Rewiring Protein Synthesis: From Natural to Synthetic Amino Acids

2.1. Selection and editing of natural amino acids by aaRSs and trans-editing factors

Picking the correct amino acid by the aaRS poses a significant challenge due to the structural similarity between some amino acids [19]. For example, valine (Val) and isoleucine (Ile) differ by a small methyl group in the side chain. The initial selection at the active site of isoleucine-tRNA synthetase (IleRS) only provides 180-fold discrimination between Val and Ile [20], which does not satisfy the average amino acid misincorporation rate of approximately 10⁻⁴ [21]. Additional proofreading is provided by an appended editing domain in IleRS, which hydrolyzes the misacylated Val-tRNA^Ile in a post-transfer editing reaction [20]. Editing domains are also present in valyl- (ValRS) [22, 23], leucyl- (LeuRS) [24], prolyl- (ProRS) [25], threonyl- (ThrRS) [26], alanyl- (AlaRS) [27], and phenylalanyl- (PheRS) [28] tRNA synthetases. The editing site specifically hydrolyzes the misacylated aa-tRNA but excludes the correct aa-tRNA, therefore preventing waste of resources. For example, the editing site of PheRS hydrolyzes tyrosyl- and meta-tyrosyl-tRNA^Phe, but not Phe-tRNA^Phe [28–30]. A conserved glutamate residue in the editing site specifically recognizes the hydroxyl group of tyrosine and meta-tyrosine for substrate selection [30–32] (Figure 2). Besides cis-editing domains within aaRSs, post-transfer editing is also achieved by free-standing trans-editing factors [33, 34]. To date, trans-editing factors that are homologous to the editing domains of ThrRS, AlaRS and ProRS have been identified [33–35]. Unlike cis-editing domains that recognize specific aa-tRNAs, trans-editing factors have recently been shown to exhibit broader substrate recognition patterns [36, 37]. For example, a homolog of the ProRS editing domain (ProXp-y) hydrolyzes Ser-tRNA^Thr and Thr-tRNA^Val produced by ThrRS and ValRS, respectively [36], indicating a generic role of trans-editing factors in the improvement of aminoacylation fidelity.

Editing mechanisms of aaRSs are mostly conserved throughout evolution. However, mitochondria and certain organisms with reduced genomes have lost some editing functions. In yeast mitochondria, both PheRS [38] and ThrRS [39, 40] have lost the entire cis-editing domains. Mitochondrial PheRS fidelity is achieved by a more selective active site [38], and mitochondrial ThrRS utilizes an alternate proofreading mechanism, namely pre-transfer editing [41], to reduce aminoacylation errors [40]. Interestingly, PheRS and ThrRS in certain mycoplasma species are also editing-defective [42, 43]. Mycoplasma PheRS and ThrRS with defunct editing sites are intrinsically error-prone and not able to support growth in Escherichia coli and Saccharomyces cerevisiae that have lost their native cytosolic counterparts [43, 44]. It is still unclear why these mycoplasma species have lost the editing function during evolution. One possibility is that reduced translation fidelity is not only tolerated, but may also be advantageous during the life cycle of mycoplasma.

2.2. Stress-induced misacylation

Aminoacylation fidelity is a prerequisite for accurate proteins synthesis. Mutations in the aaRS or tRNA could increase errors during aminoacylation [6]. More recent studies revealed that certain stress conditions also increase specific aminoacylation errors via distinct mechanisms [29, 45–48]. In mammalian cells, viral infection and oxidative stress substantially increase attachment of methionine to noncognate tRNAs [45]. Such global misacylation is induced by phosphorylation of methionyl-tRNA synthetase (MetRS), and may provide transient protection against oxidative stress thanks to reversible oxidation of Met in the proteome [49]. Oxidative stress also leads to misacylation by two other aaRSs – ThrRS and PheRS. The ThrRS editing site contains a conserved cysteine (C182 in Escherichia coli ThrRS) that is essential for the editing activity [26, 50]. We have shown that C182 is sensitive to oxidative stress and can be modified to a sulfenic acid, which can be reverted to cysteine when the oxidative stress is removed [46, 51]. Oxidation of C182 is facilitated by two conserved histidines in the editing site [51], and leads to an editing deficiency and misacylation of serine (Ser) to tRNA^Thr. C182 thus serves as a redox switch to control ThrRS editing. For PheRS, oxidized Phe is attached to tRNA^Phe when the editing activity is compromised [29].

Microorganisms frequently encounter environmental changes. In the hyperthermophilic archaeon Aeropyrum pernix, low temperature enhances misacylation of tRNA^Leu with Met by MetRS [47], although the underlying mechanism remains unclear. In E. coli, growth under anaerobic conditions suppresses a succinyl-lysine modification of MetRS that is required for high fidelity during aminoacylation, resulting in increased Met misacylation [48]. These recent advances suggest that aminoacylation fidelity is remarkably flexible under stress conditions and likely serve as novel regulatory mechanisms (see discussion below in Physiological impact of rewiring protein synthesis).

2.3. Plasticity of wild-type aaRSs towards synthetic amino acids

Quality control mechanisms in aaRSs have evolved to discriminate against amino acids and analogs that are present in the natural environments. Synthetic amino acids that are structurally similar to the cognate amino acids can be effective substrates for aaRSs. For example, the wild-type PheRS has been shown to activate several Phe analogs [52]. Screening of a synthetic amino acid library reveals that wild-type E. coli tryptophanyl- (TrpRS) and tyrosyl- (TyrRS) tRNA synthetases uses a wide range of synthetic amino acids as substrates [53]. An exciting application that takes advantage of the substrate plasticity of aaRSs is bioorthogonal noncanonical amino acid tagging (BONCAT) [54]. BONCAT uses endogenous MetRS to incorporate a Met analog azidohomoalanine (AHA) into Met positions, and further click chemistry allows fluorescent tagging or purification of the nascent proteome. This residue-specific approach has been broadly used to visualize and quantitate newly synthesized proteins, to probe the dynamics of protein modifications, and to produce new materials [14].

2.4. Engineering of aaRS to expand the genetic code with novel synthetic amino acids

As discussed above, some wild-type aaRSs are capable of activating certain synthetic amino acids. However, such synthetic amino acids need to compete with the cognate amino acid for the aaRS and are often inserted into the entire proteome with low homogeneity. In contrast to residue-specific incorporation of synthetic amino acids, a site-specific approach uses an engineered aaRS/tRNA pair to insert the synthetic amino acid only at particular locations of the target protein, typically mutated to the UAG codon [55]. The most commonly used aaRSs for such purposes are TyrRS from Methanocaldococcus jannaschii, and pyrrolysyl-tRNA synthetase (PylRS) from Methanosarcina species. These systems have been reviewed extensively elsewhere [12, 55, 56]. Recently, this site-specific approach has provided important insights into the dynamics of post-translational modifications, such as lysine acetylation and serine phosphorylation [16, 57]. Another intriguing application of site-specific incorporation of synthetic amino acids is biocontainment. Two recent studies have engineered genetically modified organisms that depend on the addition of a synthetic amino acid for growth, thanks to an engineered aaRS suppressing an in-frame stop codon of an essential gene with the synthetic amino acid [17, 18].

4.1. Recognition of natural proteinogenic amino acids by the ribosome

The ribosome uses 22 natural amino acids to synthesize proteins. When the crystal structure of the large ribosomal subunit was solved, it was found that the amino acid side chain of the aa-tRNA directly contacts the rRNA residues in the ribosomal aa-tRNA binding (A) site [66], suggesting that different amino acids may be recognized by the ribosome with distinct efficiencies. Later in vitro kinetic studies showed that amino acid side chains of aa-tRNAs do contribute to ribosome binding [61, 67]. Fast kinetic analyses of dipeptide formation revealed that the N-alkylamino acid proline (Pro) is a poor substrate for peptide elongation on the purified ribosome [68]. In vivo, a specialized elongation factor (EF-P in bacteria and eIF5A in eukaryotes) promotes efficient translation of Pro codons [69–71]. Enhancement of Pro translation by EF-P or eIF5A requires diverse post-translational modifications, including β-lysylation in E. coli and Salmonella [72, 73], rhamnosylation in Pseudomonas auruginosa [74], lysine modification with 5-aminopentanol group in Bacillus [75], and hypusine in yeast [71].

4.2. Ribosomal translation of D- and β-amino acids

Ribosomes use L-α-amino acids as building blocks for peptides. D- and β-amino acids (Figure 4) attached to the tRNA have been shown to be poor substrates for the wild-type ribosome [76]. Hecht and colleagues tested ribosome variants with mutations in the peptidyl-transferase center, and found several ribosome mutants with substantially increased efficiency to incorporate D-amino acids into a reporter protein in vitro [77]. A recent study by Englander et al. provided mechanistic insights into discrimination of D-amino acids by the wild-type ribosome [78]. It was shown that D-aa-tRNAs are accepted by the ribosomal A site and used as substrates for peptide formation. However, ribosomes with D-amino acid in the peptidyl-tRNA (P) site are partitioned into translating and arrested subpopulations, therefore impeding further peptide elongation [78].

Similar to D-amino acids, β-amino acids have also been inserted into peptides by mutant ribosomes in vitro [79, 80]. A recent study shows that some β-amino acids can even be used by the wild-type ribosome as substrates [81]. Ribosomal incorporation of β-amino acids into proteins has also been achieved in vivo [82]. Using mutant ribosomes and endogenous aaRSs and EF-Tu, Czekster et al. successfully inserted β-Phe analogs into reporter proteins in E. coli [82]. Including D- and β-amino acids in the protein alphabet will significantly expand the toolbox for de novo protein design [83]. Synthetic peptides containing D- and β-amino acids can be used as new biomaterials with novel properties, such as resistance to degradation by proteases. D-amino acids are key components of the bacterial cell wall and many antibiotics. Site-specific insertion of D-amino acids into peptides could also have potential applications in the development of novel antimicrobial strategies.

5.1. Fitness costs of rewiring protein synthesis

Rewiring protein synthesis with natural or synthetic amino acids leads to production of novel proteomes comprised of proteins with noncononical sequences. Many proteins with amino acid replacements fail to fold correctly and even form aggregates [84–86], causing both gain of toxicity and loss of function [29, 87, 88]. Reduced translational fidelity causes various cell defects, ranging from growth defects and cell death in bacteria [89–91], mitochondrial dysfunction in yeast [92], shortened life span in flies [93], and neurodegeneration and cardioproteinopathy in mammals [85, 94]. Editing defects in aaRS further increase DNA mutations in bacteria [95] and activates DNA damage response in zebrafish [96]. Recent work also suggests that a PheRS editing defect suppresses stringent response in E. coli [97], and a ribosomal error-prone mutation suppresses bacterial motility [98].

In addition to reduced translational fidelity, genome-wide codon reasignment also leads to reduced fitness. In Methanosarcina acetivorans, deleting the tRNA^Pyl gene reduces the genetic code from 21 to 20 amino acids, which causes downregulation of amino acid metabolism and growth defects in methanol [99]. This work suggests a selective advantage of an expanded genetic code in nature. In another study, evolved E. coli cells that completely replaces Trp with an analog (Tpa) in the proteome appear longer and thinner, and need to accummulate beneficial mutations to support robust growth [100].

5.2. Benefits of reduced translational fidelity

Despite the fitness cost of translational errors, reduced translational fidelity paradoxically provides benefits under certain stress conditions [101, 102]. In Candida albicans, ambiguos translation of CUG codons with Ser and Leu increases phenotypic diversity [103]. In the budding yeast, depletion of a release factor that forms prion increases resistance of cells to various stresses [104]. Further, methionine misacylation has been suggested to increase cell resistance to oxidative stress [45] and antibiotics [48], and produce alternate protein variants that better adapt to low temperatures [47]. We have also shown that reduced translation fidelity caused by a ribosomal mutation or canavanine improves bacterial tolerance to oxidative stress by activating the general stress response [105]. Reduced translational fidelity may also directly modify drug targets to increase resistance. For example, in Mycobacteria, mistranslation of glutamine (Gln) and asparagine (Asn) codons with Glu and Asp in the RpoB protein creates variant proteins that resist rifampicin [106, 107].