“Not‐so‐popular” orthogonal pairs in genetic code expansion

Abstract

During the past decade, genetic code expansion has been proved to be a powerful tool for protein studies and engineering. As the key part, a series of orthogonal pairs have been developed to site‐specifically incorporate hundreds of noncanonical amino acids (ncAAs) into proteins by using bacteria, yeast, mammalian cells, animals, or plants as hosts. Among them, the pair of tyrosyl‐tRNA synthetase/tRNA^Tyr from Methanococcus jannaschii and the pair of pyrrolysyl‐tRNA synthetase/tRNA^Pyl from Methanosarcina species are the most popular ones. Recently, other “not‐so‐popular” orthogonal pairs have started to attract attentions, because they can provide more choices of ncAA candidates and are necessary for simultaneous incorporation of multiple ncAAs into a single protein. Here, we summarize the development and applications of those “not‐so‐popular” orthogonal pairs, providing guidance for studying and engineering proteins.

1. INTRODUCTION

Site‐specific incorporation of noncanonical amino acids (ncAAs) into proteins by genetic code expansion has been widely used for protein studies and engineering (Chen et al., 2018; Chung et al., ²⁰²⁰; de la Torre & Chin, ²⁰²¹). For this purpose, a series of orthogonal pairs of aminoacyl‐tRNA synthetase (AARS) and tRNA have been engineered to recognize different ncAAs and designated codons, correspondingly. Among them, the pair of tyrosyl‐tRNA synthetase (TyrRS)/tRNA^Tyr from Methanococcus jannaschii and the pair of pyrrolysyl‐tRNA synthetase (PylRS)/tRNA^Pyl from Methanosarcina species are the most popular ones (Wan et al., 2014; Wang et al., ²⁰⁰¹). Although these two pairs can recognize a wide range of ncAAs, their substrates are mostly phenylalanine and lysine analogs. Furthermore, it is beneficial to simultaneously incorporate different ncAAs into a single protein. For example, p‐azido‐phenylalanine and N⁶‐[(2‐propynyloxy)carbonyl]‐lysine were genetically incorporated into a single protein for double‐labeling and Förster resonance energy transfer (FRET) studies by introducing both MjTyrRS/tRNA^Tyr and PylRS/tRNA^Pyl pairs (Wang et al., 2014; Wu et al., ²⁰¹²). Recently, our group utilized the pairs of acetyllysyl‐tRNA synthetase/tRNA^Pyl and phosphoseryl‐tRNA synthetase/tRNA^Sep to simultaneously incorporate acetyllysine (AcK) and phosphoserine (Sep) into a single protein to study crosstalk between posttranslational modifications (PTMs) (Venkat et al., 2018). Such strategy needs more choices of pairs which are mutually orthogonal to each other. Thus, it is necessary to reapproach those “not‐so‐popular” orthogonal pairs. Here, we summarize the development and applications of them, which include TrpRS/tRNA^Trp, PheRS/tRNA^Phe, LeuRS/tRNA^Leu, LysRS/tRNA^Lys, and SepRS/tRNA^Sep, as well as GlnRS/tRNA^Gln, GluRS/tRNA^Glu, ProRS/tRNA^Pro, HisRS/tRNA^His, and SerRS/tRNA^Ser (Table 1).

TABLE 1.

Orthogonal pairs and their origins, hosts, and incorporated ncAAs

OTSs	Origins	Hosts	Incorporated ncAAs (and corresponding AARS mutants)	References
TrpRS/tRNATrp	S. cerevisiae	E. coli	NapA (Y105L, E140P, T232V, I252V, F262Y), 1‐MTP (Y105L, E140A, T232C, I252V, F262C), BtA (Y105L, E140A, T232C, I252V, F262C), 6‐MTP (Y105V, E140P, T232C, I252C), 5‐HTP (T107C, P254T, C255A)	10, 11
	E. coli	E. coli (ATM), mammalian cells	5‐HTP, 5‐MTP, 5‐Brw, 5‐AzW, 5‐PrW, 5‐AmW (V144G, V146C for all above ncAAs)	12
PheRS/tRNAPhe	S. cerevisiae	E. coli	pFF (WT), pBrF (T415A), 6‐ClW (T415G), 6‐BrW (T415G), BtA (T415G), 2‐NapA (N412G, T415G, S418C, S437F)	15, 16, 17, 18, 19
	Human mitochondria	E. coli, mammalian cells	4‐AzF (T467G, A507A), 2‐NapA (T467G, A507A), 3‐CNF (T467G, A507A), L‐Dopa (Q356, T467, A507), 6‐CNW (E391D, T467G, A507G), 7CNW (F464V, T467G, A507G)	21
LeuRS/tRNALeu	E. coli	S. cerevisiae, mammalian cells, C. elegans	OmeY (M40L, L41E, Y499R, Y527A, H537G), AcA (M40V, L41M, Y499L, Y527L, H537G), nbC (M40W, L41S, Y499I, Y527A, H537G), DanA (M40A, L41N, T252A, Y499I, Y527G, H537T)	24, 25, 26, 27, 28
	M. thermo‐autotrophicum	E. coli	N/A	29
LysRS/tRNSLys	P. horikoshii	E. coli	hGln (E41I, Y268S)	32
SepRS/tRNASep	M. maripaludis	E. coli	Sep (K347E, N352D, E412S, E414I, P495R, I496R, L512I), serine phosphonate (E412P, E414F, P495M, I496W, F529S), pThr (G320A, L321Y)	36, 37, 38, 42, 43
GlnRS/tRNAGln	S. cerevisiae	E. coli	N/A	46
	E. coli	S. cerevisiae	N/A	47
GluRS/tRNAGlu	P. horikoshii	E. coli	N/A	30
ProRS/tRNAPro	P. horikoshii	E. coli	N/A	50
HisRS/tRNAHis	C. crescentus	E. coli	N/A	53
SerRS/tRNASer	M. mazei	E. coli	N/A	54

Abbreviation: N/A, not available.

2. ORTHOGONAL PAIRS BASED ON TRYPTOPHANYL‐TRNA SYNTHETASE (TRPRS)/TRNA^TRP

Tryptophan has the largest side chain among the set of 20 canonical amino acids, and the capacious tryptophan binding pocket of TrpRS provides potentials for incorporation of ncAAs with bulky side chains. Ellington and coworkers developed an orthogonal pair of TrpRS/tRNA^Trp from Saccharomyces cerevisiae in Escherichia coli cells (Hughes & Ellington, 2010). They found that the amber suppressor SctRNA^Trp _CUA could be mischarged by E. coli LysRS, as SctRNA^Trp _CUA has a high sequence identity (73%) and shares several identify determinants with EctRNA^Lys. To eliminate the misacylation by EcLysRS, the amber suppressor SctRNA^Trp _CUA was rationally designed by increasing G:C content of the anticodon stem to reduce its structural flexibility. The best tRNA mutant (SctRNA^Trp _CUA‐AS3.5) could reduce misacylation by nearly 50‐fold.

To use the pair of ScTrpRS/tRNA^Trp to incorporate ncAAs into proteins, Schultz and coworkers further engineered both SctRNA^Trp _CUA and ScTrpRS (Chatterjee et al., 2013). First, the acceptor stem of SctRNA^Trp _CUA‐AS3.5 was randomized and the tRNA mutant library was subjected to three rounds of alternating positive selections by suppressing a permissive amber codon in the gene of chloramphenicol acetyltransferase (CAT) and negative selections by suppressing two permissive amber codons in the gene of barnase which is toxic to E. coli cells. The best tRNA mutant (SctRNA^Trp _CUA‐H14) had a significantly enhanced suppression activity than the initial SctRNA^Trp _CUA‐AS3.5. Then, the tryptophan binding pocket of ScTrpRS (Tyr105, Glu140, Thr232, Ile252, Cys254, and Phe262) was randomized, and variant libraries were screened for an ncAA, 3‐(1‐naphthyl)‐alanine (NapA) by the similar double‐sieve selection scheme. Several NapA‐specific ScTrpRS variants were obtained (Table 1). They further tested the specificities of those variants against other tryptophan analogs. Interestingly, those ScTrpRS variants could recognize 1‐methyl‐tryptophan (1‐MTP), benzothienyl‐alanine (BtA), and 6‐methyl‐tryptophan (6‐MTP), correspondingly (Figure 1). Then, ScTrpRS variants were used to incorporate different tryptophan analogs into the site of a key tryptophan residue in the chromophore of enhanced cyan fluorescent protein (eCFP). Replacement of the tryptophan residue with different tryptophan analogs showed altered spectral features, demonstrating a potential use of the ScTrpRS/tRNA^Trp pair to monitor tryptophan structures and functions in proteins.

FIGURE 1.

NcAAs which have been incorporated by orthogonal pairs derived from TrpRS/tRNA^Trp

Later, Ellington and coworkers engineered the orthogonal pair of ScTrpRS/tRNA^Trp for genetic incorporation of 5‐hydroxy‐tryptophan (5HTP) by directed evolution with compartmentalized partnered replication (CPR), in which synthetic circuits that most efficiently drive Taq DNA polymerase production in cells can be preferentially amplified during the subsequent compartmentalized polymerase chain reaction (PCR) step (Ellefson et al., 2014). First, a library of ScTrpRS variants with randomized three residues around the tryptophan binding pocket (T107, P254, and C255) was screened for 5‐HTP by suppressing an amber codon in the gene of Taq DNA polymerase. After three rounds of selection, the best ScTrpRS variant (T107C, P254T, and C255A) was obtained, showing 1500‐fold improvement in 5‐HTP incorporation than WT ScTrpRS. Then, three libraries of SctRNA^Trp _CUA with randomized mutations at the anticodon stem, the acceptor stem, or loop sequences were subjected to the similar CPR selection, respectively. The best SctRNA^Trp _CUA mutant (U16G, G43U, and U58G) was obtained, showing 3‐fold enhancement than the parental SctRNA^Trp _CUA‐AS3.4.

The orthogonal pairs of ScTrpRS/tRNA^Trp mentioned above can use E. coli cells at the host but not eukaryotic cells, because S. cerevisiae belongs to the eukaryote, and its AARS/tRNA pairs usually cross‐react with the counterparts in the organism from the same domain. To expand the genetic code in eukaryotic cells, bacterial AARS/tRNA pairs are usually utilized due to their orthogonality to eukaryotic counterparts. However, bacterial AARS/tRNA pairs commonly cannot be evolved by using the facile E. coli‐based selection system for the same reason. To overcome this limit, Chatterjee and coworkers generated an altered translational machinery (ATM) E. coli strain in which evolved orthogonal pairs can be used in both bacteria and eukaryotes (Italia et al., 2017). First, they removed the genes of EcTrpRS and EctRNA^Trp from the E. coli genome and introduced the pair TrpRS/tRNA^Trp from S. cerevisiae to decode normal tryptophan codons in the E. coli genome. Then, the pair of EcTrpRS and EctRNA^Trp _UCA opal suppressor was put back into the ATM E. coli strain for further engineering, as it was known that the amber suppressor EctRNA^Trp _CUA could be mischarged by E. coli GlnRS (Jahn et al., 1991). Next, the tryptophan binding pocket of EcTrpRS (Phe7, Ser8, Val144, Pro145, and Val146) was randomized, and the EcTrpRS variant library was screened for 5‐HTP by the double‐sieve selection scheme mentioned above but with opal codons instead of amber codons in selection genes. The best EcTrpRS variant (V144G and V146C) was obtained. They further tested the specificities of this EcTrpRS variant for other 5‐substituted tryptophan analogs and found that it could also genetically incorporate 5‐methoxy‐tryptophan (5‐MTP), 5‐bromo‐tryptophan (5‐BrW), 5‐azido‐tryptophan (5‐AzW), 5‐propargyloxy‐tryptophan (5‐PrW), and 5‐amino‐tryptophan (5‐AmW) in both E. coli and mammalian cells (Figure 1). Among those tryptophan analogs, 5‐PrW and 5‐AzW introduce azido and alkyne groups respectively, which can be used to label proteins by click chemistry.

3. ORTHOGONAL PAIRS BASED ON PHENYLALANYL‐TRNA SYNTHETASE (PHERS)/TRNA^PHE

It has been known for a long time that certain AARSs can incorporate analogs of their cognate amino acids into proteins at corresponding codons uniformly through the proteome, such as the encoding of p‐fluoro‐phenylalanine (pFF) at Phe codons by PheRS in Phe‐auxotrophic strains (Richmond, 1962). Furter utilized this fact to site‐specifically incorporate pFF into proteins at introduced amber codons in E. coli cells (Furter, 1998). Site‐specifically fluorinated proteins are ideal for (Kwon & Lim, 2015) F‐nuclear magnetic resonance (NMR) spectroscopy analyses to monitor protein conformational changes and detect protein‐protein interactions. The incorporation of pFF can also serve as a vehicle to deliver isotopically labeled phenylalanine analogues to a position of choice for Fourier‐transform infrared (FTIR) spectroscopy experiments. To avoid the incorporation of pFF at Phe codons by the native E. coli PheRS, a pFF‐resistant E. coli strain was used. It harbors a PheRS(Ser294) variant which cannot recognize pFF (Fangman & Neidhardt, 1964). Then a plasmid with the pair of yeast PheRS and tRNA^Phe _CUA (orthogonal to E. coli counterparts) was transformed into this pFF‐resistant E. coli strain. A large excess of pFF was supplemented in the growth medium to compete with Phe for the yeast PheRS. Although the system successfully incorporated pFF at the introduced amber codon, the fidelity was not very high with pFF (~65%), Phe (~20%), and Lys (~15%).

In order to increase the fidelity, Tirrell and coworkers engineered both the yeast PheRS and tRNA^Phe _CUA for site‐specific incorporation of p‐bromo‐phenylalanine (pBrF) into proteins in E. coli cells (Kwon et al., 2006). Introduction of aryl halides such as pBrF into proteins allows site‐specific modifications via palladium‐catalyzed cross‐coupling reactions with terminal alkene or alkyne reaction partners. First, a UG base pair 30U::40G was introduced to the yeast tRNA^Phe _CUA amber suppressor to reduce the mischarging by EcLysRS. Then, a point mutation (T415A) was introduced to the yeast PheRS to exclude Trp misincorporation. Besides Phe analogs, Tirrell and coworkers utilized an engineered yeast PheRS(T415G)/tRNA^Phe _CUA pair to site‐specifically incorporate several Trp analogs, including 6‐chlorotryptophan (6‐ClW), 6‐bromotryptophan (6‐BrW), and BtA (Figure 2) (Kwon & Tirrell, 2007). However, the pair of PheRS(T415A)/tRNA^Phe still has Phe misincorporation, while the pair of PheRS(T415G)/tRNA^Phe keeps Trp misincorporation. To further increase the specificity of ncAA incorporation against canonical amino acids, Kown and coworkers created a library of yeast PheRS variants with a randomized Phe binding pocket (N412, S418, T415, and S437). Then a high‐throughput screening method based on GFP fluorescence reduction was performed for 2‐naphthylalanine (2‐NapA). Selected variants allow incorporation of either 2‐NapA or canonical amino acids at multiple nonpermissive sites of GFP, where amino acids other than Phe lead to reduction in the fluorescence of cells. The best PheRS variant (N412G, T415G, S418C, and S437F) realized high‐fidelity (98%) incorporation of 2‐NapA into a protein at an introduced amber codon (Kwon & Lim, 2015). As 2‐NapA has a larger hydrophobic side chain than any canonical amino acids, it has been used to tune the active site of proteins to modulate the binding affinity to inhibitors (Zheng & Kwon, 2013).

FIGURE 2.

NcAAs which have been incorporated by orthogonal pairs derived from PheRS/tRNA^Phe

Recently, Lin and coworkers designed a chimeric PheRS/PylRS system which has broad orthogonality in both bacteria and eukaryotes (Ding et al., 2020). First, they transplanted the key orthogonal components from the PylRS system to human mitochondrial PheRS/tRNA^Phe. Briefly, the tRNA binding domain of PylRS was added to the N‐terminal of the PheRS catalytic domain, while the PylRS binding part of tRNA^Pyl was introduced to tRNA^Phe. This chimeric PheRS/tRNA^Phe _CUA pair showed good orthogonality in both E. coli and mammalian cells (no amber suppression was detected in the presence of chPheT or chPheRS alone). Then, the chimeric PheRS (chPheRS) was evolved for different ncAAs, including both Phe analogs and Trp analogs. For Phe analogs, the chPheRS‐1 variant (T467G and A507A) could incorporate 4‐Azido‐Phe (4‐AzF), 2‐NapA and 3‐Cyano‐Phe (3‐CNF), while the chPheRS‐2 variant (Q356, T467, and A507) incorporated L‐3,4‐Dihydroxy‐Phe (L‐Dopa). For Trp analogs, the chPheRS‐3 variant (E391D, T467G, and A507G) preferred the N1 and C6 substituted Trp analogues such as 6‐Cyano‐Trp (6‐CNW), while the chPheRS‐4 variant (F464V, T467G, and A507G) exhibited higher activity toward the C7 substitute Trp analogues such as 7‐Cyano‐Trp (7CNW) (Figure 2). All these systems were able to genetically incorporate those Trp analogs into proteins by using both E. coli and mammalian cells as hosts. In the same study, such chimeric approach was also applied to HisRS and SerRS systems and showed good orthogonality toward both E. coli and mammalian cells (no amber suppression was detected in the presence of chHisRS or chSerRS alone), demonstrating great potentials of developing other E. coli–mammalian shuttle systems for genetic code expansion.

4. ORTHOGONAL PAIRS BASED ON LEUCYL‐TRNA SYNTHETASE (LEURS)/TRNA^LEU

Besides Trp, Phe, and Tyr, Leu is also a bulky hydrophobic amino acid. The binding pocket of LeuRS usually has a large cavity formed by side chains without backbone elements, (Cusack et al., 2000; Fukunaga & Yokoyama, ²⁰⁰⁵) so mutating the side chains around the binding pocket is possible to generate different shapes of cavities for accommodate a wide diversity of ncAAs. To realize this idea, Schultz and coworkers started with E. coli LeuRS/tRNA^Leu _CUA (Wu et al., 2004). This pair was first shown to be orthogonal in S. cerevisiae. Then the amino acids around the Leu binding pocket (M40, L41, Y499, Y527, and H537) were randomized and selected for three ncAAs [O‐methyl tyrosine (OMeY), α‐aminocaprylic acid (AcA), and o‐nitrobenzyl cysteine (nbC)] with distinct electronic and steric properties, separately. The selection was based on the suppression of amber codons at positions 44 and 110 in the gal4 gene which drives the expression of genomic HIS3 and URA3 reporter genes. Finally, three distinct LeuRS variants were obtained, and each one worked with a different ncAA, respectively. The one for OmeY has mutations of M40L, L41E, Y499R, Y527A, and H537G, the one for AcA harbors mutations of M40V, L41M, Y499L, Y527L, and H537G, and the one for nbC brings mutations of M40W, L41S, Y499I, Y527A, and H537G. Among these three ncAAs, nbC is a photocaged amino acid which allows photo‐regulation of proteins with key cysteine residues such as the human proapoptotic protein caspase 3 used in this work (Wu et al., 2004).

In another study, Schultz and coworkers evolved the pair of E. coli LeuRS/tRNA^Leu _CUA to genetically encode a fluorescent amino acid, 2‐amino‐3‐(5‐(dimethylamino)naphthalene‐1‐sulfonamide) propanoic acid (DanA) in S. cerevisiae (Summerer et al., 2006). DanA has a small size, relatively large Stokes shift, and high sensitivity to be an ideal fluorophore to label proteins without affecting protein structures, which is usually a concern for using fluorescent proteins (commonly >20 kDa). DanA also has a distinct electronic and steric property from those three ncAAs (OmeY, AcA, and nbC) incorporated by LeuRS variants mentioned above (Figure 3a), supporting the idea that mutating the Leu binding pocket of LeuRS is expected to accommodate various ncAAs. The LeuRS evolved for DanA has mutations of M40A, L41N, Y499I, Y527G, and H537T. To further increase the fidelity of DanA incorporation, a T252A mutation was introduced to the LeuRS variant. This mutations site is at the edit domain of LeuRS, and the T252A mutation results in efficient hydrolysis of aminoacylated tRNA^Leu (Mursinna & Martinis, 2002). Then, genetically encoded DanA was used as a probe to monitor protein unfolding. Besides S. cerevisiae, the pair of EcLeuRS/tRNA^Leu _CUA for DanA has been transplanted to mammalian cells, neurons, and Caenorhabditis elegans, providing practical applications in cell biology and neurobiology studies (Parrish et al., 2012; Wang et al., ²⁰⁰⁷).

FIGURE 3.

NcAAs which have been incorporated by orthogonal pairs derived from (a) LeuRS/tRNA^Leu; (b) LysRS/tRNA^Lys; and (c) SepRS/tRNA^Sep

On the other end, a pair of archaeal LeuRS and tRNA^Leu _CUA was developed to be orthogonal in E. coli cells (Anderson & Schultz, 2003). First, the anticodons of leucyl tRNAs from different archaeal strains (Archaeoglobus fulgidus, Halobacterium sp. NRC‐1, M. jannaschii, Pyrococcus furiosus, and Pyrococcus horikoshii) were mutated to CUA for amber suppression. Then, different combinations of tRNA^Leu _CUA amber suppressors and LeuRSs from different archaeal strains were tested for their efficiencies of amber suppression in E. coli cells. The best one was the pair of Methanobacterium thermoautotrophicum LeuRS and Halobacterium sp. NRC‐1 tRNA^Leu _CUA. Then a G37A mutation was introduced to tRNA^Leu _CUA for better amber suppression, and the acceptor stem of tRNA^Leu _CUA was optimized (a discriminator base A73 and all Watson‐Crick base pairs in the stem) for better orthogonality in E. coli cells. To expand the list of codons used for genetic code expansion, Halobacterium sp. NRC‐1 tRNA^Leu _CUA were further mutated and optimized for opal (UGA) and four‐based codon (AGGA) suppression. However, no ncAAs were incorporated into proteins by using this pair, possibly because of the relatively low aminoacylation activity (Santoro et al., 2003).

5. ORTHOGONAL PAIRS BASED ON LYSYL‐TRNA SYNTHETASE (LYSRS)/TRNA^LYS

Besides three‐base suppressor tRNAs, those with four‐base anticodons have been proposed to provide more codon choices for genetic code expansion. Schultz and coworkers selected LysRS from P. horikoshii as a candidate for a new orthogonal pair toward quadruplet codons because PhLysRS is flexible for substitutions in the anticodon loop of its cognate tRNA^Lys (Anderson et al., 2004). To solve the problem that PhLysRS is toxic to E. coli cells, serial culture of the E. coli mutator strain XL1‐red with a plasmid containing PhLysRS was performed. The resulting nontoxic variant PhLysRS∆ lost its residues after S357, which contains the anticodon recognition site (Terada et al., 2002). Then they designed a cognate suppressor tRNA for PhLysRS∆. The conserved recognition sites among all the available archaeal tRNA^Lys for the archaeal LysRS were retained, while nonessential mismatched base pairs deleterious to suppression efficiency were removed. The four‐base anticodon UCCU was added to such tRNA^Lys for decoding the AGGA quadruplet codon, but it is not orthogonal in E. coli. To obtain orthogonal tRNA^Lys _UCCU mutants, a library was constructed in which the last 4 bp of the acceptor stem were randomized, as this region is essential for the recognition of most AARSs and cognate tRNAs. The library was screened based on ampicillin resistance by suppression the AGGA codon for the position A184 of β‐lactamase. The best tRNA^Lys _UCCU mutant contains a mutated acceptor stem (G1U::C72A, C71U, and C4U::G69A) and a A37C substitution in the anticodon loop. Next, PhLysRS∆ was evolved for homoglutamine (hGln) (Figure 3b), which is structurally close to lysine. Two residues (E41 and Y268) recognizing the ε‐amino group of lysine were randomized, and the variant library was screened by a GFP‐based reporting system (Santoro et al., 2002). The PhLysRS∆ variant specific for hGln contains two mutations of E41I and Y268S. To prove the concept of simultaneous incorporation of two ncAAs into one protein, this evolved orthogonal pair of PhLysRS∆/tRNA^Lys _UCCU was used together with another orthogonal pair of M. jannaschii TyrRS/tRNA^Tyr _CUA to decode an AGGA four‐base codon and a TAG three‐base codon as hGln and OmeY in myoglobin, respectively. This work also opens an avenue of using other quadruplet codons such as CCCU and CUAG as signals for ncAA incorporation in E. coli cells.

6. ORTHOGONAL PAIRS BASED ON PHOSPHOSERYL‐TRNA SYNTHETASE (SEPRS)/TRNA^SEP

Phosphorylation is one of the most abundant PTMs in nature. Because of its importance, many groups have developed orthogonal translation systems (OTSs) for genetic incorporation of phosphoamino acids. Site‐specific incorporation of phosphotyrosine utilizes orthogonal pairs derived from MjTyrRS/tRNA^Tyr _CUA, (Fan et al., 2016; Luo et al., ²⁰¹⁷) while Sep and phosphothreonine (pThr) has been installed into proteins by orthogonal pairs derived from the pair M. maripaludis SepRS/MjtRNA^Cys _CUA (Park et al., 2011; Rogerson et al., ²⁰¹⁵; Zhang et al., ²⁰¹⁷).

It was known that certain methanogenic archaea used an unusual AARS, SepRS to catalyzes the formation of Sep‐tRNA^Cys for cysteine biosynthesis (Sauerwald et al., 2005). Söll and coworkers utilized this pathway to develop the first OTS of Sep incorporation in E. coli (Park et al., 2011). First, MjtRNA^Cys was mutated to an amber suppressor tRNA^Cys _CUA (named tRNA^Sep). An additional C20U mutation was introduced to tRNA^Sep to improve aminoacylation by SepRS (Hohn et al., 2006). The endogenous Sep phosphatase gene serB was deleted from the E. coli genome to stabilize Sep in cells. However, Sep‐tRNA^Sep cannot be delivered to ribosome efficiently by the elongation factor‐Tu (EF‐Tu), because EF‐Tu binds poorly to tRNAs carrying negatively charged amino acids (Sanderson & Uhlenbeck, 2007). Thus, the amino acid binding pocket of EF‐Tu (H67, D216, E217, F219, T229, and N274) was randomized, and the library of EF‐Tu variants was selected for Sep‐tRNA^Sep recognition based on suppression of an amber codon in the CAT gene. The resulting EF‐Tu variant (EF‐Sep) harbors mutations of H67R, D216N, E217G, F219Y, T229S, and N274W. Then, the OTS containing SepRS, tRNA^Sep, and EF‐Sep successfully incorporated Sep into a human protein MEK1, demonstrating a biological application in kinase signaling studies. Later, Park and coworkers further optimized this Sep‐OTS (Lee et al., 2013). SepRS binds to the anticodon region of tRNA^Cys, so the introduced CUA anticodon for amber codon suppression decreased the aminoacylation activity. To improve aminoacylation, the anticodon‐binding region of SepRS (E412, E414, P495, and I496) was randomized and the library of SepRS variant was selected based on suppression of an amber codon in the CAT gene. Additional DNA shuffling was conducted. The resulting SepRS variant (SepRS9) has four mutations in the anticodon‐binding region (E412S, E414I, P495R, and I496R) and three additional mutations in other regions (K347E, N352D, and L512I). Then, they identified key residues of EF‐Sep for Sep‐tRNA^Sep binding by alanine scanning and randomized N216, of which the substitution with Ala increased the binding with Sep‐tRNA^Sep. The best EF‐Sep variant (EF‐Sep21) has the mutation of N216V. Together with SepRS9 and tRNA^Sep, the optimized Sep‐OTS (SepRS9, tRNA^Sep, and EF‐Sep21) increased Sep‐incorporation nearly 10‐fold of the original Sep‐OTS. Then, Sep was genetically incorporated into H3 protein to investigate histone acetyltransferase activity of Gcn5p and the SAGA complex. Later, Rinehart and coworkers optimized Sep‐OTSs in recoded E. coli strains by increasing the copy number of tRNA^Sep from one to five in the plasmid, and the purity of Sep‐incorporation approached ~90% (Pirman et al., 2015).

Phosphorylated residues usually undergo hydrolysis by specific or non‐specific phosphatases inside cells, so it is useful to incorporate nonhydrolyzable analogs of phosphorylated amino acids into proteins to isolate and study bound phosphatases. For this purpose, Chin and coworkers genetically incorporate a nonhydrolyzable analog of Sep, serine phosphonate (replacing the O‐P bond in Sep with a C‐P bond) (Rogerson et al., 2015). First, they further engineered tRNA^sep with randomized sequences around the anticodon (G29‐U33 and G37‐C41) and SepRS with randomized anticodon‐binding region (E412, E414, K471, P495, I496, and F529). The best pair of SepRS(2)/tRNA^Sep(B4) had an additional 20‐fold than pervious Sep‐OTSs. Then, the E. coli strain was genetically modified for the incorporation of the nonhydrolyzable analog serine phosphonate. Serine phosphonate has a very similar structure to Sep, and SepRS cannot discriminate it from Sep, so the intracellular concentration of Sep needs to be decreased. To do so, SerB (the phosphatase to hydrolyze Sep) was overexpressed and SerC (the aminotransferase to produce Sep) was inactivated.

As Sep and pThr have very similar structures (Figure 3C), it is feasible to start with Sep‐OTS for pThr incorporation. However, Sep is naturally existing in cells for serine biosynthesis, while pThr is not involved in any metabolic pathways in E. coli cells. Furthermore, negatively charged compounds like pThr cannot go through the cell membrane easily, so it could not be sufficient for pThr incorporation by simply supplementing media with pThr. In Salmonella, a threonine kinase (PduX) phosphorylates free threonine to form pThr for coenzyme B₁₂ biosynthesis (Fan & Bobik, 2008). Chin and coworkers introduced PduX into E. coli cells to produce a high level of intracellular pThr (Zhang et al., 2017). Another challenge for pThr incorporation is also derived from the structural similarity between pThr and Sep. It is difficult to evolve SepRS to recognize pThr but not Sep. To solve this problem, Chin and coworkers randomized a group of amino acids (M317, N318, L319, G320, and L321) in the Sep binding pocket of SepRS which were modeled to interact with the additional methyl‐group of pThr. The SepRS variant library was screened by parallel positive selections in cells with PduX or without PduX combined with deep sequencing and statistical analysis to that compare sequencing read counts for each synthetase clone in the presence and absence of pThr. The resulting pThrRS variant has the mutations G320A and L321Y. Together with an optimized tRNA^Sep and EF‐Sep, this pThr‐OTS successfully incorporated pThr into ubiquitin and a protein kinase Cdk2 for phosphoprotein structure determination and synthetic protein kinase activation.

7. OTHER ORTHOGONAL PAIRS

Beside the five orthogonal pairs discussed above which have successfully incorporated ncAAs into proteins, several other orthogonal pairs have also been developed but without ncAA incorporation so far, providing bases for further engineering for ncAAs with distinct structures.

7.1. GlnRS/tRNA^Gln

Motivated by the fact that E. coli GlnRS does not charge S. cerevisiae tRNA^Gln, (Whelihan & Schimmel, 1997) Schultz and coworkers developed an orthogonal pair based on ScGlnRS/tRNA^Gln in E. coli cells (Liu & Schultz, 1999). First, the anticodon of SctRNA^Gln was mutated from CUG to CUA for amber suppression. An additional mutation U38A was introduced to SctRNA^Gln _CUA to increase suppression efficiency. Both suppressor tRNAs cannot be acylated by any E. coli AARSs. In addition, ScGlnRS could aminoacylate SctRNA^Gln _CUA and SctRNA^Gln _CUA (A38) but none of E. coli tRNAs, making them orthogonal pairs in E. coli cells. From the other end, RajBhandary and coworkers developed an orthogonal pair of E. coli GlnRS and a mutant human initiator tRNA_i for S. cerevisiae (Kowal et al., 2001). The mutant tRNA_i has U35A36 mutations for amber suppression and the U50G51:C63A64 mutations in the TψC stem to allows it to participate in elongation. Such mutant tRNA_i cannot be charged by any yeast AARSs but only by E. coli GlnRS, showing good orthogonality in S. cerevisiae cells.

7.2. GluRS/tRNA^Glu

Archaeal GluRS recognizes both tRNA^Glu and tRNA^Gln and charges them with Glu to form Glu‐tRNA^Glu and Glu‐tRNA^Gln, respectively. Then Glu‐tRNA^Gln is converted to Gln‐tRNA^Gln by transamidation (Tumbula et al., 2000). GluRS is able to recognize four tRNA anticodons (UUC, CUC, UUG, CUG) for both tRNA^Glu and tRNA^Gln, so it is expected to tolerate mutations in tRNA anticodon for amber suppression. In the meantime, archaeal tRNA^Glu cannot be aminoacylated by E. coli AARSs (Kwok & Wong, 1980). Based on these facts, Schultz and coworkers developed an orthogonal pair of GluRS/tRNA^Glu in E. coli cells (Santoro et al., 2003). All available sequences of archaeal tRNA^Glu were aligned, and the consensus sequences were retained. The anticodon was mutated to CUA for amber codon suppression, and the nucleotide to the 3′ side of the anticodon was replaced by A to improve suppression efficiency in E. coli. The resulting tRNA^Glu _CUA was shown to be orthogonal to E. coli AARSs. However, none of the GluRSs from selected archaea species could aminoacylate such tRNA^Glu _CUA efficiently. The consensus sequence of archaeal tRNA^Glu has a non‐canonical G10‐U28 base which could lower the aminoacylation activity, so a library of tRNA^Glu _CUA variants in positions 10 and 28 was created and screened. All the active clones have the canonical G10‐C28 base pair, which has a 5‐fold increase in amber suppression. Then GluRSs from A. fulgidus, Methanosarcina mazei, M. thermoautotrophicum, and P. horikoshii were evaluated for the orthogonality with tRNA^Glu _CUA (GC) in E. coli, individually. The best GluRS was from P. horikoshii. The pair of PhGluRS and the consensus tRNA^Glu _CUA (GC) forms an orthogonal pair for E. coli cells.

7.3. ProRS/tRNA^Pro

Proline is unique among the set of 20 canonical amino acids, because it is the only one with a secondary amine. N‐modified amino acids could endow proteins with unique properties, thus OTSs based on ProRS will be significant for protein engineering. For this purpose, Schultz and coworkers tested different combinations of ProRS and tRNA^Pro _CUA from several archaeal strains (A. fulgidus, P. horikoshii, M. mazei, Sulfolobus solfataricus, M. jannaschii, Halobacterium sp. (NRC‐1), and Thermus Thermophilus) for their amber suppression efficiencies (Chatterjee et al., 2012). The best pair appeared to be the pair of P. horikoshii ProRS and A. fulgidus tRNA^Pro _CUA. Since the anticodon is a major identify element for the tRNA^Pro recognition, the anticodon‐binding region of PhProRS was evolved. Six residues around the anticodon‐binding pocket (Y335, E338, R345, E347, D352, and K368) were randomized, and the PhProRS variant library was selected based on suppression of amber codons in both genes of GFP and CAT. The best one has mutations of Y335N, E338C, R345I, E347T, D352E, and K368S. Then, the anticodon of AftRNA^Pro was mutated to AGGG and CUAG, respectively, and the same PhProRS variant library was selected toward CCCT and CTAG suppression, correspondingly. The best ProRS variant for tRNA^Pro _AGGG only has two mutations of D352E and K368S, because tRNA_AGGG ^Pro is similar to the native tRNA^Pro _NGG. The best ProRS variant for tRNA^Pro _CUAG harbors mutations of Y335W, E338D, R345T, E347W, D352S, and K368S. Importantly, due to the strong recognition of ProRS with the anticodon of tRNA^Pro, those three pairs of ProRS variants and tRNA^Pro with different anticodons are mutually orthogonal to each other, providing possibility of simultaneous incorporation of different N‐modified ncAAs.

7.4. HisRS/tRNA^His

Different from other tRNAs, tRNA^His has an additional guanylate (G⁻¹) at its 5′ terminus, which is a major recognition element for HisRS (Francklyn et al., 1992; Himeno et al., ¹⁹⁸⁹). There are few exceptions. Söll and coworkers showed that tRNA^His from Caulobacter crescentus lacks such essential G⁻¹, and C. crescentus HisRS aminoacylates tRNA^His without G⁻¹ (Yuan et al., 2011). Instead, C. crescentus HisRS recognize the anticodon and the acceptor stem of ∆G⁻¹ tRNA^His. Since the conventional HisRS does not recognize the anticodon of tRNA^His, the pair of ∆G⁻¹ type HisRS and tRNA^His could be a potential orthogonal pair to the conventional pair of HisRS and tRNA^His. In this case, the anticodon of conventional tRNA^His can be mutated to CUA for amber suppression, while the conventional HisRS could be engineered to recognize His analogs and incorporate them into proteins at amber codons. On the other side, the host HisRS/tRNA^His can be replaced by ∆G⁻¹ type HisRS and tRNA^His to decode normal histidine codons in the genome.

7.5. SerRS/tRNA^Ser

Most of the ncAAs which have been genetically incorporated into proteins have relatively large side chains. To develop an orthogonal pair for ncAAs with small side chains, Schultz and coworkers recently started with SerRS which does not rely on anticodon recognition (Zambaldo et al., 2020). First, combinations of five different archaeal tRNA^Ser _CUA (A. fulgidus [CGA and GCU], M. jannaschii, M. mazei, and P. horikoshii) and six different archaeal SerRSs (A. fulgidus, P. horikoshii, M. mazei, S. solfataricus, M. jannaschii, Halobacterium sp. [NRC‐1], and P. horikoshii) were tested for suppression efficiency of an amber codon in the CAT gene. The best combination is MmSerRS and AftRNA^Ser _CUA. Then the AftRNA^Ser _CUA were further engineered for higher suppression efficiency by randomizing 11 nucleotides in the acceptor stem. The best variant has a substantially changed acceptor stem. An A to G mutation next the 3′ side of anticodon showed additional increase in amber suppression. The crystal structure of MmSerRS has also been solved, providing a solid base for incorporation of ncAAs with small side chains (Zambaldo et al., 2020).

8. PERSPECTIVES

Not all the existing orthogonal pairs have been used for ncAA incorporation, mostly because those AARSs cannot sufficiently aminoacylate corresponding suppressor tRNAs (Santoro et al., 2003). The amount of ncAA‐acylated suppressor tRNAs cannot fully compete with near cognate suppression from host endogenous tRNAs for designated codons. As most of AARSs recognize the anticodons of tRNAs, mutating original anticodons will affect aminoacylation significantly. Thus, further engineering the anticodon binding regions of AARSs is essential for ncAA incorporation. Another possible reason is that the expression, folding, or stability of introduced AARSs may not be optimal in host cells since they are usually from different domains. In this case, approaches to evaluate and increase their expression, folding, or stability in hosts will help for ncAA incorporation.

It is significant to simultaneously incorporate different ncAAs into a single protein to expand the applications of genetic code expansion. Although we have popular TyrRS and PylRS systems as well as those “not‐so‐popular” ones, we still need to pay attention to several issues. First, the poly‐specificity of evolved AARSs. Because most of the evolved AARSs were only screened against canonical amino acids in negative selections but not for other ncAAs, so they usually cannot distinguish ncAAs with similar or even distinct structures (Figure 4a). For example, the PylRS variant originally selected for AcK is 30‐fold more efficient with Phe derivatives such as 3‐iodo‐Phe than it is with AcK (Guo et al., 2014). Thus, selected orthogonal pairs should be first tested for their specificities for target ncAAs which will be simultaneously incorporated into a single protein, individually. If they can cross‐react with ncAAs for other pairs, a mixture of ncAAs will be incorporated at designated codons. To overcome this issue, engineered AARSs should be further evolved for high specificity toward the target ncAA but not for other ncAAs. To realize this aim, other ncAAs which will be simultaneously incorporated into the protein should be added into the medium in the negative selection to exclude AARS variants recognizing those undesirable ncAAs. Second, the orthogonality between selected pairs. Selected pairs should be mutually orthogonal to each other. Otherwise, we cannot obtain proteins with ncAAs at corresponding positions correspondingly (Figure 4b). All the combinations of evolved AARSs and suppressor tRNAs from selected orthogonal pairs should be tested for their mutual orthogonality. To date, several mutually orthogonal pairs have already been used to incorporate multiple ncAAs into the same protein, including SepRS/PylRS, mjTyrRS/PylRS, SepRS/mjTyrRS, PhLysRS/mjTyrRS, PylRS/EcLeuRS (Vargas‐Rodriguez et al., 2018). Third, the codon selection for individual ncAA. Most AARSs use anticodon recognition to distinguish noncognate tRNAs, thus mutating anticodons may affect aminoacylation significantly. Since most of orthogonal pairs were developed for amber suppression, their aminoacylation activity for other anticodons need to be measured before use. Fortunately, several AARSs mentioned above are tolerant to anticodon changes such as PylRS, PhLysRS, archaea GluRS, HisRS, and SerRS. If they are chosen for multiple ncAA incorporation, the anticodons of their cognate suppressor tRNAs could be flexible.

FIGURE 4.

Mutual orthogonality between selected pairs for (a) ncAAs and (b) tRNAs

AUTHOR CONTRIBUTIONS

Joseph Andrews: Writing – original draft (lead). Qinglei Gan: Writing – review and editing (supporting). Chenguang Fan: Writing – review and editing (lead).

CONFLICT OF INTEREST

All authors declare no conflict of interest.

Andrews J, Gan Q, Fan C. “Not‐so‐popular” orthogonal pairs in genetic code expansion. Protein Science. 2023;32(2):e4559. 10.1002/pro.4559