tRNA^Pyl: Structure, function, and applications

ABSTRACT

Pyrrolysine is the 22nd proteinogenic amino acid encoded into proteins in response to amber (TAG) codons in a small number of archaea and bacteria. The incorporation of pyrrolysine is facilitated by a specialized aminoacyl-tRNA synthetase (PylRS) and its cognate tRNA (tRNA^Pyl). The secondary structure of tRNA^Pyl contains several unique features not found in canonical tRNAs. Numerous studies have demonstrated that the PylRS/tRNA^Pyl pair from archaea is orthogonal in E. coli and eukaryotic hosts, which has led to the widespread use of this pair for the genetic incorporation of non-canonical amino acids. In this brief review we examine the work that has been done to elucidate the structure of tRNA^Pyl, its interaction with PylRS, and survey recent progress on the use of tRNA^Pyl as a tool for genetic code expansion.

KEYWORDS:

• Non-canonical amino acids
• pyrrolysine
• pyrrolysyl-tRNA synthetase
• pyrrolysyl-tRNA
• synthetic biology
• tRNA^Pyl

Introduction

Pyrrolysine (Pyl) was identified in 2002 as the 22nd proteinogenic amino acid.^1,2 This rare and highly specialized amino acid is found in a small number of methylamine metabolizing archaea as well as a few bacteria. In these organisms, Pyl is biosynthesized by three enzymes (PylB, PylC, and PylD) from two equivalents of lysine (Fig. 1A).³ It is then directly encoded into proteins in response to an amber (TAG) codon with the help of a unique pyrrolysyl-tRNA synthetase (PylRS) along with its cognate tRNA (tRNA^Pyl, Fig. 1B, C).⁴

Figure 1. (A) Pyrrolysine is synthesized from 2 equivalents of lysine in 3 steps. (B) tRNA^Pyl is charged with pyrrolysine by pyrrolysyl-tRNA synthetase. (C) Pyrrolysine is incorporated into the nascent peptide in response to amber (UAG) codons during translation.

The PylRS/tRNA^Pyl pair is a particularly interesting aminoacyl-tRNA synthetase (aaRS)/tRNA pair for two reasons. First, understanding how Pyl incorporation has evolved may shed light on the origin of the genetic code. Whether Pyl emerged in highly specialized microbes as a recent addition to the code or is a relic of an earlier, more diverse genetic code is still unclear.^5,6 In either case, answering this question will, no doubt, advance our understanding of how the code has evolved into its present form. Second, the archaeal PylRS/tRNA^Pyl pair has been widely demonstrated to be orthogonal in bacteria and eukaryotes and this, coupled with the fact that tRNA^Pyl is a naturally occurring amber suppressor tRNA, has led to the widespread use of the PylRS/tRNA^Pyl pair for the genetic incorporation of non-canonical amino acids (ncAAs).^7-9 Several features of PylRS including high substrate side chain promiscuity and a high tolerance for mutations in the substrate binding pocket make it amenable for ncAA incorporation.⁸ Indeed, wildtype and mutant PylRS variants in conjunction with tRNA^Pyl have been used for the genetic incorporation of over 100 ncAAs into proteins expressed in bacteria and eukaryotes¹⁰ including mammalian cells.^11,12 This methodology has further been applied for the genetic incorporation of ncAAs into entire organisms including Caenorhabditis elegans,¹³ Drosophila melanogaster,^14,15 and more recently Mus musculus.^16,17

Most of the work toward expanding the genetic code using the PylRS/tRNA^Pyl system has focused on engineering PylRS for an expanded amino acid substrate spectrum. In this review however, we will focus on the unique features of tRNA^Pyl, including its structure, interaction with PylRS, and its role as a genetic code expansion tool. For a detailed review of PylRS as a genetic code expansion tool see ref. 8.

The structure of tRNA^Pyl

The structure of tRNA^Pyl from Archaea

Archaea of the family Methanosarcinaceae share high sequence identity with regards to their tRNA^Pyl.^18,19 The secondary structure of tRNA^Pyl was first described by Kryzcki and coworkers in their seminal paper on the discovery of tRNA^Pyl from M. barkeri (Mb- tRNA^Pyl).¹ Mb- tRNA^Pyl (Fig. 2, center structure) contains several unique secondary structures that are found in all tRNA^Pyl of the Methanosarcinaceae. Perhaps most distinct among these features are found in the variable loop and the anticodon stem. The shortened variable loop contains only three nucleotides. This is in contrast to most known class I and class II tRNAs which contain variable loop lengths of four to five, or greater than ten nucleotides, respectively.²⁰ Further, the anticodon stem of Mb-tRNA^Pyl contains six instead of the commonly observed five nucleotides. Other structural anomalies include a shortened linkage from two to one nucleotides between the acceptor and D stems, lack of a TψC sequence in the T loop, and lack of a GG sequence in the D loop—a sequence that is widely conserved among cytosolic tRNA.^1,21 In spite of these abnormalities, Mb-tRNA^Pyl is still predicted to assume the cloverleaf conformation similar to canonical tRNAs.

Figure 2. The secondary structure of tRNA^Pyl from Candidatus Methanomethylophilus alvus, Methanosarcina barkeri Fusaro, and Desulfitobacterium hafniense. Identity elements in the tRNA^Pyl from M. barkeri and D. hafniense are shown in orange. Bases in direct contact with the enzyme in the crystal structure of PylSc from D. hafniense are bold. In the structure from Ca. M. alvus, circles represent positions that are occupied by a base in the tRNA^Pyl from other Methanomassiliicoccales while underlined bases represent those that are missing.

Although crystal structures of PylRS from M. barkeri (Mb-PylRS) and its mutants have been extensively reported, the crystal structure of Mb-tRNA^Pyl has not been determined.²² Nonetheless, the tertiary structure of Mb-tRNA^Pyl was elucidated by Rudinger-Thirion and coworkers through a series of RNA footprinting and ultraviolet melting experiments.²³ Melting curves displayed only two transitions corresponding to the opening of the tertiary and secondary structures, indicating that the tRNA primarily adopts one defined tertiary structure in solution. The RNA cleavage profile with probes specific for single and double stranded sequences experimentally confirmed the previously predicted cloverleaf structure. Further, the profile was consistent with a relaxed tertiary structure which the authors proposed assumed the canonical L-conformation. A relaxed tertiary structure could be explained by the lack of nucleotide modifications found in mature tRNA^Pyl.²⁴ Indeed, Söll et al. demonstrated that mature tRNA^Pyl extracted from M. barkeri contained only two modifications, namely, 4-thiouridine at position 8 and 1-methyl-pseudouridine at position 50.⁴ It is worth noting that the LC/ESI-MS experiments performed in this study to reveal nucleotide modifications are unable to detect pseudouridine due to the fact that the modified and unmodified nucleotides have the same mass. Thus the presence of pseudouridine modifications in Mb-tRNA^Pyl cannot be ruled out.

It has been noted that, with the aforementioned structural peculiarities, Mb-tRNA^Pyl bears a noticeable resemblance to bovine mitochondrial seryl-tRNA (Bt-tRNA^Ser) of Bos taurus.^4,23,25 The structural features shared by Mb-tRNA^Pyl and Bt-tRNA^Ser include both the shorted variable loop and elongated anticodon stem, as well as a shortened D loop and only one nucleotide between the acceptor and D stems.²⁶ The tertiary structure of unmodified Bt-tRNA^Ser has been solved through computer modeling, chemical probing, and solution NMR.^27,28 The data from these experiments support the conclusion that Bt-tRNA^Ser adopts a canonical L-shaped tertiary structure with a compact core region. This observation further supports the claim that although unusual in its secondary structure, the overall tertiary structure of tRNA^Pyl is similar to that of canonical tRNAs. Importantly, it has been noted that although Mb-tRNA^Pyl and Bt-tRNA^Ser share several unique structural elements, there is little similarity in their primary sequences.²⁵

Recent evidence has supported the existence of a seventh order of methanogenic archaea related to the Thermoplasmatales dubbed the Methanomassiliicoccales.^29,30 Analysis of the genomes of at least five identified members of this order has revealed the presence of genes encoding homologs of tRNA^Pyl and PylRS as well as all of the enzymes needed for Pyl biosynthesis.^5,31,32 Further, in all species identified, the gene encoding MtmB, a methyltransferase required for metabolizing monomethylamine, contains an in-frame amber codon. Taken together, these observations strongly support the direct genetic encoding of Pyl within the seventh order methanogens. Interestingly, the tRNA^Pyls from this order (Mmc-tRNA^Pyl) contain their own set of unique features (Fig. 2, left structure).⁵ Perhaps the most curious of these is a broken anticodon stem that forms a loop of five or seven bases, a feature not found in tRNA^Pyl from the Methanosarcineae or in any bacterial tRNA^Pyl. Similar to Mb-tRNA^Pyl, the variable loop of Mmc-tRNA^Pyl is shortened and contains only three bases. The D loop is shortened further from five bases in Mb-tRNA^Pyl to four, or even three bases in some Mmc-tRNA^Pyl. Finally, the D and acceptor stems are separated by one or two bases or in some cases not at all.⁵ While the identification of the Pyl incorporation cassette in the Methanomassiliicoccales is exciting because of its ability to shed light on the evolutionary history of Pyl incorporation and its potential application as a genetic code expansion tool, given its relatively recent discovery, very little is known about the structure and function of tRNA^Pyl and PylRS from this group.

The structure of tRNA^Pyl from bacteria

At the same time that Mb-tRNA^Pyl was discovered to code for an in-frame amber codon in M. barkeri, an investigation of available genomes revealed homologs of Mb-tRNA^Pyl and Mb-PylRS as well as several in-frame amber codons in the Gram-positive bacterium Desulfitobacterium hafniense.¹ Since then, the Pyl encoding machinery has been discovered in 19 bacterial species including 16 Firmicutes and 3 delta-proteobacteria.³³ Given the taxonomic distribution of Pyl encoding organisms and the results of recent phylogenetic analyses,^5,33 it is believed that bacteria obtained the ability to incorporate Pyl as a result of a horizontal gene transfer from the Methanomassiliicoccales who, in turn, inherited Pyl incorporation from a common ancestor with the Methanosarcinaea.³⁴

Despite apparently obtaining the Pyl incorporation machinery from Methanomassiliicoccales, the secondary structure of tRNA^Pyl from D. hafniense (Dh-tRNA^Pyl) more closely resembles that of the Methanosarcineae (Fig. 2, right structure). Although the latter two share very little sequence homology, their tRNA^Pyl contain many of the same structural elements including a lengthened anticodon stem, shortened variable region, D loop, and connection between the acceptor and D stems, and lack of several nearly universally conserved bases detailed above.³⁵ Bacterial tRNA^Pyl are, however, more diverse than the highly conserved sequences of the Methanosarcineae, sharing just under half of their sequence identity¹⁸ and very few universally conserved residues are found among the known species.

The crystal structure of the catalytic domain of PylRS from D. hafniense (Dh-PylSc) bound to Dh-tRNA^Pyl was solved by Nureki and coworkers in 2009, offering the first direct evidence for the tertiary structure of tRNA^Pyl (Fig. 3A).³⁵ The crystal structure confirmed that the tRNA adopts a canonical L-shape along with a compacted core region (Fig. 3B). Both the L-shape and compact core were previously predicted based on the results of biochemical experiments and through comparing the structure of tRNA^Pyl with that of the known Bt-tRNA^Ser. The compact core of tRNA^Pyl is a result of nonstandard tertiary base pairing which arises primarily from the shortened D and variable regions as well as a deletion of U8. The U8:A14 bond is highly conserved among canonical tRNAs and is important for maintaining the L-shape.³⁶ Despite the compacted structure, the acceptor and anticodon stems reside in positions similar to those found in canonical tRNAs allowing tRNA^Pyl to function normally during translation.³⁵

Figure 3. (A) The crystal structure of PylSc in complex with tRNA^Pyl from Desulfitobacterium hafniense (PDB: 2ZNI). (B) The crystal structure of tRNA^Pyl from D. hafniense. The secondary structure features of tRNA^Pyl are colored: green, acceptor stem; blue, D arm; brown, anticodon stem; purple, anticodon; yellow, variable loop; orange, T arm.

The interaction of tRNA^Pyl with pyrrolysyl-tRNA synthetase

The pyrrolysyl-tRNA synthetases from archaea and bacteria belong to the family of class II aaRSs.^1,35,37 The enzyme from the Methanosarcineae contains an N-terminal domain connected by a highly variable linker to a C-terminal catalytic domain. In bacteria however, homologs of the C-terminal catalytic domain (PylSc) and N-terminal domain (PylSn) are coded for by two separate genes. The N-terminal domain of PylRS from M. barkeri (Mb-PylRS) and the homologous PylSn from D. hafniense represent novel RNA-binding domains as both have been shown to specifically bind tRNA^Pyl in vitro but share little sequence identity with known RNA-binding proteins.³⁸ In the absence of PylSn, PylSc is still capable of charging tRNA^Pyl with Pyl analogs in vitro and in vivo albeit with less activity than full-length archaeal Mm-PylRS.^25,35 Likewise, Mb-PylRS containing a truncation of the N-terminal domain is also capable of charging tRNA^Pyl in vitro, however, the truncated enzyme showed no activity in vivo.³⁹ Given that PylRS from both bacteria and archaea is still active in vitro without the N-terminal domain, but only archaeal PylRS requires this domain for in vivo activity, the exact biochemical function of this novel RNA binding domain remains elusive. It has been postulated based on the higher affinity of PylSn for tRNA^Pyl compared with PylSc (K_D = 0.13 μM and 6.9 μM, respectively), that this domain may serve to recruit tRNA^Pyl and facilitate acylation of tRNA^Pyl by the catalytic domain allowing the cell to maintain a lower basal level of tRNA^Pyl.^38,39 Such a role is yet to be supported by experimental evidence.

The catalytic domain of PylRS from bacteria and archaea contains an architecture that mirrors that of other class II aaRSs with the conventional fold of a β-sheet surrounded by several α-helices.^22,37,40,41 To investigate the interaction of the catalytic domain with tRNA^Pyl, the crystal structure of Dh-PylSc was solved in complex with Dh-tRNA^Pyl by Nureki and coworkers (Fig. 4).³⁵ The structure revealed that like other class II aaRS, PylRS approaches tRNA^Pyl from the major groove side of the acceptor stem. An α-helix of the tRNA binding domain 1 and C-terminal tail of PylSc along with the bulge domain of the neighboring subunit in the PylSc dimer form a binding site that accommodates the acceptor helix of tRNA^Pyl. This interaction directs the acceptor helix to the catalytic site where, like other class II aaRS, the enzyme acylates the 3’-OH of the acceptor stem.⁴² Along with the acceptor stem binding site, a core-binding surface for the tRNA is assembled from the tRNA-binding domain 1 and C-terminal tail of PylSc, as well as an α-helix from the neighboring subunit. In total, 31 residues of Dh-PylSc are involved in hydrogen bonding or stacking interactions with 17 bases in Dh-tRNA^Pyl (Table 1). Four of these residues are from the neighboring subunit of the PylSc dimer. As noted by Nureki et al. several of the interactions between PylRS and tRNA^Pyl make use of structural elements that are unique among PylRSs. To accommodate the compacted core of tRNA^Pyl, PylRS itself has evolved a compacted binding site. This compacted binding site sterically occludes canonical tRNAs and explains the orthogonality of the PylRS/tRNA^Pyl pair with respect to endogenous tRNAs and aaRSs.³⁵

Figure 4. The crystal structure of Dh-tRNA^Pyl bound to Dh-PylSc. The 2 protomers in the Dh-PylSc dimer are colored gray and black. Residues that interact with Dh-tRNA^Pyl in the crystal structure are shown in cyan and salmon, respectively. The colors corresponding to secondary structure features of Dh-tRNA^Pyl are the same as in Figure 3. For clarity, only one Dh-tRNA^Pyl is shown bound to the dimer. (PDB: 2ZNI).

Table 1. The interactions between bases in Dh-tRNA^Pyl and amino acid residues in Dh-PylSc. Unless noted, interactions involve main chain hydrogen bonds.

tRNA Region	tRNA^Pyl Base	PylSc Residue
Acceptor Stem	G4	K16
		Q20
Accetor→D Stem	G9	R140^,
		E245
D Stem	G10	R21
		E50
		R144
		N286
D Stem	A11	E50
D Stem	U12	Q18
D Stem	C13	T15
		D43
D Stem	U24	Q47
D→Anticodon Stem	A26a	H51
		N288
Acceptor Stem	A66	K124
Acceptor Stem	C69	S278
Acceptor Stem	C70	Y279
Acceptor Stem	C71	K23
		Q164
		H269
Acceptor Stem	C72	Q164
		A166
Acceptor Stem	G73	E162
		Q164
Acceptor Stem	C74	Q117
		H168
Acceptor Stem	C75	Q117
Acceptor Stem	A76	R160
		S163
		L169
		F172
		E229
		S232
		R259

¹ Residues are in the neighboring PylSc subunit of the dimer.

² Side chain hydrogen bond.

³ Stacking interaction.

Biochemical studies measuring the activity of PylRS toward mutant tRNA^Pyls have identified several important identity elements for Mb-tRNA^Pyl and Dh-tRNA^Pyl (Fig. 2).^25,43 As with most tRNAs, the discriminator base (G73) and the first base pair (G1:C72) were identified as critical identity elements for tRNA^Pyl from archaea and bacteria. These residues are highly conserved among Pyl incorporating organisms including those from the seventh order and were shown to make direct contact with PylSc in the crystal structure.³⁵ Other identity elements for Dh-tRNA^Pyl include the G10:C25 and A11:U24 base pairs, and the base G9, all of which were shown to contact the enzyme in the crystal structure. For Mb-tRNA^Pyl the base pair G51:C63 as well as U33 and A37, the two bases adjacent to the anticodon, were shown to be important identity elements. The bases adjacent to the anticodon do not make contact with PylSc in the crystal structure from D. hafienses, which led to the speculation that these bases may be important for the interaction of tRNA^Pyl with PylSn. However, more recent evidence has shown that U33G and A37C mutations did not detectibly affect binding of Dh-tRNA^Pyl by PylSn in an electrophoretic mobility shift assay.³⁸ The fact that these residues are important for Mb-tRNA^Pyl but not Dh-tRNA^Pyl may represent differences in the tRNA^Pyl/enzyme interaction between bacterial and archaeal organisms. The possibility of differing binding modes is underscored by the fact that while Mb-tRNA^Pyl and Dh-tRNA^Pyl are both substrates for Mb-PylRS, PylSc is unable to detectably acylate Mb-tRNA^Pyl.³⁹ With few exceptions, most tRNAs use the anticodon as a major identity element.⁴⁴ However, notably, neither Mb-tRNA^Pyl nor Dh-tRNA^Pyl show significantly decreased affinity for their respective aaRS or a reduction in acylation when mutations are made in the anticodon. As will be discussed in the following sections, this has major implications for the use of tRNA^Pyl as a genetic code expansion tool.

Applications of tRNA^Pyl for genetic code expansion

Engineered tRNA^Pyl for genetic code expansion

As mentioned previously, the PylRS/tRNA^Pyl pair has been widely adopted for the site-specific incorporation of a variety ncAAs across a range of host organisms. In general, ncAAs are incorporated into recombinant proteins expressed in the presence of a desired ncAA and in organisms bearing plasmid-borne copies of PylRS/tRNA^Pyl. Schultz and coworkers have developed a powerful strategy for the directed evolution of aaRSs that can be used to identify novel aaRS mutants capable of incorporating new ncAAs.⁴⁵ Using this selection methodology PylRS mutants have been identified for the introduction of over 100 ncAAs with various applications including the introduction of biorthogonal functional groups for bioconjugation and fluorescence labeling, heavy atoms for X-ray crystallography, fluorescent residues for protein folding-unfolding dynamics and Förster resonance energy transfer (FRET), spin labels for electron paramagnetic resonance (EPR), posttranslational modifications, etc.^7,8 (and references therein). Although a powerful tool for chemical biology, a major limitation of ncAA incorporation is inefficient suppression of the stop codon which results in elongation termination. To effectively reassign amber codons, tRNA^Pyl must be able to outcompete release factor 1 (RF-1) which, in prokaryotes, is responsible for terminating protein translation in response to amber and ochre nonsense codons.⁴⁶ Various strategies for increasing ncAA incorporation have been reported for both prokaryotic and eukaryotic systems.^47,48,49 To this end, tRNA^Pyl itself has also become a target of modification.

One strategy that has been explored for increasing translation efficiency is optimizing the interaction between archaeal tRNA^Pyl and bacterial elongation factor Tu (EF-Tu). In growing bacterial cells, the GTP-dependent EF-Tu is one of the most abundant proteins, outnumbering ribosomes approximately 10:1.⁵⁰ EF-Tu is tasked with the critical role of binding aminoacyl-tRNA and directing it to the A-site of the ribosome in a codon-dependent fashion during protein translation.⁵¹ Despite its peculiar structure, hydrolysis protection experiments have revealed that tRNA^Pyl interacts with EF-Tu in the same manner as canonical tRNAs.²³ The binding affinity of an aminoacyl-tRNA to EF-Tu can have tremendous impact on translation efficiency as binding that is too weak prevents efficient complex formation and binding that is too tight can prevent dissociation of the tRNA/EF-Tu complex during translation.⁵² E. coli aminoacyl-tRNAs have evolved relatively uniform affinities for EF-Tu, with contributions to binding made from structural elements of the tRNA as well as the affixed amino acid.^53,54 Archaeal tRNAs however, have not evolved for interaction with bacterial EF-Tu, leading to speculations that inefficient interaction between these two molcules from different domains of life may decrease the efficiency of ncAA incorporation.^55,56 Indeed, recent evidence from in vitro translation experiments shows that bacterial EF-Tu binds aminoacyl-tRNA^Pyl with an affinity 25 times lower than bacterial aminoacyl-tRNA^Phe leading to reduced translation efficieny.⁵⁷ Thus, the interaction between EF-Tu and aminoacyl-tRNA^Pyl is a prime target for optimizing ncAA incorporation using the Pyl translation machinery.

Realizing this potential, Schultz et al. set out to optimize the interaction between EF-Tu and the tyrosyl-tRNA (Mj-tRNA^Tyr) from Methanocaldococcus jannaschii.⁵⁵ The archaeal Mj-TyrRS/Mj-tRNA^Tyr _CUA pair is orthogonal in E. coli which has led to its widespread use for ncAA incorporation. The crystal structure of EF-Tu from Thermus aquaticus in complex with E. coli tRNA^Cys revealed several positions in the acceptor and T stems of the tRNA that are predicted to interact with EF-Tu. From these data, a series of tRNA libraries were constructed and passed through alternating rounds of positive and negative selection based on the ability to suppress an amber codon. Remarkably, improvements in protein yield ranging from 175–2520% were reported using mutant Mj-tRNA^Tyrs demonstrating the important role that the EF-Tu/tRNA interaction can play in ncAA incorporation efficiency. Söll and coworkers, sought to improve ncAA incorporation efficiency in a similar fashion with tRNA^Pyl.⁵⁶ Iterative screening of three tRNA^Pyl libraries with mutations in the acceptor and T stems revealed a mutant tRNA^Pyl,dubbed tRNA^Pyl-Opt, which facilitated nearly a 3-fold increase in protein expression when superfolder green fluorescent protein (sfGFP) with an amber codon at position 149 was used as a reporter. Increases in expression level were reported with various ncAAs with the level of improvement showing some dependency on the identity of the ncAA. The higher efficiency of tRNA^Pyl-Opt allowed for the production of histone H3 containing acetyllsyine, a naturally occurring posttranslational modification important for gene regulation, with a yield nearly six times greater than that obtained with wildtype tRNA^Pyl.

Codon reassignment

The genetic code is composed of four bases assembled into 64 3-base codons. Although there exists some variation of the readout, the majority of known organisms use 61 of these codons to encode 20 amino acids and the remaining three signal translation termination. Therefore, one can see that there is considerable redundancy within the code with an amino acid often being encoded by multiple, synonymous codons. Early efforts to incorporate ncAAs using the Pyl incorporation machinery took advantage of the fact that tRNA^Pyl is a naturally occurring amber suppressor tRNA, suppressing TAG codons and leaving two redundant stop codons (TGA and TAA). However, the revelation that PylRS tolerates mutations that occur in the anticodon of tRNA^Pyl implies that this system could theoretically be used to incorporate an ncAA at any of the 64 codons. The desire to incorporate multiple different ncAAs into the same protein has led to the search for other codons that can be reassigned.

In 2010, Liu and colleagues showed that the permissive anticodon position of tRNA^Pyl can be reassigned allowing for the incorporation of two different ncAAs into the same protein.⁵⁸ By mutating the anticodon of tRNA^Pyl to one of the three stop codons opal (TGA), ochre (TAA), and amber codons were all efficiently suppressed in sfGFP expressed in E. coli. Opal codons displayed the highest level of suppression efficiency, however their use for ncAA incorporation should be avoided due to the possibility of wobble pairing between the opal codon and canonical tRNA anticodons. Indeed, it has been shown that significant levels of tryptophan are observed to be incorporated at opal codons in E. coli.⁵⁹ To incorporate two different ncAAs, cells are transformed with plasmids containing the mutant PylRS/tRNA^Pyl _UAA as well as the Mj-TyrRS/tRNA^Tyr _CUA pair. The former aaRS/tRNA introduces ncAAs at ochre codons with the later introducing ncAAs at amber codons (Fig. 5A, B). This methodology for expressing proteins with two different ncAAs has been used for the introduction of reactive handles for copper-catalyzed protein labeling as well as for catalyst free dual labeling of proteins.^58,60 The later method was used for the installation of a FRET pair to study protein ligand binding in E. coli glutamine binding protein (Fig. 5C).

Figure 5. (A) Mm-PylRS and Mj-TyrRS specifically charge their respective tRNAs with an ncAA. (B) Mm-tRNA^Pyl _UUA incorporates an ncAA at ochre (UAA) codons while Mj-tRNA^Tyr _CUA incorporates ncAAs in response to amber (UAG) codons. (C) Dual labeling of glutamine binding protein containing two reactive ncAAs.

While stop codon reassignment has proven to be effective for the introduction of two different ncAAs, it is limited to only three codons. More recent studies have looked at the possibility of expanding the number of codons available for ncAA incorporation further through sense codon reassignment. Given that most amino acids are redundantly encoded by multiple codons, reassignment of synonymous codons has the potential to greatly expand the number of available codons for ncAA incorporation. However, sense codon reassignment presents a unique set of challenges as, among other things, redefining a codon in the coding sequence of a gene may have detrimental effects on the host organisms—especially if that codon falls within an essential gene.^61,62 Further, just as stop codon suppression involves competition with elongation termination, incorporating ncAAs via sense codon reassignment requires that the orthogonal aaRS/tRNA pair be able to efficiently outcompete the endogenous translation system.

It is well known that usage of synonymous codons within an organism is biased with some codons occurring more frequently than others.⁶³ To minimize the adverse effects of sense codon reassignment using the PylRS/tRNA^Pyl pair, early efforts focused on reassigning rarely used codons, in particular, those encoding arginine. For example, Söll et al. set out to reassign the rare CGG codon using the PylRS/tRNA^Pyl pair in Mycoplasma capricolum.⁶⁴ M. capricolum was chosen as the host because of the scarcity of CGG codons in its genome (only 6 CGG codons were identified) coupled with the fact that it lacks the requisite tRNA_CCG for decoding CGG. Despite the ability of PylRS to charge mutant tRNA^Pyl _CCG in vitro, LC-MS/MS of β-galactosidase expressed in M. capricolum harboring plasmids encoding Mm-PylRS/tRNA^Pyl _CCG revealed incorporation of only Arg at CGG codons. It was suspected that Arg incorporation was a result of misacylation of tRNA^Pyl _CCG by endogenous ArgRS, which utilizes the tRNA anticodon as a major identity element. These suspicions were confirmed by in vitro assays which showed that tRNA^Pyl _CCG was efficiently charged with Arg by recombinant E. coli and M. capricolum ArgRSs. This work highlighted another difficulty with sense codon reassignment. Since most aaRSs utilize the anticodon as a major identity element, mutating the anticodon of tRNA^Pyl to a sense codon may decrease its orthogonality and render it a substrate for endogenous aaRSs.

More successful results were reported for reassignment of the rare AGG codon in E. coli. Liu and coworkers reported 24% incorporation of an ncAA at AGG codons in GFP_UV expressed in cells harboring plasmid-borne copies of Mm-PylRS and mutant tRNA^Pyl _CCU.⁶⁵ Incorporation efficiency could be further improved to approximately 90% using a minimal media devoid of Arg coupled with inducing the expression of PylRS/tRNA^Pyl _CCU ahead of GFP_UV induction. The later strategy allowed for the accumulation of aminoacyl-tRNA^Pyl _CCU thereby improving its ability to compete with Arg-tRNA^Arg _CCU for decoding AGG. Using wildtype Mm-PylRS/tRNA^Pyl _CCU three different ncAAs were incorporated at AGG codons with efficiencies of 72–92%. Although impressive, these results required the use of high concentrations (10 mM) of ncAA and still resulted in a heterogeneous product.

Sakamoto et al. minimized competition from endogenous tRNA by generating a tRNA^Arg _CCU (argW) knockout strain.⁶⁶ They went on to replace several AGG codons in essential genes in the E. coli genome. The codons that were not replaced were left to be decoded by a plasmid-borne copy of a synonymous Arg tRNA (tRNA^Arg _UCU), the UCU anticodon of which forms a wobble pair to decode AGG albeit at a lower efficiency than CCU. The greater efficiency of CCU at decoding AGG as compared with UCU allowed for tRNA^Pyl _CCU to outcompete tRNA^Arg _UCU during protein translation. When a peptide-SUMO fusion protein containing an AGG codon was expressed in the argW-knockout strain, in the presence of mutant tRNA^Pyl _CCU and mutant Mm-PylRS, the protein containing an ncAA at the AGG position was obtained with nearly quantitative occupancy of the ncAA. Minimal amounts of protein containing Arg at the AGG position were again attributed to misacylation of tRNA^Pyl _CCU by endogenous ArgRS as was reported previously for tRNA^Pyl _CCG.⁶⁴ At nearly the same time that this work was published, Yoo et al. reported the incorporation of a variety of ncAAs at AGG codons with undetectable Arg incorporation using a mutant Mj-TyrRS/tRNA^Tyr _CCU pair in an E. coli strain lacking argW as well as an enzyme required for Arg biosynthesis.⁶⁷ Through limiting the available Arg in the cell, ncAAs could be incorporated at up to three AGG codons with insignificant background incorporation.

In an impressive study, Chin and colleagues demonstrated that the PylRS/tRNA^Pyl pair can be used to reassign frequent codons with low efficiency in E. coli, human cells, and D. melanogaster.¹⁵ By mutating the anticodon of tRNA^Pyl to one of several frequently occurring codons, ncAAs bearing a reactive functional group were incorporated into proteins with efficiency of 0.02–0.65% per codon. While the incorporation efficiency was low, this limited the detrimental effects of ncAA incorporation on the organism and was useful for fluorescence labeling of the entire proteome. Further, by restricting the expression of PylRS and tRNA^Pyl to certain tissues of D. melanogaster, they were able to label proteins containing the ncAA in a tissue-dependent manner.

More recently, Söll et al. have explored the possibility of reassigning a frequently occurring serine codon using the PylRS/tRNA^Pyl pair.⁶⁸ Despite being significantly more abundant than AGG in the E. coli genome, reassignment of a Ser codon was chosen based on the fact that the seryl-tRNA synthetase of E. coli does not use the tRNA anticodon as a recognition element. This eliminates the possibility that the anticodon will cause misacylation by the endogenous aaRS as was observed when reassigning CGG and AGG Arg codons. Further, the Ser AGU codon is naturally decoded by tRNA^Ser _GCU which involves a wobble pair at the third position. Since mutant tRNA^Pyl _ACU could decode AGU via Watson-Crick pairing it was hypothesized that this stronger interaction would allow tRNA^Pyl _ACU to outcompete tRNA^Ser _GCU. LC-MS/MS of sfGFP that was expressed in E. coli cells harboring plasmid-borne copies of mutant tRNA^Pyl _ACU as well as a mutant PylRS revealed successful ncAA incorporation at one of two AGU codons in sfGFP. Further analysis revealed that the ncAA was incorporated at this AGU site at approximately 65% efficiency, however the ncAA was also found to be incorporated at an AGC codon_. Impressively, it was also demonstrated that a more frequent Ser codon (UCG) as well as the most abundant codon in E. coli (CUG), which encodes leucine, could also be recoded to an ncAA with varying success. While the overall efficiency of reassigning AGU was low, the fact that greater than 50% reassignment was achieved is noteworthy given that a relatively low ncAA concentration (1 mM) was used and that no deletions in the Ser biosynthesis pathway or of a synonymous tRNA^Ser were required. These observations suggest that, under optimized conditions, codon reassignment is not limited to rarely occurring codons.

Quadruplet codons

In the search for blank codons for ncAA incorporation four base codons offer a promising alternative to nonsense and sense codon reassignment with the potential to code for upwards of 250 blank codons.^7,69 Early efforts by Schultz et al. demonstrated the feasibility of decoding a four base codon in vivo using the orthogonal Pyrococcus horikoshii LysRS in conjunction with mutant tRNA^Lys _UCCU and, when coupled with the Mj-TyrRS/tRNA^Tyr _CUA pair, allowed for the production of proteins containing two different ncAAs.⁷⁰ However, the system was complicated by toxic effects of expressing Ph-LysRS in E. coli, and suffered low yield from inefficient decoding of quadruplet codons on the ribosome. Chin et al. overcame this later challenge by evolving ribosomes with enhanced ability for decoding four base codons.⁷¹ Using these evolved ribosomes in conjunction with the Mj-TyrRS/tRNA^Tyr _UCCU pair, they obtained quadruplet codon suppression with improved yield compared to the Ph-LysRS/tRNA^Lys _UCCU pair. However, Mj-TyrRS uses the anticodon of tRNA^Tyr as an identity element⁷² which limits the efficiency with which this pair can be used to decode quadruplet codons. More recently, the PylRS/tRNA^Pyl pair has become a promising option for quadruplet codon decoding given its greater tolerance for anticodon mutations.

In 2013, Guo et al. reported the use of the PylRS/tRNA^Pyl pair for incorporating ncAAs at quadruplet (AGGA) codons.⁷³ To identify tRNA^Pyl mutants with more efficient 4 base decoding on natural ribosomes, a series of tRNA^Pyl _UCCU libraries were constructed with randomized nucleotides in the anticodon stem and loop (Fig. 6A). The libraries were passed through positive selections based on their ability to suppress an AGGA codon in the chloramphenicol acetyltransferase (cat-AGGA) gene in the presence of an ncAA (Fig. 6B). If the mutant tRNA^Pyl _UCCU is able to efficiently suppress AGGA in cat-AGGA, the cells survive in the presence of chloramphenicol. Expressing GFP_UV containing an AGGA mutation in the presence of mutant tRNA^Pyls identified from selection, resulted in an increase in fluorescence signal of ∼2–4-fold as compared with wildtype tRNA^Pyl _UCCU. Using the best performing mutant, they were able to efficiently incorporate an ncAA at an AGGA codon in enhanced GFP (EGFP) expressed in 293T cells. Molecular simulations of the selected tRNA^Pyls revealed structural changes in the anticodon that likely facilitated decoding of the quadruplet codon on the ribosome.⁷³

Figure 6. (A) Randomization of the anticodon stem and loop allows for the optimization of tRNA^Pyl for decoding quadruplet codons. (B) In positive selection, tRNA^Pyl mutants that are able to suppress quadruplet codons are selected for. (C) Negative selection in the absence of an ncAA removes tRNA^Pyl mutants that are aminoacylated with canonical amino acids by endogenous aminoacyl-tRNA synthetases.

While Guo et al. reported quantitative ncAA incorporation in response to AGGA with mutant and wildtype tRNA^Pyl _UCCU, others^65,74 have reported significant Arg incorporation at AGGA sites with wildtype tRNA^Pyl. Interestingly, Liu et al. observed predominantly Arg incorporation with a sfGFP-134AGGA reporter, but predominantly ncAA incorporation with a GFP_UV-149AGGA reporter.⁶⁵ Recently, Arg incorporation at AGGA was also reported with the Mj-TyrRS/tRNA^Tyr _UCCU pair.⁷⁵ In this study, background Arg was attributed to misaminoacylation of tRNA^Tyr _UCCU by endogenous Ec-ArgRS as Ec-ArgRS uses the AGG anticodon as a major identity element. Introducing an A38C mutation in tRNA^Tyr _UCCU removes a minor identity element of Ec-ArgRS and this was shown to eliminate background Arg incorporation. This observation suggests that background suppression indeed arises from misaminoacylation of the orthogonal tRNA. However, in vitro aminoacylation assays have shown that tRNA^Pyl _UCCU is not a substrate for Ec-ArgRS,⁷⁴ suggesting that the observed suppression in the case of tRNA^Pyl is due to competition with endogenous tRNA^Arg _CCU. However, as we will discuss next, more recent data suggests that tRNA^Pyl _UCCU is indeed a substrate for Ec-ArgRS.

Chin et al. demonstrated that by including a round of negative selection, tRNA^Pyl _UCCUs that are aminoacylated by endogenous aaRS can be removed from the tRNA library which results in little observable background suppression.⁷⁶ In this study, they generated a tRNA^Pyl _UCCU library and performed selection for AGGA suppression capability on an evolved orthogonal ribosome that had previously been shown to have enhanced ability for decoding quadruplet codons.⁷¹ However, before performing a positive selection, the tRNA library was screened in the absence of an ncAA in cells harboring a copy of the toxic barnase gene containing AGGA codons. Therefore, library members that are aminoacylated by endogenous aaRSs give rise to expression of the toxic gene and cell death (Fig. 6C). This was followed by positive selection using cat-AGGA as reported by Guo et al. A resulting mutant tRNA^Pyl _UCCU was able to suppress AGGA codons in cat-AGGA to confer resistance at >500 µg·mL⁻¹ chloramphenicol in the presence of an ncAA, but was unable to grow on 25 µg·mL⁻¹ in the absence of an ncAA, indicating very little background incorporation. By comparison, the wildtype tRNA^Pyl _UCCU resulted in much higher background conferring resistance of up to 125 µg·mL⁻¹ chloramphenicol in the absence of an ncAA.⁷⁶ The observation that mutant tRNA^Pyl _UCCU gives lower background than the wildtype gives credence to the hypothesis that background quadruplet suppression is dominated by misaminoacylation of the tRNA by endogenous aaRS. If competition from endogenous tRNA^Arg _CCU were the prevailing reason for background Arg incorporation, one would expect similar levels of background in both cases. Although the evolved tRNA^Pyl _UCCU retained the A38 recognition element for Ec-ArgRS it had a greatly expanded anticodon loop compared with wildtype tRNA^Pyl _UCCU (12 bases compared with 8 bases in the wildtype). This expanded anticodon loop may prevent misaminoacylation by Ec-ArgRS.

The ability of three other tRNA^Pyls, whose anticodons (TAGA, AGTA, and CTAG) are not recognition elements for endogenous aaRSs, to decode quadruplet codons was also investigated by Chin et al.⁷⁶ Wildtype tRNA^Pyls containing these anticodons displayed virtually no background while the evolved versions facilitated more efficient ncAA incorporation. Guo et al. recently reported the ability of evolved tRNA^Pyls to efficiently deliver ncAA in response to TAGN (where N = A, U, G) codons.⁷⁷ Using the TAGN codon avoids the concern of competition with endogenous tRNAs. Further, after randomizing positions in the anticodon stem and loop they performed selection in an E. coli strain lacking RF-1 and in which all instances of TAG were replaced with TAA.⁷⁸ This eliminated competition from RF-1 during decoding. The evolved tRNA^Pyls were able to incorporate ncAAs at TAGN codons with high efficiency and with no detectable background into GFP_UV. Interestingly, the identity of the fourth base was not critical for decoding with all three anticodons capable of decoding TAGA, TAGU, and TAGG albeit at a lower efficiency than the Watson-Crick pair.⁷⁷

Conclusion

Over the past one-and-a-half decades much has been learned about the structure and function of tRNA^Pyl and its cognate aaRS. Despite enormous progress, there are still questions that remain to be answered. Most importantly, ascertaining how the ability to incorporate Pyl has emerged in Nature will provide more insight into the origins of the genetic code. As more organisms that are capable of incorporating Pyl are identified, a clearer picture of the evolutionary history of Pyl incorporation can be developed. The PylRS/tRNA^Pyl pair has emerged as an outstanding tool for genetic code expansion in bacteria, eukaryotic cells, and whole organisms. Recent progress in codon reassignment and quadruplet codon decoding has taken this pair beyond stop codon reassignment and greatly expanded the number of available codons for ncAA incorporation. As these methods are refined, it is likely that the number of unique ncAAs able to be incorporated into a single protein will increase. The PylRS/tRNA^Pyl pair will, no doubt, be a valuable tool for the progress of synthetic biology.

Funding

Work in W.R. Liu's laboratory is funded by National Institutes of Health (R01GM121584), National Science Foundation (CHE-1148684), and Welch Foundation (grant A-1715).