All proteins are composed of 20 canonical amino acids. Of course, as in all things biology, there are a few exceptions when other amino acids are found to occur naturally. Since most proteins have more than one of each amino acid, it is a problem when individual locations in the proteins have to be studied by covalent modification of an amino acid side-chain. For example, chemically labeling a cysteine in a location of interest in the protein will also result in the modification of other exposed cysteines. Such a scenario would call for the removal of all the exposed cysteines. This kind of mutation throughout the protein is likely to produce changes that might not be desirable in many cases. This is where unnatural amino acids (UAAs) come in. Incorporating a UAA with a chemically unique side chain at the desired location in the protein will render large-scale mutations unnecessary and may help preserve its structure.

UAAs add to the chemical diversity of protein, and thus can be used to create proteins with new functions. For instance, peptide inhibitors with a dithiol amino acid in the place of two consecutive cysteine residues can avoid the formation of disulfide bond isomers and render a much stronger inhibitory effect [1]. To date more than ~ 50 UAA have been incorporated into proteins produced in bacterial, yeast and mammalian systems. For reviews please refer to [2-4]. Cell-free systems have also been devised to introduce UAAs into proteins, including those from genomically recoded bacteria [5]. Biotech company Sutro Biopharma (Nasdaq:STRO) uses modified Escherichia coli extract to incorporate UAAs in defined numbers and positions within potential pharmaceutical proteins [6].

Except for the stop codons amber, ochre, and opal, all other triplet codons have been taken up by cellular systems to code for one of the 20 canonical amino acids. To encode a new amino acid in the translational machinery of an organism, one of these three degenerate stop codons is generally used. These codons are not recognized by any of the endogenous host tRNAs. A tRNA charged with the UAA of interest is engineered to recognize a stop codon, which then adds the UAA in the growing polypeptide chain by a mechanism commonly referred to as nonsense codon suppression. The most commonly used nonsense codon is the amber or TAG codon. The incorporation can happen in a codon-specific and mRNA-selective manner with the clever application of phase-transition and spatial separation [7]. The genetic codons might also be expanded to accommodate UAAs in a Hachimoji RNA (and DNA) system with eight genetic codes (A, C, G, T, B, P, S, Z) [8].

[enlarge]

Figure 1. A schematic of UAA incorporation method: Plasmids encoding the suppressor tRNA_CUA, the evolved amino acyl-tRNA synthetase (aaRS) and the target protein with the amber codon are transfected into cells. The media covering the cells is supplemented with the UAA (orange cross). The aaRS catalyses the acylation of the suppressor tRNA_CUA with the UAA. When the mRNA containing the amber codon is being translated in the ribosome (light green oval), the amber codon is recognized by the tRNA_CUA charged with the UAA and this amino acid is added to the growing polypeptide chain.

Early experiments to incorporate UAAs into recombinant proteins started with tRNAs that were already charged with the UAA by chemical methods, followed by either cell-free translation or injecting this tRNA into Xenopus oocytes along with the mRNA encoding the protein of interest containing the TAG codon [9, 10]. It is not hard to imagine that the yields of UAA incorporated recombinant proteins that were produced in this manner were very small. Since then, methods have been developed that allow the tRNA to be charged with the appropriate UAA inside a living cell, after which the UAA is directly incorporated into the protein of interest. These methods involve using a tRNA and an aminoacyl-tRNA synthetase (or aaRS; it is the enzyme that acylates the tRNA with an amino acid) pair from a species that is widely different from the host in which the recombinant protein is being produced, i.e., using a bacterial or archaeal tRNA/aaRS pair in a mammalian host or using an archaeal tRNA/ aaRS pair in a bacterial host.

As a first step, both the tRNA and the aaRS are engineered to recognize the UAA of interest [11-14]. Moreover, the engineered tRNA cannot be non-specifically acylated by any other aaRS endogenous to the host. In the same way, the engineered aaRS should not acylate endogenous tRNAs, i.e., the tRNA/aaRS pair should be orthogonal to the host system. Using an orthogonal tRNA/aaRS pair from a distant species helps satisfy this criterion. The anticodon on the tRNA is mutated in such a way that it specifically recognizes the stop codon that codes for the UAA (e.g., if the amber codon TAG is used, the tRNA anticodon is mutated to CUA).

Next, the aaRS gene is modified such that it recognizes the UAA specifically. This modification starts with a large library of aaRS active site mutants. After multiple rounds of positive and negative selection in bacteria or yeast the final aaRS is obtained. The rounds of positive selection involve a gene such as the chloramphenicol resistance gene, with inframe TAG codons. Only those bacteria that produce a full-length chloramphenicol resistance gene survive in the selection media containing chloramphenicol. In the negative selection round, the ability of an aaRS to non-specifically suppress amber codons in a gene that is toxic to the bacteria, in the absence of the UAA, results in the expression in the toxic gene and subsequent bacterial cell death and elimination of that aaRS clone. In the end, what are left are clones that suppress amber codons exclusively with the UAA.

Ideally, such selection protocols have to be carried out for each new amino acid and host. If the UAA incorporation system is being developed for yeast, the selection process is carried out in yeast as well. The long replication time of mammalian cells prevents such a selection process from being carried out in mammalian systems. Thus, amino aaRS sequences evolved in yeast or bacteria are used in mammalian systems [11].

Once the UAA incorporation is engineered into the translational machinery of the host cell, the UAA is added into the cell growth media for subsequent tRNA acylation and its incorporation into the target protein. Thus, the UAA has to be cell permeable, non-toxic and stable inside cells, to be effectively utilized by the engineered tRNA and aaRS.

Incorporation of UAA into recombinant proteins produced in live cells using an ‘evolved’ aaRS was pioneered in the lab of Peter Schultz at The Scripps Research Institute, La Jolla. O-Methyl L-tyrosine was the first UAA that was successfully incorporated into a recombinant protein in live cells using an engineered tRNA-aaRS pair. The protein was produced in E. coli using the archaeal tRNA^Tyr-tyrosyl-tRNA synthetase (TyrRS) pair from Methanococcus jannaschii [15].

UAAs have also been incorporated into proteins produced in yeast using an E. coli tRNA^Tyr-TyrRS pair [14]. The amino acids p-acetyl-L-phenylalanine, p-benzoyl-L-phenylalanine, p-azido-L-phenylalanine, O-methyl-L-tyrosine and p-iodo-L-tyrosine, all closely related in structure to tyrosine, have been added to the translational machinery of yeast using the tRNA^Tyr-TyrRS pair.

Many different laboratories have succeeded in suppressing amber/nonsense mutations within mammalian cells. Suppression has been achieved using both natural and UAAs [16]. The Schultz lab for the first time used the process of ‘evolution’ in yeast to generate many versions of the E. coli TyrRS that are specific for a variety of different amino acids, for use in mammalian cells [11]. Since the E. coli tRNA^Tyr does not have the right promoter elements for transcription in mammalian cells, tRNA^Tyr from Bacillus stearothermophilus was used instead. Earlier work by Sakamoto et al had shown that B. stearothermophilus tRNA^Tyr could be transcribed and acylated by a specific variant of the E. coli TyrRS in mammalian cells [17].

Various sets of orthogonal tRNA and aaRS pairs have been developed to add new UAAs into the protein synthesis machinery of bacteria (Mycobacterium [18] and E. coli), yeast and mammals, including photocaged-citrulline [19]. Some of these are E. coli TyrRS and B. stearothermophilus tRNA^Tyr in mammalian cells [11, 17, 20], M. jannaschii (Mj) TyrRS and tRNA^Tyr in bacterial cells [15], M. barkeri PylRS and tRNA^Pyl_CUA in yeast [21].E. coli TyrRS and tRNA^Tyr pair in yeast [14] E. coli LeuRS and tRNA^Leu pair in yeast [22], M. mazei PylRS and tRNA^Pyl_CUA in mammalian cells [23].

As a continuation of the multiple efforts to use UAAs to generate new enzymes for chemical synthesis (reviewed in [24] ), a recent report by Drienovska et al has demonstrated the successful incorporation of p-aminophenylalanine into the transcriptional regulator from Lactococcus lactis [25]. The efficiency of the artificial enzyme was achieved by insertion of p-aminophenylalanine residue into the hydrophobic pocket of the enzyme molecule, which boosted the activity of the aniline side chain.

With regard to the in vivo research, transgenic Caernorhabditis elegans has actively been used as a model for UAA studies. For example, UAAs can be inserted into C. elegans using a pyrrolysyl tRNA-synthetase and a pyrrolysyl tRNA construct. The detection of the incorporated UAAs can be performed by either Western blotting or immunofluorescence [26].

It is now possible to incorporate more than one UAA into a single protein in vivo. The Schultz lab has reported on the genetic incorporation of pAcF and azidolysine into the trastuzumab polypeptide sequence using aminoacyl tRNA synthetase / tRNA pairs specific for different UAAs in a mammalian cell system [27].

With rare exceptions, naturally occurring DNA sequences code for amino acids with codons made up of three base pairs, i.e., codons occur as triplets. Although the mechanistic details of how translations by quadruplet codons work are not known, there have been a number of efforts to use quadruplet codons to encode UAAs. Many of these initial efforts relied on using the tRNA with its 3 base anti-codon sequence switched out by a 4 base anticodon, which was then chemically charged with the UAA of interest. Work in the last eight to ten years by Jason Chin and colleagues at the MRC lab in Cambridge has resulted in the development of a quadruplet codon recognition system where the entire translational machinery including acylation of the tRNA happens inside a live cell. The strategy was to create an alternate translational machinery, orthogonal to the host cell (i.e., it does not translate native mRNA) and involved creating an alternate ribosome that would specifically recognize the mRNA with the quadruplet codon.

Using an alternate ribosome, which is also orthogonal is desirable, because it can be evolved to specifically recognize both the modified mRNA with the quadruplet codon and the tRNA with the quadruplet anticodon while leaving the host translational machinery undisturbed. It has been shown that the evolved ribosome, Ribo-Q1 can more efficiently translate both triplet and quadruplet codons than the native ribosomes [28].

Recently Niu et al have used an alternate approach for incorporating UAAs in response to a quadruplet codon [29]. Based on the idea that the tRNA^Pyl anticodon loop is not an important determining factor for the recognition of its cognate PylRS, the authors first changed the anticodon of UCCU to recognize the codon AGGA. F to UCCU to recognize the codon AGGA. Following this, a tRNA^Pyl_CUA library was generated where four other base positions of the anticodon loop were randomized. This library was evolved in vitro in the presence of the UAA, Nε-(tert-butyloxy-carbonyl)-L-lysine or Boc-Lys and the BocLysRS (a modified PylRS, originally described in [30] ) to produce a tRNA clone that recognized the quadruplet codon in the chloramphenicol acetyltransferase gene. This evolved tRNA was used then in E. coli and in mammalian cells for incorporation of Boc-Lys in response to the AGGA quadruplet codon.

The gene of interest containing the amber codon, the suppressor tRNA and the evolved aaRS are introduced into cells (Figure 1). This is followed by addition of media containing the UAA (can be obtained from Sigma) to the cells. After allowing the cells to grow for about two days, the cells are harvested and the protein is purified. It has been observed that increasing the copy number of suppressor tRNA leads to an increase in yields of the protein with the UAA [11]. Usually multiple copies of the tRNA are placed in tandem in the tRNA plasmid. It has also been observed that having low concentrations of the aaRS is useful to prevent cross-acylation of native tRNA molecules with the UAA and also to prevent read through (or suppression) of naturally existing amber stop codons marking the end of gene sequences. This problem is more acute in mammalian cells where the percentage of stop codons with the amber sequence is 23% than that in bacterial cells where only 5% of all stop codons are amber. Generally, during transfection in mammalian cells, the DNA coding for the AARS gene (gene coding for aaRS) is used at about 1/10^th the amount of DNA coding for the target gene.

Protocols describing the incorporation of UAAs can be found in [11, 31], and [32], among others. Dickey TH et al, for example, incorporated, by amber stop codon suppression with the help of pEVOL-pAzF from Addgene ( 31186), the unnatural amino acid azido-phenylalanine to enable specific click chemistry labeling with alkyne-containing fluorophores [33]. Romei MG et al incorporated site-specific substitutions of 10 non-canonical amino acids into photoswitchable green fluorescent protein Dronpa2 using five different aminoacyl-tRNA synthetases (aaRSs) and their corresponding tRNA_CUA to study photoisomerization in proteins [34]. D Cervettini et al developed a scalable approach (tRNA Extension) to discover and evolve aminoacyl-tRNA synthetase–tRNA pairs [35]. Incorporation of UAAs can also be achieved through inteins [36-38].

Recently, a method has been described, based on the use of Methanosarcina mazei pyrrolysyl-tRNA synthetase and the corresponding tRNA, for the genetic incorporation of UAAs in stable mammalian cell lines. The aminoacyl-tRNA synthetase/tRNA pair was stably integrated into the mammalian genome. The system was used to explore the effects of substituting histone lysine residues with the constitutively active UAA Nε-acetyl-lysine [39].

Substituting a native amino acid in a protein with an amino acid that can cross-link upon exposure to light can give information about interacting partners in the vicinity of that amino acid. One such amino acid is p-benzoyl L-phenylalanine (pBpa), which cross-links with nearby C-H bonds when exposed to light between 350-360 nm. An orthogonal tRNA/aaRS pair was evolved for incorporation of pBpa in proteins in E. coli [13] and yeast [14]. The system evolved in yeast was subsequently adapted in the mammalian system to incorporate pBpa into the adaptor protein human Grb2 [20] and probe its interaction with the EGF receptor. Since then, other studies have used the pBpa system to study interactions between proteins and their interacting partners inside living cells [40-42].

Another photocrosslinking amino acid 3’-azibutyl-N-carbamoyl-lysine (AbK) has been adapted in E. coli and mammalian cells using the M. barkeri tRNA^Pyl_CUA/PylRS pair [43].

p-azido-L-phenylalanine (AzPhe) has also been used as a photocrosslinking UAA to determine the binding sites of a small molecule (maraviroc) and peptide (T140) on the G-protein coupled receptors (GPCRs) CC chemokine receptor 5 (CCR5) and CXC chemokine receptor 4 (CXCR4), respectively [41, 44]. Details of methods used for incorporating photocrosslinking amino acid into GPCRs can be found in [45].

A recent publication describes the use of a genetically encoded photo-affinity-labelled UAA ((Se-(N-(3-(3-methyl-3H-diazirin-3-yl)propyl)propanamide)-3-yl-homoselenocysteine), known as DiZHSeC) for protein interaction studies that incorporated a label identifiable by mass spectroscopy (MS). The UAA was incorporated into the bait protein. Following in vivo photo-affinity-labelling of the prey protein, DiZHSeC can be oxidatively cleaved to leave an N-(4,4-bis-substituted-pentyl)acrylamide group attached to the prey protein which enables subsequent MS identification. This approach has been shown to identify in vivo protein-protein interactions in both E.coli and HEK 294T cells [46]. Very recently, the use of genetically encoded UAA to study the protein-protein interactions between SUMO and proteins containing the SUMO-interacting motif (SIM) has been described. Photoactivated UAAs were genetically incorporated into the SIM interaction groove of the SUMO1 protein [47].

Proteins can be made inactive by incorporating a light-removable protecting group within their sequence that blocks sites important for their biological activity. This blocking or ‘caging’ activity can be brought about by photocaged amino acids. When the unnatural photocaged amino acid is intact, the protein is inactive, but upon exposure to light, the functional blocking group in the amino acid is removed, leaving behind the native amino acid and a biologically active protein. Thus, photocaged amino acids can be used to tune protein activity finely. Photocaged versions of tyrosine [48-50], cysteine [22], lysine [23] and serine [51] have been genetically incorporated into proteins.

The role of active site mutations of isocitrate dehydrogenase 2 in epigenetic and metabolic changes and the development of certain cancers has recently been investigated using the genetic incorporation of a photoactivated caged lysine derivative UAA. The protein containing the caged UAA was inactive. Upon photoactivation, the lysine residue generated recapitulated the disease-causing mutation and enabled the time course of disease-relevant epigenetic and metabolic changes to be followed in HEK 293 cells [52].

Site-specific introduction of an amino acid that can fluoresce differently from native amino acids can be a very powerful tool to study conformational changes, localization and molecular interactions in a protein. L-(7-hydroxycoumarin-4-yl) ethylglycine, which has a high quantum yield, large Stokes’ shift and responds to changes in solvent polarity has been introduced in the E. coli translational machinery using an evolved MjtRNA^Tyr/TyrRS pair. This UAA was introduced into myoglobin and its unfolding was monitored using the fluorescence signal from hydroxycoumarin [53].

Another fluorescent amino acid dansylalanine, with the environmentally sensitive dansyl fluorophore has been added into the translational machinery of Saccharomyces cerevisiae [54]. This was achieved starting from a tRNA^Pyl_CUA and the LeuRS from E. coli. The authors showed that it was possible to monitor the unfolding of superoxide dismutase by measuring the fluorescence signals from the dansyl group.

Unnatural fluorescent amino acids in proteins have also been used for visualization of protein localization in cells. Another coumarin-based fluorescent amino acid (S)-1-carboxy-3-(7-hydroxy-2-oxo-2H-chromen-4-yl)propan-1-aminium (CouAA) has been incorporated into proteins in E. coli using an evolved Mj tRNA^Tyr/TyrRS pair. CouAA has been incorporated into the E. coli FtsZ and GRoEL for visualization of the respective proteins inside living bacterial cells [55, 56]. The polarity-sensitive fluorescent amino acid 3-(6-acetylnaphthalen-2-ylamino)-2-aminopropanoic acid (Anap) has been incorporated in EGFP (enhanced green fluorescent protein) produced in mammalian cells. After some modification, an E. coli tRNALeu/LeuRS pair previously evolved in S. cerevisiae was used to achieve this incorporation [57].

In addition, the recently reported generation of UAA β-(1-azulenyl)-l alanine, a synthetic deep-blue tryptophan analogue with specific fluorescence characteristics, would be beneficial for the studies visualizing intracellular targeted proteins. The authors have inserted β-(1-azulenyl)-l alanine into arginine C, one of the non-ribosomal peptides [58].

Another kind spectroscopic probe that can be used to study conformational changes in proteins is the IR-active probe, which can be monitored, among other methods, by FTIR (Fourier transform infrared spectroscopy). By incorporating AzPhe, a polarity and electric field sensing probe at specific positions in the GPCR, rhodopsin, the authors were able to detect electrostatic changes around the transmembrane helices of the protein that occurred during light-induced activation of the protein [59]. Using this probe, the authors were also able to detect smaller conformational changes in the rhodopsin helices that occurred before larger rigid body helix movements [60].

p-acetylphenylalanine can be incorporated at specific positions in a protein in response to a TAG stop codon. This UAA can then be further chemically reacted with a desired molecule containing a hydroxylamine functional group. This technique has been used to label proteins with the hydroxylamine derivative of Alexa Fluor® 488 functioning as a fluorescent dye [61], or used in developing pharmaceuticals, such as ARX788 or ARX517 by biotech company Ambrx Biopharma (Nasdaq: AMAM).

A similar approach has been used for site-specific labeling of proteins with a spin label, which then becomes a marker for conformational change and can be studied by EPR spectroscopy [62].

Another UAA that has been chemically modified after site-specific incorporation in a protein is AzPhe. In Neumann et al [28], the authors site-specifically introduced AzPhe as the first and N6-[(2-propynyloxy)carbonyl]-l-lysine (CAK) as the last amino acid in calmodulin in response to quadruplet and amber codons respectively. The resultant folded protein had an azide (AzPhe) and an alkyne (CAK) group in close proximity to one another. Then using a copper-catalyzed alkyne-azide [2+3] cycloaddition reaction or click reaction, the two moieties were linked together covalently to yield a cyclic protein. This reaction can be used to incorporate a variety of chemical probes (Table 1) into proteins that have a site-specific UAA with an azide or an alkyne group.

AzPhe incorporated into the GPCR, CCR5 was labeled with the FLAG peptide using the Staudinger ligation (between azide and phosphine groups) [63]. The FLAG peptide in this study was modified with a triarylphosphine to facilitate the Staudinger ligation.

A recent study has shown that AzPhe incorporated in a GPCR can also be bio-orthogonally labeled with fluorescent probes within mammalian cells using the Staudinger–Bertozzi ligation [64]. In another study, a protein with the UAA Propargyllysine that has an alkyne group was labeled with a commercially available fluorescent dye with an azide moiety using the click reaction [65]. Both papers give details of the labeling protocols. A variety of fluorescent probes with both azide and alkyne functional are now available from Invitrogen for use in click reactions.

Another group of dyes that can react with the azide moiety in a UAA in a copper-free reaction, and are thus more suitable for reactions on the surface of live cells or where copper adversely affects the activity of the protein being labeled are available as well.

The genetic incorporation of UAA has been used as part of a system for the in vivo generation of macrocyclic peptides in bacterial cells. The system used an engineered aminoacyl tRNA synthetase from Methanocaldococcus jannaschii to genetically incorporate the phenylalanine derivative UAA 3-(2-mercapto-ethyl)amino phenylalanine into target proteins. The system has the potential for the screening of genetic libraries of polypeptide sequences for the generation of functional macrocyclic peptides [66].

The use of UAA to site-specifically incorporate toxins and/or reporter groups into therapeutic antibodies has received significant attention. Traditional methods of tag incorporation based on the use of bi-functional amino-reactive tags generate heterogeneous labelling of antibodies leading to conjugate preparations with, potentially, a broad spectrum of biochemical and pharmacological properties. The use of monoclonal antibodies genetically engineered to incorporate UAA offers a potential way to circumvent this issue. A monoclonal antibody against Her2/neu has been engineered to incorporate the UAA N6-((2-azidoethoxy)carbonyl)-l-lysine at 4 specific sites using a mammalian expression system. The incorporation of this UAA enabled the generation of homogeneous antibody-toxin conjugates containing 4 toxin molecules per antibody via click cycloaddition chemistry [67]. A similar approach has been successfully applied to generating immune-conjugates for imaging studies; in this latter case, two different UAAs were genetically incorporated into the antibody [27].

UAAs and their derivatives may have clinical applications. Several studies have demonstrated the development of UAA-containing synthetic molecules with antibacterial and antitumor functions. In particular, naphthalene-tripeptides containing α-aminoisobutyric acid downregulated the growth of melanoma cells in vitro. Furthermore, a D-enantiomer of alanine bearing naphthalene-tripeptides suppressed the growth of Staphylococcus epidermidis [68]. Other research has generated cationic antimicrobial peptides containing UAAs, which have shown antimicrobial effects against both Gram-positive and Gram-negative bacteria [69].

Another application of UAAs is the generation of protein inhibitors, which would be effective in drug design. Several selective inhibitors of Keap1‐Nrf2 protein‐protein interaction with potential application for the treatment of neurodegenerative disorders were synthesized by inserting UAAs, such as thiazolidine-4-carboxylic acid and piperidine-2-carboxylic acid [70].