sodium selenite (Supplementary Figure S1B). These data suggest that allo-tRNA^UTu1D _UCA and tRNA^Pyl _CUA can simultaneously insert AcK at UAG and Sec at UGA into proteins efficiently (Figure 3D).

Next, we combined the AcK and Sec incorporation systems to produce a synthetic post-translationally modified selenoprotein (AcK-seleno-sfGFP). We first tested the activities of the MaAcLysRS3:tRNA^Pyl _CUA and the Sec-translation systems (Figure 4A) in the presence and absence of AcK and arabinose (for allo-tRNA^UTu1D _UCA expression). Significant sfGFP fluorescence (p < 0.001) was only observed in the presence of both AcK and arabinose but not if one or both is missing (Figure 4B), confirming the feasibility and compatibility of these two systems. Due to the possibility of Ser misincorporation by the pSecUGA system, we confirmed Sec incorporation using pB_PM18 (Figure 5A) which was engineered based on our previously developed Sec-specific sfGFP reporter (pB_04) (Chung et al., 2022). The sfGFP reporter is only functional (fluorescent) when Sec is inserted at amino acid 204 (position 1 of the M86 mini intein) to facilitate cleavage and splicing of the two sfGFP fragments (Appleby-Tagoe et al., 2011). Moreover, mutagenesis of codon 2 to UAG and the stop codon of the sfGFP to UAA allows for simultaneously monitoring the insertion of two ncAAs. By monitoring the fluorescence of the intein-sfGFP reporter, we only observed significant activity when both suppression systems and sodium selenite (Figure 5B) were present. These results corroborate that the fluorescence observed with the pB_PM17 plasmid (Figure 4B) is due to the presence of Sec. Moreover, the lack of fluorescence in the absence of AcK verifies its requirement and insertion at UAG (Figures 3A, 4B).

The sfGFP readthrough and intein assays demonstrated that AcK and Sec could be inserted into a single protein using two different stop codons. Therefore, we applied our integrated plasmid setup and growth conditions to express and purify acetylated seleno-sfGFP from E. coli. The location of the His₆-tag on the C-terminus allowed the separation of full-length protein from truncated versions due to early termination (Figure 5A). This strategy produced a robust band at roughly 25 kDa correlating to our expected protein size (28 kDa) (Figure 6). We also compared our optimized system, which involves recoding UGA with Sec and UAG with AcK, to our original system (recoding UGA with AcK and UAG with Sec) (Figure 6A). In both systems, we obtained pure protein; however, our yields were increased roughly 5-fold with our optimized codon set compared to the original (Figure 6B). Mass spectrometry analysis of the optimized codon set confirmed the presence of acetylated seleno-sfGFP. Protein mass was also detected that corresponded to Gln instead of AcK at a proportion of 67%, suggesting that near-cognate suppression by tRNA^Gln can impact the final protein purity (Figure 6C). This was further verified with MS/MS data to show that indeed AcK or Gln was inserted at position 151 while Sec was always observed at position 2 (Supplementary Figure S2).

To prove the versatility and utility of our system, we produced an authentic selenoprotein, human GPx1. GPx1 is predicted to be acetylated at different Lys residues based on murine GPx1 MS studies (Rardin et al., 2013). We individually targeted K114 and K148 together with the Sec at position 49. The protein was expressed under the same conditions with the same plasmid backbone as described for sfGFP (Figure 7A). Three bands were visible on the polyacrylamide gel, with the lower band corresponding to the expected size of monomeric GPx1 (25 kDa). In terms of protein yield, we found that GPx1_148AcK had roughly twice as much protein as GPx1_114AcK. This may be a result of codon context or the closer proximity of position 114 to 49Sec (Figure 7B). To confirm incorporation of AcK, a western blot with an anti-AcK antibody was used (Figure 7B). The western blot revealed the presence of acetylation in the higher molecular weight band which is approximately 30%–35% of the total protein (Supplementary Figure S3). We surmise that the lower molecular weight band corresponds to a GPx1 variant containing Gln (65%–70%) instead of AcK as suggested by the sfGFP MS data. This provides evidence that imply the potential role of AcK in GPx1 oligomerization, and emphasizes the necessity for a system which can study post-translationally modified selenoproteins.