Engineering Pyrrolysine Systems for Genetic Code Expansion and Reprogramming

Abstract

Over the past 16 years, genetic code expansion and reprogramming in living organisms has been transformed by advances that leverage the unique properties of pyrrolysyl-tRNA synthetase (PylRS)/tRNA^Pyl pairs. Here we summarize the discovery of the pyrrolysine system and describe the unique properties of PylRS/tRNA^Pyl pairs that provide a foundation for their transformational role in genetic code expansion and reprogramming. We describe the development of genetic code expansion, from E. coli to all domains of life, using PylRS/tRNA^Pyl pairs, and the development of systems that biosynthesize and incorporate ncAAs using pyl systems. We review applications that have been uniquely enabled by the development of PylRS/tRNA^Pyl pairs for incorporating new noncanonical amino acids (ncAAs), and strategies for engineering PylRS/tRNA^Pyl pairs to add noncanonical monomers, beyond α-L-amino acids, to the genetic code of living organisms. We review rapid progress in the discovery and scalable generation of mutually orthogonal PylRS/tRNA^Pyl pairs that can be directed to incorporate diverse ncAAs in response to diverse codons, and we review strategies for incorporating multiple distinct ncAAs into proteins using mutually orthogonal PylRS/tRNA^Pyl pairs. Finally, we review recent advances in the encoded cellular synthesis of noncanonical polymers and macrocycles and discuss future developments for PylRS/tRNA^Pyl pairs.

1. Introduction

Over the past 16 years pyrrolysyl-tRNA synthetase (PylRS)/ pyrrolysyl tRNA (tRNA^Pyl) pairs have been extensively engineered to add noncanonical α-L-amino acids (ncAAs) and other noncanonical monomers (ncMs) to the genetic code of diverse organisms. Pyrrolysine (pyl) systems have been central to essentially every major development in genetic code expansion and reprogramming.¹⁻⁶ Here we summarize the discovery of the pyl system, and its basic properties (Section 2), we then describe the key features of the initially characterized PylRS/tRNA^Pyl pairs for genetic code expansion (Section 3); these sections provide the necessary background on the properties of the pyl system that make it amenable to engineering. We then review the development and optimization of PylRS/tRNA^Pyl pairs for genetic code expansion in E. coli, and for genetic code expansion in prokaryotic and eukaryotic systems (Section 4), and the concerted biosynthesis and incorporation of ncAAs using PylRS/tRNA^Pyl pairs (Section 5). We review the scope of the ncAAs that can be site-specifically incorporated with PylRS/tRNA^Pyl pairs and summarize the types of applications they enable (Section 6). While genetic code expansion in living cells has, until recently, essentially been limited to α-L-amino acids with variant side chains or closely related α-L-hydroxy acids, we review recent progress in adding new classes of ncMs to the genetic code (Section 7). We then review the discovery of mutually orthogonal, triply orthogonal and quintuply orthogonal PylRS/tRNA^Pyl pairs from newly discovered and characterized PylRS/tRNA^Pyl classes (Section 8), and efforts to direct PylRS/tRNA^Pyl pairs to codons beyond the amber codon (Section 9). We describe progress on combining multiple engineered mutually orthogonal PylRS/tRNA^Pyl pairs that recognize distinct ncAAs and decode distinct codons for encoding multiple distinct ncAAs into proteins (Section 10), and recent progress on realizing genetically encoded cellular noncanonical polymer synthesis (Section 11). Finally, we describe the challenge and opportunities that may be addressed with pyl systems in the future (Section 12). While our focus is on genetic code expansion in living organisms, pyl systems have also contributed to in vitro code expansion⁷⁻¹² and we mention in vitro work when it has directly informed in vivo advances.

2. Discovery of the Pyrrolysine System

2.1. Pyrrolysine in Methanogens

The efficient read-through of an in-frame amber codon (TAG in genes, UAG in transcripts) in mono-, di- and trimethylamine methyltransferases (mtmB, mtbB, and mttB) in Methanosarcina barkeri (Mb) first pointed toward the expansion of the genetic code beyond the canonical 20 amino acids in some methanogenic archaea.¹³⁻¹⁵ The crystal structure of the mtmB protein of Mb revealed a modified lysine residue forming an ε-amide bond with a (4R,5R)-4-substituted-pyrroline-5-carboxylate, later termed pyrrolysine (Pyl – we use bold letters when explicitly referring to the amino acid), which was incorporated in response to an in frame amber codon (Figure 1a,b).¹⁶ The precise chemical composition of the 4-substituent of the pyrroline ring was not resolved in the initial crystal structure and the 4-methyl group, and the final structure of Pyl, was confirmed by mass spectrometry (MS).¹⁷ The unique chemistry of Pyl assists the transfer of the methyl group of mono-, di-, or trimethylamine (MMA, DMA and TMA, respectively) to the corrinoid cofactor of the corrinoid proteins MtmC, MtbC, or MttC respectively, this enables the relevant organisms to use methylamines as an energy source. Recent work has provided structural insight into the mechanism of methyl group transfer.¹⁸

Figure 1.

Encoded cellular incorporation of Pyl at amber codons, via natural genetic code expansion. a, The chemical structure of Pyl. b, The amber suppressor tRNA, tRNA^Pyl_CUA, is selectively charged by PylRS with Pyl. EF-Tu transports the aminoacylated pyl-tRNA^Pyl_CUA to the ribosome, where Pyl is site-specifically incorporated into a protein in response to an amber stop (UAG) codon in the mRNA. Adapted from Dunkelmann et al.¹⁹ – copyright © The Author(s) 2024 CCBY http://creativecommons.org/licenses/by/4.0/.

2.2. PylRS/tRNA^Pyl Pairs Direct Incorporation of Pyl

In parallel with the discovery of Pyl in the crystal structure of mtmB a highly unusual tRNA gene - pylT - encoding a tRNA with a CUA anticodon was revealed,²⁰ this gene was proximal to the mtmB gene in Mb Fusaro. The deletion of pylT from the genome of Methanosarcina acetivorans (Mac) rendered the archaea unable to grow on methylamine substrates.²¹PylS, a gene directly adjacent to pylT, encoded an aminoacyl-tRNA synthetase (aaRS) bearing low sequence similarity to canonical aaRS enzymes. The aaRS enzyme, termed pyrrolysyl-tRNA synthetase (PylRS), belongs to class IIc synthetases and is likely to have evolved from phenylalanyl-tRNA synthetase (PheRS). The MacPylRS/MactRNA^Pyl_CUA pair, when expressed in E. coli containing MtmB202TAG, led to Pyl-dependent synthesis of full length MtmB containing Pyl at position 202, as judged by tandem mass spectrometry (MS/MS) of a tryptic fragment; this provided evidence that MacPylRS aminoacylates MactRNA^Pyl_CUA with Pyl to enable the cotranslational incorporation of this amino acid in response to the amber codon (Figure 1b).²²

2.3. Pyl Biosynthesis

The gene cluster encoding mtmB, mtbB, mttB, and the genes for the MbPylRS/MbtRNA^Pyl_CUA pair in Mb Fusaro also contains pylB, pylC, and pylD. The sequence similarity of these genes to known metabolic enzymes directly indicated a role in the biosynthesis of Pyl. Interestingly, the phylogenetically distant Gram-positive bacterium Desulfitobacterium hafniense (Dh), the genome sequence of which was analyzed after the identification of an mtmB gene with an in frame amber codon, encodes homologous genes to pylT and pylS together with pylB, pylC, and pylD (Figure 2a).²⁰ The occurrence of the same gene cluster in distantly related organisms suggested that the pylTSBCD gene cluster is a self-contained biosynthetic pathway, transferred through horizontal gene transfer, that directs the incorporation of Pyl in response to an amber codon. Indeed, expression of Mac pylTSBCD in E. coli resulted in the genetic encoding of Pyl in response to amber codons.²³

Figure 2.

Pyl biosynthesis is mediated bypylBCD. a, Operon structure of the gene cluster pylTSBCD in the archaeon Methanosarcina acetivorans and the bacterium Desulfitobacterium hafniense. b, Biosynthetic pathway of Pyl from two molecules of lysine mediated by pylBCD. First, the radical S-adenosyl methionine (SAM) enzyme PylB converts lysine into (3R)-3-methyl-D-ornithine (3MO). Subsequently PylC ligates a second lysine to 3MO and PylD oxidizes the terminal amine of the conjugate to an aldehyde. The pyrroline ring is then spontaneously formed by a condensation reaction.

The biosynthetic pathway of Pyl (Figure 2b) was elucidated by stable isotopic labeling, MS, and genetics of E. coli cells transformed with variants of the pylTSBCD operon from Mac and the Mb gene mtmB. Two molecules of lysine were found to be the precursor of Pyl.²⁴ These mechanistic studies elucidated the function of PylC and PylD and provided hints toward the function of PylB; the mechanism of PylB was confirmed in subsequent experiments.²⁵ In the pathway, PylB, which is a radical iron–sulfur-S-adenosyl-L-methionine protein, converts lysine into 3-methyl-D-ornithine (3MO) through a rearrangement of the carbon backbone; this is consistent with prior work showing that addition of D-ornithine (DO) to E. coli cells harboring pylTSBCD boosted amber suppression efficiency 7-fold.²⁶ PylC, a member of the carbamoyl phosphate synthetase family, ligates 3MO to the ε-nitrogen of lysine forming an amide bond in an ATP-dependent manner. Finally, PylD, a nicotinamide adenine dinucleotide dependent dehydrogenase, oxidizes the δ-amine of the 3MO residue of 3-methyl-D-ornithyl-N^ε-L-lysine and catalyzes the ring closure at the C-5 position of the 3-methyl-D-ornithyl group by dehydrogenation. Crystal structures of all three enzymes have given further insights into the mechanism of Pyl biosynthesis.^27,28 Additional experiments have shown that the pyl biosynthetic pathway can function in all domains of life.²⁹ Interestingly, in Acetohalobium arabaticum the expression of pylTSBCD, and therefore the incorporation of Pyl in response to amber codons, is regulated by the growth condition of the bacterium.³⁰ When grown on pyruvate, in the absence of TMA, the genes in pylTSBCD are not sufficiently expressed and amber codons are not suppressed. However, when TMA is added to the growth medium the pyl pathway is expressed and Pyl incorporated into mttB and other proteins.

The genes of the pylTSBCD operon were initially believed to be stereotypically ordered in archaea, with pylT being followed by pylS, pylB, pylC, and pylD, and in a more variable, but compact, configuration in bacteria.³¹ Furthermore, only rare examples of organisms harboring the full pyl system together with genes other than mtmB, mtbB, and mttB with in-frame amber codons have been reported.^32,33 Recent data from culture and metagenomic approaches hint at much wider distribution of PylRS/tRNA^Pyl pairs in archaea and an astonishing sequence diversity within the isoacceptor class.³⁴⁻⁴²

3. PylRS/tRNA^PylPairs

Pyrrolysine systems were initially identified from select methanogenic archaea from the order of Methanosarcinales, and from certain methanogenic bacteria. However, culture and metagenomic based approaches have provided a basis for substantially expanding our understanding of pyl system diversity.³⁵⁻³⁹ Known PylRS enzymes display three distinct architectures. In the first architecture, the C-terminal domain (PylRSc) is covalently connected via a flexible linker to a highly basic, yet hydrophobic, N-terminal domain (PylRSn). In the second architecture–initially described for bacterial PylRS systems, but now also identified in archaeal genomes–the PylRSn and PylRSc are expressed as separate polypeptides. In the third architecture–found in Methanomassiliicoccales and other archaea–no N-terminal domain has been identified, and the protein functions as a stand-alone C-terminal domain.^{35,36,38−40,42}

In this section we introduce the nomenclature used throughout this review for PylRS/tRNA^Pyl pairs (Section 3.1) and then describe insights into the intrinsic properties of PylRS/tRNA^Pyl pairs derived from the characterization of initially discovered systems. Most of this work focused on the archaeal systems from Mb, and Methanosarcina mazei (Mm) and the bacterial system from Dh. The insights into the active site of the C-terminal catalytic domain in these PylRS enzymes (Section 3.2) appear to broadly translate to other PylRS systems that have been investigated. Studies on the N-terminal domain of PylRS enzymes (Section 3.3) provide information about systems containing this domain. Detailed insights into the interface between PylRS and tRNA^Pyl (Section 3.4) are likely to be system specific, but the overall compact structure of tRNA^Pyl and the topology of the complexes is thought to be a general feature of pyl systems. The observation that the active site of PylRS enzymes can accommodate Pyl analogs (Section 3.5) seems to hold for other pyl systems tested. The observation that PylRS enzymes do not recognize the anticodon of tRNA^Pyl (Section 3.6) also appears to extend to other pyl systems where this has been tested.

3.1. Nomenclature of PylRS/tRNA^Pyl Pairs

For clarity we use the following nomenclature to accurately, and unambiguously describe PylRS enzymes and pyl tRNAs. First, we subdivide the PylRS/tRNA^Pyl pairs into groups (Figure 3a). Three major groups can be defined based on the architecture of the pyl system: the + N group (where PylRSn is covalently linked to PylRSc), the ΔN group (which lacks the PylRSn), and the sN group (where PylRSn and PylRSc are expressed in trans from separate genes). The definition of groups is solely based on the architecture, inferred from genomic sequence, of the PylRS enzyme. We then define five pyl classes (A, B, C, N and S) based on the sequence and function of a PylRS/tRNA^Pyl pair; the five PylRS/tRNA^Pyl classes are discussed in detail in Section 8 (Figure 3a). When referring to a PylRS enzyme we use the letter of the class and specify if PylRSc is present in combination with PylRSn (in trans, or covalently linked) by adding the suffix ‘⁺’, or lacks PylRSn by adding the suffix “^Δ” after the letter assigning the class. As an example, MmPylRS is a member of class N and is referred to as N⁺-MmPylRS, whereas its tRNA is referred to as N-MmtRNA^Pyl (Figure 3b). DhPylRS is a member of class S, and if DhPylRSc is expressed in the presence of the DhPylRSn we refer to it as S⁺-DhPylRS and if it is expressed in absence of DhPylRSn we use S^Δ-DhPylRS, the cognate tRNA^Pyl is referred to as S-DhtRNA^Pyl. We refer to all distinct tRNA mutants by adding a descriptive suffix after the suffix “^Pyl”. Furthermore, we refer to cognate interactions as those between PylRS/tRNA^Pyl pairs of the same class, and noncognate interactions as those between PylRS/tRNA^Pyl pairs of distinct classes. Finally, when referring to the nucleotide number in pyl tRNAs we use the numbering system outlined (Figure 3c).

Figure 3.

PylRS/tRNA^Pylpair nomenclature. a, Division of PylRS/tRNA^Pyl pairs into three groups and five classes. The groups are defined by the architecture of the PylRS enzyme. The + N group contains PylRS enzymes where PylRSn and PylRSc are covalently connected by a flexible linker, the Δgroup is comprised of PylRS enzymes lacking PylRSn in their host genome, and the sN group is composed of PylRS enzymes where PylRSn and PylRSc are produced in trans from distinct genes. The classes (N, A, B, C, and S) represent a finer subdivision of the pyl system based on sequence identity clustering of PylRS-, and tRNA^Pyl sequences and the aminoacylation specificity of the PylRS/tRNA^Pyl pairs with respect to each other. b, Nomenclature used for PylRS enzymes and pyl tRNAs in this review. The tRNA nomenclature is in line with International Union of Pure and Applied Chemistry (IUPAC) rules and extended to include pyl class information, as well as tRNA^Pyl variant information. We note that when referring to tRNA^Pyl in plural, we write pyl tRNAs, in accordance with IUPAC rules. c, Numbering of residues in tRNA^Pyl. The numbering is in line with the general convention for tRNAs according to Sprinzl et al.⁴³ However, some common nucleotides are missing in pyl tRNAs (9, 16, 17, 18), and some unusual nucleotides are present (25a, 25b, 42a). Nucleotides in dark gray are present in all described pyl tRNAs, nucleotides in light gray are present in some pyl tRNAs.

3.2. The C-Terminal Catalytic Domain of PylRS Enzymes

The C-terminal domain of N⁺-PylRS and S⁺-PylRS contains all the sequence motifs that define the catalytic domains of class IIc aaRS enzymes. Insight into the structure of PylRS resulted from the X-ray crystal structure of the C-terminal catalytic domain of N⁺-MmPylRS (N-MmPylRSc270, residues 185–454).^44,45 As expected from sequence homology, N-MmPylRSc270 closely resembles the structure of other class II synthetases in which a β-sheet core is surrounded by several long helices.⁴⁶ All three sequence motifs of class II synthetases are present; sequence motif 1 is responsible for the dimerization of the N⁺-MmPylRS and sequence motifs 2 and 3 build the nucleotide interface.⁴⁷ The apo structure of S-DhPylRSc is very similar to that of the C-terminal domain of the archaeal PylRS.⁴⁸ Structures of additional PylRS systems have revealed broadly similar active site structures.^7,9,49−59 Phylogenetic and structural analysis suggests that PylRS enzymes arose from PheRS – by gene duplication and neo-functionalization–before the last universal common ancestor (LUCA).^42,45

The crystal structures of the catalytic domain of N⁺-MmPylRS (N-MmPylRSc270) alone, and in complex with (i) the nonhydrolyzable adenosine triphosphate (ATP) analog adenylyl imidodiphosphate (AMP-PNP), (ii) the Pyl analog N^ε-((cyclopentyloxy)carbonyl)-L-lysine (CycK) with ATP and, (iii) Pyl-AMP, provided structural insights into substrate-binding in the catalytic domain.^44,45 A notable feature of these structures is the deep hydrophobic pocket in which the amino acid substrate is bound.

A comparison of the crystal structures of the catalytic domain of N⁺-MmPylRS in its apo form and in complex with AMP-PNP, suggests that ATP binding leads to substantial changes in the architecture of N⁺-MmPylRS.⁵⁹ In contrast, the structure of the backbone of N-MmPylRSc only changes minimally between the three structures with bound substrates. Two direct hydrogen bonds are formed with the amino acid substrates: R330 forms a hydrogen bond with the primary (backbone) carbonyl, and N346 forms a hydrogen bond with the secondary (side chain) carbonyl. Besides these two hydrogen bonds, the positioning of the substrate is mainly mediated through nondirected van der Waals interactions (Figure 4a).^45,59 The relaxed substrate recognition could stem from the absence of Pyl-like metabolites in methanogenic bacteria and archaea, limiting the evolutionary pressure for a tight active site fit of the substrate. In accordance with this hypothesis, PylRS enzymes do not contain a substrate editing domain.

Figure 4.

Pyl PNP-AMP and Pyl-AMP binding in the active site of N⁺-MmPylRS. a, Binding of Pyl and PNP-AMP in the deep hydrophobic pocket of the active site of N-MmPylRSc (PDB 2ZCE).⁵⁹ Direct hydrogen bonds are formed between the primary (backbone) carbonyl of Pyl and R330 as well as the secondary carbonyl (side chain) of Pyl and N346. The α-amine forms a hydrogen bond with a coordinated water molecule. Pyl and the interaction partners are shown as sticks representation, N-MmPylRSc is shown as a transparent electrostatic surface (red negatively charged, white noncharged, blue positively charged). b, Recognition of Pyl-AMP by N-MmPylRSc (2Q7H).⁴⁵Pyl-AMP forms the same direct hydrogen bonding network with N-MmPylRSc as observed for Pyl in the structure shown in panel a with an additional hydrogen bond being formed between the α-amine of Pyl and Y384. Y384 is part of a flexible loop which closes the active site and was not visible in the crystal structure depicted in panel a. Pyl-AMP and the interacting amino acids within PylRS are shown in stick representation, PylRS is shown as a transparent electrostatic surface (red negatively charged, white noncharged, blue positively charged).

Interestingly, the α-amine of the amino acid substrate is not involved in a direct hydrogen-bonding network with essential residues in N⁺-MmPylRS⁺. Although the hydroxy group of Y384 is involved in a hydrogen bond with the α-amine in the crystal structure of the complex of N-MmPylRSc270 and Pyl-AMP (Figure 4b),⁴⁵ a Y384F mutation did not impede the reactivity of N⁺-MmPylRS.⁵⁸ On the contrary, the Y384F mutation enhanced N⁺-MmPylRS activity.^58,60 The absence of α-amine binding by PylRS is in stark contrast to its structurally closely related homologue PheRS, which forms a tight hydrogen-bonding network with the α-amine.⁶¹⁶²

3.3. The N-Terminal Domain of PylRS Binds tRNA^Pyl

When discovered, the N-terminal domain of PylRS bore no significant sequence similarity to known-families of RNA-binding proteins and its function was unclear.⁶³ Early studies suggested that the archaeal N⁺-MbPylRS enzyme required the N-terminal domain to support amber suppression activity in vivo, as several N-terminal truncations of N⁺-MbPylRS did not produce measurable amber suppression when paired with its tRNA^Pyl in E. coli.⁶⁴ However, it is unclear whether increasing the levels of N-MbPylRSc would have led to in vivo activity in these assays. Interestingly, N-MmPylRSc retained some in vitro acylation activity.⁵⁹

S^Δ-DhPylRS (the C-terminal domain of S⁺-DhPylRS) also displayed some activity in vitro in the absence of its N-terminal domain (S-DhPylRSn). Initial observations pointed toward low activity of S⁺-DhPylRS in vivo in E. coli when tested with the large Pyl analog CycK (activity was only detectable with a highly sensitive genetic assay).⁵⁷ However, when assayed with the smaller Pyl analog, N^ε-((allyloxy)carbonyl)-L-lysine (AllocK), S^Δ-DhPylRS alone led to enhanced amber suppression in E. coli. The preference of S⁺-DhPylRS for smaller Pyl analogues is consistent with it having a smaller substrate binding pocket, than archaeal N⁺-PylRS enzymes.⁶⁵

The N-terminal domains of archaeal and bacterial PylRS enzymes bind tRNA^Pyl, and the affinity of S-DhPylRSn for S-DhtRNA^Pyl was at least an order of magnitude higher than the affinity of S^Δ-DhPylRS for S-DhtRNA^Pyl.^63,64 A mutational screen of S-DhtRNA^Pyl suggested that S-DhPylRSn bound the D- and T-stem as well as the T- and variable loops of S-DhtRNA^Pyl_.⁶³ Our understanding of N-terminal domain binding has been augmented by a crystal structure of the N-terminal domain of N⁺-MmPylRS bound to N-MmtRNA^Pyl (see Section 3.4).⁶⁰ Additional experiments suggested that the N-terminal domain of PylRS system can contribute to tRNA^Pyl specificity as well as affinity.⁴⁰

These studies established PylRSn as a previously unknown RNA binding domain that increases the affinity, and may alter the specificity, of certain PylRS systems for tRNA^Pyl. It remains unclear whether there is allostery between the C-terminal domain and N-terminal domain in the catalytic cycle for aminoacylation.

3.4. PylRS Form Unique Interfaces with tRNA^Pyl

N-MmtRNA^Pyl and S-DhtRNA^Pyl form canonical cloverleaf secondary structures. They also form L-shaped tertiary conformations and, like canonical tRNAs, are likely to interact with elongation factor thermo unstable (EF-Tu) in the translation cycle.^63,66,67 As with all tRNAs, the tertiary core of tRNA^Pyl is formed by the interaction between nucleotides, in the D- and T- loops (Figure 5a). However, several unique features–mainly a short variable and D- loop (three and five nucleotides, respectively) – result in the tRNA^Pyl core being exceptionally compact.⁵⁷ Furthermore, unusually few nucleotide modifications (4-thiouridine at position 8 and 1-methyl-pseudouridine at position 50) have been identified in tRNA^Pyl to date.⁶⁸

Figure 5.

The PylRS:tRNA^Pylbinding interface. a, Crystal structure of N-MmPylRSn bound to N-MmtRNA^Pyl (PDB 5UD5).⁶⁰ N-MmPylRSn interacts with the variable loop (dark blue), D-stem (cyan), T-loop (purple), and T-stem (light blue) of N-MmtRNA^Pyl. b, Crystal structure of S^Δ-DhPylRS in complex with S-DhtRNA^Pyl (PDB 2ZNI).⁵⁷ S^Δ-DhPylRS forms a dimer in the crystal structure and in vivo where each protomer (colored in two shades of green) predominantly interacts with one tRNA^Pyl, while forming some interactions with the second tRNA^Pyl.

S^Δ-DhPylRS forms a dimer in the crystal structure and in solution.⁵⁷ In the crystal structure, the asymmetric unit contains two S^Δ-DhPylRS and two S-DhtRNA^Pyl molecules (Figure 5b). Although each tRNA^Pyl predominantly interacts with one protomer, the synthetase dimer forms a concave structure that complements the shape of the acceptor helix and directs the 3′ end of the tRNA to the catalytic site.

As with all class II synthetases,⁶⁹ S^Δ-DhPylRS approaches S-DhtRNA^Pyl from the major groove. However, it has evolved several distinctive features to recognize the unusual shape of its cognate S-DhtRNA^Pyl. For instance, G8 of S-DhtRNA^Pyl, which is idiosyncratically flipped out of the tRNA body, is accommodated by a unique cation-π interaction from R140 of the synthetase. Moreover, the tRNA-binding domain 1, which is exclusive to PylRS enzymes, makes conserved specific interactions with the compact tertiary core of tRNA^Pyl. As a result, tRNAs from other iso-acceptor classes, which possess bulkier cores, are unlikely to be sterically compatible with PylRS, which helps explain the orthogonality of the PylRS/tRNA^Pyl pair.

The crystallization of the N-terminal domain of N⁺-MmPylRS resulted in a more complete understanding of the structure of the PylRS-tRNA^Pyl complex (Figure 5a).⁶⁰ The N-terminal domain folds into a compact globular protein which binds one zinc ion. The fold is unique among known synthetases. The N-terminal domain tightly fits into the concave surface generated by the T-loop and the variable loop. The compact fit prevents canonical tRNAs with larger variable loops from being efficiently recognized. The N-terminal domain and C-terminal domain surround the tRNA^Pyl and form the largest interaction surface known for an aaRS-tRNA complex.⁶⁰

3.5. The Active Site of PylRS Accepts Pyl Analogs

Early observations demonstrated that N⁺-MbPylRS can activate a variety of different Pyl analogs;^70,71 these experiments were performed with analogs, as access to Pyl was limited due to its challenging chemical synthesis.⁷²⁻⁷⁴ Later, it was shown that the active site of S^Δ-DhPylRS is also promiscuous, but restricted to smaller Pyl analogs.⁶⁵ These observations are in line with S^Δ-DhPylRS having a smaller active site, as L309 in N⁺-MmPylRS corresponds to W139 in S^Δ-DhPylRS.

The promiscuity of PylRS enzymes is essentially limited to changes in the substituent at the secondary carbonyl.⁵⁸ The relaxed fit of Pyl in the large hydrophobic pocket, in which the side chain binds mainly through nondirected van der Waals interactions, is consistent with the observed promiscuity. Thus, PylRS predominantly recognizes amino acid substrates composed of a secondary carbonyl (which may engage in hydrogen bonding with N346⁴⁵) linked to large hydrophobic side chains.

3.6. PylRS Does Not Recognize the Anticodon of tRNA^Pyl

Biochemical characterization of S⁺-DhPylRS/S-DhtRNA^Pyl demonstrated that the anticodon is not a recognition element of PylRS enzymes.⁶⁷ The anticodon of S-DhtRNA^Pyl could be switched without significantly affecting S-DhPylRSn binding or S^Δ-DhPylRS acylation efficiencies. Numerous experiments demonstrated that PylRS enzymes do not require conserved nucleotides in the anticodon stem of their pyl tRNAs for efficient aminoacylation.^{57,59,60,63,75−80} Structural analysis of the PylRS-tRNA^Pyl complex substantiated previous observations that PylRS does not interact with anticodon stem-loop.^59,75

While mutations in the anticodon stem loop have little effect on aminoacylation, ribosomal translation in E. coli is more efficient with tRNAs bearing specific nucleotides adjacent to the anticodon;⁸¹ therefore, translational readthrough of the amber codon in E. coli is sensitive to the identity of nucleotides in the anticodon stem and loop. This explains why mutagenesis of nucleotides adjacent to the anticodon (U33 and A37) in N⁺-MbtRNA^Pyl decreased translational readthrough of amber codons in E. coli.⁷⁵ Similarly, while some archaeal pyl tRNAs have C37 in their natural anticodon stem loop,^35,38−40 the C37A mutants of these pyl tRNAs led to more efficient amber suppression than the native sequence in E. coli.³⁹

4. Genetic Code Expansion with PylRS/tRNA^Pyl Pairs

A number of aaRS/tRNA pairs have been developed for genetic code expansion.^{20,22,35,38−40,82−92} However, the PylRS/tRNA^Pyl pair has several properties that make it ideal for genetic code expansion: (1) The PylRS/tRNA^Pyl pair is orthogonal in E. coli (Section 4.1),⁸² (2) the pair does not use canonical amino acids and can be engineered to accept a wide-range of new substrates (Section 4.2), (3) the pair is a natural amber suppressor, (4) the anticodon of tRNA^Pyl can be altered to decode diverse codons, and (5) the PylRS/tRNA^Pyl pairs tested are orthogonal in all kingdoms of life,^{3,20,82,93−108} such that variants evolved for new substrates in E. coli can be transplanted to other prokaryotes, eukaryotic cells, plants, and animals (Section 4.3). A number of innovative approaches have been developed to increase the efficiency of ncAA incorporation with PylRS/tRNA^Pyl systems in both prokaryotic and eukaryotic systems (Section 4.4). Because of their unique advantages PylRS/tRNA^Pyl pairs have rapidly become the most widely used systems for genetic code expansion and reprogramming.

4.1. PylRS/tRNA^Pyl Pairs Are Orthogonal in E. coli

Orthogonal aaRS enzymes aminoacylate their cognate (orthogonal) tRNA but minimally acylate other tRNAs in the cell of interest. Similarly orthogonal tRNAs are substrates for their cognate (orthogonal) aaRS but not efficient substrates for any endogenous synthetases in the cell of interest. An orthogonal pair is composed of an orthogonal aaRS and orthogonal tRNA. The N⁺-MbPylRS/N-MbtRNA^Pyl pair (and the analogous Mm pair) are orthogonal in E. coli;⁸² N-MbtRNA^Pyl is minimally acylated by endogenous synthetases in this host.^22,68,82 N⁺-MbPylRS and its active site variants direct the incorporation of their substrates in response to the amber codon without directing measurable misincorporation of their substrates, in competition with canonical amino acids, at sense codons.⁸²

4.2. Engineering PylRS Enzymes for the Incorporation of ncAAs

With an orthogonal aaRS/tRNA pair in hand, the next challenge was to engineer the synthetase such that it uses a desired ncAA and no canonical amino acids. The most common way to engineer PylRS to use ncAAs uses double-sieve selection on the amber suppression activity of PylRS/tRNA^Pyl_CUA pairs (Section 4.2.1). Recent work has also explored the use of chimeras between PylRS and other synthetases as a starting point for generating enzymes for ncAA incorporation (Section 4.2.2).

4.2.1. Selections for PylRS Variants That Direct ncAAs into Proteins

The most established and general method to alter the specificity of orthogonal aaRS enzymes to selectively incorporate a ncAA of interest relies on double-sieve selections on a library of synthetase mutants; these libraries commonly use saturation mutagenesis at five or more positions in the region of the gene corresponding to the active site of an enzyme, though small intelligent libraries that mutate several positions to a smaller subset of codon possibilities have also proved useful.¹⁻⁶ Double-sieve selections for ncAA incorporation were first successfully demonstrated for the evolution of Methanococcus jannaschii tyrosyl–tRNA synthetase (MjTyrRS) to direct the incorporation of the photo-cross-linker para-benzoyl-L-phenylalanine (BpA) in response to the amber codon in E. coli.¹⁰⁹ In this approach (Figure 6), synthetase variants that load an amino acid (the ncAA or a canonical amino acid) are selected in a positive selection step; this step uses a positive selection marker (e.g., chloramphenicol acetyl transferase (cat)) containing an amber codon at a permissive site in its gene, and selects for synthetases that acylate their cognate tRNA and enable the production of the positive selection marker in the presence of the added ncAA. Synthetase genes that survive the positive selection are then subjected to a negative selection step; this step uses a negative selection marker (e.g., barnase) containing one or more amber codons at permissive sites in its gene, and selects against synthetases that acylate their cognate tRNA and enable the production of the negative selection marker in the absence of the added ncAA.⁸² Double-sieve selections have been extended to select for ncAA specific synthetases in eukaryotic cells.¹¹⁰

Figure 6.

Double-sieve selection strategy for directed evolution of ncAA specificity in orthogonal aaRS enzymes. Aminoacyl-tRNA synthetase libraries are first submitted to a round of positive selection in the presence of the target ncAA. In this step, the acylation of a suppressor tRNA and ribosomal translation through an amber codon in a positive selection marker mRNA (frequently chloramphenicol acetyltransferase) is linked to cell survival. Next, surviving library members are submitted to a negative selection step in the absence of the ncAA, where acylation of a suppressor tRNA and ribosomal translation through an amber codon in a negative selection marker (frequently barnase) is linked to cell death. Aminoacyl-tRNA synthetase variants that selectively charge the target ncAA, and no canonical amino acids, onto their cognate tRNA survive both steps of selection. For this selection approach to work, ncAAs must function with the ribosome and other translation factors. Additional rounds of selection can be performed. Adapted with permission from Chin et al.³ – copyright © 2014 Annual Reviews.

In some cases, positive and negative screens, using fluorescence-based readouts have been used in place of selections,^111−116 and these approaches have been extended to in vitro screening in liposomes.¹¹⁷ Negative screens have the advantage of allowing the experimenter to more precisely gate the amount of synthetase activity permissible in the absence of the ncAA (something that can in principle be achieved by tunable negative selections);¹¹⁸ this is potentially advantageous for identifying the most active and sufficiently selective synthetases,¹¹⁹ since the fidelity of the genetic code is in part controlled by competition.^120,121 However, screening approaches generally come at the cost of lower throughput.

The double-sieve selection was adapted to select N⁺-MbPylRS/N-MbtRNA^Pyl variants that direct the cotranslational incorporation of N^ε-acetyl-L-lysine (AcK), a key post translational modification (PTM), in response to the amber codon in E. coli.⁸² This demonstrated that PylRS could be evolved in the laboratory to incorporate ncAAs. Over the past 16 years PylRS/tRNA^Pyl pairs have been evolved and engineered⁵⁸ to incorporate hundreds of ncAAs with numerous applications (Section 6), including numerous and diverse acylated lysine derivatives.

Furthermore, the active site of PylRS enzymes can easily be remodeled to accommodate residues with side chains containing aromatic groups.⁵⁶ A number of phenylalanine derivatives are substrates for engineered PylRS enzymes, and PylRS may also direct the incorporation of larger aromatic ring systems.^{19,120,122−132} PylRS enzymes for histidine,^19,133,134 tyrosine,^135−138 or tryptophan^139−141 analogs have also been developed. We note that a manually curated databank of PylRS active site variants and the ncAAs that these variants have been reported to incorporate (as measured by intact MS verification of the modified protein) has been generated.¹⁴²

Parallel positive selections coupled to deep sequencing, developed to generate a phosphothreonine specific aaRS from the phosphoseryl-tRNA synthetase (SepRS) of Methanococcus maripaludis, have also been used to evolve PylRS enzymes in proof-of-principle experiments in E. coli.¹⁴³ In this method, two or more positive cat selections are run in parallel in the presence and absence of the desired ncAA. All surviving colonies are subsequently isolated and the gene pool is subjected to next-generation sequencing (NGS). Analysis of the selected synthetase sequences leads to the identification of aaRS variants that are enriched in the plus ncAA sample with respect to the minus ncAA sample.

Until recently, all selections for synthetases for ncAA incorporation relied on protein synthesis-based read-outs and consumed substantial quantities of ncAA. Recent work, aimed at incorporating ncMs that may not function efficiently in protein synthesis or be efficient ribosomal substrates, developed tRNA display¹⁹ – a direct selection for synthetases that aminoacylate their cognate tRNAs with ncMs, whether or not the resulting acylated tRNAs function in translation. tRNA display has been used to evolve N⁺-MmPylRS variants for eight ncAAs and eight ncMs, including six β-amino acids, one α,α-disubstituted-amino acid and one β-hydroxy acid. This approach is discussed in detail in Section 7.

4.2.2. PylRS Chimeras for ncAA Incorporation

Recent work has attempted to leverage the unique characteristics of the PylRS/tRNA^Pyl pair by generating chimeras that combine the tRNA binding abilities of the N-terminal domain of PylRS enzymes with the substrate scope of the catalytic domains of aaRS/tRNA pairs for canonical amino acids (Figure 7).¹⁴⁴ In a proof-of-concept experiment, the catalytic domain of E. coli histidyl-tRNA synthetase (EcHisRS) was linked through a flexible linker to N-MbPylRSn. The architecture of EcHisRS, with a distinct N- and C-terminal domain, in which the N-terminal domain performs catalysis and the C-terminal domain is responsible for anticodon recognition, facilitated the design of the fusion protein. The catalytic domain of EcHisRS predominantly binds the acceptor arm of its cognate tRNA and N-MbPylRSn mainly interacts with the D- and T-stem and the variable and T-loop of tRNA^Pyl. Therefore, tRNA chimeras composed of the body of N⁺-MbtRNA^Pyl and the acceptor arm of EctRNA^His were generated. A further mutated version of the pair led to 60% amber suppression efficiency for incorporating histidine when compared to wild type (wt) green fluorescent protein (GFP) production.^144,145

Figure 7.

Generation of chimeric aminoacyl-tRNA synthetase/tRNA (chRS/chtRNA) pairs for genetic code expansion. The tRNA binding function of PylRS enzymes can be coupled to the catalytic domain of certain canonical aaRS enzymes forming orthogonal chRS/chtRNA pairs. Like N⁺-MmPylRS, E. coli histidyl-tRNA synthetase (EcHisRS) has two distinct domains connected by a flexible linker. One domain is responsible for tRNA binding (C-terminal domain - HisRSc), and one for the catalytic activity (N-terminal domain - HisRSn). The catalytic domain (CD) predominantly interacts with the acceptor stem of EctRNA^His and the tRNA binding domain (TBD) predominantly interacts with the anticodon stem and loop of EctRNA^His. A chRS is generated through the fusion of the TBD of N⁺-MmPylRS (which predominantly interacts with the T-, and D-stem, as well as the T- and variable loop of N-MmtRNA^Pyl) with the CD of EcHisRS. The combination of the chRS with the engineered chtRNA, where the acceptor stem in N-MmtRNA^Pyl was replaced with the acceptor stem of EctRNA^His, resulted in an orthogonal chRS/chtRNA pair. This pair combined the aminoacylation specificity of EcHisRS with the tRNA recognition and orthogonality of N⁺-MmPylRS and could be used in prokaryotic and mammalian cells. A chimeric PheRS/tRNA^Phe pair was also engineered. Adapted from Ding et al.¹⁴⁴ – copyright © The Author(s) 2020 CCBY http://creativecommons.org/licenses/by/4.0/.

As aaRS/tRNA pairs for other canonical amino acids do not commonly have easily separatable N- and C-terminal domains, with distinct catalytic or tRNA binding functions, generalizing this approach to produce efficient chimeras for other chemotypes has proven challenging. Nonetheless, the chimera between the N-MbPylRSn and the catalytic domain from human mitochondrial PheRS, led to an orthogonal chimeric pair (chPheRS/chtRNA^Phe) with 6% amber suppression efficiency when compared to wt GFP production. This pair was extensively engineered and evolved using double-sieve selections, and versions of the pair were developed to incorporate a number of phenylalanine, tyrosine, and tryptophan analogs.¹⁴⁴ Extensive engineering and evolution of this system enabled the efficient and selective incorporation of p-azido-L-phenylalanine (p-AzF), and the authors generated an E. coli strain that was synthetically auxotrophic for this ncAA.¹⁴⁵ The highly engineered chPheRS/chtRNA^Phe was also used for the genetic encoding of tryptophan analogues that can be deprotected in vivo to reveal tryptophan residues.¹⁴⁶

4.3. Transplanting PylRS Systems for ncAA Incorporation to Other Organisms

PylRS/tRNA^Pyl pairs have been used to expand the genetic code across all domains of life and PylRS variants developed in E. coli, where directed evolution and engineering are most straightforward, have now been transplanted to a wide range of other organisms (Figure 8). The underpinnings of this approach have been extensively reviewed,³ and this section provides a brief summary and update.

Figure 8.

PylRS/tRNA^Pylpairs are orthogonal and have been used for ncAA incorporation across all domains of life. PylRS/tRNA^Pyl pairs can be engineered for ncAA specificity in E. coli cells, using directed evolution approaches, and then used for genetic code expansion in diverse host organisms including bacteria, archaea, eukaryotic cells, a plant species, and several animals.

4.3.1. Genetic Code Expansion in Other Prokaryotes

PylRS/tRNA^Pyl pairs are orthogonal, and have been used for genetic code expansion in diverse prokaryotes including Rhodobacter sphaeroides,⁹⁹Neisseria meningitidis,¹⁰⁰cyanobacteria,^101,147Pseudomonas aeruginosa,¹⁰²Bacillus subtilis,¹⁴⁸Lactococcus lactis,¹⁰³Streptomyces albus,¹⁰⁴Shigella and Salmonella.¹⁰⁵

4.3.2. Genetic Code Expansion in Eukaryotic Cells

PylRS/tRNA^Pyl pairs are orthogonal in eukaryotic cells. However, the expression, processing, and nuclear export of heterologous tRNAs poses specific challenges for genetic code expansion in eukaryotic cells. Eukaryotic tRNA genes contain internal A-and B-box elements that are required for their RNA polymerase III-dependent transcription. Genes encoding tRNA^Pyl lack these A- and B-box elements. Therefore, extragenic RNA polymerase III promoters, which contain their own A- and B-box sequences, have been used to transcribe tRNA^Pyl.^93,149 This approach has the additional advantage of disentangling the sequence of the tRNA from its ability to be transcribed, enabling the sequence of the tRNA to be optimized for intrinsic function. Although the addition of the 3′CCA end is part of tRNA maturation in eukaryotes,¹⁵⁰ the tRNA^Pyl gene is usually introduced with the 3′CCA end included.^93,149

In Saccharomyces cerevisiae (S. cerevisiae), placing eukaryotic tRNA genes with their own A- and B-box sequences upstream of the N-MmtRNA^Pyl gene, and expressing N⁺-MmPylRS from standard promoters enables the creation of functional genetic code expansion systems. An early system used a human leucine tRNA gene as a promoter for the N-MmtRNA^Pyl gene, and reported preliminary growth phenotypes consistent with weak ncAA dependent amber suppression.⁹³ Using the yeast tRNA^Arg_UCU gene as a promoter for the N-MmtRNA^Pyl gene led to high level expression of MmtRNA^Pyl, this system was used to demonstrate that N⁺-MmPylRS and N-MmtRNA^Pyl (but not N-MbtRNA^Pyl) are orthogonal in yeast. Using this system five different ncAAs were site-specifically incorporated into recombinant proteins, and ncAA incorporation was characterized by MS and MS/MS.¹⁵¹ Recent work has demonstrated that the SNR52 promoter (which also contains A- and B-box elements) can be used to express Methanomethylophilus alvus (alv) A-alvtRNA^Pyl in S. cerevisiae; this enabled ncAA incorporation with the A^Δ-alvPylRS/A-alvtRNA^Pyl pair in S. cerevisiae;¹⁵² these experiments demonstrated how the development of orthogonal PylRS/tRNA^Pyl pairs in E. coli can lead to rapid advances in other organisms.

PylRS/tRNA^Pyl pairs are also orthogonal in mammalian cells, where the expression of tRNA^Pyl is commonly driven by a U6 promoter (which contains A- and B-box sequences) with or without a cytomegalovirus (CMV) enhancer.^{91,93,144,145,153−156}

A systematic optimization of ncAA incorporation efficiency with the N⁺-MmPylRS/N-MmtRNA^Pyl pair in mammalian cells, revealed that tRNA^Pyl concentration is a key factor in increasing the amount of ncAA containing recombinant protein produced.¹⁴⁹ N-MmtRNA^Pyl levels were optimized by using eight copies of the N-MmtRNA^Pyl gene, each under a U6 promoter; this led to substantially more tRNA production than a CMV enhancer U6 promoter system, and the increase in tRNA levels was correlated with a ten to 20-fold increase in ncAA incorporation efficiency.

Transient transfection of PylRS/tRNA^Pyl genes in mammalian cells leads to a heterogeneous population of cells with great variability in the levels of PylRS and tRNA^Pyl between cells. This results in variability in stop codon read through levels, and therefore variability in ncAA incorporation efficiency, between cells.¹⁰⁶ By using a PiggyBac system, eight copies of a gene encoding tRNA^Pyl, a single copy of a gene encoding PylRS, and genes of interest containing amber stop codons at the desired positions, were introduced into the genomes of diverse mammalian cell lines.¹⁰⁶ Stable integration of the genes encoding a PylRS/tRNA^Pyl pair variant permitted the site-specific installation of preacylated lysine residues in histones, providing an orthogonal means to study the consequence of a PTM at a specific site in a protein without the pleotropic effects that may result from manipulating acetyl-transferases or deacetylases. These experiments also defined genes that are up or down regulated as a result of amber suppression in mouse embryonic stem cells. Other stable cell lines for PylRS/tRNA^Pyl pairs have since been developed,^157−159 and PylRS/tRNA^Pyl pairs have also been used to incorporate ncAA into viruses,^160−162 insect cells¹⁶³ and brain organoids.¹⁶⁴

PylRS/tRNA^Pyl pairs have also been used to expand the genetic code of plants and animals. The first genetic code expansion in a multicellular system was demonstrated in the nematode Caenorhabditis elegans (C. elegans).^95,98,165 The orthogonality of the PylRS/tRNA^Pyl pair, and ncAA incorporation, was further demonstrated in several model organisms, including the plant Arabidopsis thaliana,⁹⁷ and animals: Drosophila melanogaster,^94,166,167Mus musculus,^108,168 and Danio rerio.^107,169

4.4. Improving the Efficiency of ncAA Incorporation with PylRS/tRNA Pairs

PylRS/tRNA^Pyl variants that enable ncAAs incorporation have provided a starting point for work aimed at improving the efficiency of ncAA incorporation through further engineering of PylRS and tRNA^Pyl. These approaches complement strategies to enhance ncAA incorporation at a target codon by minimizing competition with translation factors that otherwise decode that codon, or improving strains for ncAA amino acid incorporation.^{77,149,152,170−173} Here we focus on strategies for engineering the intrinsic properties of PylRS/tRNA^Pyl systems, rather than the expression level of the pair, to increase ncAA incorporation, and on approaches where PylRS/tRNA^Pyl systems were central to the development of strategies for improving ncAA incorporation.

4.4.1. Improving the Efficiency of ncAA Incorporation in E. coli

A continuous evolution approach, based on phage assisted continuous evolution (PACE),¹⁷⁴ was investigated to improve the activity of a chimeric PylRS (N⁺-chPylRS - residues 1–149 of N⁺-MbPylRS fused to residues 185–454 of N⁺-MmPylRS) with N^ε-(tert-butoxycarbonyl)-L-lysine (BocK). This approach selected for the ability of a N⁺-chPylRS/N-tRNA^Pyl pair to read through amber stop codons in the M13 pIII gene (or to read through amber stop codons in a T7 RNA polymerase gene, in a less stringent system where T7 RNA polymerase transcribed the pIII gene), in mutagenic E. coli provided with BocK. The pIII protein is required for generating infective phage particles. Active N⁺-chPylRS/N-tRNA^Pyl pairs led to production of infective phage carrying the corresponding N⁺-chPylRS gene, but inactive N⁺-chPylRS/N-tRNA^Pyl pairs did not. The phage carrying active N⁺-chPylRS, and bearing the pIII protein could infect fresh mutagenic cells where they were subject to further mutations in N⁺-chPylRS. This approach led to four consensus mutations in the N-terminal domain of the chimera (V32I, T56P, H62Y, and A100E). The combination of these mutations in N⁺-chPylRS led to a 1.5-fold increase in the production of GFP from a gene containing an amber stop codon (from 40% to 60% of a wt GFP control) and an about 4-fold increase in GFP from a gene containing three amber stop codons. Transfer of the mutations to PylRS variants for other ncAAs also led to increases in incorporation efficiency. Experiments that mutated PylRS, its N-terminal domain, or the linker connecting the N- and C-terminal domains, led to similar increases in efficiency.^{111,117,175−177} N-terminal solubility tags have also been used to increase the amber suppression efficiency of the N+-MbPylRS/MbtRNA^Pyl pair several fold.¹⁷⁸ Additionally, a mutation in the N-terminal domain of N⁺-MmPylRS may increase stability to proteolysis and enhance the performance of the enzyme.¹⁷⁹

Intriguingly, PACE also generated split N⁺-chPylRS enzymes, where mutations generated a stop codon separating N- and C-terminal domains into two fragments (an in sequence AGT codon enabled the translation initiation of the C-terminal domain). Maximal amber suppression activity required the presence of both domains. Continuous evolution approaches have the potential to improve the ncAA dependent activity of PylRS/tRNA^Pyl pairs.^180−185

One strategy to improve ncAA incorporation with PylRS/tRNA^Pyl pairs has focused on the mutation of nucleotides in the T-stem and acceptor stem of N-MmtRNA^Pyl (49:65, 50:64, 6:67 and 7:66). These nucleotide positions are known to be important for recognition of E. coli tRNAs by endogenous EF-Tu.¹⁸⁶ The variant tRNA^Pyl, named N-MmtRNA^Pyl-opt led to a 3-fold increase in AcK incorporation in response to one amber codon and a 5-fold increase in incorporation in response to two amber codons, when compared to the parent tRNA^Pyl. Although N-MmtRNA^Pyl-opt improved the incorporation of multiple chemically distinct ncAAs, the extent of the increase in yield varied substantially; this observation is consistent with the thermodynamic compensation between amino acid and tRNA binding to EF-Tu¹⁸⁷ and suggests that tRNA^Pyl may need to be independently optimized for each ncAA.

4.4.2. Improving the Efficiency and Specificity of ncAA Incorporation with PylRS/tRNA^Pyl Pairs in Eukaryotic Systems

Archaeal pyl tRNAs have secondary structures which are highly divergent from canonical mammalian tRNAs, but are shared by mitochondrial tRNA^Ser_UGA. Researchers attempted to improve overall tRNA^Pyl activity in mammalian cells by substituting single bases or base pairs in N-MmtRNA^Pyl with the corresponding bases from human tRNAs. In a series of experiments, they inserted tRNA^Pyl recognition elements for PylRS enzymes from group +N and sN into Bos taurus (Bt)tRNA^Ser_UGA and assessed the amber suppression activity of all hybrid tRNAs in mammalian cells (in prior work, the sequences recognized by N⁺-MbPylRS within N-MbtRNA^Pyl had been transplanted into the mitochondrial tRNA from BttRNA^Ser_UGA to generate chimeras, some of which were active in E. coli(75)). Mutants bearing the highest number of tRNA^Pyl bases generally performed best. One chimera based on N-MmtRNA^Pyl - termed N-MmtRNA^Pyl-M15 - led to a roughly 2.5-fold improvement of BocK incorporation when paired with a N⁺-MbPylRS variant in mammalian cells.¹⁸⁸ Interestingly, the most active mutants in mammalian cells acquired a canonical B-box in the T arm, which is absent in native tRNA^Pyl. The study confirmed that tRNA^Pyl engineering can increase the efficiency of ncAA incorporation in eukaryotic cells. However, the increase in activity is limited to incorporating some ncAAs, and maximal incorporation efficiency may require a new tRNA^Pyl for each ncAA.¹⁸⁸

Since tRNA^Pyl needs to be transcribed, processed, modified, and exported from the nucleus and function with the endogenous translational machinery, it is not straightforward to predict what sequence changes may lead to optimized ncAA incorporation in mammalian cells. These considerations suggested that there may be value in developing methods for the directed evolution of tRNAs in these cells. “Virus-assisted directed evolution of tRNAs” (VADER) has been developed for directed evolution of tRNAs that direct more efficient ncAA incorporation in mammalian cells (Figure 9).¹⁸⁹

Figure 9.

Virus-assisted directed evolution of tRNAs (VADER) in mammalian cells: tRNA^Pyl libraries are encoded in the DNA of adeno-associated viruses 2 (AAV2), such that each virus (hexagon) only carries one tRNA variant. The replication of the virus is coupled to amber suppression. N⁺-MmPylRS dependent amber suppression leads to the incorporation of an azide functionality on the surface of the virus. Viruses harboring a selective and active N-MmtRNA^Pyl can be isolated on streptavidin beads by bio-orthogonal labeling and either submitted to additional rounds of evolution, or further characterization by single colony sequencing or NGS. Adapted from Jewel et al.¹⁹⁰ – copyright © The Author(s) 2020 CCBY-NC-ND https://creativecommons.org/licenses/by-nc-nd/4.0/.

In VADER, tRNA gene libraries were encoded in the DNA of adeno-associated viruses 2 (AAV2), such that one virus carried one tRNA genotype, and the replication of the virus was rendered dependent on the suppression of amber codons introduced into the gene for a virus capsid protein. Mammalian cells were infected with the virus, and transfected with the gene for a N⁺-MbPylRS variant that–when provided with a cognate N-MmtRNA^Pyl – directed the incorporation of an azide containing ncAA in response to amber codons in the viral gene; cells also contained the other genes necessary for virus replication.

Inactive N-MmtRNA^Pyl variants did not make viral particles, and active N-MmtRNA^Pyl variants led to the production of viral particles. Cells which contained active N-MmtRNA^Pyl variants, that are aminoacylated by the N⁺-MbPylRS variant, produced virus particles that displayed an azide on their capsid and contained the gene for the N-MmtRNA^Pyl variant in their DNA. These viruses were labeled with a biotinylated alkyne probe, and affinity purified for sequencing or further rounds of evolution.

The most active hit (N-MmtRNA^Pyl-A2.1), selected from a library of acceptor stem mutants, led to a 3-fold increase of the incorporation of the azide containing lysine derivative in mammalian cells. Similar increases in ncAA incorporation were observed for other ncAAs when using N-MmtRNA^Pyl-A2.1. There was no increase in ncAA incorporation with N-MmtRNA^Pyl-A2.1 in E. coli, suggesting that the observed increases in ncAA incorporation were host cell specific. Recent improvements in VADER have enabled the identification of more active N-MmtRNA^Pyl variants in mammalian cells.¹⁹⁰ As this approach relies on the use of a particular ncAA, it may be challenging to adapt it to select tRNAs that are optimized for different ncAA chemotypes.

Analysis of the spatial distribution of an active site mutant of N⁺-MmPylRS in mammalian cells suggested that the mutant enzyme predominantly clustered in the nucleus. Sequence analysis identified a putative nuclear localization signal (NLS) near the N-terminus of N⁺-MmPylRS. The authors theorized that a large proportion of the mutant N⁺-MmPylRS/N-MmtRNA^Pyl pair was segregated from cytosolic translation, hampering the overall efficiency of ncAA incorporation in mammalian cells.¹⁹¹ In these experiments, attaching a nuclear export signal (NES) to this mutant redirected the enzyme to the cytosol. The authors reported a 15-fold increase in amber suppression efficiency with the mutant N⁺-MmPylRS enzyme bearing an NES and used their system for super resolution imaging by ‘points accumulation for imaging in nanoscale topography’ (PAINT).¹⁰⁸ Prior work had shown that some N-MbtRNA^Pyl was in the nucleus and used a system without an NES for super resolution imaging by ‘stochastic optical reconstruction microscopy’ (STORM).¹⁹² It is unclear to what extent the effect of the NES on ncAA incorporation efficiency reported in the later work was a function of the specific PylRS mutant and the expression level of the PylRS/tRNA^Pyl pair.

The use of PylRS enzymes from thermophilic archaea (Methanosarcina thermophila (Mth – 50 °C) and Methanosarcina flavescens (Mfl −45 °C)), which live at temperatures significantly above those observed for N⁺-MbPylRS (37 °C) and N⁺-MmPylRS (35 °C), has shown promise in optimizing ncAA incorporation experiments in mammalian cells.¹⁹³ The wt N⁺-MthPylRS, and N⁺-MflPylRS promoted the incorporation of BocK at an amber codon to an equivalent or lower level than N⁺-MbPylRS; however, a series of PylRS variants (derived from active site transplants from evolved PylRS mutants) were more active in the N⁺-MthPylRS and N⁺-MflPylRS scaffolds, than in the N⁺-MbPylRS scaffold. A potential reason for this difference could be the higher tolerance of the more thermally stable enzymes to active site perturbations. All measured active site transplants led to enzymes with lower thermal stability than their scaffolds, but starting with a thermostable scaffold led to more stable transplants.

A eukaryotic release factor 1 mutant (eRF1 - E55D) was discovered that maintains efficient termination on UGA and UAA codons, but facilitated increased ncAA incorporation at amber codons with N⁺-MmPylRS/N-MmtRNA^Pyl pairs. This system was used to produce GFP bearing a ncAA in yields rivaling wt protein production, and faciliated the incorporation of two or three copies of a ncAA into GFP in mammalian cells.¹⁴⁹ The use of tailored expression and delivery systems, as well as eRF1 mutants, provided additional examples of optimized systems for PylRS/tRNA^Pyl pair performance in mammalian cells.^194−196

An important factor affecting the efficiency of ncAA incorporation via amber suppression is the sequence context in which the ncAA is to be encoded. Recent work has employed the piggyBac system expressing the PylRS/tRNA^Pyl pair in combination with stochastic orthogonal recoding of translation with enrichment (SORT-E, see Section 9.4.1)^166,167 to capture and quantify polypeptides resulting from read through of endogenous amber codons. Using this data, the authors generated a model to predict sequence contexts that favor amber suppression and validated this model experimentally.¹⁷³

Despite the large number of endogenous amber codons in mammalian cells, ncAAs are selectively incorporated into target proteins in response to in frame amber codons introduced into transgenes, and there is minimal steady state incorporation of ncAAs at endogenous termination codons.^192,197

Investigators have colocalized genetic code expansion components to membraneless, phase separated, spatially localized, subcellular compartments with the goal of enhancing the specificity with which a stop codon of interest is decoded to incorporate a ncAA.¹⁹⁸ To achieve this, target mRNAs bearing a stop codon were equipped with an ms2 tag that is specifically bound by major capsid protein (MCP). N⁺-MmPylRS and MCP were fused to proteins that undergo phase separation in cells, and to kinesin motor proteins (which are specifically enriched at microtubule plus ends). The resulting localized membraneless organelles supported protein translation with an efficiency of about 40% that of cytoplasmic translation. Translation through an amber codon in an ms2 tagged message (localized to a subcompartment) was favored over translation through an amber codon in non-ms2 tagged message (localized to the cytosol) by a factor of 5. In an extension of this work, PylRS variants for distinct ncAAs were directed to distinct sub cellular compartments, and distinct mRNAs bearing amber stop codons were directed to each compartment. This enabled the incorporation of a different ncAA in response to the amber codon in each message.¹⁹⁹ This work increased the efficiency of translation in sub compartments to about 80% of cytoplasmic translation. It also increased the specificity for translating through an amber codon in a sub compartment to translating through and amber codon in the cytosol to 20-fold. These interesting approaches have thus far been investigated using transient transfection to introduce the genes of interest. In future work it will be interesting to investigate what happens to these systems during cell division and in long-term cultures. If synthetic subcompartments can be stably embedded in cells, without adversely perturbing natural biology, they may provide a powerful approach for generating compartmentalized genetic codes; this strategy could contribute to efforts to study in vivo biology with genetic code expansion, without effects that may arise from global amber suppression.

5. Biosynthesis of ncAA for Incorporation with PylRS/tRNA^Pyl Systems

A discrete body of work has focused on coupling strategies for the cellular biosynthesis (or semisynthesis) of ncAAs to the incorporation of the biosynthesized ncAAs into proteins using orthogonal aaRS/tRNA pairs in E. coli.^26,200−204 Most recent work with the pyl systems has focused on exploiting the permissiveness of the natural pyl biosynthetic enzymes to analogues of their natural substrates, or engineering the pyl biosynthetic enzymes to accept new substrates; PylRS or its variants are then used to incorporate the biosynthetic products into proteins in E. coli. Endogenous metabolic enzymes have also been used to generate substrates for engineered PylRS variants.

PylC can direct the formation of an isopeptide bond between lysine and DO (when added to E. coli in place of the 3MO, the natural substrate of PylC) and PylD converts the product to desmethylpyrrolysine (dPyl).^24,29 Since dPyl is a substrate for N⁺-MmPylRS, feeding DO to E. coli expressing PylC, PylD and PylRS/tRNA^Pyl pairs led to the incorporation of this ncAA in response to amber codons.²⁶ The electrophilic imine of dPyl in proteins was used to conjugate proteins with diverse molecules through 2-amino-benzaldehyde (2-ABA) or 2-amino-acetophenone (2-AAP) reagents (Figure 10a).²⁰² Similarly, the substrate promiscuity of PylC, PylD, and PylRS subsequently enabled in vivo production of proteins with an alkyne bio-orthogonal handle from E. coli cells transformed with pylTSCD and supplemented with 3-S-ethynyl-D-ornithine (EO) (Figure 10a).²⁰³

Figure 10.

Exploiting substrate promiscuity and engineering the pyl pathway for ncAA incorporation. a, PylC and PylD can accept D-ornithine (DO), or S-ethynyl-D-ornithine (EO) forming desmethylpyrrolysine (dPyl) with or without an alkyne substituent on the pyrroline ring. The imine of dPyl can be functionalized with 2-amino-benzaldehyde (2-ABA) or 2-amino-acetophenone (2-AAP) derived compounds. Furthermore, the alkyne can be labeled in bio-orthogonal reactions resulting in the double labeling of recombinant proteins. b, PylC can be engineered to accept d-cysteine forming D-cysteinyl-ε-lysine (CεK), which can be incorporated into recombinant proteins by an engineered N⁺-MbPylRS/N-MbtRNA^Pyl pair and used to cyclize proteins which contain intein-derived C-terminal thioesters.

Recent work has aimed to engineer the substrate specificity of both PylC and PylRS to expand the range of ncAAs that can be biosynthesized in E. coli and incorporated into proteins (Figure 10b). In the first step, a N⁺-MbPylRS/N-MbtRNA^Pyl pair was engineered to incorporate D-cysteinyl-ε-lysine (CεK) into proteins in response to the amber codon, upon addition of the ncAA to E. coli.²⁰⁴ In the second step, several positions in the active site of PylC were subject to saturation mutagenesis, and cells containing the resulting PylC library, and an engineered N⁺-MbPylRS/N-MbtRNA^Pyl variant for CεK, were selected for on the basis of their ability to support amber suppression upon addition of d-cysteine to cells. This led to the discovery of a PylC mutant (S177N, E179P, D233S, and T256 V) that used added d-cysteine and endogenous lysine for the in vivo synthesis of CεK. Cells that biosynthesized CεK supported amber suppression by the engineered N⁺-MbPylRS/N-MbtRNA^Pyl variant at comparable levels to cells to when 4 mM, chemically synthesized, CεK was added.

Components of the pyrrolysine biosynthesis pathway, notably PylB, are suboptimally expressed in E. coli.²³ Linking the N-terminus of PylB to small ubiquitin-related modifier (SUMO) increased solubility enabling purification and crystallization.²⁵ Phage assisted noncontinuous evolution (PANCE) on a SUMO-PylB, PylC, PylD operon led to improved production of Pyl in E. coli.²⁰⁵ Most mutations accumulated in the SUMO-pylB gene, and were ascribed to increasing the solubility as well as the protease resistance of SUMO-PylB. When combined with the PylRS/tRNA^Pyl pair the evolved pathway yielded up to 32 times higher yields of proteins encoding Pyl than the starting pathway.

Other work has used enzymes that are endogenous to E. coli to assemble substrates for engineered N⁺-MmPylRS variants. Addition of allyl mercaptan to E. coli led to the synthesis of S-allyl-L-cysteine via the reaction with endogenous O-acetyl-L-serine, a metabolic intermediate in cysteine biosynthesis. This reaction was catalyzed by endogenous O-acetyl-serine sulfhydrylase enzymes. A N⁺-MmPylRS variant was evolved to direct the incorporation of S-allyl-L-cysteine.²⁰⁶

Given the range of chemotypes that can now be incorporated by PylRS variants it seems likely that both further engineering of the Pyl biosynthesis pathway as well as further engineering of other biosynthetic pathways will yield scalable and sustainable routes to the synthesis or semisynthesis of monomers that can be incorporated by variant PylRS/tRNA^Pyl pairs.

6. Applications of ncAAs Genetically Encoded by PylRS/tRNA^Pyl Pairs

Genetic code expansion permits the site-specific introduction of new chemical functionalities into proteins in live cells and animals. This has enabled the scalable production and purification of recombinant proteins bearing defined modifications, and the development of approaches for imaging and controlling protein function in live cells and animals. Excellent overviews of applications based on the genetically encoded site-specific introduction of ncAAs into proteins are available,^{1,3,4,62,207−210} and are beyond the scope of this review. Here we focus on classes of ncAAs and applications that have been uniquely enabled by PylRS/tRNA^Pyl pairs. These applications include: (1) encoding ncAAs corresponding to PTMs, particularly lysine PTMs (Section 6.1); (2) caging canonical amino acids for triggered activation of protein function, where PylRS systems have expanded both the range of canonical amino acids that can be caged and, the caging modalities and the range of organisms where caging can be easily applied (Section 6.2); (3) encoding bio-orthogonal groups, where PylRS systems have provided access to aliphatic long chain ncAAs that undergo rapid bio-orthogonal labeling and thereby enabled rapid site-specific protein labeling, and dual labeling, in cells and organisms (Section 6.3); (4) strategies for reversible photocontrol and installing biophysical probes, enabled by rapid bio-orthogonal chemistry or the encoding of photoswitches and probes with PylRS/tRNA^Pyl pairs (Section 6.4, 6.5); (5) genetically encoding ncAAs for identifying protein interactions, where PylRS systems provide facile access to cross-linkers in diverse cells and organisms, access to more flexible cross-linking ncAAs that may capture protein interactions more efficiently, and access to new functional groups for cross-linking and new cross-linking modalities (Section 6.6); (6) encoding ncAAs that can be used as mechanism based traps of enzyme substrates, and to augment enzyme function, where PylRS systems provide access to unique chemical functionality (Section 6.7, Section 6.8); (7) strategies for translational control in cells and animals, where PylRS systems benefit from bioavailable ncAAs and enables the approaches to be used in diverse cells and organisms (Section 6.9).

6.1. Post Translational Modifications

Genetic code expansion permits the direct installation of ncAAs that correspond to the post-translationally modified forms of canonical amino acids, commonly described as “genetically encoded PTMs”. This enables the study of the structural, mechanistic and functional consequence of PTMs at specific sites in proteins. The majority of this work has focused on expressing and purifying recombinant proteins bearing defined PTMs that can be used for structural or mechanistic studies; this approach, along with complementary approaches,^211,212 addresses the challenge of making recombinant proteins bearing PTMs at defined sites when the enzyme that naturally installs the modification is unknown (common for PTMs identified by proteomics) or modifies other sites in the protein in vitro.^{82,106,137,143,213−219} A smaller, but important, body of work has genetically installed PTMs at specific-sites in proteins produced in physiologically relevant hosts, to address the functional consequences of these modifications in vivo;^{94,95,219−221} these approaches commonly aim to address the functional consequences of these modifications, without the pleiotropic effects that result from manipulating the enzymes that naturally add or remove the modification. PylRS/tRNA^Pyl pairs have provided access to PTMs of lysine (Figure 11), which cannot be accessed by other orthogonal aaRS/tRNA pairs, as well as some PTMs of other canonical amino acids.

Figure 11.

PylRS-mediated genetic encoding of ncAAs corresponding to post-translationally modified forms of canonical amino acids. a, Acylated derivatives of lysine for which PylRS variants have been evolved. The structures include non-natural PTM mimetics; trifluoroacetyl-lysine (not shown) can also be incorporated. b, Strategy to genetically direct the succinyl-lysine (SucK) and glutaryl-lysine (GluK) in proteins. c, Protected versions of N^ϵ-methyl-L-lysine (MeK), which have been genetically encoded into proteins. Following incorporation, the ncAAs have been deprotected to produce proteins with monomethylated lysine residues at specific sites in proteins. d, e, Strategies to generate proteins bearing dimethylated lysine residues. f to i, Strategies to generate site-specifically ubiquitinated proteins containing one to two non-natural linkages between lysine and ubiquitin. j, Strategy to genetically direct a natural lysine-ubiquitin linkage.

The first directed evolution of a PylRS enzyme toward a new substrate focused on introducing AcK into proteins in E. coli.⁸² The genetic encoding of AcK has enabled the consequences of lysine acetylation, on the structure and function of recombinant proteins, to be studied.^220−231 The genetic encoding of AcK has been extended to various hosts, including live animals, enabling studies on the functional consequences of acetylation in a cellular or organismal context.^{94,95,106,219−221} Genetically encoded nonhydrolyzable analogues of AcK have provided a stable version of the modification in cells.^232−234 PylRS variants have been engineered and evolved to encode many other lysine acylations including formylation,²³⁵ propionylation,^236,237 butyrylation,^219,236,237 2-hydroxyisobutyrylation,^238,239 β-hydroxybutyrylation,²³⁹ crotonylation,^236,237,240 lactylation,²³⁹ lipoylation,²³⁹ benzoylation,^241−243 and threonylation.²⁴⁴ Acylated lysine derivatives with long-chain terminal olefins have also been encoded using engineered PylRS enzymes.^239,245 These approaches have provided a wealth of insight into the diverse roles of lysine acylations (Figure 11a).^{219,235−246} Recently, lysine derivatives bearing protected succinyl- or glutaryl-groups were genetically encoded and deprotected on the protein to succinyl-lysine (SucK) and glutaryl-lysine (GluK); this approach was used to study the function of these negatively charged PTMs (Figure 11b).²⁴⁷ The genetic encoding of an azido-norleucine, followed by a traceless Staudinger ligation, enabled the generation of various acylated lysine derivatives at site-specific positions in ubiquitin and histone H3 with modest efficiency.²⁴⁸ Genetic code expansion-based approaches to study lysine acylations have recently been extensively reviewed.²⁴⁹

Lysine methylations (mono-, di- and trimethylation) are another important class of PTM that have been tackled by genetic code expansion with PylRS/tRNA^Pyl pairs. It was challenging to create an active site that would distinguish between lysine and methylated lysines. For monomethylated lysine researchers addressed this challenge by encoding protected versions of N^ϵ-methyl-L-lysine (MeK), in which the protecting group further differentiated the structure of the ncAA from lysine (Figure 11c).²⁵⁰ This concept was demonstrated using a tert-butyloxycarbonyl (Boc) protecting group, and used to make histone H3 MeK9; binding to the methylation reader protein HP1 was demonstrated. A variety of other protected ncAAs have been used to encode MeK, and these approaches now enable in vitro,^251−253 or in vivo deprotection.^254,255 Di- and trimethylation have been more challenging to encode, but N⁺-PylRS/N-tRNA^Pyl pairs have been used as part of strategies to selectively install dimethyl-lysine in histone H3^256,257 and p53 proteins (Figure 11d,e).²⁵⁷

Ubiquitin and ubiquitin-like proteins constitute an important class of lysine modification that cannot be directly genetically encoded. Several approaches have used N⁺-PylRS/N-tRNA^Pyl pairs to encode ncAAs that can be linked to ubiquitin; these approaches complement a variety of other methods developed to access ubiquitinated proteins.^258−262 Several of these efforts lead to non-native sequence in or around the linkage. Genetically encoding CεK followed by reaction with a C-terminal thioester of ubiquitin 1–75, produced protein-ubiquitin conjugates (Figure 11f). However, the native chemical ligation reaction (NCL) resulted in a nonstandard linkage in which G76 of ubiquitin was replaced by cysteine.²⁶³ Additionally, the genetic encoding of protected ϵ-aminooxy-L-lysine derivatives enabled the production of recombinant, isosteric and nonhydrolyzable ubiquitin conjugates (Figure 11g).²⁶⁴ An emerging method to generate site-specifically ubiquitinated, and SUMOylated proteins uses an engineered N⁺-MbPylRS/N-MbtRNA^Pyl pair to encode N^ϵ-((2-azidoacetyl)glycyl)-L-lysine (AzGGK) into proteins (Figure 11h). Following reduction of the azide to an amine, the protein of interest is reacted with ubiquitin, bearing mutations at positions 72 and 74 that generate a sortase recognition motif, in a sortase-mediated transpeptidation. The resulting protein conjugate has a native isopeptide bond, but contains two mutations in ubiquitin. This approach has been used to make ubiquitinated proliferating cell nuclear antigen (PCNA) and has been extended to SUMOylation, and to ubiquitination in mammalian cells. The mutations in the C-terminus of ubiquitin confer resistance to some deubiquitinating enzymes, but appear to have minimal effects on the binding of ubiquitin binding proteins tested. In addition, these linkages can be formed under nondenaturing conditions.²¹⁶ By using orthogonal sortase enzymes, that recognize distinct C-terminal mutants of ubiquitin, this approach has been extended to the assembly of ubiquitin chains (Figure 11i).²⁶⁵ Recently, this strategy has been further expanded to asparagine endopeptidase mediated ligation, which requires only one mutation in the ubiquitin C-terminus.²⁶⁶

Several methods have achieved entirely native linkages to ubiquitin. The genetic encoding of protected δ-thiol-L-lysine into proteins was achieved with an evolved N⁺-MbPylRS/N-MbtRNA^Pyl pair. In combination with chemistry first demonstrated in synthetic peptides,²⁶⁷ this enabled the traceless site-specific ubiquitination of recombinant proteins (Figure 11j).²¹⁷ Furthermore, the coordinated use of chemical protection/deprotection schemes, enabled by N⁺-MbPylRS/N-MbtRNA^Pyl pair-mediated ncAA incorporation, led to the synthesis of site-specifically linked ubiquitin chains with native linkages; this resulted in the structure of K6 linked diubiquitin and revealed a K29 specific deubiquitinating enzyme.²¹⁸

Several other approaches have been used to generate a variety of native and non-native PTMs. The reaction of nucleophiles with dehydroalanine is a well-established route to installing a wide-variety of PTMs into proteins.^268,269Se-alkylselenocysteine has been encoded using an engineered N⁺-MmPylRS/N-MmtRNA^Pyl pair. The encoded ncAA has been converted to dehydroalanine and used to generate various PTM mimetics.²⁷⁰ A challenge with dehydroalanine-based approaches is that they commonly lead to racemization of the α carbon to generate a mixture of proteins that includes a stereochemically non-native backbone. The site-specific incorporation of a protected phospho-tyrosine precursor, which can be converted to phospho-tyrosine in vitro, under denaturing conditions, enabled the site-specific genetic encoding of the modification into recombinant proteins produced in E. coli; this method requires the protein to be refolded.¹³⁷ Halo-tyrosine derivatives have also been site-specifically encoded.¹³⁵ Proteins can also be modified by bio-orthogonal chemistry (Section 6.3) and these approach has been used to generate non-native PTMs, including glycosylation.^271,272

PylRS systems have been crucial for encoding diverse PTMs, most notably lysine acylations and ubiquitination. The tools provided by these approaches have provided unique insight into the roles of PTMs and combinations of PTMs. Future work may focus on genetically encoding other key modifications including glycosylation and tyrosine phosphorylation, and on the use of genetically encoded PTMs and their nonremovable analogs to interrogate and control the state of living cells and organisms.

6.2. Controlling Protein Function by Caging Canonical Amino Acids

Activatable proteins can be created by the site-specific introduction of protected versions of canonical amino acids, in place of the corresponding canonical amino acids, at positions that mediate an important function of a protein (Figure 12a). The protected amino acid is converted to a canonical amino acid in the protein, by optical, chemical or enzymatic activation, to restore the native protein sequence and function. These experiments are commonly performed using transgenes and are commonly applied to proteins that provide new activities to the cell (e.g.: cre recombinase or TEV protease), or constitutively active versions of native proteins (e.g.: constitutively active kinase mutants).^273−275 Non-pyl aaRS/tRNA pairs have been used to encode photocaged amino acids and generate photoactivatable proteins: photocaged tyrosine and cysteine have been encoded in E. coli, photocaged serine has been encoded in yeast, and photocaged glutamate has been encoded in yeast and in mammalian cells.^276−279 The development of PylRS/tRNA^Pyl pairs for genetic code expansion enabled facile photocaging, and photoactivation, of proteins in mammalian cells, and in some cases animals, for lysine, tyrosine, cysteine, aspartic acid and histidine residues.^{136,138,153,280−284} Optical decaging is commonly fast, can be executed with millisecond pulses of (blue (410 nm) or UV (365 nm)) light at powers that do not trigger optical DNA damage responses, and can be spatially controlled. Optical decaging may be difficult to achieve in deep tissue samples, but in some cases 2-photon approaches have been developed to address this challenge. PylRS/tRNA^Pyl pairs have also enabled the caging and chemical decaging of lysine, tyrosine, selenocysteine and tryptophan residues. Chemical decaging commonly occurs over time scales of minutes to hours and may not go to completion, is commonly not spatially confined, but may be easier to effect in deep tissues.

Figure 12.

Non-canonical amino acids, corresponding to caged canonical amino acids, that can be genetically encoded into proteins and deprotected to the corresponding canonical amino acids. a, Schematic representation of the deprotection of ncAAs, corresponding to caged canonical amino acids, to canonical amino acids. The deprotection can be done with light (orange structures in panels b-h), by the addition of a small molecule (blue structures in panels b-h, with deprotecting agent in gray), or enzymatically (purple structures). b to h Genetically encoded ncAAs that can be deprotected to lysine, tyrosine, (homo)-cysteine, selenocysteine, aspartic acid, histidine and tryptophan derivatives, respectively.

6.2.1. Optical Decaging of ncAAs Encoded Using PylRS/tRNA^Pyl Pairs

The N⁺-MbPylRS/N-MbtRNA^Pyl pair was evolved to incorporate a photocaged derivative of lysine (Figure 12b); this ncAA was first used in place of key lysine residues in a nuclear localization sequence to enable the optically triggered nuclear localization of target proteins from the cytosol to the nucleus of mammalian cells.¹⁵³ The genetic encoding of photocaged lysine has been extended to generate photoactivatable enzymes, including kinases,^285,286 polymerases,²⁸⁷ DNA helicases,²⁸⁸ and recombinases^98,165,289 for spatially and temporally controlled activation, and lysine derivatives that can be deprotected with two-photon optics have been developed.²⁸⁰ A recent exciting application of genetically encoded photocaged lysine was the optical activation of cre recombinase in one of a pair of bilaterally symmetric neurons in Caenorhabditis elegans (C. elegans); these neurons cannot be genetically distinguished.⁹⁸ This enabled the cre-mediated activation of expression of channelrhodopsin in a single neuron and subsequent optogenetic stimulation of a single neuron in freely moving animals.⁹⁸ These experiments provided an approach that enabled the unique contributions of single neurons, within bilaterally symmetric pairs, to be determined.

The N⁺-MbPylRS/N-MbtRNA^Pyl pair has been evolved to incorporate photocaged derivatives of tyrosine (Figure 12c).^136,138 Encoding a photocaged tyrosine in place of tyrosine at a phosphorylation site in STAT1 in mammalian cells, enabled the tyrosine phosphorylation site to be blocked. The phosphorylation site was revealed by optical activation leading to light-induced phosphorylation of the tyrosine residue. Photocaged tyrosine has also been used to activate kinases,²⁹⁰ and proteases,¹³⁶ including caspases.²⁹⁰ Anthrax toxin component lethal factor has been caged for optical pro-drug activation.²⁹⁰

N⁺-MbPylRS enzymes have also been engineered for genetically encoded incorporation of photocaged cysteine and homocysteine (Figure 12d).^281,282 Encoding photocaged cysteine enabled the creation of photoactivated TEV protease in mammalian cells. The development of a photocaged selenocysteine enabled the incorporation of selenocysteine into proteins without using the selenocysteine insertion sequence (SECIS) normally required in mRNAs for the incorporation of selenocysteine (Figure 12e).²⁹¹

Engineered N⁺-MbPylRS/N-MbtRNA^Pyl pairs have been used to encode a photocaged aspartic acid for optically activating kinase and GTPase activity in mammalian cells (Figure 12f).²⁸³ Engineered N⁺-MbPylRS/N-MbtRNA^Pyl pairs have also been developed to encode a photocaged histidine, which was used to optically activate luciferase in mammalian cells (Figure 12g);²⁸⁴ this is an exciting development and, following refinement of the current system, this advance may enable the photocaging of many enzyme activities and binding activities mediated by histidine residues.

6.2.2. Chemical Decaging of ncAAs Encoded Using PylRS/tRNA^Pyl Pairs

PylRS/tRNA^Pyl pairs have been used to site-specifically incorporate ncAAs in which a lysine residue is connected through a carbamate linkage to functional groups to form a protecting group; the protecting group reacts with added chemicals to reveal the lysine residue (Figure 12b). Several chemistries have been used to effect this deprotection including palladium catalyzed reactions with encoded alkynes,²⁹² the reaction of tetrazines with encoded trans-cyclooctenes (TCOs),^293−297 Staudinger reductions with encoded ortho-azido benzyl groups,²⁹⁸ the reaction of TCOs with encoded para-azido benzyl groups,²⁹⁹ the aza-Cope reaction of N^ϵ-((prop-2-yn-1-yloxy)carbonyl)-L-lysine (PrAK) with formaldehyde–providing the basis for a formaldehyde sensor,³⁰⁰ and the reaction of N^ϵ-(((2-methylbut-3-yn-2-yl)oxy)carbonyl)-L-lysine (DMProcK) with Cu(I)-BTTAA.³⁰¹ With the exception of the copper-triggered approach, which was only used to activate proteins on cell surfaces, these approaches have been used to activate protein function in live cells.

N⁺-MmPylRS/N-MmtRNA^Pyl pairs have been used to site-specifically incorporate ncAAs corresponding to protected tyrosine residues (Figure 12c). An allene ether derivative of tyrosine was incorporated into proteins and deprotected to tyrosine using palladium reagents in mammalian cells.³⁰² Similarly, a disubstituted propargyl group caged tyrosine was incorporated into proteins and deprotected to tyrosine with Cu(I)-BTTAA.³⁰¹ The palladium-triggered approach has been used to activate protein function in live cells, while the copper-triggered approach was demonstrated on cell surfaces. An engineered N⁺-MbPylRS/N-MbtRNA^Pyl pair has enabled the palladium mediated deprotection of site-specifically installed Se-allyl-selenocysteine in E. coli.³⁰³

Recently a N¹-vinyl tryptophan was encoded using a derivative of chPheRS/chtRNA^Phe.¹⁴⁴ This ncAA was deprotected in cells through an inverse electron demand Diels–Alder reaction with tetrazines (Figure 12h).¹⁴⁶ This approach was used to regulate a variety of protein activities and protein interactions.

Enzymatic routes have also been explored for the deprotection of ncAAs (Figure 12b,d).^304,305 A phenylacetamidomethyl protected derivative of homocysteine was site-specifically incorporated into proteins using an engineered N⁺-MbPylRS/N-MbtRNA^Pyl pair and deprotected using penicillin G acylase, as part of a strategy for protein dual labeling.³⁰⁵ While enzymatic deprotection has the potential to be substantially faster than current small molecule based deprotections on proteins it may be more restricted in the sites on proteins at which it can mediate deprotection.

In future work it will be interesting to see if chemical deprotection can be accelerated to control processes that occur on a wider range of biological time scales, and to develop approaches to spatially localize chemical deprotection.

6.3. Genetically Encoding Bio-orthogonal Groups for Site-Specific Labeling of Proteins

Genetic code expansion enables the site-specific installation of ncAAs bearing bio-orthogonal groups that can be selectively labeled with molecules bearing a bio-orthogonal reaction partner (Figure 13a to f). This paradigm substantially expands the range of chemical functionalities that can be attached to proteins and has provided routes to labeling proteins with molecules for imaging, controlling, and augmenting protein function; these approaches have provided new insight into basic biology and new approaches to generating defined protein modifications for a variety of applications, including the generation of potential therapeutics.^{207,208,306−313} PylRS/tRNA^Pyl pairs have enabled the encoding of ncAAs with aliphatic side chains bearing bio-orthogonal groups (Figure 13g).

Figure 13.

Bio-orthogonal handles that have been genetically encoded using (engineered) PylRS/tRNA^Pylpairs for protein labeling. a to e, Schematic representations of some commonly used bio-orthogonal reactions for genetic code expansion mediated site-specific labeling of proteins. (a) CuAAC, (b) SPAAC, (c) KHC, (d) ACC, and (e) iEDDA. f, Structure of the some of the most-important bio-orthogonal handles used to date. Noncanonical amino acids bearing Nor, TCO, Cyp, as well as BCN groups have all been genetically encoded using engineered PylRS/tRNA^Pyl pairs and extensively used for iEDDA-based labeling, and SPAAC-based labeling. g, List of lysine derivatives bearing a variety of bio-orthogonal handles including azides, ketones, tetrazine, strained alkene, and (strained) alkynes. h, As in (g) but for alanine derivatives. i, As in (g) but for phenylalanine derivatives. Parts of this figure are reprinted with permission from Lang, K.; Chin, J. W. Bioorthogonal Reactions for Labeling Proteins. ACS Chem. Biol. 2014, 9, 16–20 - copyright © 2014 American Chemical Society.

PylRS/tRNA^Pyl pairs have been used to genetically encode lysine derivatives bearing azides and alkynes.^58,314−320 Some labeling of azides can be achieved using Staudinger ligations, but these are slow and often do not go to completion. Azides and alkynes can be labeled by Cu(I) mediated chemistry, this is faster and can be driven to completion, but the reaction conditions may cause oxidative damage of proteins (Figure 13a,g,i).^{58,314−319,321,322} Encoded azides may also be labeled with ring strained alkynes,³²⁰ but these reactions are again slow (Figure 13b,g).^207,323,324 Genetically encoded alkynes have been used for palladium mediated protein labeling with iodophenyl reagents,^325,326 and ruthenium(II) mediated hydrosilylation.³²⁷ Genetically encoded fluorosulfates have been developed for palladium mediated protein labeling by Suzuki cross-couplings.³²⁸ Genetically encoded alkynes have been used for thiol–yne reactions with reactive thiols,³²⁹ and genetically encoded alkenes have been labeled with thiol–ene reactions;^330,331 the thiol reactants in these reactions are commonly not bio-orthogonal. A ncAA in which an alkyne is linked to lysine through a hydrolyzable ester has enabled the capture and release of a protein in which it is encoded.³³² Noncanonical amino acids bearing ketones,²³⁴ and (caged) aldehydes,^333,334 have been incorporated with PylRS/tRNA^Pyl pairs and used for labeling with hydroxylamine reagents, and acrylamide functionalities have also been encoded for polymerization (Figure 13c,g,i).⁸⁸ Genetically encoded cyclopropenones have been encoded for the reaction with phosphines.³³⁵ A quadricyclane containing ncAA has been encoded and reacted with nickel bis(dithiolene) in a formal [2σ + 2σ + 2π] cycloaddition.³³⁶ Biotin analogs have also been encoded.³³⁷

The rate and specificity of bio-orthogonal reactions is crucial for labeling in living systems.²⁰⁸ Encoding components of fast, metal free, bio-orthogonal reactions has been essential for achieving efficient and quantitative labeling of recombinant proteins and for achieving labeling on and in live mammalian cells. Labeling proteins within cells is generally considered more challenging than labeling on the cell surface, in part because the fluorophore label must enter the cell, but excess fluorophore needs to be washed out. Engineered PylRS/tRNA^Pyl pairs have been central to encoding ncAAs bearing components of rapid bio-orthogonal reactions with on protein rate constants greater than 1 M^–1 s^–1, and have enabled the site-specific labeling of proteins in diverse systems, including mammalian cells.

1,2-Aminothiols³³⁸ have been genetically encoded with N⁺-MmPylRS/N-MmtRNA^Pyl pair variants, and used for rapid labeling with cyanobenzothiazoles (CBTs); this chemistry uses free thiols and its primary utility is for in vitro labeling (Figure 13d,g). Components of inverse electron demand Diels–Alder reactions between strained (or otherwise activated) alkenes or alkynes and tetrazines have been genetically encoded (Figure 13e,g,h,i). Noncanonical amino acids containing norbornenes (Nors),^{53,323,339−341} cyclopropenes (Cyps),^166,342,343 bicyclononynes (BCNs)^197,343,344 and TCOs,^{197,323,341,343−347} other reactive alkenes,^348−350 and isocyanides³⁵¹ have been genetically encoded using PylRS/tRNA^Pyl pairs (Figure 13f,g,i). Genetically encoding a Nor containing ncAA with a N⁺-MbPylRS/N-MbtRNA^Pyl pair enabled the first labeling of a genetically encoded ncAA in mammalian cells.³³⁹

Genetically encoded BCNs and TCOs enabled rapid labeling of proteins in E. coli and mammalian cells with tetrazine probes, that exhibit turn-on fluorescence, in seconds to minutes.¹⁹⁷ These approaches demonstrated that, despite the large number of endogenous amber codons in mammalian cells, target proteins could be specifically labeled.^192,197 The bio-orthogonal labeling of genetically installed Cyp, BCN and TCO ncAAs have been further improved, and extensively used to image and control protein function (Figure 13g,i) (refs (157, 161, 166, 167, 192, 271, 309, 344, and 352−383)).

Reactive tetrazines have recently been encoded using engineered N⁺-MbPylRS/N-MbtRNA^Pyl pairs, enabling this class of ncAAs to be encoded in E. coli and mammalian cells (Figure 13g,h).^384−387 Encoded tetrazines may also be reacted with cyclopropenone-caged dibenzoannulated bicyclononynes (photo-DMBO), following ultraviolet (UV) illumination.³⁸⁴ Noncanonical amino acids bearing strained alkenes have also been encoded, using engineered PylRS/tRNA^Pyl pairs, for “photoclick reactions”: 2 + 3 cycloadditions with hydrazonoyl chlorides generated in situ by illumination of tetrazoles with UV light. Photoclick reactions have been demonstrated with encoded Nors, Cyps, (spirocyclic) alkenes,^97,388,389 and the phototransducing dibenzo[b,f][1,4,5] thiadiazepine.³⁹⁰

Strategies for cyclizing proteins by encoding two distinct ncAAs with compatible bio-orthogonal groups (Figure 14a), labeling recombinant proteins at two to three distinct sites (Figure 14b), encoding one PTM together with a ncM for labeling (Figure 14c), as well as the labeling of two distinct proteins within a cell, have commonly used PylRS to encode a bio-orthogonal group at one site.^{76,88,155,305,338,344,380,391−407} Single ncAAs containing two distinct functionalities (an amine and an azide, or an azide and a tetrazine, respectively) (Figure 14d) have also been encoded and used for dual,⁴⁰⁸ or triple (together with a genetically encoded ketone bearing ncAA)⁴⁰⁹ labeling of proteins. Strategies to site-specifically dual label proteins by genetic code expansion have recently been reviewed.⁴¹⁰

Figure 14.

Site-specific double labeling of proteins. a, Schematic representation of a protein cyclization mediated by encoded ncAAs containing bio-orthogonal groups, together with the chemistry used in the first genetically programmed, ncAA-mediated, protein cyclization. b, Schematic representation of protein double labeling, distinct encoded ncAAs (stars) are labeled with complementary functional groups in sequential or one pot, concerted labeling reactions. Examples of protein double labeling are shown. The bio-orthogonal handles are colored with respect to the bio-orthogonal reaction for which they were used in the initial examples: ACC (gray), CuAAC (light blue), KHC (purple), iEDDA (red), SPAAC (blue), and CRACR (orange). Labeling was sequential (S) or concerted (C). The exact reaction partners and conditions for each reaction are provided in the indicated references and we note that the mutual orthogonality of many reactions will rely on the exact molecules used, and the reaction conditions. For sequential labeling the order of labeling is indicated. Citations labeled in describe reactions performed in mammalian cells. c, Schematic representation of the encoding of a PTM together with a bio-orthogonal handle; the ncAA corresponding to a post translationally modified canonical amino acid (green star) can bind its specific readers (green blob) and the PTM containing protein can be enriched using bio-orthogonal reactions. A specific example, which was established for potential applications in mammalian cells is shown. d, Genetically encoding single ncAAs bearing two bio-orthogonal handles for protein double, and triple labeling. The bio-orthogonal handles are colored with respect to the bio-orthogonal reaction for which they were used: KHC (purple), iEDDA (red) and SPAAC (blue). Labeling order of functional groups is specified when not concerted (C).

6.4. Optical Switching of Protein Function in Live Cells

The optical switching of protein function provides a route for reversible control. Noncanonical amino acid–based approaches to optical switching have focused on attaching photoisomerizable groups (notably azobenzene) to proteins, and coupling isomerization of the azobenzene to a change in the state of the protein. In photobio-orthogonal ligand tethering (photo-BOLT), a BCN containing ncAA was encoded proximal to the active site of a kinase expressed in mammalian cells. A tetrazine-azobenzene-kinase inhibitor conjugate was tethered to the protein through a bio-orthogonal reaction, and illumination at different wavelengths of light was used to repeatedly turn the kinase on and off in live cells. Building on prior work,⁴¹¹ the direct genetic encoding of ncAAs containing azobenzenes has also been achieved using engineered PylRS/tRNA^Pyl pairs.^412−414 Azobenzenes bearing halo-alkane substitutions have also been encoded; these have been reacted with a cysteine residue proximal to the site of encoding to form a covalent bridge, and photoisomerization allowed conformational switching in vitro.^413,415 Red shifted azobenzenes containing ncAAs were developed and,⁴¹⁴ several of these azobenzenes were directly encoded and used to reversibly control luciferase activity in live mammalian cells.⁴¹² These experiments highlighted the challenges of identifying sites in the protein at which isomerization of the azobenzene side chain led to marked changes in protein activity; these challenges might be addressed by developing computational approaches to predict positions in proteins at which photoswitchable amino acids could be used to control protein activity. The encoding of photoswitchable ncAAs has the potential to enable the reversible control of diverse protein activities in vitro and in vivo.

6.5. Biophysical Probes

Biophysical probes, including coumarin and acridonyl-based fluorophores.^280,416,417 and nuclear magnetic resonance (NMR),^139,418,419 electron paramagnetic resonance (EPR),⁴²⁰ and Raman spectroscopy³⁴⁹ probes have been site-specifically incorporated using PylRS/tRNA^Pyl pairs. Noncanonical amino acids have also been incorporated into nanopores to alter their sensing properties.⁴²¹ These approaches complement and extend the functionalities that can be added to proteins by installing ncAAs bearing bio-orthogonal groups and subsequent derivatization (Section 6.3). The direct installation of fluorophores typically requires high concentrations of fluorescent ncAAs to be added to cells, and the fluorophores that can be encoded are commonly excited with light in the blue region of the electromagnetic spectrum.^280,416,417 The range of fluorophores that can be attached by rapid bio-orthogonal chemistry, and the lower concentrations of added fluorophores required for bio-orthogonal labeling, have made bio-orthogonal labeling of proteins the preferred route for most ncAA-mediated fluorescent labeling of proteins in cells;²⁰⁸ this approach has enabled the labeling and imaging of proteins that cannot be tagged in a functional form with fluorescent protein fusions, precise labeling of proteins for super resolution imaging, and the construction of fluorescence resonance energy transfer (FRET) probes for following protein conformational changes in live cells.^{192,309,323,339,343,344,370,399,400} Fluorinated amino acids have been encoded to provide a unique signal for F19 NMR and a TMS containing ncAA has been encoded.^139,418,419

6.6. Genetically Encoding Cross-Linkers

The site-specific installation of ncAAs with side chains bearing photo-cross-linking functionalities has provided powerful approaches to map protein–protein interactions in vitro and in vivo,^{209,422−424} and to stabilize protein complexes for structural studies (Figure 15a).^425,426 These ncAAs contain functional groups (aryl azides, diazirines, benzophenones) that, upon UV illumination, can be converted to reactive species (nitrenes, carbenes, triplet diradicals) that react (through insertion into C–H, or heteroatom-H bonds) with adjacent molecules to form covalent bonds; the cross-linking is commonly analyzed by following the gel electrophoretic mobility shift of the protein containing the cross-linking moiety, and mass spectrometry.

Figure 15.

Protein cross-linking using genetically encoded ncAAs. a, Schematic representation of method to identify protein–protein interactions by genetic code expansion mediated site-specific installation of cross-linking ncAAs. Proteins bearing an either photo- (orange), or proximity- (purple) inducible cross-linking ncAA are used in cells to capture interaction partners. The cross-linked proteins are analyzed (by gel electrophoresis or mass spectrometry-based methods) to define interaction partners of the target proteins and the sites of interaction. In some cases cross-links between specific proteins have been used to stabilize complexes for structural studies. b, Phenylalanine derived cross-linking ncAAs incorporated by engineered PylRS variants. c, Lysine derived cross-linking ncAAs incorporated by engineered PylRS variants. d, Schematic representation of the bait and pray concept for a trifunctional ncAA; the bait protein, yellow, contains a ncAA with a cleavable linker, a photo cross-linking functionality and a bio-orthogonal handle. Illumination activates the cross-linker to cross-link to a prey protein, the bait protein is released to facilitate analysis of the resulting stump on the prey protein, and the bio-orthogonal handle is used to pull down the covalent bait-prey complex for analysis by MS. Examples of ncAAs used for this approach, in some cases the site of cleavage at Se is also used for capture and enrichment. e, Schematic representation of a bifunctional ncAA containing both a photo-cross-linking functionality and a PTM; the PTM interacts with a binding protein which is covalently captured, upon illumination, by cross-linking. f, Distinct ncAAs bearing photo-cross-linkers and PTMs of canonical amino acids can be encoded in a single protein; this provides an alternate route to capturing PTM specific interactions.

Bpa, and p-AzF were incorporated with engineered MjTyrRS/MjtRNA^Tyr pairs in E. coli and with engineered EcTyrRS/EctRNA^Tyr pairs in eukaryotic cells (Figure 15b);^109,110,427 these ncAAs have subsequently been incorporated using N⁺-MmPylRS variants.^127,130 PylRS/tRNA^Pyl pairs have uniquely enabled the incorporation of lysine derivatives bearing diazirines (Figure 15c).^{105,428−431} Unlike benzophenones, which can be reversibly photoexcited and have minimal reaction with solvent, the carbenes resulting from photoactivation of diazirines are formed irreversibly and may react nonproductively with solvent. Nonetheless, lysine derivatives bearing diazirines have proved popular because the cross-linking group is small and may minimally perturb protein interactions, and because the longer, flexible lysine side chain may enable the cross-linker within the bait protein to reach suitable reactive groups on the interacting prey proteins. Following the first genetic encoding of a diazirine containing lysine derivative, and the demonstration of its utility for photo-cross-linking protein interactions in vitro and in vivo,⁴²⁸ diazirine containing lysine derivatives have been extensively used for the study of protein interactions.^{166,167,431−437} A diazirine containing ncAA with a shorter side chain has been genetically encoded, and may be used in future cross-linking studies.⁴³⁸ Lysine derivatives bearing tetrazoles have also been used for cross-linking,⁴³⁹ and lysine derivative bearing furans have been reported to cross-link to RNA following red light activation.⁴⁴⁰ An o-nitrobenzyl alcohol derivative of lysine has been encoded with an engineered N⁺-MmPylRS/MmtRNA^Pyl pair for the residue-selective-photo-cross-linking of proteins.⁴⁴¹

Identifying cross-linked proteins, and the sites of cross-linking, by MS can be challenging. Therefore, researchers have aimed to develop methods that, following cross-linking, allow cleavage of the covalent linker between the bait and prey proteins and thereby label the prey protein with a mass tag (Figure 15 d). To achieve this, researchers have created lysine derivatives bearing a diazirine, in which selenium atoms were used to replace the gamma or delta carbon atom of lysine. Following photo-cross-linking, selenium–carbon bonds were oxidatively cleaved and the resulting selenic acid was used as a handle for enriching modified bait proteins. In an alternative design, oxidative cleavage leads to a N-(4,4-bis-substituted-pentyl)acrylamide (NPAA) moiety in the prey protein, which can be readily identified by MS.^442−445 Alternatively, the cross-linker was extended to include an alkyne functionality, which enabled cleaved bait proteins to be enriched through bio-orthogonal reactions with azide-biotin.⁴⁰⁷ Bifunctional ncAAs containing a diazirine and an alkyne moiety were used for photo-cross-linking and labeling of proteins⁴⁴⁶ and to cross-link proteins to RNA.⁴⁴⁷

Identifying PTM dependent protein interactions is an outstanding challenge. To address this challenge for lysine crotonylation, researchers developed ncAAs in which the diazirine functionality was attached to the gamma carbon of crotonyl lysine, distal from the crotonyl group (Figure 15e). This ncAA (and a matched protected lysine control) was encoded using a N⁺-MmPylRS/MmtRNA^Pyl pair and used to trap protein interactions.⁴⁴⁸

PTMs and cross-linkers have also been encoded in two distinct ncAAs in a protein in E. coli (Figure 15f).⁴⁴⁹ Another study incorporated BpA and a photocaged tyrosine into an epidermal growth factor receptor (EGFR)-targeting antibody fragment (7D12). The binding of 7D12 to EGFR was dependent on the optical decaging of photocaged tyrosine, and the covalent linkage of 7D12 to EGFR was also induced by light.⁴⁵⁰

Noncanonical amino acids bearing electrophilic chemical moieties that undergo reactions with nearby nucleophiles provide a complement to photo-cross-linking approaches (Figure 15a,b,c).^209,451 In contrast to photo-cross-linking approaches, these approaches require the ncAA to be incorporated in proximity to a nucleophile with which the cross-link is formed. Engineered PylRS enzymes have enabled the site-specific genetic encoding of (activatable) Michael acceptors,^452−454 aryl carbamates,⁴⁵⁵ dienes,⁴⁵⁶ and various halides.^{413,415,457−463} Early examples that used halides for cross-linking focused on intramolecular protein stapling and cross-linking high affinity protein–protein interactions;^457−459 later a bromo alkane derivative of phenylalanine was used to study interfaces within protein complexes.^464−466 Lysine derivatives, with longer chain bromo alkane appendages of C4–C7, enabled the in vivo stabilization of a protein complex with a dissociation constant of 10–30 μM, and enabled structural studies.^460,467,468 Aryl fluorosulfate containing ncAAs, initially developed for the proximity induced reaction with lysine, histidine and tyrosine residues⁴⁶¹ and used to covalently label protein drugs,⁴⁶⁹ undergo proximity-induced sulfur-fluoride exchange with the 2’ hydroxyl of the RNA backbone, and display some selectivity for RNA in cells.⁴⁶² These ncAAs also show some selectivity for glycans on cell surfaces.⁴⁷⁰

6.7. Mechanism Based Traps for Interrogating Enzyme Function

Many enzymes, including serine hydrolases, cysteine proteases and the ubiquitination machinery, react with their substrates through serine or cysteine nucleophiles to form acyl-enzyme intermediates. These ester or thioester intermediates are unstable and commonly have half-lives of seconds to minutes (Figure 16a). In contrast to esters or thioesters, amides are exceptionally stable (Figure 16b). It was hypothesized that replacing serine or cysteine residues with 2,3-diaminopropionic acid (Dap), in which an amino group replaces the hydroxyl or thiol group, would create “enzymes” that could form a stable amide link with substrates. Dap would be hard to genetically encode, due to its similarity to serine and cysteine, therefore a photocaged version of Dap (pcDap) was designed, and a N⁺-MbPylRS/N-MbtRNA^Pyl pair evolved to incorporate this ncAA (Figure 16c).⁴⁷¹ The incorporation of pcDap into recombinant proteins (and the post-translational deprotection of pcDap to Dap) has been used to generate stable intermediates for structural studies. Incorporating Dap into the thioesterase domain of a nonribosomal peptide synthase enabled a series of stable acyl-enzyme “intermediates” to be captured and structurally characterized, providing structural insight into how the nonribosomal peptide synthetase controls peptide elongation and cyclization to make a defined macrocycle (Figure 16d).⁴⁷¹ By adding a Dap containing protease to a cell lysate, putative substrates of the protease were selectively captured. Genetically encoded pcDap has also been used to capture hydrolase substrates in live mammalian cells (Figure 16d). Following the encoding of pcDap in place of the catalytic residue of a membrane embedded protease (RHBDL4), an enzyme trap was photoactivated in mammalian cells and RHBDL4 substrates were identified. Encoding pcDap into RBBP9, an orphan hydrolase in mammalian cells, was used to discover that it is an aromatic amino-peptidase.⁴⁷² Extensions of this approach enabled protein fusions to a newly discovered PETase to be coupled to polyethylene terephthalate (PET).⁴⁷³ It seems likely that this approach will be extended to discover the substrates of diverse hydrolases and to make further useful protein conjugates.

Figure 16.

Trapping acyl-enzyme intermediates by PylRS mediated genetic encoding of photocaged 2,3-diaminopropionic acid (pcDap). a, Active site serines or cysteines react with the carbonyl groups of their target forming an acyl-enzyme intermediate. The active enzyme is regenerated by the nucleophilic substitution of the intermediate with a hydroxyl, amine, or thiol functionality (R3). b, By introducing Dap, instead of the catalytic cysteine or serine, a cleavage resistant acyl-intermediate may be formed. c, Light activation of a genetically encoded pcDap to Dap. d, Genetically encoding pcDap and conversion to Dap in target proteins enables the N-terminal fragment of protein and peptide substrates (or the analogous portion of other classes of substrate molecules) to be covalently captured. In vitro experiments with defined substrates enables structural studies of acyl-enzyme intermediates. Experiments in cell lysates or live mammalian cells, using tagged hydrolases, enable substrate identification by MS. Panels a and b adapted with permission from Huguenin-Dezot et al.⁴⁷¹ - copyright © 2018 Springer Nature Limited.

6.8. Altering Enzyme Function

Genetic code expansion has been used to augment and probe enzyme function.^{133,474−486} The genetic encoding of proximal ligands with new-to-nature chemical properties in heme proteins has emerged as an efficient method to tune the enzymatic properties of heme enzymes. The genetic encoding of 3-N-methyl histidine (NmH) into an engineered ascorbate peroxidase, using a N⁺-MbPylRS/N-MbtRNA^Pyl pair increased the turnover number of the enzyme up to 5-fold.⁴⁸² Genetically encoding NmH in place of an iron coordinating histidine in myoglobin increased peroxidase activity 4-fold, and subsequent directed evolution and screening led to a protein with a peroxidase activity a thousand-fold higher than myoglobin. Myoglobins containing NmH also enabled cyclopropanation of styrene in the absence of reductant and under aerobic conditions, and a variety of histidine analogs were incorporated into myoglobin, using N⁺-MbPylRS/N-MbtRNA^Pyl pairs, to further tune reactivity.^133,483,484 The genetic encoding of NmH as proximal ligand in a myoglobin scaffold enabled the capture of a reactive heme-carbenoid complex.⁴⁸⁴

Introducing NmH in place of a catalytic histidine residue, in a previously reported computational enzyme design,⁴⁸⁵ generated an enzyme that performed hydrolysis of a model ester. The parent enzyme rapidly formed an acyl enzyme intermediate between the histidine residue and the substrate, but hydrolysis of this intermediate was slow, leading to burst phase kinetics for product formation. The NmH-containing enzyme did not show burst phase kinetics, suggesting that the acyl-enzyme intermediate was more rapidly hydrolyzed. Directed evolution and screening of the NmH-containing enzyme led to a 15-fold improvement in catalytic efficiency.⁴⁸⁶

In future work it will be interesting to encode a wider range of ncAAs with the potential to expand the catalytic power of proteins. It will also be interesting to leverage emerging approaches for encoding combinations of ncAAs into a single proteins, to discover enzymes in which new functions emerge from combinations of ncAAs.

6.9. PylRS/tRNA^Pyl Pairs for Translational Control of Gene Expression

PylRS/tRNA^Pyl pairs have been central to the development of strategies for the translational control of gene expression, in which the production of proteins from genes bearing amber codons is conditional upon addition of a ncAA substrate. The facile development of PylRS/tRNA^Pyl pairs for eukaryotic cells and animals and the bioavailability of ncAA substrates (including an alkyne lysine derivative) for PylRS in multicellular systems have enabled these approaches.^3,96

Translational control has been used for conditional production of attenuated viruses for immunization (Figure 17a).^{160,488−490} In this approach amber (UAG) stop or quadruplet (UAGA) codons are introduced into the protein coding genes in a viral genome to make the production of viral proteins, and therefore the virus, dependent on the presence of a PylRS/tRNA^Pyl pair and cognate ncAA. In the cells used to produce the virus the PylRS/tRNA^Pyl pair and ncAA are provided, but in cells or animals challenged with the resulting virus the PylRS/tRNA^Pyl pair is absent and the ncAA is withheld; as a result, virus reproduction is attenuated in these cells or animals. This approach has been investigated as a strategy for immunization in animal models for Influenza A, HIV-1 and the RNA virus Enterovirus 71.^160,489,490 For mice bearing the PylRS/tRNA^Pyl pair, and infected with Enterovirus 71 bearing amber codons in their genome, it was demonstrated that ncAA addition could be used to elicit a dose dependent increase in viral RNA and a corresponding increase in antibody response.⁴⁹⁰ This provides an intriguing, and potentially generalizable, strategy to tune the level of viral attenuation for immunization.

Figure 17.

Genetic code expansion mediated translation control for the production of attenuated viruses and the ncAA induced restoration of circadian rhythms. a, Viral genomes were engineered to contain UAG stop codons in essential viral protein coding genes. Viruses were produced in host cells encoding PylRS/tRNA^Pyl pairs. The production of the viruses was dependent on the presence of the PylRS/tRNA^Pyl as well as its ncAA substrate. The viruses were then harvested and used to vaccinate animals, because the animals do not contain amber suppressor tRNAs the viruses cannot reproduce in the animal; this approach provides a strategy for generating attenuated viruses for immunization. Adapted with permission from Chin et al.⁴ - copyright © 2017 Springer Nature Limited. b, By making the production of a regulatory protein of circadian rhythms (Cry1) dependent on the ncAA induced N⁺-MmPylRS/N-MmtRNA^Pyl pair mediated translation, the circadian rhythm of otherwise arhythmic (Cry1/2 null) mice, as measured by their wheel running behavior, can be induced by providing the mice with the ncAA in their drinking water. When the ncAA is withdrawn, the circadian rhythm is switched off again. The data shows a circadian reporter (per2-luciferase) as measured by luciferase levels in suprachiasmatic nucleus slices derived from Cry1/2 null mice. When the ncAA was added to the culture medium the Cry1-dependent molecular clockwork was initiated. Adapted from Maywood et al.⁴⁸⁷ - copyright © The Author(s) 2018 CCBY http://creativecommons.org/licenses/by/4.0/.

Translational control in live mice has provided a powerful approach for reversibly controlling behavior. Building on the first approaches for genetic code expansion in live mice using the N⁺-MmPylRS/N-MmtRNA^Pyl pair,⁹⁶ AAVs were used to deliver the genes for a N⁺-MmPylRS/N-MmtRNA^Pyl pair, with N⁺-MmPylRS expressed from cell type specific promoters, along with a Cry1 gene bearing an amber codon, to the brains of Cry1/2 deficient mice. These mice lack circadian rhythms. Addition of a ncAA to the drinking water of the mice turned on the circadian behavior, and removal of the ncAA turned off circadian behavior (Figure 17b). This approach provided a translational switch to study circadian biology and has led to several previously inaccessible insights.^487,491 Translational control using N⁺-MmPylRS/N-MmtRNA^Pyl pairs has also been used to restore dystrophin expression in mice bearing amber codons in the dystrophin gene.⁴⁹²

Split aminoacyl-tRNA synthetases genes, which produce protein fragments that can assemble to produce a functional synthetase, provide a potential approach to refining the temporal and spatial regulation of translational control.^493,494 Recent work has shown that A^Δ-alvPylRS can be split into two polypeptides at several positions, and fusing the genes for interacting polypeptides to the genes encoding the resulting N-, and C-terminal fragments enables reconstitution of A^Δ-alvPylRS function. This approach was demonstrated with interacting coiled-coils, and small molecule dependent dimerization domains, and used as a basis for screening protein–protein interactions.⁴⁹⁵

7. Genetically Encoding ncMs with PylRS/tRNA^Pyl Pairs

A body of work has focused on genetically encoding ncMs beyond α-L-amino acids. Alpha-hydroxy acids are substrates for ribosomal translation in E. coli(496−503) and PylRS/tRNA^Pyl pairs have been generated to incorporate α-hydroxy acids (Section 7.1). The permissiveness of PylRS, and designed mutants, for ncMs has been explored (Section 7.2), primarily in vitro, where weak and nonspecific activities can be measured. Powerful selections have been developed to discover PylRS variants that selectively acylate tRNA^Pyl with a desired ncM and ncAA in vivo, regardless of whether the ncM is a ribosomal substrate,¹⁹ these approaches have enabled the addition of new ncMs to the genetic code of E. coli (Section 7.3).

7.1. Adding α-Hydroxy Acids to the Genetic Code of E. coli with PylRS Systems

PylRS/tRNA^Pyl pairs provide unique advantages for incorporating α-hydroxy acids with both aliphatic and aromatic side chains into proteins.

7.1.1. Genetically Encoding Hydroxy Acids with Aliphatic Side Chains

N⁺-MmPylRS can acylate its cognate N-MmtRNA^Pyl with the hydroxy acid analogue of BocK – (S)-6-(((tert-butoxy)carbonyl)amino)-2-hydroxyhexanoic acid (BocK–OH), and this enables the incorporation of this ncM into proteins in response to the amber codon, as confirmed by MS and selective, base-mediated cleavage, of the resulting ester bond. Glutathione S-transferase (GST) incorporating BocK–OH in response to an amber codon was produced at approximately 10% of the wt protein yield.⁴⁹⁹ The incorporation of BocK–OH into the leader sequence of a lanthipeptide enabled removal of the leader sequence by alkaline hydrolysis of the resulting backbone ester.⁴⁹⁸ To investigate the oligomerization state of lysosomal-associated membrane protein type 2A (LAMP2A) in mammalian cells, (S)-6-(((allyloxy)carbonyl)amino)-2-hydroxyhexanoic acid AllocK–OH was introduced at defined positions in one monomer of the protein bearing distinct N and C terminal tags. Upon photo-cross-linking with a second monomer of LAMP2A that site specifically incorporated BpA, and hydrolytic cleavage of the ester bond, the sites of cross-linking could be localized to the N- or C-terminal side of the hydroxy acid by gel electrophoresis and immunoblotting against the N- and C-terminal tags.⁵⁰³ This approach defined protein interfaces and provided insight into the in vivo geometry and multimerization state of the LAMP2A complex.

Active site mutants of N⁺-PylRS enzymes were permissive to (S)-2-hydroxy-6-((S)-tetrahydrofuran-2-carboxamido)hexanoic acid THFK–OH, (S)-6-(((benzyloxy)carbonyl)amino)-2-hydroxyhexanoic acid CbzK–OH, AllocK–OH, (S)-3-(3-bromophenyl)-2-hydroxypropanoic acid m-BrF–OH and (S)-2-hydroxy-3-(4-(prop-2-yn-1-yloxy)phenyl)propanoic acid AlkyneY–OH (Figure 18).^497,498 Cellular incorporation of BocK–OH into a protein, followed by hydrazine mediated cleavage of the resulting ester, generated a protein fragment bearing a C-terminal hydrazide; this was chemically ligated to a protein fragment bearing an N-terminal cysteine.⁴⁹⁷

Figure 18.

Hydroxy acids incorporated by PylRS/tRNA^Pylpairsin vivo: Chemical structures of hydroxy acids that have been genetically encoded into proteins with PylRS/tRNA^Pyl pairs. Only examples for which the incorporation has been confirmed by MS are listed. The hydroxy acids BocK–OH, AllocK–OH, AlkynK–OH, ButK–OH, PenK–OH, NorK–OH, CbzK–OH and AcK–OH are all substrates for wt N⁺-MmPylRS. The wt A^Δ-1r26PylRS has a similar substrate scope, but does not recognize NorK–OH, and AcK–OH. The engineered N⁺-MbPylRS(L274A, C313 V) charges THFK–OH, CbzK–OH, AllocK–OH, and BocK–OH. The promiscuous PylRS enzymes N⁺-MbPylRS(L274A, C313 V), and A^Δ-alvPylRS(N166A, V168A), charge the aromatic residues m-BrF–OH and Alkyn-F–OH, or m-CF₃F–OH, respectively. PylRS enzymes were specifically evolved to charge aromatic hydroxy acids, remarkably those enzymes do not show a measurable background incorporation with aromatic canonical amino acids.¹³² The two aromatic hydroxy acid selective mutants N⁺-MmPylRS(M300S, A302H, M344L, N346A) and N⁺-MmPylRS(M300S, A302H, M344L, N346A, C348S, V401L, W417T) both charge F–OH, p-IF–OH and NapA–OH, with the latter being more active with the bulkier substrates, and the former more active with F–OH.

Recent work investigated the metabolism of α-hydroxy acids and their incorporation into proteins in E. coli, using N⁺-MmPylRS/N-MmtRNA^Pyl pairs, in more detail.¹³² When 2 mM BocK–OH was added to E. coli, a substantial fraction was converted to BocK. However, N⁺-MmPylRS/N-MmtRNA^Pyl selectively directed the incorporation of the hydroxy acid into proteins. Remarkably, even upon addition of equimolar amounts of a hydroxy-acid (BocK–OH, AllocK–OH or, CbzK–OH) and an amino acid with an identical side chain, N⁺-MmPylRS selectively mediated the incorporation of the hydroxy acid into proteins in E. coli; there was no detectable amino acid incorporation, as assayed by MS and an ester bond-hydrolysis assay, into the proteins produced in these experiments.

7.1.2. Genetically Encoding Hydroxy Acids with Aromatic Side Chains

Hydroxy acids with aromatic side chains, (S)-2-hydroxy-3-(4-iodophenyl)propanoic acid (p-IF–OH) and (S)-2-hydroxy-3-(naphthalen-2-yl)propanoic acid (NapA–OH), were also substantially converted to the corresponding amino acids when added to E. coli. The deletion of transaminases that may convert a fraction of added hydroxy acids to amino acids, while apparently sufficient to support the incorporation of phenyllactic acid (F–OH) with an engineered MjTyrRS/MjtRNA^Tyr pair,⁵⁰² does not provide a general strategy for ablating the conversion of hydroxy acids to amino acids.¹³² Tyrosyl-tRNA synthetase/tRNA^Tyr pairs, unlike PylRS, recognize the α-amine in their active sites, and they commonly lead to incorporation of amino acids when cells are provided with the corresponding hydroxy acid.¹³² As a result, hydroxy acids bearing aromatic side chains have been challenging to genetically encode using orthogonal TyrRS/tRNA^Tyr pairs.¹³² Recent work leveraged the preference of PylRS enzymes for hydroxy acids over amino acids to evolve N⁺-MmPylRS variants that selectively acylate N-MmtRNA^Pyl with hydroxy acids bearing aromatic side chains (p-IF–OH, NapA–OH, and even F–OH, which only differs from phenylalanine by replacement of the α-amine with a hydroxy group) in E. coli.¹³²

The evolved PylRS enzymes for hydroxy acids with aromatic side chains were selective for these hydroxy acids with respect to hydroxy acids with aliphatic side chains. Similarly, the wt PylRS, and engineered versions of PylRS, for hydroxy acids with aliphatic side chains were selective for these hydroxy acids with respect to hydroxy acids with aromatic side chains. Thus, the active sites of PylRS variants for hydroxy acids with aliphatic side chains and the active sites of PylRS variants for hydroxy acids with aromatic side chains are commonly mutually orthogonal in their hydroxy acid substrate recognition.

A nonpeer reviewed preprint built on the development of a N⁺-MbPylRS variant that encodes pcDap and its deprotection to Dap as a route to trapping hydrolase substrates.^471,472 Building on the preference of PylRS systems for hydroxy acids,^132,499 the authors used the N⁺-MbPylRS variant for pcDap to encode the hydroxy analogue of pcDap, pcDap–OH. Upon photo deprotection of incorporated pcDap–OH, the resulting free amine reversibly attacked the ester bond through a kinetically favored five membered transition state.^504−506 The equilibrium for this reaction favors the formation of the thermodynamically favored peptide bond, and leads to the formation of a protein containing a β-amino acid linkage with a hydroxy substituent on the 2 position. This linkage is formed in low yield, due to low ncM incorporation efficiencies of pcDap–OH (1.4-fold over background misincorporation when incorporated into GFP, and 5.2 fold over background when assayed in a more sensitive NanoLuc expression system).

7.2. In Vitro and Nonspecific Activities of PylRS and Its Designed Variants with ncMs

Evidence that N⁺-MmPylRS is able to aminoacylate N-MmtRNA^Pyl with 6-((tert-butoxycarbonyl)amino)hexanoic acid (BocAhx) - a BocK derivative lacking the α-amine, N^ε-(tert-butoxycarbonyl)-N^α-methyl-L-lysine (MeBocK),⁴⁹⁹BocK–OH and N^ε-(tert-butoxycarbonyl)-D-lysine (D-BocK) was provided by in vitro aminoacylation assays with each compound. These assays used the electrophoretic mobility shift of N-MmtRNA^Pyl upon aminoacylation to follow the acylation reaction, and did not explicitly identify the species that is acylated on to N-MmtRNA^Pyl. Therefore, it remained possible that in some cases the acylation resulted from a contaminant in the reaction; later work did not detect activity for other PylRS enzymes with D-amino acids.⁵⁰⁷ Nonetheless, these experiments provided the first compelling evidence that the weak recognition of the α-amine by PylRS could be exploited to acylate tRNA^Pyl with alternative substrates.

The A^Δ-alvPylRS/alvtRNA^Pyl pair has recently been developed for genetic code expansion (Section 8).³⁸In vitro experiments with this pair, and three of its active site variants (N166A V168L (A^Δ-alvPylRS(1)), N166A V168 K (A^Δ-alvPylRS(2)), and N166A V166 V (A^Δ-alvPylRS(3))), demonstrated that A^Δ-alvPylRS derived enzymes can acylate A^Δ-alvtRNA^Pyl with α-hydroxy-acids (wt and all three mutants), N-formyl-L-α-amino acids (A^Δ-alvPylRS(1) and (2)), as well as α-carboxy acid monomers (all three mutants). A^Δ- alvPylRS(1) and (2) also showed a low, but measurable, acylation activity with α-thio acids and N-methyl-L-α-amino acids. A MS-based assay corroborated the identity of each acylated monomer. A crystal structure of A^Δ-alvPylRS(3) with the α-carboxy monomer 2-(3-(trifluoromethyl)benzyl)malonic acid (m-CF₃BME), and AMP-PNP provided insight into monomer recognition.⁵⁰⁷ The A^Δ-alvPylRS(3) enzyme appeared to display modest selectivity for α-carboxy acid derivatives, over natural amino acids and most of the monomers tested were not loaded onto A^Δ-alvtRNA^Pyl at levels substantially above background mis-acylation with natural amino acids, notably phenylalanine. Each class of monomer (with the exception of N-methyl-L-α-amino acids) was genetically encoded in vitro at the first position of a peptide in an in vitro translation reaction based on start codon skipping and devoid of phenylalanine. This approach ensured that the ribosome did not need to use the variant functional groups in a bond forming reaction, and allowed weak the activities of the synthetases to be utilized in the absence of competition with phenylalanine. The mutant synthetases do not appear to be specific or active enough with these monomers (except the hydroxy acid) to support their site-specific in vivo incorporation.

In another nonpeer reviewed preprint (published in its final form after this review was submitted),⁵⁰⁸ it was reported that wt A^Δ-alvPylRS could aminoacylate its cognate tRNA with (S)-β², and (R)-β²BocK–OH derivatives in vitro. The A^Δ-alvPylRS(3) mutant was reported to accept the (R)-β² hydroxy acid - (R)-3-hydroxy-2-(3-(trifluoromethyl)benzyl)propanoic acid ((R)-β²-m-CF₃F–OH) - with low efficiency, and the addition of the (S)-β² enantiomer did not yield any acylated A^Δ-alvtRNA^Pyl. The wt A^Δ-alvPylRS/A-alvtRNA^Pyl pair facilitated the incorporation of some (S)-6-((tert-butoxycarbonyl)amino)-2-(hydroxymethyl)hexanoic acid ((S)-β²-Bock–OH) monomer at one to two positions in GFP (position 3, and an insertion between positions 213 and 214 into a loop of GFP) in E. coli. However, a substantial amount of glutamine–minimally 30% – was incorporated at the same position as the hydroxy acid monomers, and the overall yield was low. The nonspecific activities reported in these experiments further highlighted the challenges in moving from in vitro acylation–where low and nonspecific synthetase activities can be detected–to in vivo experiments where more active and specific variants are required to outcompete misincorporation processes.

7.3. Evolving PylRS Enzymes for Selective In Vivo Acylation

The most powerful approaches for discovering aaRS/tRNA pairs for new ncAAs have used directed evolution to select synthetase variants from a library of active site mutants (Section 4). These approaches have relied on translational read outs and therefore required the monomers to be ribosomal substrates, frequently at specific positions in the reporter protein. This limitation is not a problem for most α-L-amino acids with variant side chains, which are generally well-tolerated by the ribosome and the rest of the translation machinery. However, many ncMs may not be good in vivo substrates for the translational machinery of cells⁵⁰⁹ and therefore double-sieve selections that use translational readouts are likely of limited utility for discovering orthogonal aaRS/tRNA pairs for ncMs. An evolutionary deadlock has been identified for translation-based selections, including double-sieve selections: ribosomes, and other translational components, cannot be evolved to polymerize ncMs that cannot be acylated onto tRNAs, and aaRS enzymes cannot be evolved to acylate tRNAs with ncMs that are not substrates for translation. Recent work has broken this deadlock by developing direct selections for tRNA acylation (Section 7.3.1),¹⁹ these selections enabled the discovery of PylRS variants for several classes of ncMs (Section 7.3.2), and the addition of several ncMs to the genetic code of E. coli (Section 7.3.3).

7.3.1. The Development of a Scalable tRNA Display Platform

tRNA display enables the direct selection of aaRS enzymes that selectively and efficiently acylate their cognate tRNA with a desired ncM, regardless of whether the ncM, when loaded on to the tRNA, is a substrate for ribosomal polymerization.¹⁹ To develop tRNA display, methods were created to (1) selectively isolate acylated tRNAs with respect to tRNAs that were not acylated, (2), link the in vivo acylation of a tRNA to the sequence of the aaRS mutant responsible for the acylation, and (3) identify aaRS sequences that selectively acylate their tRNA with the ncM of interest rather than a canonical amino acid. These methods link the phenotype (acylation with a ncAA or ncM) to genotype (sequence of the aaRS that performs the acylation).

To selectively isolate the acylated form of N-MmtRNA^Pyl from cells the authors developed–based on the previously developed tRNA extension (tREX) protocol⁸³ (Figure 19a,b) – fluorescent tRNA extension (fluoro-tREX) (Figure 19a,c), and biotin tRNA extension (bio-tREX) (Figure 19a,d). In bio-tREX, tRNAs are isolated from cells, and oxidized with sodium periodate. During oxidation, the acyl-group of the 3′ ribose of acylated tRNAs acts as chemical protecting group, preventing the conversion of the diol functionality into a dialdehyde (preserving the ribose for enzymatic extension). Subsequently, the tRNAs are deacylated, and a DNA probe containing a 5′ overhang with a terminal poly G stretch and sequence complementary to the 3′ end of N-MmtRNA^Pyl is hybridized to the 3′ end of N-MmtRNA^Pyl. The overhang is then extended with the DNA polymerase fragment Klenow (exo-) using a nucleotide mixture containing biotinylated deoxycytidine triphosphates (bio-dCTPs), instead of dCTPs. Only the formerly acylated tRNAs are biotinylated, and can be isolated on streptavidin beads. Bio-tREX provided an efficient method to selectively isolate acylated tRNAs. In fluoro-tREX, Cy5-dCTPs are used instead of bio-dCTPs and the extension product is directly visualized following gel electrophoresis.

Figure 19.

Transfer RNA extension protocols for the analytical, and physical separation of acylated tRNAs from free tRNAs. a, Isolated tRNAs are oxidized with sodium periodate, during which the diol functionality of the 3′-ribose of acylated tRNAs is protected from oxidation to the aldehyde. A DNA probe bearing a 3′Cy3 label is annealed to the tRNA and the tRNA is extended by Klenow (exo-) DNA polymerase fragment. For fluoro- and bio-tREX a DNA probe that has a 5′poly G stretch, and is otherwise devoid of G, is used together with an extension mix that contains dNTPs without dCTPs and is supplemented with either biotinylated- or Cy5 labeled dCTPs. b, In tREX, the difference in mass between the extended (previously acylated) and nonextended (previously nonacylated) tRNAs is visualized by gel electrophoresis to identify acylated tRNAs. c, In fluoro-tREX the previously acylated tRNAs are labeled with Cy5 and can be visualized by gel electrophoresis. d, In bio-tREX, the previously acylated tRNAs are biotinylated, bound to streptavidin beads and can then be eluted and analyzed by gel electrophoresis. Adapted from Dunkelmann et al.¹⁹ – copyright © The Author(s) 2024 CCBY http://creativecommons.org/licenses/by/4.0/.

To link the in vivo acylation of N-MmtRNA^Pyl to the sequence of the N⁺-MmPylRS mutant responsible for the acylation the authors developed several new approaches. First, they showed that the RNA sequence of N-MmtRNA^Pyl can be split at the anticodon to create a 5′ half and a 3′ half, and that these two halves of N-MmtRNA^Pyl can be produced, through in vivo RNA processing, from a single transcript (Figure 20a). The two halves assemble in vivo to form a split tRNA, which is acylated by N⁺-MmPylRS as efficiently as the parent tRNA in E. coli. Next, they covalently linked the mRNA of N⁺-MmPylRS to the 5′ end of the 3′ half of the split tRNA to create a split tRNA-mRNA fusion (stmRNA). The stmRNA encodes the PylRS enzyme (genotype), while also containing the substrate (tRNA^Pyl). Within each cell the activity of the PylRS enzyme determines the extent to which its encoding stmRNA is acylated (Figure 20b).

Figure 20.

Methodological basis of the tRNA display platform. a, Assembly, maturation and acylation of split tRNAs in vivo. Pyrrolysyl tRNA can be split, at the anticodon, into two halves and expressed as circularly permutated tRNA from one construct in cells. The split tRNA is recognized and acylated by PylRS in vivo. b, Split tRNA-mRNA fusions (stmRNAs) can be produced from one transcript by circular permutation of the split tRNA and attaching the PylRS mRNA to the 5′ end of the 3′ half of the tRNA. Split tRNA-mRNA fusions serve as the mRNA for the production of PylRS enzymes, as well as tRNA substrates of PylRS, thereby connecting genotype (PylRS mRNA) to phenotype (acylated tRNA^Pyl). c, Biotin mRNA extension (bio-mREX) leads to the selective isolation of the DNA sequences of active PylRS enzymes. Split tRNA-mRNA fusions are oxidized with sodium periodate. During the oxidation the 3′ end of acylated stmRNAs is protected, while the 3′ end of free stmRNAs is inactivated. A DNA probe is annealed to the stmRNA, extended and the stmRNAs with intact 3′ ends are biotinylated. Therefore, only formerly acylated stmRNAs get biotinylated. The biotinylated stmRNAs are isolated on streptavidin beads, reverse transcribed, and either submitted to quantitative PCR (qPCR) or cloned into a new backbone for multiple rounds of selection. Adapted from Dunkelmann et al.¹⁹ – copyright © The Author(s) 2024 CCBY http://creativecommons.org/licenses/by/4.0/.

Several additional advances enabled researchers to identify aaRS sequences that selectively acylate their tRNA with a monomer of interest. First, the bio-tREX protocol was adapted for the selective capture and reverse transcription of stmRNAs that were acylated in the cell (creating biotin mRNA extension (bio-mREX)) (Figure 20c). Bio-mREX recovered 300-fold more stmRNA molecules encoding wt N⁺-MmPylRS than encoding an attenuated activity mutant of N⁺-MmPylRS, when cells were provided with BocK. Second, the expression level of N⁺-MmPylRS from the stmRNA was tuned such that the recovery of stmRNAs molecules, encoding N⁺-MmPylRS variants for a ncAA, by bio-mREX was well correlated with the amber suppression efficiency of the corresponding N⁺-MmPylRS/N-MmtRNA^Pyl pair with the respective ncAA.

tRNA display used stmRNA libraries (which vary the sequence of N⁺-MmPylRS). These libraries were subject to parallel bio-mREX-based selections in the presence and absence of ncMs and, the N⁺-MmPylRS sequence within the captured stmRNAs was determined by NGS (Figure 21a). A measure of enrichment, a proxy for acylation activity, was provided by comparing the abundance of a sequence after selection in the presence of a ncM to its abundance in the input. A measure of selectivity, a proxy for acylation specificity, was provided by comparing the abundance of a sequence after selection in the presence of the ncM to the abundance of a sequence following selection in absence of the ncM. Desired sequences were enriched and selective.

Figure 21.

Evolution of β-aminoacyl-, β -hydroxyacyl-, and α-, α-disubstituted aminoacyl-tRNA synthetases by tRNA display. a, Schematic representation of the tRNA display protocol. Split tRNA-mRNA fusions encoded PylRS active site libraries are submitted to two parallel bio-mREX experiments, in presence and absence of the target ncM. Isolated cDNA is submitted to NGS. Two parameters are determined: (1) Enrichment–a proxy for acylation activity–which is calculated as the ratio of the relative abundance of a sequence after the selection, divided by the relative abundance of the same sequence in the input library. (2) Selectivity–a proxy for acylation specificity–which is calculated as the relative abundance of a sequence after selection in the presence of the ncM, divided by the relative abundance of the same sequence after selection in absence of the ncM. Enrichment and selectivity are plotted in spindle plots, and highly enriched and selective sequences further characterized. b, Schematic representation of substrates for evolved N⁺-MmPylRS variants. Substrates (S)-β³-m-BrF, (S)-β³-m-CF₃F, (S)-β³-p-BrF, and (S)-α-Me-p-IF–OH were site-specifically genetically encoded in a protein. c, Crystal structure (PDB 8OVY) of β-amino acid (S)- β³-m-BrF at position 150 in green fluorescent protein incorporated with an evolved N⁺-MmPylRS/N-MmtRNA^Pyl pair. Adapted from Dunkelmann et al.¹⁹ – copyright © The Author(s) 2024 CCBY http://creativecommons.org/licenses/by/4.0/.

Experiments with N⁺-MmPylRS libraries and ncAAs that are known ribosomal substrates demonstrated the power of tRNA display. Active and selective sequences were obtained directly from the sequencing data and there was a strong correlation between the enrichment of PylRS sequences in tRNA display and the activity of the resulting N⁺-MmPylRS/N-MmtRNA^Pyl pair in amber suppression-dependent protein expression. N⁺-MmPylRS variants for eight ncAAs were discovered from several large libraries by tRNA display. Highly active and selective pairs, with convergent N⁺-MmPylRS sequences, were discovered directly by sequence-based criteria. These experiments also used 50–100 times less compound than previous double sieve-based selection methods.

A nonpeer reviewed preprint reported a translation independent method for PylRS enzyme engineering (termed “selection for tRNA-acylation without ribosomal translation” (START); this paper was published after submission of this review).⁵¹⁰ Transfer RNA display and START both rely on the periodate oxidation step to chemically differentiate between charged and free tRNAs, and probe-mediated extension of the previously acylated tRNA, as previously described in tREX.^83,511 However, START relies on cutting out the gel band generated by the original tREX protocol to isolate previously acylated tRNAs,⁸³ while tRNA display uses selective biotinylation of the formerly acylated tRNA and subsequent pulldowns. START aims to indirectly correlate genotype and phenotype by adding a barcode to the tRNA anticodon stem loop and sequencing plasmids which encode the barcoded tRNAs gene together with the gene for the active site variants of PylRS. In contrast, tRNA display directly links the tRNA substrate to the mRNA of the synthetase and so directly connects the genotype to the phenotype.

tRNA display leads to 30-fold higher enrichments of active over attenuated PylRS variants (300-fold vs 11-fold), and tRNA display enables effective selections from libraries that are 3 orders of magnitude larger than those interrogated by START (10⁸ member vs 10⁵ members). Furthermore, with tRNA display the active mutants are directly isolated and can be submitted for further rounds of evolution, while START can only be run in one step. While START has been used to evolve active site variants for ncAAs which have previously been incorporated, only tRNA display has yielded PylRS enzymes, which are active and selective with new monomers (Section 7.3.2).¹⁹

7.3.2. PylRS Variants for ncMs from tRNA Display

One or two rounds of tRNA display selection, with highly diverse active site libraries of N⁺-MmPylRS, led to the discovery of active and selective PylRS mutants for eight ncMs. These ncMs included six β³-amino acids, a β³-hydroxy acid, and an α-, α-disubstituted amino acid (Figure 21b). This work provided the first active and selective aaRS enzymes for these three classes of monomers.

Each evolved PylRS variant exhibited ncM dependent in vivo acylation of N-MmtRNA^Pyl, as directly characterized by fluoro-tREX. Importantly, the identity of the ncM loaded onto N-MmtRNA^Pyl was directly confirmed by a MS-based assay. These assays provide important approaches to characterizing the specificity of aaRS systems that load ncMs onto their tRNAs, regardless of whether the acylated tRNAs function in translation.

An important aspect of tRNA display is the wealth of sequence information derived from the analysis of each selection. The observation that N⁺-MmPylRS enzymes accepting β³-amino acids predominantly share a M300D mutation and a A302H mutation, provides insight into sequence motifs that may support recognition of this class of substrates. Interestingly, the aspartic acid in position 300, which may be in proximity to the β³ amine changes to an uncharged asparagine in the case of the mutants observed for β³-hydroxy acid, while the A302H mutation is conserved between both.

The generation of efficient and selective synthetases for some ncMs, was sufficient to enable their incorporation into proteins (Section 7.3.3). However, other ncM that were efficiently and specifically acylated were not incorporated into proteins; the ability to efficiently and selectively acylate tRNAs with these ncMs provides a starting point for using translational selections to evolve translational components (eg: tRNAs, EF-Tu, EF-P, ribosomes) to facilitate their genetically encoded incorporation.

We anticipate that direct selections for acylation will enable the generation of efficient and selective synthetases for diverse new classes of ncMs. It may also allow selection for other enzymatic activities that can be linked to tRNA acylation.

7.3.3. Adding ncMs to the Genetic Code of E. coli Using PylRS/tRNA Pairs from tRNA Display

N⁺-MmPylRS variants discovered by tRNA display, when combined with N-MmtRNA^Pyl in cells containing a GFP reporter with an amber codon at position 150 and added ncM led to clean, site-specific genetic encoding of three β³-amino acids: (S)-3-amino-3-(3-bromophenyl)propanoic acid ((S)-β³-m-BrF), (S)-3-amino-3-(4-bromophenyl)propanoic acid ((S)-β³-p-BrF), (S)-3-amino-3-(3-(trifluoromethyl)phenyl)propanoic acid ((S)-β³-m-CF₃F), and one α-, α-disubstituted amino acid (S)-2-amino-3-(4-iodophenyl)-2-methylpropanoic acid ((S)-α-Me-p-IF) into GFP at position 150, as judged by whole protein mass spectrometry and tryptic MS/MS.

This work constituted the first expansion of the genetic code for the incorporation of these classes of monomers into a protein in a living organism. Protein yields were 3–35 mg L^–1 of culture and the high efficiency of ncM incorporation permitted the production of the first crystal structure of a β³-amino acid containing protein through expansion of the genetic code (Figure 21c).

We anticipate that the ability to genetically encode these new classes of ncMs will enable translational selections to enhance the efficiency with which they are incorporated at diverse sites in proteins. By combining these advances with advances in encoding noncanonical polymers (Section 11) it may be possible to genetically encode the cellular synthesis of noncanonical polymers composed of ncMs.

8. Discovery and Engineering of Orthogonal and Multiply Orthogonal PylRS/tRNA^Pyl Pairs

Recent work has demonstrated that the sequence diversity between and within PylRS/tRNA^Pyl groups discovered by culture and metagenomic approaches^35,37 (Section 3) can be exploited as a starting point for discovering new orthogonal pairs (Section 8.1), sets of mutually orthogonal pairs (Section 8.2), triply orthogonal pairs (Section 8.3), and even quintuply orthogonal pairs (Section 8.4). We define two pairs X and Y as mutually orthogonal if the aaRS of X acylates the tRNA of X, but not the tRNA of Y, and the aaRS of Y acylates the tRNA of Y, but not the tRNA of X. Thus, we define mutual orthogonality in terms of acylation specificity. Triply and quintuply orthogonal pairs extend the idea of mutually orthogonal pairs to three or five pairs. In order to be used to uniquely direct the incorporation of distinct monomers, mutually orthogonal pairs may need further engineering to generate mutually orthogonal active sites–such that the active site of the aaRS of X recognizes its substrate and not the substrate of the aaRS of Y, and vice versa. In order to be used to decode distinct codons, the tRNA of X and the tRNA of Y must be altered, such that they uniquely decode distinct codons.

8.1. A^Δ-PylRS/tRNA^Pyl Pairs Are Active and Orthogonal in E. coli

Pioneering efforts focused on characterizing five N^Δ-PylRS enzymes (from alv, Methanogenic archaeon ISO4-G1 (g1), Methanogenic archaeon ISO4-H5 (h5), Methanonatronarchaeum termitum (term), and Methanomassiliicoccus luminyensis (lum1), respectively), for which tRNA^Pyl like sequences could be identified, in E. coli. All five pyl tRNAs were predicted to fold into a canonical cloverleaf structure. The most notable features of the pyl tRNAs from group ΔN, when compared to pyl tRNAs from group + N, are a shortened D-loop, the absence of the U8 nucleotide between acceptor stem and D-arm from the sequence (with the exception of lum1tRNA^Pyl), as well as a unique bulge or loop in the anticodon stem (with the exception of g1tRNA^Pyl); these features had not been observed in any other tRNAs sequences at the time this work was carried out.^35,38,512

All five N^Δ-PylRS/tRNA^Pyl pairs were active and orthogonal in E. coli. This demonstrated that the N^Δ-PylRS genes encoded functional enzymes that did not require other components to acylate their tRNAs. Compared to the widely used N⁺-MmPylRS/N-MmtRNA^Pyl pair, A^Δ-alvPylRS/A-alvtRNA^Pyl led to higher amber suppression activity, while A^Δ-g1PylRS/A-g1tRNA^Pyl showed comparable activity; the other pairs were much less active. The low activity of some of the systems in E. coli was later rescued by installing a translationally favored A37 nucleotide into the anticodon loop of the pyl tRNAs. A^Δ-g1PylRS, and A^Δ-alvPylRS both belong to the class A (as defined later, Section 8.3) of group ΔN.

The high activity of the A^Δ-alvPylRS/A-alvtRNA^Pyl pair correlates with a very low number of amber stop codons in the parent organism (1.6% of stop codons for alv, and 11% of stop codons for lum1, respectively) as well as a high number of in frame amber codons in proteins (19 for alv and three for lum1, respectively).³⁶ However, in vitro aminoacylation kinetics for N⁺-MmPylRS/N-MmtRNA^Pyl, N⁺-MbPylRS/N-MbtRNA^Pyl, and A^Δ-alvPylRS/A-alvtRNA^Pyl pairs with Pyl suggested that the k_cat/K_m for the A^Δ-alvPylRS/A-alvtRNA^Pyl pair is 3-fold lower than for the N⁺-MmPylRS/N-MmtRNA^Pyl, and 11-fold lower than for the N⁺-MbPylRS/N-MbtRNA^Pyl pair.⁴² It is not uncommon for in vitro acylation kinetics to correlate poorly with in vivo incorporation measurements; there are several possible reasons for this: (1) in vitro measurements commonly use in vitro transcribed tRNAs–which may not reflect the modification, processing or folding state of tRNAs in cells–and (2) in vitro acylation measurements report on only one step in in vivo ncAA incorporation, which includes ncAA uptake and availability, EF-Tu binding, and ribosomal decoding.

As it was commonly assumed that PylRS systems required a N-terminal domain (in cis or trans) to function efficiently in vivo, the discovery and characterization of numerous PylRS systems that do not require an N-terminal domain and are very active and orthogonal was a surprising and important advance.

8.2. Mutually Orthogonal PylRS/tRNA^Pyl Pairs

A body of work has investigated the discovery and generation of mutually orthogonal PylRS/tRNA^Pyl pairs. Natural mutually orthogonal pairs have been discovered from distinct organisms and within a single organism (Section 8.2.1), and optimized mutually orthogonal pairs have been evolved and engineered in E. coli (Section 8.2.2), and subsequently been developed in mammalian cells (Section 8.2.3).

8.2.1. Natural Mutually Orthogonal PylRS/tRNA^Pyl Pairs

A screen of the amber suppressor activities of intergroup and intragroup combinations of PylRS enzymes and pyl tRNAs provided pivotal insights into the preferences of PylRS enzymes from groups + N and ΔN for different pyl tRNAs.³⁸ Both N^Δ-PylRS enzymes, A^Δ-alvPylRS as well as A^Δ-g1PylRS, led to almost no amber suppression when paired with group + N tRNA N-MmtRNA^Pyl. This observation constituted the first example of complete in vivo orthogonality of PylRS enzymes toward pyl tRNAs, which are active with at least one alternative PylRS enzyme. Remarkably, the screen resulted in the discovery of a naturally mutual orthogonal PylRS/tRNA^Pyl pair, composed of the group ΔN pair (A^Δ-g1PylRS/A-g1tRNA^Pyl) and a group + N pair (N⁺-MmPylRS/N-MmtRNA^Pyl). However, the lower activity of A^Δ-g1PylRS in comparison to A^Δ-alvPylRS combined with a small, but measurable, cross-reactivity of N⁺-MmPylRS with A-g1tRNA^Pyl incentivized the development of more active mutually orthogonal pairs based on the highly active A^Δ-alvPylRS/A-alvtRNA^Pyl pair (Section 8.2.2).

Recent work has discovered genes for two sets of ΔN group PylRS/tRNA^Pyl pairs encoded in the genome of a single organism, the extremely halophilic euryarchaeal methanogen Candidatus Methanohalarchaeum thermophilum (therm).⁵¹³ The two C^Δ-thermPylRS pairs (1 and 2–the numbering is in accordance with the characterization of thermtRNA^Pyl(1) in E. coli in parallel work⁴⁰) are formed from distinct C^Δ-PylRS enzymes and C-pyl tRNAs. An amino acid deletion in sequence motif 2 of C^Δ-thermPylRS1 distinguishes it from C^Δ-thermPylRS2 (the enzymes have a sequence identity of 53%), and the most notable difference between the C-pyl thermtRNAs appears to be the change of the discriminator base from the canonical G in C-thermtRNA^Pyl(2) to A in C-thermtRNA^Pyl(1) (Figure 22a). Characterization of the C^Δ-thermPylRS/C-thermtRNA^Pyl pairs in the model halophilic archaeon Haloferax volcanii (no activity was measured in amber suppression experiments in E. coli) revealed that the two pairs are mutually orthogonal, and that the mutual orthogonality is a direct consequence of both the discriminator base differences and amino acid deletion in motif 2 (Figure 22b). This work demonstrated the occurrence of naturally, mutually orthogonal aaRS/tRNA pairs of the same iso-acceptor class, derived from the same host species.

Figure 22.

Naturally mutually orthogonal PylRS/tRNA^Pylpairs in archaea. a, The archaeon Candidatus Methanohalarchaeum thermophilum (therm) harbors two distinct PylRS/tRNA^Pyl pairs in its genome. The two pyl tRNAs carry distinct discriminator bases (DBs). b, The intraorganism mutual orthogonality between therm(1)PylRS/thermtRNA^Pyl(1) pair and therm(2)PylRS/thermtRNA^Pyl(2) is derived from the combination of an unusual DB in the thermtRNA^Pyl(1) which is only recognized therm(1)PylRS that has a shortened motif 2, and is orthogonal to thermtRNA^Pyl(2). Experiments defining mutual orthogonality were carried out in the model archaeon Haloferax volcanii.

Pyrrolysine systems are the only isoacceptor systems for which orthogonal and mutually orthogonal pairs have been discovered within an isoacceptor class (and both inter and intra organism examples have been reported). This may be a function of: (1) the sequence and structural differences of pyl pairs from aaRS/tRNA pairs for canonical amino acids that favors orthogonality of pyl pairs to canonical pairs, and (2) the large sequence diversity of pyl pairs that has arisen through natural evolution. We note that these two points may be related, as the sequence and structural differences of pyl systems to canonical pairs may make their evolutionary trajectories less constrained–mutations in pyl systems that lead to toxic cross reactivity with canonical systems may be rarer than mutations in canonical systems that lead to toxic cross reactivity with each other.

8.2.2. Engineered Mutually Orthogonal PylRS/tRNA^Pyl Pairs

Since the combination of N⁺-MmPylRS and A-alvtRNA^Pyl led to high levels of amber suppression, the binding interface between N⁺-MmPylRS and A-alvtRNA^Pyl was targeted for disruption in an effort to generate a highly active mutually orthogonal PylRS/tRNA^Pyl pair. The distinct architectures of N⁺-PylRS and N^Δ-PylRS enzymes implied distinct tRNA^Pyl binding modes, with the absence of the N-terminal domain presumably weakening T-stem and D-stem, and T-loop and variable loop recognition by N^Δ-PylRS enzymes (Figure 23a).⁶³ The authors hypothesized that extension of the variable loop of A-alvtRNA^Pyl might be tolerated by N^Δ-PylRS enzymes but not N⁺-PylRS enzymes (Figure 23b).³⁸

Figure 23.

Engineering mutually orthogonal PylRS/tRNA^Pylpairs inE. coli. a, Structure of N-MmPylRSn (pink) bound to MmtRNA^Pyl (gray) (5UD5).⁶⁰ The interaction interface of N-MmPylRSn with the variable loop (blue) is dependent on the presence of the N-terminal domain. The absence N-PylRSn and therefore variable loop recognition in group ΔN PylRS enzymes provides a direct means to engineering PylRS specificity by variable loop extension, impeding the interaction interface with N⁺-MmPylRS. b, Depiction of A-alvtRNA^Pyl libraries with extended anticodon loops. Nucleotides marked in blue were randomized, and changes to the nucleotides marked in yellow emerged as unprogrammed mutations during the selection. c, Schematic representation of the strategy employed to generate mutually orthogonal PylRS/tRNA^Pyl pairs formed of N⁺-MmPylRS/N-MmtRNA^Pyl and A^Δ-alvPylRS/A-alvtRNA^Pyl by (1) identifying A-alvtRNA^Pyl variants with an expanded variable loop that are active with A^Δ-alvPylRS, and (2) screening against cross-reactivity with N⁺-MmPylRS.

To engineer orthogonal A-alvtRNA^Pyl variants, libraries with randomized variable loops of three to six nucleotides were subjected to a round of positive selection for activity with A^Δ-alvPylRS and a negative screen for absence of activity with N⁺-MmPylRS. As hypothesized, A^Δ-alvPylRS enzymes tolerated variable loops of up to six nucleotides in A-alvtRNA^Pyl while retaining high activity. On the other hand, variable loop extensions in A-alvtRNA^Pyl completely abolished the activity with the N⁺-MmPylRS enzyme (Figure 23c). Notably, unprogrammed compensatory mutations at position G60A, T, or C, or C21T of A-alvtRNA^Pyl emerged spontaneously in functional hits with extended variable loops; G60 is in direct contact with the variable loop and subsequent work showed that the mutation of G60 is necessary for enhanced activity and orthogonality of at least two extended variable loop sequences.¹⁵⁵

Subsequent experiments have probed the nucleotides in A^Δ-alvtRNA^Pyl that may confer specificity for A^Δ-alvPylRS.⁵¹⁴ The base pairs G3:C70, G5:C68, and the wobble pair G28:U42 in A^Δ-alvtRNA^Pyl were replaced with the corresponding base pairs from N-MmtRNA^Pyl, and the canonical U8 base from N-MmtRNA^Pyl was inserted between the D loop and acceptor stem in A^Δ-alvtRNA^Pyl. tRNAs bearing these substitutions and insertions were acylated less efficiently with A^Δ-alvPylRS, which led to the hypothesis that the nucleotides at these positions of A^Δ-alvtRNA^Pyl may be identity elements for A^Δ-alvPylRS. Transplanting the putative identity elements of A^Δ-alvtRNA^Pyl into N-MmtRNA^Pyl led to its acylation by A^Δ-alvPylRS. The introduction of a G:C base pair in place of A26:U44 in N-MmtRNA^Pyl further increase the activity with A^Δ-alvPylRS both, in vitro and vivo.

Superimposing the structures of A^Δ-alvtRNA^Pyl and N-MmtRNA^Pyl, obtained from cryogenic electron microscopy, provided a picture of how the absence of U8, and presence of the G28:U42 wobble pair, in A^Δ-alvtRNA^Pyl may alter the fold of this tRNA with respect to N-MmtRNA^Pyl.¹²⁴ Molecular dynamics simulations suggested that G-C base pairs in the acceptor stem (G3:C70 and G5:C68) rigidifies A-alvtRNA^Pyl with respect to N-MmtRNA^Pyl which contains A-U pairs at these positions. The authors suggested that tRNA shape and rigidity may contribute to synthetase specificity.

Introducing mutations that confer specificity for N^ε-((((1R,8S,9r)-bicyclo[6.1.0]nonan-9-yl)methoxy)carbonyl)-L-lysine (BCNK) or CbzK in N⁺-MmPylRS into the active site of A^Δ-alvPylRS generated BCNK or CbzK specific version of the enzyme; this demonstrated that, in some cases, the specificity of new PylRS systems may be altered to recognize new ncAAs by transplanting mutations from previously engineered PylRS systems into the homologous active sites of new PylRS systems. Additional studies of the A^Δ-alvPylRS/A-alvtRNA^Pyl pair have further investigated its substrate scope, and the scope of active site transfer for generating new ncAA specificity.^{9,49,241,501,515−517}

N⁺-MbPylRS systems had also previously been discovered with mutually orthogonal active sites. A CbzK specific mutant recognized CbzK but not N^ε-(((2-methylcycloprop-2-en-1-yl)methoxy)carbonyl)-L-lysine (CypK), and wt N⁺-MbPylRS recognized CypK but not CbzK. Transplanting the CbzK specific mutations into A^Δ-alvPylRS generated a A^Δ-alvPylRS variant that directed the incorporation of CbzK but not CypK, while N⁺-MmPylRS directed the incorporation of CypK but not CbzK.³⁸ This demonstrated that active sites of mutually orthogonal PylRS/tRNA^Pyl pairs can be diverged to generate mutually orthogonal ncAA substrate specificity.

To direct the incorporation of two distinct ncAAs into a protein in E. coli the anticodon of N-MmtRNA^Pyl was expanded to decode a quadruplet codon (AGGA). The resulting N⁺-MmPylRS/N-MmtRNA^Pyl_UCCU and engineered A^Δ-alvPylRS/A-alvtRNA^Pyl(6)_CUA pair were used to incorporate CypK and CbzK in response to AGGA and UAG codons on an orthogonal message decoded by an orthogonal quadruplet decoding ribosome.³⁸ Overall, this work engineered highly active mutually orthogonal PylRS/tRNA pairs to incorporate distinct ncAAs into proteins in response to distinct codons.

8.2.3. Mutually Orthogonal PylRS/tRNA^Pyl Pairs in Mammalian Cells

The mutually orthogonal N⁺-MmPylRS/N-MmtRNA^Pyl and A^Δ-alvPylRS/A-alvtRNA^Pyl pairs developed in E. coli(38) were also orthogonal and mutually orthogonal in mammalian cells.⁹¹ This demonstrated that directed evolution and screening platforms in E. coli could be exploited to create mutually orthogonal PylRS/tRNA^Pyl pairs for use in higher organisms. Independent work generated mutually orthogonal PylRS/tRNA^Pyl pairs from N⁺-MmPylRS/N-MmtRNA^Pyl and A^Δ-alvPylRS/A-alvtRNA^Pyl by testing variable loop mutants of A-alvtRNA^Pyl in mammalian cells.¹⁵⁴ This route was lower throughput than the selection in E. coli, but still led to solutions with some activity in mammalian cells. However, low throughput screening in mammalian cells could not discover the spontaneous mutations that enhanced the activity of the mutually orthogonal A-alvtRNA^Pyl variants from the E. coli selection.³⁸

The insights gained during the E. coli evolution of A-alvtRNA^Pyl permitted the rational engineering of improved mutual orthogonality for the N⁺-MmPylRS/N-MmtRNA^Pyl and A^Δ-g1PylRS/A-g1tRNA^Pyl pairs in mammalian cells. The natural mutual orthogonality of these pairs had previously been established in E. coli,³⁸ but low-level acylation of A-g1tRNA^Pyl by N⁺-MmPylRS remained.

To minimize this mis-acylation, the acceptor stem sequence of A-g1tRNA^Pyl was converted to the acceptor stem sequence of A-alvtRNA^Pyl, and the G60A mutation and extended variable loop sequence discovered in selections with A-alvtRNA^Pyl were transplanted into A-g1tRNA^Pyl. The resulting A-g1tRNA^Pyl derivative showed improved orthogonality with respect to N⁺-MmPylRS, while retaining most of its activity with A^Δ-g1PylRS. These pairs were further engineered to recognize distinct ncAAs and decode distinct stop codons in mammalian cells, providing the basis for an approach for dual fluorescent labeling of proteins in mammalian cells.¹⁵⁵

8.3. Discovery and Engineering of Triply Orthogonal PylRS/tRNA^Pyl Pairs

Interrogation of the sequence and function of 11 PylRS/tRNA^Pyl pairs from the ΔN group (including newly identified PylRS and tRNA^Pyl genes) led to the discovery of triply orthogonal PylRS/tRNA^Pyl pairs.³⁹

Two distinct clusters emerged when the 11 tRNA^Pyl sequences from the ΔN group were clustered. One cluster (termed sequence class A – containing A-alvtRNA^Pyl, an engineered version of which together with A^Δ-alvPylRS forms a mutually orthogonal pair with the class N pair N⁺-MmPylRS/N-MmtRNA^Pyl) contained four pyl tRNAs. The other cluster (termed sequence class B) contained seven pyl tRNAs. Interestingly, the bases conserved within the tRNA^Pyl sequences of each sequence class mainly occurred in the T-stem, T-loop, and acceptor stem; the latter two sequence motifs form important contacts with PylRS enzymes. When using the ΔN-PylRS sequences for hierarchical clustering, the same clusters emerged as seen for pyl tRNAs.³⁹ These observations suggested that aaRS and tRNA sequences in the two classes might have distinct molecular recognition properties.

The measured specificities of the synthetases and tRNAs were generally well correlated with the observed clustering of their sequences: class A synthetases preferentially acylated class A pyl tRNAs over class B pyl tRNAs, and class B synthetases preferentially acylated class B pyl tRNAs over class A pyl tRNAs. Synthetases and pyl tRNAs were grouped into functional classes A and B on the basis of their activity. The N^Δ-termPylRS/termtRNA^Pyl pair was an interesting exception; the PylRS of this pair is exceptionally orthogonal and only interacts with its cognate termtRNA^Pyl, while the termtRNA^Pyl of this pair is a substrate for all tested N^Δ-PylRS enzymes. The screen resulted in the discovery of 18 naturally mutually orthogonal PylRS/tRNA^Pyl pairs, each composed of an A^Δ-PylRS/A-tRNA^Pyl pair, and a B^Δ-PylRS/B-tRNA^Pyl pair.³⁹

The correlation between classes derived from the hierarchical clustering of the tRNA^Pyl sequences, and the functional analysis of the N^Δ-PylRS/tRNA^Pyl pairs hints at distinct identity elements for each class of pyl tRNAs. Those identity elements likely facilitate the class-specific tRNA^Pyl recognition by the cognate N^Δ-PylRS enzymes, which in itself will be encoded in the amino acid sequence of the synthetases.³⁹

It was postulated that triply orthogonal PylRS/tRNA^Pyl pairs could be engineered by combining the mutually orthogonal pairs derived from class N and class A, with a pair from the newly discovered class B. By screening several natural class N pyl tRNAs the authors discovered four N-pyl tRNAs which retained activity with the N⁺-MmPylRS enzyme, while being orthogonal to all class A and B N^Δ-PylRS enzymes. The cognate pair N⁺-MmPylRS/ Methanosarcina spelaei (spe) N-spetRNA^Pyl resulted in high activity and formed the first pair of the triply orthogonal set. Interestingly, N-spetRNA^Pyl contains only one change in sequence with respect to N-MmtRNA^Pyl, the Watson–Crick pair C6:G67 is converted wobble pair U6:G67; this change is sufficient to render N-spetRNA^Pyl entirely orthogonal to class A and B N^Δ-PylRS enzymes (Figure 24a).³⁹

Figure 24.

Engineering triply orthogonal PylRS/tRNA^Pylpairs inE. coli. a, The screen of natural variance in N-pyl tRNAs led to the discovery that N-spetRNA^Pyl is for class N tRNA^Pyl triply orthogonal sets. b, Depiction of previously engineered A-alvtRNA^Pyl variable loop extension mutants.³⁸ Three mutants fulfilled the requirements to form the A-pyl tRNAs in triply orthogonal sets. c, Directed evolution strategy for identifying B-inttRNA^Pyl variants completing the triply orthogonal set. Acceptor stem and variable loop libraries were run independently and the most orthogonal B-inttRNA^Pyl variants combined into hybrid B-pyl inttRNAs. Multiple B-pyl inttRNAs fulfilled the orthogonality requirement to form triply orthogonal sets when paired with a B^Δ-PylRS, a select class A N^Δ PylRS/tRNA^Pyl pair, and N+MmPylRS/N-spetRNA^Pyl. d, Summary of the experimental strategy to generate triply orthogonal PylRS/tRNA^Pyl pairs including a depiction of all interactions that were controlled in the process (red arrows indicate undesired activity, gray dashed arrows orthogonality). Adapted with permission from Dunkelmann et al.³⁹ - copyright ©2020 Nature Springer Limited.

To generate triply orthogonal class A pyl tRNAs, high cross reactivity with class N PylRS was controlled. Three A-alvtRNA^Pyl variants with expanded variable loops fulfilled the criteria for orthogonality,³⁸ providing A-tRNA^Pyl variants which were only active with class A N^Δ-PylRS enzymes and orthogonal to PylRS enzymes from class B, and class N, as required for the triplet set (Figure 24b).

Finally, B-pyl tRNAs were engineered to control potential cross reactivities with both, class N, and class A PylRS enzymes. This was achieved through parallel screening of acceptor stem and variable loop libraries, followed by the combination of the best hits from each library into hybrid B-pyl tRNAs. This led to the identification of several triply orthogonal B-Candidatus Methanomassiliicoccus intestinalis (int)tRNA^Pyl variants (Figure 24c). Overall, this work generated 12 sets of engineered, triply orthogonal PylRS/tRNA^Pyl pairs. Each set of triply orthogonal pairs was composed of pairs derived from class A, class B and class + N (Figure 24d).³⁹

The triply orthogonal pairs: N⁺-MmPylRS/N-spetRNA^Pyl_CUA, Candidatus Methanomethylophilus sp. 1r26 (1r26) A^Δ-1r26PylRS(CbzK)/A-alvtRNA^Pyl-8_UACU, and B^Δ-lum1PylRS(NmH)/B-inttRNA^Pyl-a17-vC10_UCCU, in which the synthetases had been altered to recognize distinct ncAAs and the anticodons had been altered to read distinct quadruplet codons or the amber codon, were used to incorporate three distinct ncAAs into a protein. These experiments were performed in E. coli cells bearing an orthogonal message, O-_StrepGFP(40TAG, 136AGGA, 150AGTA)_His6, and an orthogonal ribosome (ribo-Q1) that reads this message and efficiently decodes amber and quadruplet codons.³⁹ When an engineered set of triply orthogonal PylRS/tRNA^Pyl pairs was later paired with computationally designed, highly active orthogonal mRNAs 2.6 ± 0.4 mg L^–1 of GFP protein containing three distinct ncAAs was isolated (approximately 9% of wt GFP yield).⁵¹⁸

Furthermore, the triply orthogonal set composed of N⁺-MmPylRS/N-spetRNA^Pyl_UCUA, A^Δ-g1PylRS(CbzK)/A-alvtRNA^Pyl-8_UACU, and Methanomassiliicoccales archaeon RumEn M1 (rum) B^Δ-rumPylRS(NmH)/B-inttRNA^Pyl-a17-vC10_UCCU, combined with the orthogonal Archaeoglobus fulgidus (Af) AfTyrRS(PheI)/AftRNA^Tyr-A01_CUAG, enabled the production of a GFP protein with four distinct ncAAs in its sequence. Each ncAA was incorporated in response to a quadruplet codon in a computationally designed O-mRNA read by ribo-Q1, thereby realizing a 68-codon code to incorporate four distinct ncAAs.⁵¹⁸ The protein yield was 0.41 ± 0.03 mg L^–1 (approximately 2% of wt GFP yield); this corresponds to an average quadruplet decoding efficiency of 38% per quadruplet, highlighting the suitability of engineered pyl tRNAs for decoding quadruplet codons.

8.4. Discovery and Engineering of Quintuply Orthogonal PylRS/tRNA^Pyl Pairs

The sequence-activity data gathered from screening the amber suppression efficiency of PylRS/tRNA^Pyl pairs, in which the (class A or class B) PylRS enzyme and tRNA^Pyl are from different organisms, revealed clear sequence thresholds for orthogonality. When two PylRS C-terminal domain sequences from different organisms share a sequence identity of 55% or more, the likelihood of a PylRS enzyme being active with the tRNA^Pyl from the other organism is approximately 90%. PylRS enzymes with lower sequence identity than 55% in their C-terminal domain may or may not form active pairs with the other tRNA^Pyl. Similarly, when two pyl tRNAs from different organisms share a 75% or higher sequence identity, the corresponding PylRS enzymes will aminoacylate the tRNA^Pyl from the other organism in approximately 90% of cases. The PylRS enzymes of pyl tRNAs which share lower than 75% sequence identity, may or may not aminoacylate the other tRNA^Pyl.

PylRS C-terminal domain sequences, in a database of 351 sequences, were clustered according to the experimentally determined sequence identity thresholds for synthetase orthogonality, leading to 37 distinct clusters. Twenty-seven clusters belonged to the Ns group (25 of bacterial origin, and two of archaeal origin) and seven belonged to the archaeal ΔN group. Pyrrolysyl tRNAs, from the same organism as representative PylRS sequences, were identified for 95% of the PylRS clusters. These tRNAs were clustered using the experimentally determined threshold for tRNA orthogonality, leading to two new pyl classes. Class C, which constituted a loosely related group of six pyl tRNAs from the ΔN group, and class S, which was composed of all 25 bacterial members of the N group. Sixteen pyl tRNAs (and their synthetases) – including both representative members of all tRNA clusters and tRNAs selected for their unusual structural features (Figure 25) – were selected for experimental characterization. Out of the 16 PylRS/tRNA^Pyl pairs, three originated from the previously characterized classes A, B, and N, seven from the bacterial class S, and six from the archaeal class C.

Figure 25.

The archetypical tRNA structures for pyl tRNAs of classes N, A, B, C, and S. Important nucleotides and features are highlighted in the class-specific colors and labeled. Adapted with permission from Beattie et al.⁴⁰- copyright © 2023 Nature Springer Limited.

The PylRSn domain of multiple class S PylRS enzymes remained recalcitrant to cloning, and even in the instances of successful class S⁺-PylRS assembly into expression constructs, the transformed E. coli strains displayed attenuated growth phenotypes. To remove the toxicity effects, the researchers engineered the synthetic class S^Δ for all S⁺-PylRS enzymes by removing the N-terminal domains from the constructs.

The amber suppression activity of the selected pyl tRNAs with selected synthetases were measured in an activity screen. Fifteen out of 16 pyl tRNAs resulted in AllocK dependent amber suppression, when paired with at least one PylRS enzyme. Thirteen out of 20 (including four instances where group S PylRS enzymes were tested with and without N-terminal domain) tested PylRS enzymes led to amber decoding dependent GFP production at a level at least 30% of the wt GFP. Forty-six naturally mutually orthogonal pairs emerged from these experiments. Mutually orthogonal pairs that used the same set of PylRS enzymes but different pyl tRNAs were defined as belonging to the same family. There were 15 families of mutually orthogonal pairs, and one family of triply orthogonal pairs. Overall, class S^Δ PylRS enzymes did not form an independent reactivity pattern, but were classified as B-like (S^ΔB), or C-like (S^ΔC).

To engineer quintuply orthogonal sets of PylRS/tRNA^Pyl pairs from the five pyl classes, 25 pairwise interactions between PylRS and pyl tRNAs needed to be controlled (five cognate interactions, 20 noncognate interactions). At the interclass and intraclass level, 20 of desired pairwise interactions were exemplified in the activity screen, while in the five pairwise interactions that were not fully exemplified at the class level from the screen, only one of two interactions remained to be controlled. This data did not provide specific pairs within each class that exemplified all the properties that could be found within the class as a whole.

A set of PylRS/tRNA^Pyl pairs, with one pair from each of the five classes was defined as the starting point for creating a set of quintuply orthogonal pairs (Figure 26). To control all pairwise interactions, within and between five specific PylRS/tRNA^Pyl pairs a sequence of screens and directed evolution experiments were performed (Figure 26). This process led to the discovery of 924 mutually orthogonal pairs in 22 families, 1,324 triply orthogonal pairs in 18 families, 128 quadruply orthogonal pairs in 7 families, and 8 quintuply orthogonal pairs in 1 family. All of the newly engineered PylRS/tRNA^Pyl sets, met or exceeded, the orthogonality of all previous triply orthogonal pairs. It is remarkable that while no mutually orthogonal pair has been discovered within any other isoacceptor class the pyl isoacceptor class has provided quintuply orthogonal pairs. Future work may aim to develop these pairs to decode distinct codons and incorporate distinct ncAAs or ncMs.

Figure 26.

Engineering quintuply orthogonal PylRS/tRNA^Pylpairs inE. coli. Summary of the experimental strategy to generate quintuply orthogonal PylRS/tRNA^Pyl pairs including a depiction of all interactions that were controlled in the process (red arrows indicate undesired activity, gray dashed arrows orthogonality). Starting from a rationally chosen set of five PylRS/tRNA^Pyl pairs composed of one pair of each pyl class N, A, B, C and S, a series of screens and directed evolution experiments resulted in the generation of quintuply orthogonal sets. Quintuply orthogonal pyl tRNAs for B^Δ-PylRS and S⁺-PylRS, respectively, were identified by a round of positive selection of the depicted tRNA^Pyl library, which is based on the S-i2tRNA^Pyl scaffold, in the presence of either B^Δ-PylRS or S⁺-PylRS followed by negative screens against all other classes of PylRS enzymes (N, A, C and S or N, A, B and C, respectively).

9. Directing pyl Systems to New Codons

PylRS enzymes do not recognize the anticodon of pyl tRNAs (Section 3.6),^60,63,75 which has enabled pyl tRNAs to be directed to different codons, including stop codons (Section 9.1) and quadruplet codons (Sections 9.2), codons with noncanonical bases (Section 9.3) and sense codons (Section 9.4). Directing PylRS/tRNA^Pyl to sense codons provides a basis for proteome labeling strategies (Section 9.4.1). These approaches–in combination with strategies for generating mutually orthogonal synthetases and tRNAs (Section 8), strategies for generating orthogonal ribosomes that more efficiently read quadruplet codons and stop codons, and the synthesis of genomes with compressed genetic codes–have enabled the incorporation of multiple distinct ncAAs into a protein (Section 10) and the encoded synthesis of entirely non canonical polymers (Section 11).

9.1. Directing ncAA Incorporation at a Stop Codon

The absence of anticodon recognition by PylRS enzymes permitted the facile reassignment of tRNA^Pyl_CUA to the two alternative stop codons, the opal (UGA) and ochre (UAA) codon in E. coli.⁷⁵

9.2. Directing ncAA Incorporation at a Quadruplet Codon

As part of an in vitro screen for quadruplet decoding tRNAs, chemically acylated N-MactRNA^Pyl variants with extended anticodons were used to decode quadruplet codons.^519,520 Later, evolution of N-MmtRNA^Pyl_UCCU variants for improved AGGA decoding in cells demonstrated that directed evolution experiments could improve quadruplet decoding of N-MmtRNA^Pyl_UCCU, and the selected N-MmtRNA^Pyl_UCCU variants enabled quadruplet decoding in mammalian cells.⁸⁰ Recently, a quadruplet decoding tRNA^Pyl variant has been used to incorporate ncAAs in response to a quadruplet codon introduced into C.elegans(165) and quadruplet decoding N-pyl MmtRNAs have been optimized for the decoding of UAGA, AGGA, AGUA, and CUAG on wt messages in mammalian cells.⁵²¹ The reliance of these approaches on the natural ribosome, which does not efficiently decode quadruplet codons, may have limited the discovery of highly active, quadruplet decoding pyl tRNAs in these systems.⁷⁶

N-MbtRNA^Pyl was systematically evolved for enhanced decoding of UAGA, AGGA, AGUA, and CUAG codons on orthogonal messages by the orthogonal quadruplet decoding ribosome RiboQ1 (O-RiboQ1) in E. coli.⁴⁰⁶ These quadruplet codons were chosen because the triplets that form the first three bases of the quadruplet are not recognition elements for endogenous aaRS enzymes in E. coli. Eight bases at positions flanking the extended anticodon, in the anticodon stem and anticodon loop of N-MbtRNA^Pyl_XXXX, (tRNAs with fixed sequence extended anticodons) were randomized. The resulting libraries were subjected to a double-sieve selection to select against tRNAs that did not function with endogenous synthetases, and to select for tRNAs that decoded their cognate codon when provided with N⁺-MbPylRS. For each quadruplet codon, an evolved N-MbtRNA^Pyl_XXXX mutant that worked with Ribo-Q1 to decode its cognate codon on an orthogonal message was identified.

9.3. Directing ncAA Incorporation at Codons with Noncanonical Bases

A body of work has developed non-natural “bases pairs” that do not pair with natural bases, and interact with each other in a variety of ways, including hydrophobic stacking and cross intercalation.² The NaM:TPT3 pair (Figure 27a) was incorporated into a plasmid at the second position of the anticodon of N-MmtRNA^Pyl gene and in the second position of a codon in a GFP gene. The (deoxy) “nucleotide” triphosphates of NaM and TPT3 were taken up by E. coli, provided with a heterologous nucleotide triphosphate transporter, and the deoxyNaMtriphosphate (dNaMTP) and deoxyTPT3triphosphate (dTPT3) were used to replicate the plasmid containing the non-natural base pair.⁷⁸ While substantial loss of the base pair occurred through replication and DNA repair, this loss could be counteracted to some extent by targeting the sequences resulting from repair for Cas9 cleavage.⁷⁸ T7 transcription through the non-natural base pair (using NaMTP and TPT3) produced N-MmtRNA^Pyl with a non-natural base at position 2 of its anticodon and a GFP mRNA with the complementary non-natural base at position 2 of a codon. Translation with the N⁺-MmPylRS, the engineered N-MmtRNA^Pyl, and engineered GFP mRNA enabled protein production; 90% of the resulting protein incorporated a ncAA substrate of PylRS in response to the codon with the noncanonical base (Figure 27b). The authors have explored putting non-natural bases at different positions of the codon anticodon interaction using N-MmtRNA^Pyl variants; the NaM:TPT3 pair was not tolerated at the first or third position of the codon, but a NaM:NaM homopair between the third position of the codon and the third position of the anticodon was tolerated in some contexts.⁷⁹

Figure 27.

Transcription and translation with synthetic bases. a, Structure of synthetic base pair (d)NaM(dX):(d)TPT3(dY). b, A heterologous nucleotide transporter enables the uptake of the triphosphate of dX, dY, X and Y in E. coli. A plasmid harboring a tRNA gene with an anticodon containing a synthetic codon, as well as a gene using a synthetic codon in its sequence is transformed and maintained in E. coli. The gene and tRNA harboring synthetic codons are transcribed and used by the ribosome in translation. Adapted with permission from de la Torre et al.² - copyright © 2021 Nature Springer Limited.

The orthogonality of codon anticodon interactions with non-natural bases at position 2 or 3 of the codon, and different natural bases at the remaining positions the codon, were explored using variant N⁺-MmPylRS/N-MmtRNA^Pyl pairs and an azido containing pyl analogue as a substrate. This screen identified a triply orthogonal set of codon anticodon interactions: AXC:GYT; GXT:AYC; AGX:XCT, where X= NaM and Y= TPT3, and the sequences are written codon 5′-3′:anticodon 5′-3′.⁷⁹

9.4. Efforts to Direct ncAAs at Sense Codons

In vitro studies demonstrated that the anticodon of N⁺MbtRNA^Pyl can be changed to various sense codons without apparent effects on aminoacylation efficiency by PylRS enzymes.^58,75 Sense codons cannot commonly be reassigned to arbitrary ncAAs in cells for long periods. The sustained efficient reassignment of genomic sense codons to ncAAs leads to proteome mis-synthesis, and target genes containing sense codons are decoded by essential cellular tRNAs. Efforts to reassign the rare arginine codon AGG to homoarginine in E. coli, combined an engineered N⁺-MmPylRS/N-MmtRNA^Pyl_CCU with temperature-dependent depletion of the endogenous EctRNA^Arg_UCU that normally reads genomic AGG codons, and the recoding of 38 AGG codons in 32 essential genes to synonymous codons. Since homoarginine was somewhat tolerated in place of arginine in the genomically encoded proteins, this approach permitted the incorporation of homoarginine into a target protein.⁵²² However, this approach is not efficient or generalizable. Efforts to reassign sense codons with PylRS/tRNA^Pyl pairs in wt cells have not yielded homogeneous proteins.^523,524 The synthesis of an E. coli genome that compressed the number of sense codons used to encode the canonical amino acids, and enabled deletion of the tRNAs that normally decode the removed synonyms, has been used to cleanly encode ncAAs in response to sense codons in synthetic genes.^77,525 This work is discussed in the context of incorporating multiple distinct ncAAs (Section 10).

9.4.1. Stochastic Orthogonal Recoding of Translation

Methods for selectively tagging and identifying the proteome synthesized in specific cells and/or at specific times have proved very useful for deciphering a range of biological processes.^{166,167,526,527} The N⁺-MmPylRS/N-MmtRNA^Pyl pair has been adapted to direct the incorporation of ncAAs at substoichiometric levels into the proteome in response to sense codons. Since PylRS does not recognize the anticodon of its cognate tRNA, ncAAs can be directed in response to diverse sense codons. This approach, known as stochastic orthogonal recoding of translation (SORT), enables ncAA-dependent tagging of the proteome in response to diverse codons. Stochastic orthogonal recoding of translation complements approaches that use endogenous methionyl synthetases (or their mutants) and close analogs of methionine to label the proteome at methionine codons by selective pressure incorporation-based methods, and has several potential advantages over these methods. First, SORT uses an orthogonal synthetase and therefore ncAA incorporation efficiency, and the chemical structure of the ncAAs used, is not limited by the active sites of natural synthetases; this means that SORT can label proteomes with diverse chemistry and does not require minimal media or starvation. Second, SORT can be directed to diverse codons and therefore proteome coverage may be enhanced. Third, because PylRS/tRNA^Pyl pairs are orthogonal in all domains of life the approach is portable to many organisms.

Two versions of SORT were demonstrated in E. coli; SORT with chemo-selective modification (SORT-M) tags the newly synthesized proteome with ncAAs that are then labeled with fluorophores to visualize proteome synthesis, while SORT-E tags the newly synthesized proteome with ncAAs that are then labeled with biotin for selective enrichment and identification by MS. SORT-M and SORT-E have been implemented via incorporation of an alkyne containing ncAA or a cyclopropene containing ncAA and labeling with azide or tetrazine based probes, respectively.^96,166

SORT-M was extended to D. melanogaster, where cell-type specific promoters were used to drive N⁺MmPylRS to tag the proteome and visualize newly synthesized proteins in specific cells at specific developmental stages.¹⁶⁷ Adeno-associated virus based expression of the SORT components, with N⁺MmPylRS driven by cell-type specific promoters, enabled the use of SORT-M and SORT-E to investigate the proteomes of neuronal cultures, brain slices, and spatially and genetically defined regions of the brains of live mice.

Recent developments in the generation of mutually orthogonal PylRS/tRNA^Pyl pairs (Section 8),³⁸ may enable further advances in SORT. Mutually orthogonal chemical handles may be encoded in response to distinct codons and in distinct tissues at distinct times. These advances should enable multicolor pulse-chase proteome labeling, and multicolor cell-type specific labeling.

10. Incorporating Multiple Distinct ncAAs into Proteins with pyl Systems

Strategies for incorporating distinct ncAAs in response to distinct codons have extensively utilized engineered PylRS/tRNA^Pyl pairs in combination with other orthogonal aminoacyl-tRNA synthetase/tRNA pairs. The codons used include a quadruplet codon and a stop codon, up to three stop codons, sets of quadruplet codons (Section 10.1), codons containing noncanonical bases (Section 10.2), and sense codons (Section 10.3). An extensive review on genetic code expansion mediated double incorporations has recently been published.⁴¹⁰

10.1. Incorporation of ncAAs with pyl Systems at Quadruplet Codons and Stop Codons

The N⁺-MbPylRS/N-MbtRNA^Pyl pair and MjTyrRS/MjtRNA^Tyr_CUA pair were shown to be mutually orthogonal in their aminoacylation specificity.⁴⁰⁶ The N⁺-MbPylRS/N-MbtRNA^Pyl pair and an engineered MjTyrRS/MjtRNA^Tyr_UCCU pair were used to incorporate two distinct ncAAs in response to an amber and quadruplet codon on an orthogonal message, decoded by O-RiboQ1 in E. coli. Encoding azide and alkyne containing ncAAs into calmodulin at specific sites enabled copper catalyzed protein cyclization.

The N⁺-MmPylRS/N-MmtRNA^Pyl_UUA pair was subsequently combined with active site variants of the MjTyrRS/MjtRNA^Tyr_CUA pair to direct the incorporation of two distinct ncAAs at an ochre and amber codon in E. coli.^92,528 This approach has been extended to mammalian cells, where the N⁺-MmPylRS/N-MmtRNA^Pyl_UUA pair was partnered with derivatives of the mutually orthogonal EcTyrRS/EctRNA^Tyr_CUA pair to read the ochre and amber stop codons.⁴⁰¹ This approach has also been extended to all three stop codons, using a system where the opal stop codon (UGA) was read by an engineered EcTrpRS/tRNA^Trp_UCA, in E. coli;⁵²⁹ in this system the endogenous EcTrpRS/tRNA^Trp_CCA pair is replaced by the corresponding yeast pair. Other work has used the N⁺-MmPylRS/N-MmtRNA^Pyl_UUA pair in combination with an engineered A^Δ-alvPylRS/A-alvtRNA^Pyl_CUA, these previously described mutually orthogonal pairs (Section 8)³⁸ were used to direct the incorporation of two ncAAs in response to the cognate stop codons, and a mutant MjTyrRS/MjtRNA^Tyr_AUA pair that acts as an initiator tRNA was used to direct a third ncAA at position 1 of GFP.⁵³⁰ In general, efforts to reassign multiple stop codons resulted in low protein yields,⁷⁶ and this approach is toxic for cells as the termination of endogenous proteins is impaired.

Evolved N⁺-MbPylRS/N-MbtRNA^Pyl_XXXX-derived pairs with extended anticodon tRNAs that decode UAGA, AGGA, AGUA, and CUAG codons (Section 9) were combined with evolved MjTyrRS/MjtRNA^Tyr_CUA pairs, to encode 12 distinct pairs of ncAAs, using O-RiboQ1 and a cognate orthogonal message bearing the amber codon and a quadruplet codon of interest.⁷⁶ This approach was used to label proteins at defined sites for FRET studies of protein conformational change,⁷⁶ and formed the basis of an approach for the concerted, rapid, and quantitative dual-labeling of proteins, using two highly active mutually orthogonal chemistries.³⁹¹ The combination of the evolved N⁺-MbPylRS/N-MbtRNA^Pyl_CUAG pair with an engineered MjTyrRS/MjtRNA^Tyr_CUA pair was used to encode two distinct ncAAs in an O-RiboQ1 dependent manner into a phage-displayed single-chain antibody variable fragment, expanding the chemical scope of phage display.³⁸⁰

Engineered mutually orthogonal PylRS/tRNA^Pyl pairs (Section 8) have also been used to incorporate pairs of ncAAs in response to an amber and quadruplet codon. An engineered N⁺-MmPylRS/N-MmtRNA^Pyl_UCCU pair was combined with an engineered A^Δ-alvPylRS/A-alvtRNA^Pyl_CUA pair to direct the incorporation of distinct ncAAs in response to an AGGA and CUA codon on an orthogonal message decoded by O-RiboQ1.³⁸ In extensions of this approach, triply orthogonal PylRS/tRNA^Pyl pairs (Section 8) were used to incorporate three distinct ncAAs in response to two quadruplet codons and an amber codon on an orthogonal message.^39,518

Sets of quadruplet codons have also been used to incorporate multiple distinct ncAAs into proteins. In one example, an engineered MjTyrRS/MjtRNA^Tyr_UCCU pair was combined with an engineered N⁺-MbPylRS/N-MbtRNA^Pyl_UCUA pair to produce a protein incorporating two distinct ncAAs in response to two quadruplet codons on an orthogonal message read by O-RiboQ1 in E. coli.⁷⁶ In recent work, engineered triply orthogonal PylRS/tRNA^Pyl pairs were combined with an engineered orthogonal AfTyrRS/AftRNA^Tyr to incorporate four distinct ncAAs in response to four quadruplet codons on an orthogonal message read by O-RiboQ1 in E. coli.⁵¹⁸ To express all four tRNAs and aaRS enzymes in E. coli, an automated tRNA operon generator was developed, and strategies to express multiple aaRS as polycistronic operons from a single transcript were also developed. This work expanded the genetic code from 64, to 68 codons (Figure 28).

Figure 28.

Genetic encoding of four distinct ncAAs using a 68-codon genetic code. The genetic incorporation of four distinct ncAAs requires the control of the orthogonality of the engineered translational machinery on several levels. First, active sites need to be engineered for each aaRS which are selective for the target ncAA, and exclude all other ncAAs as well as the canonical amino acids. Second, multiply orthogonal aaRS/tRNA pairs need to be engineered, which are compatible with each other. Third, each tRNA needs to be addressed to a distinct codon in the genetic code. Finally, the ribosome may need to be engineered to read alternative codons, or polymerize novel ncMs. By combining engineered triply orthogonal PylRS/tRNA^Pyl pairs with an orthogonal AfTyrRS/AftRNA^Tyr pair, addressing each active site to a specific ncAA, and addressing each tRNA to a specific quadruplet codon, four distinct ncAAs were successfully encoded in E. coli using an evolved quadruplet decoding ribosome (RiboQ1). Adapted with permission from Dunkelmann et al.⁵¹⁸ - copyright © 2021 Nature Springer Limited.

10.2. Incorporation Two Distinct ncAAs at Codons Containing Synthetic Bases

The triply orthogonal set of codon anticodon interactions: AXC:GYU; GXU:AYC; AGX:XCU (where X= NaM and Y= TPT3, and the sequences are written codon 5′-3′:anticodon 5′-3′) discovered using N⁺-MmPylRS/N-MmtRNA^Pyl pairs (Section 9) were leveraged to incorporate two distinct ncAAs and incorporate serine in response to a noncanonical codon.⁷⁹ To achieve this a variant of MjtRNA^Tyr was engineered with a GYU anticodon, N-MmtRNA^Pyl was engineered with a XCU anticodon, and EctRNA^Ser was engineered with an AYC anticodon. Cells containing these tRNAs, their cognate synthetases and a GFP gene with the cognate codons were used to direct two ncAAs and serine into GFP (Figure 29). Ninety six percent of the isolated protein contained all three amino acids, as judged by protein intact MS of the protein.⁷⁹

Figure 29.

Genetic encoding of two ncAAs using two mutually orthogonal synthetic codons. Three synthetic codons are orthogonal to each other GYU:AXC, AYC:GXU, and XCU:AGX. The combination of the engineered MjTyrRS/MjtRNA^Tyr_GYU and N⁺-MmPylRS/N-MmtRNA^Pyl_XCU pairs together with an EcSerRS enzyme and an EctRNA^Ser_AYC variant permits the site-specific incorporation of two ncAAs and serine at synthetic-codon defined positions in a protein. In EcSerRS, Ser has been typeset in non-italic.

10.3. Incorporation of Three Distinct ncAAs at Sense Codons

Recent work freed three codons, two sense codon (TCG and TCA) and the TAG codon for reassignment to ncAAs in E. coli. To achieve this a version of the four mega bases E. coli genome was synthesized in which every annotated occurrence of TCA and TCG codons was replaced with defined synonymous codons (AGC and AGT respectively); the TAG stop codon was also replaced with the TAA stop codon. The resulting strain, called syn61, contained over 18, 000 codon changes and used a compressed genetic code composed of 61 codons to encode the proteome of the cell.⁵²⁵ The endogenous tRNAs (serT and serU) that normally decode UCA and UCG codons, as well as release factor 1 (prfA) that normally terminates protein synthesis at TAG codons were deleted in an evolved version of Syn61, creating syn61Δ3.⁵²⁵ This freed these codons for reassignment to new monomers (Figure 30a).

Figure 30.

Genetic encoding of three distinct ncAAs at sense codons in syn61Δ3. a, Schematic showing the codon compression scheme used to generate syn61 and the steps taken to reassign all three free codons to new ncAAs. b, Genetic encoding of three distinct ncAAs at three distinct sense codons in syn61Δ3. Two mutually orthogonal PylRS/tRNA^Pyl pairs from classes N and A were used together with the AfTyrRS/AftRNA^Tyr pair. Parts of this figure are adapted with permission from Robertson et al.⁷⁷ copyright © 2021, some rights reserved; exclusive licensee American Association for the Advancement of Science.

Synthetic genes–that used the compressed genetic code for the canonical amino acids and the freed codons (TCA, TCG and TAG) to define the sites of ncAA incorporation–were introduced into syn61Δ3. A set of engineered triply orthogonal pairs, based on A^Δ-1r26PylRS/A-alvtRNA^Pyl_CGA, N⁺-MmPylRS/N-MmtRNA^Pyl_CGA, and AfTyrRS/AftRNA^Tyr (Section 8), were used to encode several combinations of ncAA into proteins with very high efficiency and fidelity in response to the freed codons (Figure 30b).^38,39,77,83

11. Genetically Encoded Noncanonical Polymer and Macrocycle Synthesis

By writing synthetic genetic sequences composed of the freed codons, syn61Δ3 was leveraged for the genetically encoded synthesis of noncanonical polymers composed entirely of ncAAs. This required the efficient, scalable, sequential decoding of freed codons to incorporate ncAAs, which has not been achieved by other approaches used to encode ncAAs in proteins.

The encoded synthesis of noncanonical polymers composed of two distinct ncAAs or ncMs (A and B) requires the ribosome to accept each ncAA (or ncM) in four elementary polymerization steps (A+B → AB, B+A → BA, A+A → AA, B+B → BB); these steps are not investigated when incorporating a ncAA or ncM into a protein. These elementary reactions were demonstrated in response to four dicodon combinations inserted into GFP; this was done using three pairs of ncAAs and components of the triply orthogonal pairs based on A^Δ-1r26PylRS/A-alvtRNA^Pyl_CGA, N⁺-MmPylRS/N-MmtRNA^Pyl_CGA, and AfTyrRS/AftRNA^Tyr (Section 8). This work was further developed to encode six non-natural tetrameric sequences, and a hexameric sequence for each of the three pairs of ncAAs (Figure 31a). In addition, an octameric polymer sequence composed of CbzK and AllocK was encoded. All the encoded noncanonical polymer sequences generated were confirmed by MS.⁷⁷ While noncanonical polymers were initially made as GFP fusions, tetramers and hexamers were subsequently produced as free molecules (Figure 31b).

Figure 31.

Synthesis of noncanonical polymers in cells. a, Synthesis of noncanonical polymers as GFP fusions. Synthetic genes for noncanonical polymers were expressed as N-terminal fusions to GFP; the order of codons in the sequence defined the pattern of monomer building blocks in the resulting polymer, one example of a synthetic gene is shown, and reassignmentschemes (r.s. 1–3) define the identity of the monomers. Two mutually orthogonal PylRS/tRNA^Pyl pairs from classes N and A were used for r.s. 1 and of one of the pyl pairs (from either class N or A, respectively) in combination with the AfTyrRS/AftRNA^Tyr pair for r.s. 2 and 3. b, Synthetic genes for the encoded synthesis of free non canonical polymers. An example of a noncanonical hexamer synthesized in cells using recoding scheme 1 is shown. c, Synthetic genes for the encoded synthesis of a noncanonical macrocycle. An example of a cell-based noncanonical macrocycle synthesis using recoding scheme 1 is shown. Figure adapted with permission from Robertson et al.⁷⁷ copyright © 2021, some rights reserved; exclusive licensee American Association for the Advancement of Science.

Extensions of this approach enabled the genetically programmed synthesis of noncanonical macrocycles. The first such macrocycles was formed entirely from ncAAs cyclized through a cysteine (Figure 31c).⁷⁷ Subsequent work has demonstrated the synthesis of 37 genetically programmed noncanonical peptide or depsipeptide macrocycles containing diverse ncAAs or hydroxy acids at programmed sites (Figure 32).¹³² In future work it may be possible to directly synthesize libraries of genetically encoded noncanonical macrocycles in cells that enable direct selection for molecules that perform a specific function.

Figure 32.

Cell-based synthesis of macrocyclic (depsi)-peptides. a, Strategy for the encoded cell-based synthesis of artificial macrocycles. The indicated synthetic genes were used with the indicated reassignmentschemes (r.s.). Two mutually orthogonal PylRS/tRNA^Pyl pairs from classes N and A as well as the AfTyrRS/AftRNA^Tyr (Y) pair were used in different combinations; r. s. 1,2, 4, 6 used a class N pyl pair and a Y pair: r. s. 2, 5, 7, 8 used a a class N pyl pair and a class A pyl pair. b, The ten core structures of 37 macrocyclic products from the encoded cell based synthesis that were isolated and characterized by MS. For each core structure the different recoding schemes, according to which the macrocyclic product were synthesized, are indicated. Adapted with permission from Spinck et al.¹³² - copyright © The Author(s) 2023 CCBY http://creativecommons.org/licenses/by/4.0/.

The work on genetically programmed noncanonical polymer synthesis exemplifies how sense codon reassignment can be used to make entirely noncanonical polymers and macrocycles composed of distinct building blocks; in these molecules the order of chemical building blocks is defined by the sequence of codons in a synthetic gene, and the identity of the chemical building blocks is defined by the monomers directed to the ribosome by the mutually orthogonal synthetases and tRNAs that read the distinct codons. In future work, strategies for encoding noncanonical polymers will be combined with strategies for expanding the chemical scope of cellular translation to ncMs (Section 7). This should enable the encoded cellular synthesis of an even wider range of noncanonical polymers.

12. Conclusion and Future Challenges

In the last two decades PylRS/tRNA^Pyl pairs have emerged as the most widely used orthogonal aaRS/tRNA pairs for genetic code expansion.² PylRS/tRNA^Pyl pairs have enabled the incorporation of ncAAs with a range of chemical structures,⁶ the incorporation of ncAAs in all domains of life,³ and the incorporation of ncAAs in response to diverse codons.² These advances build on several properties that are unique to the pair: the unique ncAAs accommodated by PylRS and the malleability of its active site,^{45,58,70,82,93} the unique sequence and structure of PylRS and tRNA^Pyl with respect to other aaRS/tRNA pairs,^20,57,60 and the absence of anticodon recognition in all PylRS/tRNA^Pyl pairs where this has been investigated.^59,75

Engineered PylRS/tRNA pairs, in combination with creative chemistry, have enabled new approaches for studying, and rapidly manipulating biological systems, and these approaches have led to numerous new insights.¹⁻⁶ We anticipate that PylRS/tRNA^Pyl pairs will continue to be the systems of choice for adding ncAAs with new and useful functionalities to the genetic code.

The surprising discovery that different PylRS/tRNA^Pyl pairs can be mutually orthogonal in their acylation specificity (as well as orthgonal to endogenous pairs for canonical amino acid) has led to many new directions.³⁸ Approaches for mining the astonishing diversity of PylRS/tRNA^Pyl pairs, in combination with engineering approaches has led to sets of up to five quintuply orthogonal PylRS/tRNA^Pyl pairs.^{35,36,38−40} Understanding the sequence variations, and molecular mechanisms, underpinning the orthogonality of PylRS/tRNA^Pyl pairs from distinct classes may provide further insight into how to engineer the next generation of aaRS/tRNA pairs for genetic code expansion. As more sequences become available it will be interesting to see whether even more sets of mutually orthogonal pairs can be discovered, or engineered de novo.

The combination of mutually orthogonal PylRS/tRNA^Pyl pairs that recognize distinct ncAAs and are directed to distinct codons has enabled the incorporation of several distinct ncAAs into proteins in response to diverse codons, including quadruplet codons, codons containing synthetic bases, and sense codons in organisms with compressed genetic codes.² This has facilitated the labeling of proteins with multiple distinct functionalities, including: pairs of fluorophores for FRET studies, and cross-linkers and PTM analogs to trap PTM specific interactions.⁴¹⁰ In future work, it will be interesting to explore the generation of emergent properties from combinations of ncAAs. This might include the generation of new enzyme active sites, where catalysis results from combinations of ncAAs, and this direction holds promise for generating new to nature catalytic function.

The combination of mutually orthogonal PylRS/tRNA^Pyl pairs, that recognize distinct ncAAs and are directed to distinct sense codons, with a cell run on a synthetic genome that uses a compressed genetic code, has enabled the encoded synthesis of polymers and macrocycles composed entirely of ncAA.^77,132,525 In future, work it will be interesting to expand the length of genetically encoded polymers that can be synthesized in cells to enable the synthesis of ncAA based foldamers. This may require strategies for generating more efficient synthetase/tRNA pairs, and strategies for continuous evolution may be particularly valuable for addressing this challenge.^180−185

Recent work has developed direct selections for PylRS variants that acylate their cognate tRNAs with ncMs (including β-linked monomers, and α-,α-disubstituted monomers) beyond α-L-amino acids,¹⁹ and this has enabled an expansion of the classes of monomers that can be added to the genetic code of cells. It will now be possible to use translation-based selections, including double-sieve selections, to optimize other components of the translational machinery–including tRNAs, E-Tu and other translation factors, and ribosomes–to further increase the efficiency and specificity of ncM incorporation. Future work will also expand the range of ncMs, and the classes of ncMs that can be genetically encoded in cells.

The cellular genetic encoding of ncMs into proteins may provide new ways to control the secondary and tertiary structure of proteins and to generate protease resistant proteins;^531,532 this may enable diverse applications, including the scalable biosynthesis of modified protein therapeutics and the discovery of new therapeutics. The genetic encoding of ncMs may also enable the discovery and optimization of macrocycles with improved cellular uptake and stability.^77,132,210

By combining strategies for encoding ncMs, with the development of mutually orthogonal pairs, and strategies for generating free codons it may be possible to genetically encode the cellular synthesis of polymers composed entirely of ncMs, with complete control over sequence and composition. This may enable the directed evolution of polymers with diverse backbone structures to generate foldamers⁵³³ that can replace and expand the functions carried out by natural proteins.

Biographies

Daniel Dunkelmann is a Junior Research Fellow at Magdalene College, Cambridge,and a Branco Weiss Fellow at the Max Planck Institute of Molecular Plant Physiology, Potsdam-Golm.

Jason Chin is currently a Programme Leader at the Medical Research Council Laboratory of Molecular Biology (MRC-LMB), where he is also a Head of the Centre for Chemical & Synthetic Biology (CCSB) and joint Head of the Division of Protein and Nucleic Acid Chemistry. He is a Professor of Chemistry and Chemical Biology at the University of Cambridge Department of Chemistry. He is a fellow in Natural Sciences at Trinity College, Cambridge. Jason is a member of EMBO, a Fellow of the Academy of Medical Sciences, and a Fellow of The Royal Society.

Author Contributions

CRediT: Jason W Chin conceptualization, investigation, project administration, writing-original draft, writing-review & editing; Daniel Lorenz Dunkelmann conceptualization, investigation, visualization, writing-original draft, writing-review & editing.

The authors declare the following competing financial interest(s): J.W.C. is a founder of Constructive Bio. D.L.D is a consultant for Constructive Bio.

Special Issue

Published as part of Chemical Reviewsspecial issue “Noncanonical Amino Acids”.