FastDigest restriction enzymes were purchased from Fermentas (Burlington, Canada). The EasyGeno Assembly Cloning kit was purchased from Tiangen Biotech (Beijing, China). The 2× Taq PCR Master Mix was purchased from Beijing Solarbio Science & Technology Co., Ltd. (Beijing, China). KOD-Plus-Neo DNA polymerase was purchased from Toyobo Co., Ltd. (Osaka, Japan). Oligonucleotide primers and genes were synthesized by GENEWIZ Bio Inc. (Suzhou, China). All other chemicals and reagents used were of analytical grade.

Detailed information on the strains used in this study is presented in Table 1. Escherichia coli EcNR2, E. coli C321.ΔA, and E. coli C321.ΔA.exp were obtained from Addgene (Watertown, MA, United States). Escherichia coli DH10B and E. coli BL21(DE3) were purchased from Thermo Fisher Scientific (Rockford, IL, United States) and Novagen (Darmstadt, Germany), respectively. Unless otherwise indicated, bacteria were grown in LB medium at 37 or 30°C (for E. coli EcNR2 and E. coli C321.ΔA only) supplemented with kanamycin (30μg/ml), spectinomycin (50μg/ml), chloramphenicol (34μg/ml), and/or ampicillin (50μg/ml) when needed.

After the OD₆₀₀ values of the culture reached 0.6~0.8, the cells were harvested by centrifugation and washed three times with sterile phosphate buffer at 4°C. Total RNA was extracted from the pellet cells using the RNeasy Plant Mini Kit (Qiagen, CA, United States) according to the manufacturer’s protocol. The purity and concentration of the resulting RNA were determined using a NanoDrop instrument (Thermo Scientific, Wilmington, DE, United States), and the integrity was analyzed using 1% agarose gel electrophoresis. RNA integrity number values were calculated using the Agilent Bioanalyzer 2,100 system (Agilent Technologies, Santa Clara, United States). Ribosomal RNA (rRNA) was depleted from the samples using the Illumina Ribo-Zero rRNA removal kit, and the resulting mRNA was fragmented with an average insert length of approximately 200bp. After construction of the cDNA library using the TruSeq Stranded Total RNA Library Prep Kit (Illumina, San Diego, CA, United States), the sequences were determined using the Illumina HiSeq platform at Majorbio Bio-Pharm Technology Co., Ltd., Shanghai, China. To gain statistical confidence, RNA-seq experiments were performed in three independent biological replicates. The datasets generated for this study can be found in the NCBI with an accession number of PRJNA748036.

The quality of the raw reads was filtered and controlled using SeqPrep¹ and Sickle software (Version 1.33). The resulting valid sequences were then mapped and aligned to the reference genome (E. coli C321.ΔA)² using BLAST⁺ (version 2.9.0) and Bowtie2 (version 2.3.5).³ The rRNA contamination assessment was performed by randomly selecting 10,000 raw reads in each sample and aligning them to Rfam database⁴ with blast method. Based on the annotation results, percentage of rRNA in each sample is calculated, which is estimated as rRNA contamination. The bioinformatics analyses were performed using the free online platform of Majorbio Cloud Platform⁵ from Shanghai Majorbio Bio-pharm Technology Co., Ltd. The gene expression level analyses were performed using RSEM tool.⁶ FPKM and TPM method are used to calculated expression level, FPKM represents fragments per kilobase of exon model per million mapped reads, and TPM represents transcript per million mapped reads. The FPKM and TPM are able to eliminate the influence of different gene length and sequencing discrepancy on the calculation of gene expression. Therefore, the calculated gene expression can be directly used for comparing the difference of gene expression among samples. Principal coordinate analysis (PCoA) was completed using edgeR.⁷ The differentially expressed genes (DEGs) and transcripts in each sample were determined using the Salmon (version 0.14.1) and DESeq2 (version 1.24.0)⁸ program with default settings. Significant differences in gene expression were evaluated using one-way ANOVA with a set of value of p<0.05, and fold-change ≥2. Goatools⁹ is exploited to identify statistically significantly enriched Gene Ontology (GO) term using Fisher’s exact test. The purpose of performing FDR correction is to reduce the Type-1 error by Bonferroni, Holm, BY, BH (multiple hypothesis test method). After multiple testing correction, GO terms with corrected value of p≤0.05 are significantly enriched in DEGs. KOBAS 2.0¹⁰ is exploited to identify statistically significantly enriched pathway using Fisher’s exact test. The purpose of performing FDR correction is to reduce the Type-1 error by Bonferroni, Holm, BY, BH (multiple hypothesis test method). The calculating formula of value of p and corrected value of p is similar with that in GO analysis. After multiple testing correction, we choose pathways with value of p<0.05 that are significantly enriched in DEGs. Functional analyses of the DEGs were performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG), GO, Swiss-Prot (version 2019.7.1), and Pfam (version v32.0) databases.

After washing three times with phosphate buffer at 4°C, the pelleted cells were resuspended in 2ml of lysis buffer (20mM HEPES, 10mM NaOH, pH 7.5) containing 150mM NaCl and 1% protease inhibitor cocktail (GE Healthcare, Pittsburgh, United States) and sonicated on ice. After centrifugation at 15,000×g for 20min to remove cell debris, the protein samples were precipitated using trichloroacetic acid at a final concentration of 20% (w/v) and washed with ice-cold acetone. After air drying, the precipitated proteins were stored at −80°C until use. Total proteins were analyzed by SDS-PAGE, and the concentrations were determined using the Bradford method with bovine serum albumin as a standard. Before mass spectrometric analysis, the protein samples were dissolved in 8M urea/100mM NH₄HCO₃ and subjected to reductive alkylation and trypsin hydrolysis. Three biological replicates were performed for each strain in the proteomic analyses.

The Mascot and Proteome Discoverer software was used to search the raw LC–MS/MS data against the UniProt E. coli database. Significant differences in protein expression were evaluated using one-way ANOVA with a set of value of p<0.05 and fold-change ≥2. Differentially expressed proteins were analyzed by KEGG pathway analysis and GO analysis using the Swissport database and the Pfam database to identify significantly enriched KEGG pathways and GO functions. Based on the GO and KEGG categories, proteins were classified according to cellular components, molecular functions, and biological processes.

Detailed information on the plasmids used in this study is presented in Table 1. The plasmids pCAS, pTargetF, pULTRA-CNF, and pEVOL-pBpF were obtained from Addgene. The plasmids pBAD24-GFP and pET26b-GFP were previously used in our laboratory. The gene of a previously reported orthogonal tRNA-synthetase pair that incorporated 5-hydroxytryptophan (5-HTP) into proteins was chemically synthesized and cloned into a pULTRA plasmid to generate the plasmid pULTRA-5HTP. The primers used in this study are listed in Table 2. The primer pairs GFP-Y39TAG-F and GFP-Y39TAG-R, GFP-Y151TAG-F and GFP-Y151TAG-R, and GFP-Y182TAG-F and GFP-Y182TAG-R were used to construct amber mutations at the permissive sites of GFP using the QuikChange site-directed mutagenesis method (Papworth, 1996).

T7 RNA polymerase was integrated into the genome of E. coli C321.ΔA.exp using the CRISPR-Cas9 system (Jiang et al., 2015). The T7 RNA polymerase gene was first amplified by PCR from the genome of E. coli BL21(DE3) using the primer T7-C321-F/T7-C321down-R pair, and the homology arms were amplified using the genome of E. coli C321.ΔA.exp as a template with the primer C321-UP-F/C321-UP-R pair. The two resulting PCR products were used to amplify the T7 RNA polymerase expression cassettes by overlapping PCR with the primers C321-UP-F/T7-C321down-R. Genome editing experiments were performed as described by Jiang et al. (2015). Briefly, E. coli C321.ΔA.exp was first transformed with pCas and induced with arabinose (10mM final concentration) to express recombination-related proteins. After induction, the harvested cells were used to generate competent cells. Then, 100ng of pTarget-lacZ plasmid and 400ng of T7 RNA polymerase expression cassette DNA were co-electroporated into competent cells. After recovery at 30°C for 1h, the cells were spread onto LB agar containing kanamycin and spectinomycin and incubated overnight at 30°C. Transformants were identified by colony PCR with the primer C321-JD-F/C321-JD-R pair, and DNA sequencing with the primer T7 RNApoly-F/T7 RNApoly-R pair.

To cure the gene editing-related plasmids, the verified edited colony was inoculated into 5ml of LB medium containing IPTG and kanamycin and cultured at 30°C for 12h. The culture was spread onto LB plates containing kanamycin and incubated at 30°C overnight. Then, the colonies were confirmed to be cured of plasmid pTarget-lacZ by determining their sensitivity to spectinomycin. The pCas plasmid was cured by incubating the colonies without plasmid pTarget-lacZ in LB medium overnight at 37°C. Finally, E. coli C321.ΔA.expT7 with a free plasmid was verified by colony PCR.

A single colony of each E. coli strain was picked and grown in LB medium at 37°C overnight. The resulting cultures were normalized and diluted to an OD₆₀₀ of 0.05, in 50ml of fresh LB medium. The strains were incubated at 37°C, and the OD₆₀₀ values were measured every 30min for 12h.

The E. coli strains harboring two plasmids, one containing different orthogonal tRNA-synthetase pairs and the other containing GFP variant genes, were grown at 37°C in 50ml of LB supplemented with antibiotics. After the OD₆₀₀ values of the cultures reached 0.4, 1mM of ncAAs was added to the cultures and shaken for another 40min. Then, the inducer IPTG or/and arabinose with final concentrations of 0.4 and 10mM was added to the medium to induce protein expression. After induction at 30°C for 10h, 1ml of the culture was harvested by centrifugation at 8,000×g for 10min. The cell pellets were washed three times and suspended in 1ml of phosphate buffer. After normalization, the GFP fluorescence of the cells was measured using a plate reader (E_x/E_m=488/513nm). Notably, the cells used to express 5HTP containing proteins were cultured in M9 medium.

Recombinant GFP variants containing different ncAAs were produced in E. coli strains, as described above. The harvested cells were suspended in phosphate buffer and sonicated on ice. After centrifugation at 13,000×g for 10min at 4°C, the soluble fraction containing the recombinant proteins was recovered and purified using a Ni²⁺-charged chelating Sepharose Fast Flow column (GE Healthcare, Uppsala, Sweden). The purified proteins were analyzed by SDS-PAGE, and the concentrations were measured using the Bradford method with bovine serum albumin as a standard (Bradford, 1976).

All data in this study were evaluated using one-way ANOVA in SPSS version 20.0. Statistical significance was set at p<0.05.

The strain E. coli EcNR2 is the parent strain of E. coli C321.ΔA (Lajoie et al., 2013). To reveal the potential effects of UAG codon reassignment on E. coli, the transcriptomic information of E. coli C321.ΔA and E. coli EcNR2 was identified and compared using transcript sequencing. After removing the low-quality reads, the number of clean reads obtained from each sample ranged from 31.25 to 35.48 million, with genome mapped ratios from 97.50 to 99.31%. PCA analysis and correlation coefficient analysis revealed that greater variability existed between the two E. coli strains than within each biological replicate (Supplementary Figures S1, S2). The results suggested that the replicates in each group had reproducible transcriptome profiles, and UAG codon reassignment could alter the transcriptomic pattern of E. coli. The total number of expressed genes in E. coli C321.ΔA and E. coli EcNR2 was 3,722 and 3,676, respectively (Figure 1A). A total of 132 genes were uniquely expressed in E. coli C321.ΔA, whereas 86 genes were uniquely expressed in E. coli EcNR2. The threshold for identification of significantly DEGs was set as at least twofold-change and p<0.05. Compared with E. coli EcNR2, 394 genes were upregulated, and 389 genes were downregulated in E. coli C321.ΔA (Figure 1B). Thus, DEGs accounted for approximately 21.04% of the total number of expressed genes. A scatter plot was constructed to represent the overall changes in the DEGs between E. coli C321.ΔA and E. coli EcNR2 (Figure 1C). Detailed information on the upregulated and downregulated genes, including gene names, fold changes, and function categories, is shown in Supplementary Tables S1 and S2.

A total of 783 DEGs identified from the transcripts were assigned to 19 clusters of orthologous groups (COG; Figure 1D). According to the COG classification analysis, 231 genes were classified into the category “unknown function, [S],” accounting for 29.50% of the total number of DEGs. The function class containing the relatively most significantly up-expressed genes was “Cell wall/membrane/envelope biogenesis, [M],” whereas the function class containing the relatively most significantly down-expressed genes was “Cell motility, [N].”

To further assess the effects of the transcript changes caused by UAG codon reassignment on E. coli, we performed GO annotation and KEGG pathway analysis to identify the functional classification of the DEGs. As shown in Figure 1E, GO analysis mapped all of the DEGs into biological processes, cellular components, and molecular functions via biological process functional categorization. Within the “biological process” function class, the largest numbers of DEGs were related to the “cellular process” subclass, whereas within the “cellular component” class, the “cellular anatomical entity” subclass contained the most DEGs. Within the “molecular function” function class, the “binding” and “catalytic activity” subclasses had the largest number of DEGs. The metabolic pathways affected by the DEGs were revealed by KEGG pathway analysis. The 20 most abundant metabolic pathways in the DEGs are shown in Figure 1F, and four significantly (p<0.05) enriched KEGG pathways were “Flagellar assembly,” “Bacterial chemotaxis,” “Arginine and proline metabolism,” and “Two-component system.” The metabolic pathway “Two-component system” contained the largest number of DEGs when compared with other enriched KEGG pathways.

To determine the potential effects of UAG codon reassignment on the proteome of E. coli, the protein synthesis profiles of E. coli C321.ΔA and E. coli EcNR2 were identified and compared using label-free proteomic sequencing analysis. The whole cell proteins of E. coli C321.ΔA and E. coli EcNR2 were first analyzed by SDS-PAGE (Figure 2A). According to the identified proteomic information, the total number of expressed proteins in E. coli C321.ΔA and E. coli EcNR2 was 2,089 and 2,221, respectively (Supplementary Figure S3). The number of unique proteins identified in E. coli C321.ΔA and E. coli EcNR2 was 31 and 63, respectively. The threshold for identification of significantly differentially expressed proteins was set as at least twofold-change, and p<0.05. The overall significant changes in the proteome of E. coli C321.ΔA were represented by a differential volcano plot (Figure 2B). Compared with E. coli EcNR2, 256 proteins were downregulated, and 125 proteins were upregulated. Thus, the significantly differentially expressed proteins accounted for roughly 18.24% of the total expressed proteins in E. coli C321.ΔA. The locations of the differentially expressed proteins in E. coli are shown in Figure 2C, and 91.60% of the identified differentially expressed proteins, including 111 upregulated and 238 downregulated proteins, were cytoplasmic.

The significantly differentially expressed proteins were analyzed by GO enrichment and KEGG pathway analyses. According to GO categorization, the 10 most relatively abundant down-expressed groups were “intracellular pH elevation,” “pH elevation,” “putrescine catabolic process,” “glucarate catabolic process,” “glucarate metabolic process,” “D-glucarate catabolic process,” “D-glucarate metabolic process,” “glycine decarboxylation via glycine cleavage system,” “cellular response to acidic pH,” and “polyamine catabolic process,” whereas the top 10 relatively abundant up-expressed groups were “galactose transmembrane transport,” “methylgalactoside transport,” “glycoside transport,” “L-lyxose metabolic process,” “methyl accepting chemotaxis protein complex,” “transmembrane signaling receptor activity,” “porin activity,” “wide pore channel activity,” “hemotaxis,” and “taxis” (Figures 2D,E). As shown in Figure 2F, a large proportion of the upregulated proteins was enriched in seven KEGG pathways, including “Global and overview maps,” “Carbohydrate metabolism,” “Membrane transport,” “Signal transduction,” “Cell motility,” “Cellular community – prokaryotes,” and “Amino acid metabolism.” The down-expressed proteins were abundant in “Global and overview maps,” “Carbohydrate metabolism,” “Amino acid metabolism,” “Metabolism of other amino acids,” “Signal transduction,” “Cellular community – prokaryotes,” “Metabolism of cofactors and vitamins,” “Energy metabolism,” “Membrane transport,” and “Drug resistance: antimicrobial” (Figure 2G).

Without the competition of RF1, the orthogonal tRNA-synthetase pair could efficiently decode the amber UAG codon and remarkably improve the yields of ncAA-containing proteins in E. coli C321.ΔA (Lajoie et al., 2013). However, the fitness of E. coli C321.ΔA is far lower than that of its non-recoded ancestor, which limits its application in many fields, including industrial-scale protein production, metabolic engineering, and live cell imaging (Wannier et al., 2018). Using multiplex genome engineering and predictive modeling, Kuznetsov et al. (2017) recovered the fitness deficit of E. coli C321.ΔA and generated the strain E. coli C321.ΔA. exp. Escherichia coli C321.ΔA. exp showed slightly enhanced ncAA-dependent protein expression. To facilitate the incorporation of ncAAs into proteins using the pET series plasmids in E. coli C321.ΔA. exp and further improve the yields of ncAA-containing proteins, we tried to integrate the T7 RNA polymerase gene into the genome using the CRISPR-Cas9 system in this study.

The T7 RNA polymerase expression cassette was amplified from the E. coli BL21(DE3) genome and inserted into the genome of E. coli C321.ΔA. exp to replace the lac Z gene via the CRISPR-Cas9 system, as described in the “Materials and Methods” section (Figure 3A). Genomic PCR screening and sequencing verified that T7 RNA polymerase was integrated into the genome of E. coli C321.ΔA. exp (Figure 3B). The T7 RNA polymerase inserted strain was named E. coli C321.ΔA. exp T7. We tested whether insertion of the T7 RNA polymerase gene affected the growth of E. coli C321.ΔA. exp. According to Figure 3C, the growth rate of E. coli C321.ΔA. exp T7 was similar to that of E. coli C321.ΔA. exp, indicating that integration of the T7 RNA polymerase gene had limited effects on the strain growth. Next, we investigated whether the inserted T7 RNA polymerase gene could work well in E. coli C321.ΔA. exp by performing recombinant protein expression using pET derivative plasmids. As shown in Figure 3D, both E. coli C321.ΔA. exp and E. coli C321.ΔA. exp T7 could detect strong GFP fluorescence when harboring the pBAD24-GFP plasmid after induction with arabinose, but only E. coli C321.ΔA.exp T7 showed remarkable GFP fluorescence when both strains harbored the pET26b-GFP plasmid after induction with IPTG. Thus, we experimentally verified that pET derivative plasmids could not be used in E. coli C321.ΔA.exp directly because of the lack of T7 RNA polymerase, but could work well in our strain E. coli C321.ΔA.exp T7.

To test the ability of incorporation of different ncAAs at multiple UAG sites in E. coli C321.ΔA.exp T7, we used two widely reported orthogonal plasmids (pEVOL and pUltra derivatives) to incorporate six ncAAs into GFP with amber mutations. For comparison, we also performed the same experiments in E. coli BL21(DE3), E. coli DH10B, and E. coli C321.ΔA.exp. The structures of ncAAs and the main features of the plasmids used in this experiment are shown in Figures 4A,B. According to the growth curves, E. coli BL21(DE3) grew faster in the exponential phase than E. coli C321.ΔA.exp and E. coli C321.ΔA.exp T7, but the same biomass accumulated in the stationary phase. Escherichia coli DH10B showed a similar growth rate to E. coli C321.ΔA.exp and E. coli C321.ΔA.exp T7, but had a slightly lower biomass in the stationary phase (Figure 3C).

We first compared the GFP expression ability of the four strains when using pIoF as a building block. As shown in Figure 5A, when using pET26b as a vector for expression of the GFP Y151TAG gene, only strains E. coli BL21(DE3) and E. coli C321.ΔA.exp T7 had strong in-cell GFP fluorescence when supplemented with 1mM pIoF. The strains E. coli DH10B and E. coli C321.ΔA.exp did not show any visible fluorescence due to the lack of T7 RNA polymerase and the inability to use the T7 promoter in the pET26 plasmid. In E. coli strain C321.ΔA.exp T7 harboring plasmid pET26b-GFPY151TAG, a relatively high GFP fluorescence was also detected in the absence of pIoF. We speculated that the reason for this phenomenon is that UAGs in the GFP gene could be misread by endogenous tRNAs after RF1 removal in E. coli C321.ΔA.exp T7, as described by Johnson et al. (2012). When using plasmid pBAD24 as a vector for expression of the GFP Y151TAG gene, all four strains showed remarkable in-cell GFP fluorescence in the presence of 1mM pIoF. Escherichia coli DH10B showed stronger fluorescence than E. coli BL21(DE3), which may be due to the inability of E. coli DH10B to catabolize the inducer L-arabinose and could express more intact pIoF-containing GFP variants. Escherichia coli BL21(DE3) showed weaker fluorescence than E. coli C321.ΔA.exp and E. coli C321.ΔA.exp T7 may be attributed to the competition between RF1 and premature translation termination in E. coli BL21(DE3). We also determined the ability of the four strains to incorporate pIoF at the three UAG sites in GFP. As shown in Figure 5B, E. coli C321.ΔA.exp T7 harboring plasmid pET26b-GFPY39TAG/Y151TAG/Y182TAG had stronger fluorescence than the other three strains, indicating that the yields of the pIoF-containing protein could be remarkably improved with the expression of proteins with multiple UAG sites in E. coli C321.ΔA.exp T7.

Next, we compared the ability of the four strains to incorporate five other ncAAs at three UAG sites in GFP. As shown in Figure 6, regardless of the ncAAs and orthogonal tRNA-synthetase pairs used, E. coli C321.ΔA.exp T7 harboring the plasmid pET26b-GFPY39TAG/Y151TAG/Y182TAG showed the highest fluorescence among the samples. This result demonstrates that pET derivative plasmids are better than pBAD derivative plasmids in improving the yields of ncAA-containing proteins in E. coli C321.ΔA derivative strains. The purified GFP with different ncAAs expressed in E. coli C321.ΔA.exp T7 harboring plasmid pET26b-GFPY39TAG/Y151TAG/Y182TAG is shown in Supplementary Figure S3. Escherichia coli C321.ΔA.exp and E. coli C321.ΔA.exp T7 harboring pBAD24-GFPY39TAG/Y151TAG/Y182TAG also showed relatively strong in-cell fluorescence, but was lower than that of E. coli C321.ΔA.exp T7 harboring the plasmid pET26b-GFPY39TAG/Y151TAG/Y182TAG. We speculate that this may be because E. coli C321.ΔA.exp and E. coli C321.ΔA.exp T7 degrade and reduce the concentration of the inducer L-arabinose during recombinant protein expression, thus, decreasing the yield of the protein products.