Split aminoacyl-tRNA synthetases for proximity-induced stop codon suppression

Significance

The ability to control gene expression in living cells has many useful applications in biotechnology and medicine. In this study, we report the design and engineering of split orthogonal aminoacyl-tRNA synthetases (o-aaRSs) that can be used to tightly control gene expression in bacteria and in human cells. Furthermore, we show that these split o-aaRSs can serve as useful tools for detecting protein–protein interactions, including those involved human disease.

Abstract

Synthetic biology tools for regulating gene expression have many useful biotechnology and therapeutic applications. Most tools developed for this purpose control gene expression at the level of transcription, and relatively few methods are available for regulating gene expression at the translational level. Here, we design and engineer split orthogonal aminoacyl-tRNA synthetases (o-aaRS) as unique tools to control gene translation in bacteria and mammalian cells. Using chemically induced dimerization domains, we developed split o-aaRSs that mediate gene expression by conditionally suppressing stop codons in the presence of the small molecules rapamycin and abscisic acid. By activating o-aaRSs, these molecular switches induce stop codon suppression, and in their absence stop codon suppression is turned off. We demonstrate, in Escherichia coli and in human cells, that split o-aaRSs function as genetically encoded AND gates where stop codon suppression is controlled by two distinct molecular inputs. In addition, we show that split o-aaRSs can be used as versatile biosensors to detect therapeutically relevant protein–protein interactions, including those involved in cancer, and those that mediate severe acute respiratory syndrome-coronavirus-2 infection.

The ability to modulate gene expression in living cells is important for developing synthetic gene circuits and has many biotechnology and biomedical applications (1). Extrinsic control of gene expression is often achieved at the level of transcription, where the concentration of a protein within a cell is tuned by controlling the rate at which the encoding gene is transcribed into mRNA. Many technologies that rely on transcriptional control have been described in the literature (2, 3). These technologies modulate gene expression by, for example, controlling the ability of RNA polymerase to bind DNA, or by regulating the activity of the polymerase itself (1, 4, 5). Gene expression can also be controlled at the level of translation by controlling the rate at which mRNA is decoded on the ribosome. Toehold switches, which regulate the ability of the ribosome to bind mRNA, are an example of technology that enables translational control of gene expression (6). Compared to transcriptional control, translational control has a faster response time, i.e., the delay between induction and protein production is shorter (7–9). The faster response time of translational control is desirable for many applications (7); however, technologies that rely on transcriptional control have been more thoroughly developed (2).

One strategy that has been used to translationally control gene expression is to control stop codon suppression within a gene (9–12). Most often, this approach relies on orthogonal aminoacyl-tRNA synthetase (o-aaRS) and suppressor tRNA pairs to suppress stop codons in the mRNA when gene expression is desired. The o-aaRSs used for this purpose are those that have been engineered to recognize noncanonical amino acids (ncAAs) as substrates (13). In the absence of an ncAA, mRNA translation is disrupted by the in-frame stop codon. To induce protein synthesis, an exogenous ncAA is provided—the o-aaRS attaches the ncAA onto the suppressor tRNA, which suppresses the in-frame stop codon, allowing full translation of the target gene (9–11, 14, 15). This type of extrinsic control of gene expression has been used for various applications including biocontainment of genetically modified organisms (16–18), to develop genetically encoded digital logic gates (19, 20), and to control the expression of therapeutic proteins in living animals (7).

While o-aaRSs can be used to control gene expression through conditional stop codon suppression, this technology has significant limitations. First, the promiscuous activity of o-aaRSs can lead to mischarging of suppressor tRNAs with natural amino acids, causing significant background translation of the target gene (17, 21, 22). Although strategies have been developed to limit tRNA mischarging (17, 23), even low levels of background translation can be prohibitive for some applications, such as when stop codon suppression is used for biocontainment or to control the expression of toxic genes. To mitigate this limitation, we sought to engineer an o-aaRS whose activity could be strictly regulated, such that the o-aaRS itself is activated when stop codon suppression is desired. A second limitation of conditional stop codon suppression is that this technique is limited to the use of ncAAs as inducers of gene expression. We reasoned that the ability to control stop codon suppression using more diverse signals, such as protein–protein interactions (PPIs) or structurally distinct small molecules, would greatly expand the utility of this technology.

To engineer o-aaRSs for better control of gene expression, we turned our attention to split protein methods, which have been used to control the activity of enzymes with switch-like precision (24). In this approach, an enzyme is split into two inactive fragments and these fragments are genetically fused to chemically inducible dimerization domains. The activity of the split enzyme can then be regulated using a small-molecule ligand that induces dimerization of these domains (24, 25). This strategy has been used to control the activity of diverse enzymes including esterases (26), proteases (27), RNA polymerases (4, 5), engineered base editors (28, 29), and Cas9 (30) for a multitude of synthetic biology applications. Split aaRSs were previously developed for controlling ncAA insertion into proteins in response to reassigned sense codons (31, 32); however, this technology has not yet been used to control gene expression through stop codon suppression.

Here, we describe the design and engineering of split o-aaRSs that can be used to conditionally control stop codon suppression in bacteria and eukaryotes. These split o-aaRS systems are based on the archaeal pyrrolysyl-tRNA synthetase (PylRS) and tyrosyl-tRNA synthetase, enzymes which have been extensively engineered to recognize diverse ncAA substrates (13, 33, 34). We demonstrate that the activity of split o-aaRSs can be modulated by PPIs or small-molecule ligands that induce fragment dimerization. Furthermore, we show that split o-aaRSs can serve as unique tools to detect therapeutically relevant PPIs and to control gene expression in both Escherichia coli and human embryonic kidney (HEK293) cells.

Results

Design and Validation of a Split PylRS.

To develop a split o-aaRS based on PylRS, we opted to use the PylRS from the archaeon “Ca. Methanomethylophilus alvus.” Compared to other PylRS orthologs that are routinely used for genetic code expansion, the “Ca. M. alvus” PylRS displays significantly greater solubility and activity, making it an ideal choice for developing into a split enzyme (35–37). For our initial studies, we used a variant of PylRS containing an N166S mutation in the enzyme’s amino acid-binding pocket. We previously engineered this PylRS variant to recognize meta-iodo-l-phenylalanine (1, mIF; Fig. 1A) as a substrate (38, 39). First, using the crystal structure of the “Ca. M. alvus” PylRS as a guide (35, 40), we selected seven potential split sites on PylRS that were located at surface-exposed, unstructured regions, distal to the tRNA-binding interface (Fig. 1B). Next, PylRS was bisected into N-terminal (PylRS_N) and C-terminal (PylRS_C) fragments at each of these split sites, and the enzyme fragments were fused, via a flexible linker, to a pair of interacting coiled-coil peptides (SYNZIP17 and SYNZIP18) (41). To screen the bisected PylRS variants for activity in E. coli, the SYNZIP-fused fragments were coexpressed along with the pyrrolysine tRNA (tRNA^Pyl) and a green fluorescent reporter protein (sfGFP) harboring a TAG stop codon at position two (sfGFP-2TAG; Fig. 1C) (42). As we anticipated that the bisected enzymes would have lower activity than the full-length PylRS, we included in this system two redundant copies of the gene encoding tRNA^Pyl. We reasoned that increasing the expression level of tRNA^Pyl in this way might improve our ability to detect bisected enzyme variants with very low activity. We measured sfGFP fluorescence in cells grown in the presence and absence of 2 mM mIF. Under these conditions, sfGFP production should occur only if the bisected PylRS is able to aminoacylate tRNA^Pyl with mIF. We observed significant sfGFP production in cells expressing five of the seven bisected PylRS variants (Fig. 1D). Gratifyingly, with each of these variants, sfGFP production was only observed in the presence of mIF, demonstrating that the bisected PylRS retains aminoacylation activity and specificity for the ncAA substrate. The most active of the bisected PylRS variants were those split at position Q23 and position D137.^*

Fig. 1.

Development of a split “Ca. M. alvus” PylRS. (A) Structures of ncAAs used in this study. (B) The crystal structure of the PylRS from “Ca. Methanomethylophilus alvus” (PDB: 6ezd). The seven sites at which the enzyme was split into N-terminal and C-terminal fragments are labeled. (C) The three-plasmid system used to identify active split PylRS variants. The N-terminal (PylRS_N) and C-terminal (PylRS_C) fragments were fused to interacting peptides SYNZIP17 (SZ17) and SYNZIP18 (SZ18), respectively. (D) sfGFP expression in E. coli cells coexpressing various SYNZIP-fused PylRS_N and PylRS_C fragments in the presence (blue bars) and absence (orange bars) of the PylRS substrate mIF. Data are displayed as the mean ± SEM for three biological replicates.

The above data demonstrate that bisected PylRS retains aminoacylation activity in E. coli; however, it is possible that both halves of the enzyme are not required for this activity. This is especially true for variants that are split into unequal fragments of significantly different sizes. For example, the most active bisected PylRS was the Q23 variant which is split into a PylRS_N fragment of 23 amino acids, and a PylRS_C fragment of 252 amino acids. In this case, the C-terminal fragment contains the lion’s share of the enzyme, and it is possible that this fragment alone is capable of aminoacylating tRNA^Pyl. To determine whether both the N-terminal and C-terminal fragments are required for aminoacylation, we compared sfGFP production in cells expressing either PylRS_N or PylRS_C alone, to cells expressing both of these fragments. With each bisected PylRS variant, sfGFP production was only detected when PylRS_N and PylRS_C were coexpressed (SI Appendix, Fig. S1). These data indicate that both fragments of the enzyme are required to reconstitute the activity of bisected PylRS and suggest that these fragments are assembling in cellulo to form the catalytically competent enzyme.

Next, we asked whether the activity of bisected PylRS is dependent on the interacting SYNZIP peptides. To determine whether SYNZIP17 and SYNZIP18 were required for PylRS activity, we systematically deleted the peptides from PylRS_N and PylRS_C, and then measured sfGFP production in cells coexpressing these truncated fragments. We limited this experiment to the two most active bisected PylRS variants, namely Q23 and D137. With both of these variants, sfGFP production was highest in cells expressing SYNZIP-fused PylRS fragments and no fluorescence was detected when both SYNZIP peptides were deleted (Fig. 2A). However, significant sfGFP production was observed with the Q23 variant when only PylRS_N was fused to a SYNZIP peptide. These data suggest that the Q23 variant can self-assemble into an active enzyme and does not necessarily require fusion to the interacting SYNZIP peptides. In contrast with the D137 variant, sfGFP production was only detected when both PylRS_N and PylRS_C were fused to SYNZIP peptides, demonstrating that the interaction between these peptides is required to restore PylRS activity. To confirm the results of the sfGFP assay, we coexpressed the SYNZIP-fused PylRS fragments in E. coli that were simultaneously expressing a chloramphenicol acetyltransferase reporter gene containing an in-frame stop codon at position 112 (CmR-112TAG; Fig. 2B). We then challenged these cells to grow on media containing mIF and varying concentrations of the antibiotic chloramphenicol. Under these conditions, cells should only grow on chloramphenicol if the bisected PylRS is capable of aminoacylating tRNA^Pyl (43). Consistent with the sfGFP assay, we found that cells expressing the Q23 variant could grow on chloramphenicol with only PylRS_N fused to an SYNZIP peptide (Fig. 2C). In contrast, cells expressing the D137 variant displayed robust growth on up to 100 μg/mL Cm only when both PylRS_N and PylRS_C were fused to SYNZIP peptides (Fig. 2D and SI Appendix, Fig. S2). To determine how bisecting PylRS impacts its activity, we compared sfGFP production in cells expressing either the bisected or full-length enzyme. sfGFP production in cells expressing the PylRS variant bisected at position D137 was 14% that of cells expressing full-length PylRS as determined by in-cell fluorescence (SI Appendix, Fig. S3), demonstrating that the bisected enzyme retains significant aminoacylation activity. Collectively, these data demonstrate that the activity of bisected PylRS can be efficiently restored by fusing the two PylRS fragments to interacting polypeptides.

Fig. 2.

Split PylRS must be fused to interacting polypeptides for activity. (A) sfGFP expression in E. coli cells coexpressing PylRS_N and PylRS_C fragments with and without fusion to SYNZIP peptides. Data were collected in the presence (blue bars) and absence (orange bars) of 2 mM mIF and are displayed as the mean ± SEM for three biological replicates. (B) The three-plasmid system to detect split PylRS activity using a chloramphenicol acetyltransferase reporter. SYNZIP-fused PylRS fragments were coexpressed with a chloramphenicol acetyltransferase gene (CmR) containing an in-frame TAG codon. (C and D) Cells were challenged to grow on media containing mIF and 50 μg/mL chloramphenicol. The PylRS variant split at position Q23 does not require SYNZIP18 for growth (C), whereas, the variant split at position D137 requires both SYNZIP17 (SZ17) and SYNZIP18 (SZ18) (D).

Design and Validation of a Split Tyrosyl-tRNA Synthetase.

Given our success in developing a split o-aaRS based on PylRS, we next asked whether the widely used tyrosyl-tRNA synthetase from Methanocaldococcus jannaschii (MjTyrRS) could also be developed into a unique split reporter. Based on the crystal structure of MjTyrRS in complex with tRNA^Tyr (44), we identified five potential cut sites at which to split the enzyme into N-terminal (TyrRS_N) and C-terminal (TyrRS_C) fragments (Fig. 3A). These fragments were genetically fused to SYNZIP17 and SYNZIP18, and the abovementioned sfGFP expression assay was used to monitor the activity of the bisected enzyme (Fig. 3B). For this assay, we used an MjTyrRS variant known as AzFRS.2.t1, which was thoroughly engineered for high substrate specificity and activity in E. coli (45). We measured MjTyrRS activity using the ncAA substrate para-iodo-l-phenylalanine (2, pIF, Fig. 1A). With four of the bisected MjTyrRS variants, we observed robust sfGFP production when the cell growth media was supplemented with 2 mM pIF (SI Appendix, Fig. S4), indicating that the bisected enzymes retained aminoacylation activity. With each of these variants, sfGFP production required simultaneous coexpression of both TyrRS_N and TyrRS_C, demonstrating that the N-terminal and C-terminal fragments of MjTyrRS are both required for the aminoacylation activity (SI Appendix, Fig. S5). The most active MjTyrRS variants were those bisected at position K99 and N268. To confirm that the activity of the bisected enzymes relies on the interacting SYNZIP peptides, we measured the activity of these MjTyrRS variants with and without fusion to SYNZIP17 and SYNZIP18. For the K99 variant, neither SYNZIP17 nor SYNZIP18 was absolutely required for activity (Fig. 3C). In contrast, with the N268 variant, sfGFP production was only detected when both TyrRS_N and TyrRS_C were fused to SYNZIP peptides, demonstrating that this PPI is required to restore the activity of split MjTyRS. Cells expressing the MjTyrRS variant bisected at position N268 produced 39% as much sfGFP as cells expressing the full-length MjTyrRS, as determined by in-cell fluorescence measurements (SI Appendix, Fig. S6).

Fig. 3.

Developing a split M. jannaschii tyrosyl-tRNA synthetase (MjTyrRS). (A) The crystal structure of the MjTyrRS (PDB: 1j1u). The five sites at which the enzyme was split into N-terminal and C-terminal fragments are labeled. (B) The three-plasmid system used to identify active split MjTyrRS variants. The N-terminal (TyrRS_N) and C-terminal (TyrRS_C) fragments were fused to interacting peptides SYNZIP17 (SZ17) and SYNZIP18 (SZ18), respectively. MjTyrRS activity was monitored by measuring sfGFP fluorescence in cells expressing a reporter sfGFP gene harboring an in-frame TAG codon. (C) sfGFP expression in E. coli cells coexpressing various SYNZIP-fused TyrRS_N and TyrRS_C fragments in the presence (blue bars) and absence (orange bars) of the MjTyrRS substrate pIF. Data are displayed as the mean ± SEM for three biological replicates.

Detecting Therapeutically Relevant PPIs Using Split PylRS.

The above data demonstrate that the D137-bisected PylRS and the N268-bisected MjTyrRS variants can be used to detect the interactions between synthetic coiled-coil peptides. Next, we asked whether bisected o-aaRSs could be used as general tools to detect PPIs between important therapeutic targets. Our goal was to develop a universal tool that can be used to detect diverse PPIs in both bacteria and in human cells. Since MjTyrRS is not orthogonal in eukaryotic cells, we focused our efforts on the split PylRS. First, we asked whether split PylRS could be used to detect interactions between members of the Bcl-2 family of proteins, which are intricately involved in regulating apoptosis (46, 47). Inhibitors of interactions between various Bcl-2 family proteins have been extensively targeted for the development of anticancer therapeutics (48, 49). We focused on detecting interactions between the BH3 domain of the proapoptotic protein tBID and the antiapoptotic proteins Mcl-1 and Bcl-2. It has been shown that Mcl-1 and Bcl-2 bind strongly to tBID to inhibit apoptosis, and aberrant overexpression of Mcl-1 or Bcl-2 is observed in many cancer types (48). We cloned PylRS_N as a genetic fusion to Mcl-1 or Bcl-2 and cloned PylRS_C as a fusion to the BH3 domain of tBID or a control peptide (Dead BID) which lacks the BH3 domain that is required for binding (26). These split PylRS fusion proteins were then coexpressed, along with tRNA^Pyl and the sfGFP-2TAG reporter, and we measured sfGFP fluorescence in the presence and absence of mIF (Fig. 4A). We observed robust sfGFP production when tBID-fused PylRS_C was coexpressed with either Mcl-1- or Bcl-2-fused PylRS_N (Fig. 4 B and C). Gratifyingly, sfGFP fluorescence was only detected in the presence of mIF, suggesting that protein production is a direct result of PylRS activity. As expected, no sfGFP production was detected with Dead-BID-fused PylRS_C since Dead BID is not expected to interact with either Mcl-1 or Bcl-2. These results demonstrate that split PylRS can be used detect the interaction between tBID and Mcl-1 or Bcl-2.

Fig. 4.

Detecting therapeutically relevant PPIs using split PylRS. (A) The three-plasmid system used to detect the interaction between tBID and Bcl-2 or Mcl-1. The N-terminal PylRS fragment (PylRS_N) was fused to tBID or a negative control peptide (Dead BID). The C-terminal PylRS fragment (PylRS_C) was fused to Bcl-2 or Mcl-1. PylRS activity was monitored by measuring sfGFP fluorescence in cells expressing a reporter sfGFP gene harboring an in-frame TAG codon. (B and C) sfGFP expression in cells expressing PylRS_N-tBID and Bcl-2-PylRS_C (B) or Mcl-1-PylRS_C (C). (D) A model of the 5-HB from SARS-CoV-2 (based on PDB: 6lxt). The HR1 and HR2 helices are shown in gray and orange, respectively. The 5-HB contains an empty HR2 binding site formed by adjacent HR1 helices. (E) The three-plasmid system used to detect the interaction between the 5-HB and HR2. The 5-HB and an HR2 peptide were fused to PylRS_N and PylRS_C, respectively. PylRS activity was monitored by measuring sfGFP fluorescence in cells expressing a reporter sfGFP gene harboring an in-frame TAG codon. (F) sfGFP expression in cells coexpressing PylRS_N-5-HB and HR2-PylRS_C. All data were collected in the presence (blue bars) and absence (orange bars) of 2 mM mIF. Data are displayed as the mean ± SEM for three biological replicates.

Next, we asked whether split PylRS could be used to detect PPIs that mediate the entry of severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) into host cells. Specifically, we sought to detect the interaction between the heptad repeat 1 (HR1) and heptad repeat 2 (HR2) motifs of the spike (S) protein of SARS-CoV-2. During viral entry, the S protein inserts itself into the host’s membrane, thereby tethering the virion to the host (50, 51). Afterward, the HR1 and HR2 motifs of the S protein interact to form a six-helix bundle comprised of three HR1 helices and three HR2 helices. Within the six-helix bundle, each HR2 helix binds to a highly conserved hydrophobic groove formed by two HR1 helices (SI Appendix, Fig. S7) (50–52). This interaction brings the virus and host membranes into proximity, enabling subsequent membrane fusion. Because of its critical role in mediating membrane fusion, the HR1–HR2 interaction is the primary target for many fusion inhibitors that have been developed as potential antiviral therapeutics to combat the COVID-19 pandemic (53–55). To detect the HR1–HR2 interaction using split PylRS, we designed a five-helix bundle (5-HB) (56) comprised of two HR2 helices and three HR1 helices connected by flexible linkers (SI Appendix, Fig. S7). This 5-HB has an exposed hydrophobic groove to which an HR2 helix can bind (Fig. 4D) (55). We then genetically fused the 5-HB to PylRS_N and fused an HR2 helix to PylRS_C (Fig. 4E). When the 5-HB- and HR2-fused PylRS fragments were coexpressed, along with tRNA^Pyl and sfGFP-2TAG, we observed robust sfGFP production in the presence of mIF (Fig. 4F). In the absence of either HR2 or the 5-HB, no sfGFP fluorescence was detected, indicating that the observed activity of PylRS is a consequence of the interaction between these polypeptides. Together, these data show that split PylRS can be used as a general tool to detect therapeutically relevant PPIs.

Small-Molecule–Induced Stop Codon Suppression Using Split aaRSs.

The above data demonstrate that PPIs can be used to activate split aaRSs and induce stop codon read-through. Next, we asked whether split o-aaRSs could also be activated by chemically induced dimerization, enabling small-molecule control of gene expression through stop codon suppression. We envisioned that such a system would be a useful tool for controlling gene translation in living cells. To test this idea, we genetically fused the N- and C-terminal fragments of bisected MjTyrRS to the FKBP and FRB proteins of the rapamycin (RAP)-induced dimerization system (Fig. 5A). In this system, the receptor proteins FKBP and FRB form a stable dimer only in the presence of their ligand RAP (57, 58). We coexpressed the FKBP- and FRB-fused MjTyrRS fragments along with tRNA^Tyr, and sfGFP-2TAG, in E. coli and monitored sfGFP production in the presence and absence of both RAP and the MjTyrRS substrate pIF. In the presence of both RAP and pIF, sfGFP production was >twofold greater than that in the presence of RAP or pIF alone (Fig. 5B). Notably, the efficiency of stop codon suppression using this system was dependent on the concentration of RAP, with sfGFP production levels ranging from 1.3-fold higher than background at 10 μM RAP, to 2.3-fold higher than background at 100 μM RAP (SI Appendix, Fig. S8). These data demonstrate that the level of stop codon read-through can be controlled by RAP concentration using split MjTyrRS.

Fig. 5.

Rapamycin (RAP)-induced stop codon suppression using split MjTyrRS. (A) The N-terminal (TyrRS_N) and C-terminal (TyrRS_C) MjTyrRS fragments were fused to the RAP-binding proteins FRB and FKBP, respectively. MjTyrRS activity was monitored by measuring sfGFP fluorescence in cells expressing an sfGFP reporter gene harboring an in-frame TAG codon. (B) Robust sfGFP production required the addition of both pIF and RAP to the cell growth media. Data are displayed as the mean ± SEM for three biological replicates.

Encouraged by the results of the RAP-controlled MjTyrRS, we shifted our focus to developing a chemically controlled o-aaRS based on PylRS. With PylRS, we opted to use the abscisic acid (ABA)-induced dimerization system instead of the RAP-based system. ABA is a plant hormone that induces dimerization of the ABI and PYL proteins and, like the RAP-based system, the ABA system has been widely used to induce protein dimerization for synthetic biology applications (24, 59). We genetically fused ABI to PylRS_N and fused PYL to PylRS_C, and then measured ABA-induced stop codon suppression using the sfGFP-2TAG reporter (Fig. 6A). We observed robust sfGFP production in cells grown in media containing both ABA and the PylRS substrate mIF; however, in the absence of either ABA or mIF, sfGFP production was only slightly higher than the background signal (Fig. 6B). This observation suggests that ABA can be used to activate PylRS, thereby inducing stop codon suppression. To confirm the results of the sfGFP assay, we coexpressed the ABI- and PYL-fused PylRS fragments, along with tRNA^Pyl and the CmR-112TAG reporter gene, in E. coli (Fig. 6D). We then challenged these cells to grow on media containing chloramphenicol and mIF in the presence and absence of ABA. Under these conditions, cells were able to grow on media containing chloramphenicol, only when ABA was also present in the growth media (Fig. 6E), further demonstrating ABA-induced stop codon suppression using split PylRS. Together, these data demonstrate that split o-aaRSs can be used as unique tools for small-molecule–induced stop codon suppression to control gene expression in E. coli. These split o-aaRSs also serve as unique genetically encoded AND gates wherein two inputs (RAP and pIF or ABA and mIF) are required to achieve a desired output (stop codon read-through). Such genetically encoded digital logic gates are useful tools for engineering signaling pathways for synthetic biology (19, 60, 61).

Fig. 6.

Abscisic acid (ABA)-induced stop codon suppression using split PylRS. (A) The three-plasmid system used to detect ABA-induced suppression of TAG and TAA stop codons. The N-terminal (PylRS_N) and C-terminal (PylRS_C) PylRS fragments were fused to ABI and PYL, respectively. PylRS activity was monitored by measuring sfGFP fluorescence in cells expressing an sfGFP reporter gene harboring an in-frame TAG or TAA codon. (B and C) sfGFP fluorescence in cells expressing an sfGFP reporter harboring an in-frame TAG codon (B) or TAA codon (C). Cells were grown in the presence and absence of 2 mM mIF and 100 μM ABA. (D) The three-plasmid system used to detect ABA-induced stop codon suppression using a chloramphenicol acetyltransferase (CmR) reporter. PYL- and ABI-fused PylRS fragments were coexpressed with a CmR gene containing an in-frame TAG codon. (E) Cells were challenged to grow on media containing 2 mM mIF in the presence and absence of chloramphenicol and ABA. (F) Optimizing the linkers between PylRS fragments and PYL/ABI. All data are displayed as the mean ± SEM of three biological replicates.

Controlling TAG and TAA Stop Codon Suppression Using Split aaRSs.

An advantage of using the PylRS and tRNA^Pyl pair for genetic code expansion applications is that, unlike many o-aaRSs, PylRS does not interact with the anticodon of its cognate tRNA (34). Therefore, tRNA^Pyl can be mutated to recognize codons other than TAG without impacting its recognition by PylRS. Our previous experiments focused on using split MjTyrRS and PylRS to suppress TAG codons, and we were curious whether the split PylRS could also be used for small-molecule–induced suppression of TAA stop codons. To measure TAA suppression, the PylRS_N-ABI and PYL-PylRS_C fragments were coexpressed in E. coli along with an sfGFP reporter harboring a TAA codon at position two, and a tRNA^Pyl mutant bearing a corresponding UUA anticodon (Fig. 6A). We then measured sfGFP production in these cells in the presence and absence of ABA and the PylRS substrate mIF. In the presence of either ABA or mIF alone, sfGFP production was not significantly greater than the background signal. In contrast, we observed robust sfGFP production in cells grown in media containing both ABA and mIF (Fig. 6C), demonstrating that the split PylRS can also be used for small-molecule–induced suppression of TAA stop codons.

Controlling Gene Expression in Mammalian Cells Using Small-Molecule–Induced Stop Codon Suppression.

Given our success using ABA-induced stop codon suppression to control gene expression in E. coli, we next asked whether this system could also be used to control gene expression in mammalian cells. First, we sought to optimize the activity of PylRS in the ABA-induced system, by optimizing the linkers between PylRS fragments and the ABI/PYL proteins. We constructed a panel of seven fusion proteins in which the PylRS fragments were connected to ABI and PYL by flexible linkers of the form G(GSG)_nG, where n = 0 to 6. We then used the sfGFP-based stop codon read-through assay to quantify the activity-bisected PylRSs with these different linkers. In general, we found that longer linkers afforded greater PylRS activity, with the 20-amino acid (n = 6) linker providing an sfGFP fluorescence signal 20-fold greater than the two-amino acid (n = 0) linker. However, the increase in overall activity was accompanied by an increase in stop codon read-through in the absence of an added ncAA (Fig. 6F). This background signal might be the result of low-level mis-aminoacylation of tRNA^Pyl with a canonical amino acid by the PylRS N166S mutant and might be reduced by using a PylRS variant that is more selective than the N166S mutant. We chose the bisected PylRS with the n = 4 linker as our optimized variant, as this variant afforded a fluorescence signal 15-fold greater than the n = 0 variant, while maintaining a relatively low background signal. sfGFP production in cells expressing this optimized ABA-controlled PylRS was 31% that of cells expressing full-length PylRS as determined by in-cell fluorescence measurements (SI Appendix, Fig. S9).

Next, we sought to demonstrate ABA-induced stop codon suppression in HEK293 cells using the optimized split PylRS. To simultaneously express the PylRS fragments under a single promoter, we constructed a plasmid encoding the PylRS_N-ABI and PYL-PylRS_C gene fragments connected by a self-cleaving T2A peptide (62). This fusion gene was coexpressed with tRNA^Pyl and a secreted embryonic alkaline phosphatase reporter gene (63) containing a TAG codon at position 41 (SEAP-41TAG, Fig. 7A). SEAP-41TAG was previously used as a convenient reporter for o-aaRSs in HEK293 cells (7). For this assay, we used the wild-type PylRS as the activity of the N166S variant has not yet been demonstrated in mammalian cells. HEK293 cells were cultured in media supplemented with the PylRS substrate N^ε-boc-l-lysine (3, BocK, Fig. 1A) and varying concentrations of ABA, and stop codon suppression was monitored by quantifying SEAP activity. Consistent with ABA-induced stop codon suppression, we found that SEAP expression increased >10-fold when cells were cultured in media containing both BocK and ABA, compared to when cells were cultured in the presence of either BocK or ABA alone (P = 0.005, Fig. 7B). Furthermore, SEAP expression increased with ABA in a concentration-dependent manner. The SEAP signal was 10.5-fold greater than background at 10 μM ABA and increased to 53.3-fold greater than background at 500 μM ABA (Fig. 7B). Notably, in the absence of ABA or BocK, SEAP expression was not significantly greater than the background signal. These data demonstrate that split PylRS can be used for small-molecule control of gene expression in mammalian cells. Furthermore, these data show that split PylRS functions as a unique genetically encoded AND gate for mammalian cells, wherein two inputs (ABA and BocK) are required to generate an output signal (stop codon suppression, Fig. 7C).

Fig. 7.

Small-molecule–induced stop codon suppression in HEK293 cells. (A) The two-plasmid system used to measure ABA-induced stop codon suppression in HEK293 cells. The N-terminal (PylRS_N) and C-terminal (PylRS_C) fragments of PylRS were fused to ABI and PYL, respectively. ABI- and PYL-fused PylRS fragments were expressed from the same promoter using a self-cleaving T2A peptide. Stop codon suppression was determined by measuring SEAP activity in media of cells expressing a SEAP reporter gene harboring a TAG codon at position 41. (B) Relative SEAP activity in cells grown in the presence of BocK and varying concentrations of ABA. Data are displayed as the mean ± SEM for four biological replicates (**P = 0.005, ns = not significant, paired t test). (C) A schematic of the ABA/BocK AND gate.

Discussion

We have successfully developed split o-aaRSs for controlling stop codon suppression in bacteria and eukaryotes. These split enzymes are based on the two most widely used o-aaRSs for genetic code expansion, namely PylRS and MjTyrRS (33, 64). We demonstrated that, like with other split protein systems (24), the bisected o-aaRSs are inactive; however, the activity of these enzymes can be restored by induced proximity of the two o-aaRS fragments. Here, we showed that the proximity of the o-aaRS fragments to restore aminoacylation activity can be achieved using diverse protein–protein and protein–small-molecule interactions.

Several o-aaRS and suppressor tRNA pairs have been used to regulate gene expression in bacteria (16, 17), eukaryotes (18, 65), and whole animals (7, 66), for diverse synthetic biology applications. These o-aaRSs regulate protein production by conditionally suppressing a stop codon within a gene of interest. In the absence of an exogenous ncAA, the stop codon terminates translation, leading to production of a truncated protein. In the presence of an ncAA, the o-aaRS aminoacylates the suppressor tRNA which suppresses the in-frame stop codon. These systems rely on the assumption that in the absence of an ncAA, translation will not proceed past the stop codon; however, low-level activity of o-aaRSs with natural amino acids can lead to background translation of the target gene (17). Here, we developed split o-aaRSs whose aminoacylation activity is controlled by the small molecules ABA and RAP in a concentration-dependent manner. In the absence of these molecular switches, the o-aaRSs are inactive and, therefore, display lower background aminoacylation in the absence of an ncAA (cf. Fig. 6F). This additional level of control over stop codon suppression will likely be useful for sensitive applications where background translation is highly detrimental, such as when stop codon suppression is used for biocontainment or to control therapeutic protein expression in animals.

o-aaRSs provide a versatile means for controlling gene expression in cells. In this proof-of-concept study, we limited our experiments to only one ncAA substrate for MjTyrRS (pIF) and two ncAA substrates for PylRS (BocK and mIF); however, both of these o-aaRSs have been extensively engineered by directed evolution and rational design to recognize a multitude of distinct ncAAs as substrates (13, 33, 34, 67). PylRS can be used to control gene expression using any stop codon, and even four-base codons (34). Furthermore, MjTyrRS and PylRS are mutually orthogonal and can be used together in the same cell to simultaneously install two distinct ncAAs into proteins (38, 42, 68). We envision that split o-aaRSs based on MjTyrRS and PylRS could likewise be used together in the same cell to control the expression of multiple genes, or for multifaceted control of a single or multiple genes.

Finally, we showed that, in addition to controlling gene expression, split o-aaRSs can serve as unique tools to detect PPIs in living cells. We demonstrated this by using split PylRS to detect PPIs with therapeutic relevance. These included interactions between Bcl-2 family proteins and interactions between proteins that mediate host cell infection by SARS-CoV-2. The interacting proteins are genetically fused to the split o-aaRS fragments, and the interaction between the proteins can be detected by reporter gene expression. Expression of the reporter gene arises from stop codon suppression as a result of the proximity-induced activity of the o-aaRS. Here, we used sfGFP and chloramphenicol acetyltransferase as reporters of stop codon suppression; however, this system is compatible with virtually any gene that can be disrupted by an in-frame stop codon. Tools such as these are valuable for studying the complex biology of PPIs and for identifying drugs to pharmacologically control these interactions.

Materials and Methods

General.

All ncAAs were purchased from Chem-Impex International and used without further purification. Oligonucleotides were synthesized by the W. M. Keck Biotechnology Resource Laboratory at Yale University. Synthetic genes were purchased from Integrated DNA Technologies. DNA sequencing services were provided by the W. M. Keck Biotechnology Resource Laboratory at Yale University, Quintara Biosciences, and Plasmidsaurus. Enzymes and reagents for molecular cloning were purchased from New England Biolabs and Takara Bio. The original plasmid for expressing the “Ca. M. alvus” PylRS and tRNA^Pyl in mammalian cells was a gift from Simon Elsässer (Addgene plasmid #140011; http://n2t.net/addgene:140011; RRID: Addegene_14001) (69). All plasmids used in this study are listed in SI Appendix, Table S1. Complete plasmid sequences are provided in the SI Appendix.

sfGFP Expression Assay.

E. coli DH10B containing the indicated plasmids (SI Appendix, Table S1) were grown to saturation in 2xYT media supplemented with the appropriate antibiotics for plasmid selection. Saturated cultures (5 μL) were used to inoculate 150 μL of defined media (42) supplemented with 0.1 mM IPTG, 0.2% arabinose, 2 mM of the indicated ncAA, the indicated concentration of RAP or ABA (both in DMSO), and the appropriate antibiotics for plasmid selection. Cultures were incubated in 96-well black clear-bottom plates in a Synergy HTX microplate reader (BioTek) at 37 °C with 12 min of continuous shaking for every 15 min. Fluorescence intensity (λ_ex = 485 nm, λ_em = 535 nm) and OD₆₀₀ were measured every 15 min for 24 h. All data are reported as the mean of the fluorescence intensity divided by the OD₆₀₀ for three biological replicates, and error bars correspond to the SEM. In general, data were normalized by setting the condition that achieved maximum signal within a given experiment equal to 100%.

CmR Assay.

E. coli DH10B containing the indicated plasmids (SI Appendix, Table S1) were grown overnight in 2xYT supplemented with the appropriate antibiotics for plasmid selection. Overnight cultures were diluted with fresh 2xYT to OD₆₀₀ = 1 ± 0.1 and then further 10-fold serial dilutions were performed in 2xYT. Diluted cultures (5 μL) were spotted on LB agar plates containing the appropriate antibiotics for plasmid selection, 100 μM ABA (when indicated), 2 mM of the indicated ncAA, and the indicated concentration of chloramphenicol. Plates were incubated at 37 °C overnight before being imaged using a Bio-Rad ChemiDoc XRS+.

SEAP Expression Assay.

HEK293 cells were cultured in 24-well plates in DMEM, with Glutamine XL (Quality Biological), containing 10% FBS (Sigma Aldrich) and penicillin-streptomycin (American Type Culture Collection). Transfections were performed using FuGene 6 (Promega) and 0.5 µg total DNA, with a 6:1 FuGene:DNA ratio. After 4 h, the media was replaced with fresh media containing 2 mM BocK, ABA (10, 200, or 500 µM), or vehicle control (15% DMSO/0.2 M NaOH), and the cells were cultured at 37 °C (5% CO₂) for 2 d. A sample of the media from each well was taken for the SEAP activity assay (Phospha-Light™ SEAP Reporter Gene Assay System, Applied Biosystems), and the assay was performed according to the manufacturer’s protocol. Luminescence was measured (1 s/well) using a Synergy HTX microplate reader (BioTek) and graphed using GraphPad Prism. Data are displayed as the mean ± SEM of four biological replicates. Data were normalized by setting the condition that achieved maximum signal equal to 100%.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.