A Facile Platform to Engineer Escherichia coli Tyrosyl-tRNA Synthetase Adds New Chemistries to the Eukaryotic Genetic Code, Including a Phosphotyrosine Mimic.

The Escherichia coli tyrosyl-tRNA synthetase (EcTyrRS)/tRNAEcTyr pair offers an attractive platform for genetically encoding new noncanonical amino acids (ncAA) in eukaryotes. However, challenges associated with a eukaryotic selection system, which is needed to engineer the platform, have impeded its success in the past. Recently, using a facile E. coli-based selection system, we showed that EcTyrRS could be engineered in a strain where the endogenous tyrosyl pair was substituted with an archaeal counterpart. However, significant cross-reactivity between the UAG-suppressing tRNACUA EcTyr and the bacterial glutaminyl-tRNA synthetase limited the scope of this strategy, preventing the selection of moderately active EcTyrRS mutants. Here we report an engineered tRNACUA EcTyr that overcomes this cross-reactivity. Optimized selection systems based on this tRNA enabled the efficient enrichment of both strongly and weakly active ncAA-selective EcTyrRS mutants. We also developed a wide dynamic range (WiDR) antibiotic selection to further enhance the activities of the weaker first-generation EcTyrRS mutants. We demonstrated the utility of our platform by developing several new EcTyrRS mutants that efficiently incorporated useful ncAAs in mammalian cells, including photoaffinity probes, bioconjugation handles, and a nonhydrolyzable mimic of phosphotyrosine.

Introduction

The site-specific incorporation of noncanonical amino acids (ncAAs) into proteins in living cells is a popular technology with numerous enabling applications.[1−5] The ncAA is typically encoded by a repurposed nonsense codon, and an engineered nonsense-suppressing aminoacyl-tRNA synthetase (aaRS)/tRNA pair is used to deliver it during translation. To maintain the fidelity of translation, the ncAA-selective aaRS/tRNA pair must not cross-react with its host counterparts. Typically, such an “orthogonal” pair is imported into the host cell from a different domain of life. For example, bacteria-derived pairs are typically suitable for ncAA incorporation in eukaryotes, while eukaryote- and archaea-derived pairs are used in bacteria.[1,3−5] Although multiple bacteria-derived pairs have been developed for such application in eukaryotes,[6−10] the ncAA toolbox therein has been disproportionately reliant on the unique archaea-derived pyrrolysyl pair, which is orthogonal in both prokaryotes and eukaryotes.[1−4,11] The success of this pair stems partly from the innate plasticity of the pyrrolysyl-tRNA synthetase (PylRS), but another major advantage is the ability to engineer the PylRS using a facile Escherichia coli-based selection system.[4] In contrast, other eukaryote-compatible aaRS/tRNA pairs that are derived from bacteria must be engineered using a cumbersome yeast-based selection system.[6,8] Despite its remarkable success, the pyrrolysyl pair provides access to only certain structural classes of ncAAs. Access to additional readily evolvable pairs is critical to expand the structural diversity of the genetically encoded ncAAs in eukaryotic cells. Indeed, engineering bacteria-derived pairs such as leucyl[12] and tryptophanyl[9] or novel chimeric ones developed from the unique pyrrolysyl pair[13] has enabled the introduction of structurally diverse ncAAs to the eukaryotic genetic code.

The E. coli-derived tyrosyl-tRNA synthetase (EcTyrRS)/tRNAEcTyr pair was first established nearly two decades ago to incorporate ncAAs into eukaryotes.[6,7] Engineering this pair can provide access to ncAAs that are currently unavailable for incorporation in eukaryotes, such as those that mimic the polar post-translational modifications of tyrosine.[10,14] Indeed, the archaea-derived TyrRS/tRNA pair, which has a similar active site architecture, has been engineered to encode over a hundred different ncAAs in prokaryotes[2,3] but cannot be used in eukaryotes due to cross-reactivity. Its success highlights the untapped potential of the (EcTyrRS)/tRNAEcTyr pair for building a comparable toolbox for application in eukaryotic cells. However, to date only a small set of ncAAs have been genetically encoded using this platform, which can be partially attributed to the challenges associated with the yeast-based selection system needed to engineer it.

To overcome this limitation, we recently developed novel “altered translational machinery tyrosyl” (ATMY) E. coli strains in which the endogenous EcTyrRS/tRNA pair was functionally substituted with an archaeal counterpart (Figure S1).[10,15] We further demonstrated that the “liberated” EcTyrRS/tRNAEcTyr pair can be reintroduced into ATMY strains as a nonsense suppressor and its substrate specificity can be engineered using the facile E. coli-based selection system. Although we have demonstrated that this platform could be used to genetically encode previously inaccessible ncAAs in eukaryotes, such as p-boronophenylalanine and O-sulfotyrosine,[10,14] its scope was significantly limited by the significant cross-charging of tRNACUAEcTyr by the endogenous glutaminyl-tRNA synthetase (GlnRS).[10] Here we show that this cross-reactivity compromises the ability to identify moderately active EcTyrRS mutants from naïve mutant libraries using the first-generation selection system. To overcome this limitation, we used directed evolution to develop an engineered tRNACUAEcTyr that lacked this cross-reactivity. An optimized selection system based on this tRNACUAEcTyr enabled the facile enrichment of ncAA-selective EcTyrRS mutants from naïve libraries, including those with weaker activities. We further developed a selection strategy to improve the performance of first-generation EcTyrRS mutants with weaker activities to create robust ncAA incorporation platforms in mammalian cells. This optimized selection system was used to develop several different EcTyrRS mutants: (1) one that charges p-benzoylphenylalanine (pBPA 1, Figure ), a popular photoaffinity probe[6,16] with a high efficiency in mammalian cells; (2) a highly polyspecific mutant that enables the incorporation of a wide range of structurally similar ncAAs with useful handles (e.g., bioconjugation, electrophilic groups that enable proximity-dependent cross-linking, etc.); and (3) a unique mutant that enables the incorporation of p-carboxymethylphenylalanine (pCMF 12), a nonhydrolyzable mimic of phosphotyrosine.

Figure 1

Structures of ncAAs used in this study.

Results and Discussion

Development of an Orthogonal Mutant tRNACUAEcTyr in ATMY E. coli

In our ATMY E. coli strains, the endogenous tyrosyl pair is replaced with an archaeal counterpart; consequently, these strains were suitable hosts for engineering the EcTyrRS/tRNAEcTyr pair (Figure S1).[10] The first step in developing this directed evolution platform involves establishing the EcTyrRS/tRNAEcTyr pair in the ATMY strain as an active but orthogonal nonsense suppressor. However, we found previously that E. coli GlnRS weakly recognizes the TAG-suppressing tRNACUAEcTyr mutant.[10] This cross-reactivity is driven largely by the well-established affinity the E. coli GlnRS shows for the engineered CUA anticodon.[9,10,17,18] Background activity from this unexpected cross-charging interferes with the selection of ncAA-specific EcTyrRS variants. Previously, we reduced the impact of this cross-reactivity by optimizing the relative expression levels of tRNACUAEcTyr and tRNAGln as follows. A strain (ATMY3) was developed with a single genomic copy of tRNACUAEcTyr to minimize its expression level, while tRNAGln was overexpressed from a plasmid to outcompete tRNACUAEcTyr from being charged by GlnRS.[10,18] The resulting reduction in cross-reactivity allowed us to establish a directed evolution platform for engineering EcTyrRS in ATMY3 and developing mutants that charge previously inaccessible ncAAs in eukaryotic cells.[10,14] However, we soon realized that even though our first-generation selection system could discriminate highly active EcTyrRS mutants from the background cross-activity to facilitate their enrichment, it was unable to do the same for less-efficient EcTyrRS mutants. For example, attempts to use the first-generation selection scheme to enrich EcTyrRS mutants that charge pBPA 1 were unsuccessful, even though such a mutant (pBPARS)[6] was isolated previously using a yeast-based selection system. Furthermore, the selection of an active site library that was spiked on purpose with this known pBPARS (1:1000 ratio) mutant still failed to enrich the pBPA-selective mutants; none of the 192 surviving clones that were screened postselection showed pBPA-selective activity. A careful analysis of the behavior of the established pBPARS mutant under our selection condition, which couples its activity to the expression of a chloramphenicol-acetyl transferase (CAT-98-TAG) reporter, revealed the reason behind this failure: the cells harboring pBPARS survived only up to 15 μg/mL of chloramphenicol when pBPA was supplemented in the growth medium, while the cross-reactivity of the tRNA with the endogenous GlnRS already allowed the cells to survive up to 10 μg/mL of chloramphenicol (Figure S2). Such a small difference between the background activity and the target activity is inadequate to serve as the basis of an effective selection system. It is important to note that the ability to identify moderately active first-generation aaRS mutants is critically important for genetically encoding ncAAs with challenging structural features for which it can be difficult to directly obtain a highly efficient aaRS mutant from a naïve library. In such cases, the moderately active first-generation mutants can serve as “stepping stones” to develop more efficient variants through further evolution.[19−22] To enable such engineering efforts on EcTyrRS, we sought to develop a truly orthogonal tRNACUAEcTyr mutant that would serve as the foundation for a more versatile selection system.

The acceptor stems of tRNAs have often been targeted to improve the orthogonality or activity of tRNAs with much success.[23−26] We fully randomized five base pairs in the acceptor stem of tRNACUAEcTyr to generate a library of roughly 106 mutants (Figure A). The first base pair in the acceptor stem was not randomized, given that it was an important identity element. This library was subjected to a round of positive selection in the presence of a EcTyrRS mutant that we had previously developed to charge O-methyltyrosine.[10] In this step, active tRNACUAEcTyr mutants are enriched based on their ability to express a TAG-inactivated CAT reporter. Next, these mutants are subjected to a negative selection in the absence of a cognate EcTyrRS, where cross-reactive tRNACUAEcTyr mutants are eliminated through the expression of a TAG-inactivated toxic gene (barnase). After two rounds of positive and negative selection, the surviving tRNACUAEcTyr clones were screened for their activity and orthogonality. One particular clone (tRNACUAEcTyr-h1; Figure B) was highly promising. When tested with pBPARS, this tRNA supported growth at concentrations up to 30 μg/mL chloramphenicol in the presence of pBPA but not even 5 μg/mL (lowest concentrations tested) in its absence (Figure C). In contrast, the original tRNACUAEcTyr supported growth at concentrations up to 60 μg/mL chloramphenicol both in the presence and absence of pBPA. Note that this assay did not include the overexpression of tRNAGln, which was previously necessary to suppress the cross-reactivity of the original tRNACUAEcTyr, further underscoring the superior orthogonality of tRNACUAEcTyr-h1. Interestingly, in addition to a more A-U-rich acceptor stem, tRNACUAEcTyr-h1 encodes an additional fortuitous G24A mutation in the D-stem (Figure B).

Figure 2

Development of an orthogonal tRNACUAEcTyr variant through directed evolution. (A) The sequence of the original tRNACUAEcTyr, where the highlighted segment in the acceptor stem was randomized to all possible combinations. (B) Selecting this library of mutants led to the identification of tRNACUAEcTyr-h1, which showed dramatically attenuated cross-reactivity in ATMY. The mutations relative to its precursor are shown in red. (C) These assays were performed in an ATMY5 strain that did not encode any tRNACUAEcTyr, which was instead expressed from the pRepTrip2.3 plasmid that also harbored the CAT-TAG reporter. pBPARS was used as the cognate EcTyrRS, and the activity was measured in the presence and absence of pBPA. These assays were performed without overexpressing tRNAGln, which was previously necessary to reduce the cross-reactivity of the original tRNACUAEcTyr. Consequently, tRNACUAEcTyr shows significantly higher background activity (survival at concentrations up to 60 μg/mL chloramphenicol) in the absence of pBPA that cannot be differentiated from the activity of pBPARS in the presence of pBPA (top panel). In contrast, tRNACUAEcTyr-h1 shows no activity in the absence of pBPA and survival at concentrations up to 30 μg/mL chloramphenicol in the presence of pBPA, showcasing its high orthogonality and the ability to adequately discern the weak activity of pBPARS from the background.

Next, we sought to demonstrate that a selection system employing the orthogonal tRNACUAEcTyr-h1 can enable the enrichment of pBPA-selective mutants from a naïve EcTyrRS library, which we were unable to accomplish using the first-generation selection system. A library of ∼107 EcTyrRS mutants was constructed by randomizing six residues in the EcTyrRS active site (Figure A). Additionally, the previously described beneficial mutation D265R in the anticodon binding domain,[10,27] which presumably enhances interactions with the non-native CUA anticodon of the cognate tRNA, was also included in this library. Subjecting this library to an established double-sieve selection system led to the identification of three clones, including the previously reported pBPARS-1,[6] that selectively charged pBPA with comparable efficiencies (Figure A and B). While these results confirmed that the new selection system was capable of enriching weakly active EcTyrRS mutants, the poor efficiency of the pBPARS mutants underscored the need to develop a strategy to further improve the activities of such first-generation mutants. In particular, a more efficient mammalian incorporation system for pBPA is expected to be useful, as pBPA is a popular photoaffinity probe.[16,28−33]

Figure 3

Directed evolution of highly active pBPA-selective EcTyrRS mutants. (A) The active site of EcTyrRS. The bound substrate is shown in magenta, and the key active site residues that were subjected to randomization are highlighted. Mutations in the isolated pBPARS clones are also shown in the table below. (B) Activity of the pBPARS mutants in the ATMY E. coli strain. The activity was measured through the expression of the sfGFP-151-TAG reporter. (C) The scheme for the WiDR antibiotic selection to identify pBPARS mutants with higher activities from a random mutagenesis library. (D) Mutations associated with enhanced pBPARS mutants (shown in red). (E) Activity of the new pBPARS mutants in the ATMY E. coli strain. The activity was measured through the expression of the sfGFP-151-TAG reporter. (F) Activity of the pBPARS mutants in the HEK293T cells. The activity was measured through the expression of the EGFP-39-TAG reporter. (G) Representative fluorescence images of the HEK293T cells from panel F expressing the EGFP-39-TAG reporter.

To evolve next-generation pBPARS mutants with enhanced activities, we needed a selection system that could discriminate between mutants exhibiting moderate and high activities. Although fluorescent protein reporters have been used in the past for this purpose,[21,22] such selections typically have a lower throughput. In contrast, survival-based selections using antibiotic reporters are straightforward and provide a significantly higher throughput; however, they have not been used to evolve highly active aaRS mutants from weaker precursors. To explore if an antibiotic-based survival selection was suitable for this purpose, we used two known EcTyrRS mutants that charge O-methyltyrosine (OMeYRS) either weakly or efficiently (Figure S3A).[10] We found that the CAT-based selection system, which is traditionally used for aaRS evolution, was not able to adequately discriminate between these two EcTyrRS mutants (Figure S3B). A screen for additional antibiotic resistance reporters offering a wider dynamic range (WiDR) identified a TAG-inactivated β-lactamase reporter that allowed the survival of strong OMeYRS at concentrations up to 1200 μg/mL ampicillin but only allowed that of weak OMeYRS at concentrations up to 400 μg/mL (Figure S3B). This WiDR selection window provided an opportunity to rapidly screen libraries of pBPARS mutants to identify more active variants. Subjecting a library of pBPARS-1 mutants constructed through error-prone PCR to this WiDR antibiotic selection (Figure C) led to the identification of several clones with roughly twofold higher activities (Figure D and E), wherein the 265 position reverted back to Asp (pBPARS-2.1). Another round of WiDR evolution using pBPARS-2.1 as the starting point resulted in two new clones with a further ∼2–2.5-fold increase in activity (pBPARS-3.1 and 3.2; Figure D and E). Further evaluation in mammalian cells (Figure F) revealed that the pBPARS-3.1 mutant facilitates significantly improved pBPA incorporation (>40% wild-type reporter expression) relative to the original pBPARS-1 (<15% wild-type reporter expression). These results show that our optimized selection scheme now enables the identification of weakly active EcTyrRS mutants and that the activities of such mutants can subsequently be further improved using the WiDR antibiotic selection. In addition, the development of pBPARS-3.1, which enables the incorporation of pBPA in mammalian cells and is remarkably more efficient than to the existing incorporation system, would improve the utility of this useful photoaffinity probe.

We also explored the mechanism underlying the improved activity of pBPARS-3.1. Reverting either I7F or G180S that appear in pBPARS-3.1 (Figure S4A) was found to be detrimental, suggesting that both mutations contribute to the enhanced activity (Figure S4B). We recently reported that mutants of EcTyrRS often exhibit low thermostabilities.[34] In particular, pBPARS-1 was found to be largely insoluble under ambient conditions, which likely explains its poor activity. Interestingly, a large fraction of pBPARS-3.1 was found in the soluble fraction of the E. coli cell-free extract, while nearly all of pBPARS-1 was found in the insoluble fraction (Figure S4C). A thermostability assay in the E. coli cell-free extract further showed that pBPARS-3.1 has a significantly higher thermostability (soluble at up to 50 °C) than pBPARS-1 (insoluble even at ambient temperature) (Figure S4D). These observations suggest that the mutations acquired during directed evolution stabilize pBPARS, which likely underlies its improved performance.

To further test the scope of our optimized selection system, we sought to develop EcTyrRS mutants that charge additional useful ncAAs. Some of the most popular ncAAs are those with bioorthogonal conjugation handles such as an azide. EcTyrRS has been engineered to charge p-azidophenylalanine,[6,10] but this aromatic azide is susceptible to endogenous reduction.[10,35,36] Engineering EcTyrRS to charge ncAAs with aliphatic azides will be beneficial, since these are significantly less prone to such an undesirable loss. We pursued the selection of our EcTyrRS library using two different ncAAs, 2 and 3 (Figure ), that contained alkyl azides. Intriguingly, both selections yielded an overlapping set of mutants pAAFRS-6, pAAFRS-9, and pAAFRS-11 (Figure A), indicating that these mutants are likely polyspecific, i.e, able to charge structurally similar ncAAs. Indeed, further characterization revealed that pAAFRS-9 can selectively charge both 2 and 3 (Figure B).

Figure 4

A highly polyspecific EcTyrRS mutant capable of charging ncAAs 2–10. (A) Mutations associated with the EcTyrRS variants identified from the selection. (B) Activity of pAAFRS-9 in the ATMY E. coli strain. The activity measured using sfGFP-151-TAG as the reporter. (C) Incorporation of ncAAs 2–10 in HEK293T cells. EGFP-39-TAG was used as the reporter. Isolated yields for the reporter proteins and the observed molecular weight of each are shown. (D) Representative fluorescence images of the HEK293T cells from panel expressing the EGFP-39-TAG reporter.

Polyspecific aaRS mutants are valuable, as they enable the rapid expansion of the ncAA toolbox with many new members without the need to perform individual selections for each.[10,37] To take advantage of the potential polyspecificity of pAAFRS-9, we synthesized several ncAAs that were structurally similar to its efficient substrate pAAF (Figure ). These included additional bioconjugation handles such as alkyne and alkene (4–7) and electrophilic amino acids (8–10), which have been useful for both forging proximity-dependent cross-links between interacting proteins[38,39] and the macrocyclization of genetically encoded peptides using a proximal cysteine residue.[40,41] These substrates were found to be charged efficiently by pAAFRS-9 in ATMY E. coli (Figure B), which was observed using a sfGFP-151-TAG reporter. We also tested the activity of pAAFRS-9 in mammalian cells using a EGFP-39-TAG reporter, and all but ncAA 10 were incorporated efficiently (Figure C and D). Although ncAA 10 could be used in ATMY E. coli, it showed a high toxicity toward mammalian cells, preventing its incorporation therein. The incorporation of ncAAs 2–9 was confirmed by isolating the resulting EGFP-39-TAG reporter from the HEK293T cell-free extract using immobilized metal-ion chromatography followed by mass-spectrometry. Of particular note is the observation that the aliphatic azide pAAF 3 shows no evidence of post-translational reduction (Figure S5A), providing a more robust bioconjugation handle relative to p-azidophenylalanine in eukaryotes. Also notable is the MS analysis of the reporter protein that encodes ncAA 8, which contains an electrophilic acrylamide group, that shows the presence of a major glutathione adduct (Figure S5B). Similar glutathione adducts have been observed before when the same ncAA was incorporated into reporter proteins expressed in E. coli using an engineered archaeal TyrRS/tRNA pair.[38] These enabling ncAAs are expected to significantly enhance the impact of the toolbox empowered by EcTyrRS engineering. Moreover, additional ncAAs with similar structures would likely be accepted by pAAFRS-9, which we will continue to explore.

Tyrosine phosphorylation is an important post-translational modification in eukaryotes.[42,43] In E. coli, nonhydrolyzable structural mimics of phosphotyrosine such as p-carboxymethyl-phenylalanine (pCMF 12; Figure A) can be incorporated site-specifically using an engineered archaea-derived TyrRS/tRNA pair.[44] Site-specific pCMF incorporation has been used in many studies to interrogate the structural and functional consequences of phosphorylating particular tyrosine residues, underscoring its utility.[44−50] However, this ncAA cannot currently be incorporated into proteins expressed in eukaryotes (the archaeal Tyr pair is not orthogonal in eukaryotes). The ability to do so will be valuable, providing opportunities to interrogate the consequences of the tyrosine phosphorylation of eukaryotic proteins in their native environment. Genetically encoding a polar amino acid (such as pCMF) through engineering the pyrrolysyl pair has been challenging, which is most widely used for ncAA incorporation in eukaryotes. In contrast, EcTyrRS represents a great starting point for developing a pCMF-specific aaRS given the success of its archaeal counterpart with a similar active site.[44] To explore this possibility, we subjected the aforementioned EcTyrRS library to our optimized selection system to potentially identify a pCMF-specific mutant. Screening of the surviving clones following the double-sieve selection led to the identification of a unique mutant (Figure B) that enables the expression of the sfGFP-151-TAG reporter in ATMY E. coli only in the presence of pCMF (Figure C). Key mutations in pCMFRS include D182G, which opens up room for the carboxymethyl group, and Y37H, which likely interacts with the negatively charged carboxylate. pCMFRS also selectively facilitated the efficient expression of the EGFP-39-TAG reporter in HEK293T cells in the presence of pCMF (Figure D and E). The reporter protein was isolated from mammalian cells by IMAC in a good yield (45 μg per 107 cells) and characterized by SDS-PAGE and mass-spectrometry to confirm the incorporation of pCMF (Figure F).

Figure 5

Genetically encoding pCMF in eukaryotes. (A) pCMF mimics a phosphorylated tyrosine residue. (B) Mutations associated with pCMFRS, which were identified through the selection of a EcTyrRS library. (C) In ATMY E. coli, pCMFRS facilitates the selective expression of the sfGFP-151-TAG reporter in the presence of pCMF. (D) Activity of pCMFRS in HEK293T cells, demonstrating selective reporter expression in the presence of pCMF (fluoresence in the cell-free extract). Activities were measure using the EGFP-39-TAG reporter. (E) Representative fluorescence microscopy images of HEK293T cells expressing the EGFP-39-TAG reporter. Images were taken using pCMFRS in both the presence and absence of pCMF. (F) SDS-PAGE and ESI-MS analysis of the EGFP-39-pCMF reporter isolated from HEK293T cells.

STAT (signal transducers and activators of transcription) proteins are well-studied transcription factors in mammalian cells that are activated in response to signaling by certain extracellular stimuli through the phosphorylation of a key tyrosine residue.[51,52] Upon tyrosine phosphorylation, the STAT proteins dimerize, which is mediated by the reciprocal recognition of the phosphotyrosine residue of one monomer by the SH2 domain of the other. This dimer activates the transcription of target genes by translocating to the nucleus and binding specific DNA sequences. Schultz et al. previously expressed a truncated STAT1 protein in E. coli, which was modified with pCMF at the key Y701 residue, and showed that it binds its target DNA sequence with a high affinity in vitro, corroborating the successful phosphotyrosine mimicry by pCMF.[44] The ability to now incorporate this ncAA into proteins in mammalian cells provides an opportunity to explore whether pCMF can also trigger a relevant physiological response in living cells. Darnell et al. previously reported a facile assay to monitor STAT3 activity in mammalian cells based on the expression of a luciferase reporter expressed from a STAT3-regulated promoter.[53] We adopted this assay to determine if the incorporation of pCMF at the Tyr705 residue (phosphorylation target) can activate STAT3. To confirm our ability to incorporate pCMF into STAT3, we coexpressed STAT3-705TAG in HEK293T cells with our pCMF-specific EcTyrRS/tRNACUAEcTyr pair. Western blot analysis confirmed the successful expression of the STAT3-705TAG protein in HEK293T cells only when the media was supplemented with pCMF (Figure A). Next, we expressed both STAT3-705-pCMF and the STAT3-activated luciferase reporter in HEK293T cells.[53] As controls, we also evaluated wild-type (WT) STAT3, STAT3-Y705F (inactive), and a constitutively active mutant of STAT3.[53] The expression of each construct was optimized to achieve similar expression levels in HEK293T cells (Figure C). As expected, the Y705F mutant and the constitutively active mutant showed low and high luciferase expression, respectively (Figure B). WT STAT3 showed somewhat elevated activity relative to that of Y705F, likely due to an inadvertent activation of native signaling by a media component (e.g., bovine serum). However, the STAT3-705-pCMF mutant triggered significantly elevated levels of luciferase expression, suggesting the successful phosphotyrosine mimicry of this ncAA in mammalian cells. The ability to site-specifically incorporate this nonhydrolyzable mimic of phosphotyrosine into proteins expressed in live eukaryotic cells will be a valuable tool to probe the consequences of tyrosine phosphorylation.

Figure 6

Phosphotyrosine mimicry by pCMF activates STAT3 in mammalian cells. (A) The transfection of STAT3–705-TAG in HEK293T cells in the presence of the pCMF-selective EcTyrRS/tRNACUAEcTyr pair results in the pCMF-dependent expression of STAT3. (B) A STAT3-dependent luciferase reporter was used to evaluate STAT3 activity in HEK293T cells, showing the significantly higher activity of STAT3-705-pCMF relative inactivated STAT3 (Y705F) and WT STAT3. Activities were normalized relative to the inactive STAT3 (Y705F) control. (C) The expression of each STAT3 construct was optimized to achieve similar expression levels.

In summary, we have developed an optimized E. coli-based selection that enables the facile engineering of EcTyrRS to introduce new chemistries into the genetic code of eukaryotes. Key to this selection system is an engineered tRNACUAEcTyr mutant, which we developed through directed evolution to exhibit little cross-reactivity in ATMY E. coli. We further developed a straightforward survival-based WiDR antibiotic selection that could differentiate between strongly and weakly active EcTyrRS mutants and used it to significantly improve the activity of a poorly active first-generation pBPARS mutant through directed evolution. Together, our optimized selection system provides an opportunity to unlock the untapped potential of the EcTyrRS/tRNAEcTyr pair to introduce previously inaccessible ncAAs into the genetic code of eukaryotes. We demonstrated the utility of our selection system by developing several new EcTyrRS mutants that allowed the efficient incorporation of several novel ncAAs, which will be useful for several different applications. These include (1) a highly efficient system for incorporating pBPA, which is a popular photoaffinity probe, in mammalian cells; (2) several new ncAAs to facilitate bioconjugation, including pAAF, which is incorporated in mammalian cells with remarkable efficiency and does not undergo reduction; (3) a series of electrophilic amino acids that can be used for proximal covalent cross-linking between interacting proteins as well as for macrocyclization using a proximal cysteine residue; and finally (4) a nonhydrolyzable analog of phosphotyrosine, which will be useful for probing the consequences of this important post-translational modification. Finally, our work provides general lessons that will be valuable for developing additional engineered aaRS/tRNA pairs to expand the genetic code of eukaryotes using a similar strategy.