An Expanded Eukaryotic Genetic Code

See allHide authors and affiliations

Science  15 Aug 2003:
Vol. 301, Issue 5635, pp. 964-967
DOI: 10.1126/science.1084772


We describe a general and rapid route for the addition of unnatural amino acids to the genetic code of Saccharomyces cerevisiae. Five amino acids have been incorporated into proteins efficiently and with high fidelity in response to the nonsense codon TAG. The side chains of these amino acids contain a keto group, which can be uniquely modified in vitro and in vivo with a wide range of chemical probes and reagents; a heavy atom–containing amino acid for structural studies; and photocrosslinkers for cellular studies of protein interactions. This methodology not only removes the constraints imposed by the genetic code on our ability to manipulate protein structure and function in yeast, it provides a gateway to the systematic expansion of the genetic codes of multicellular eukaryotes.

Although chemists have developed a powerful array of methods and strategies to synthesize and manipulate small-molecule structures (1), our ability to rationally control protein structure and function is still in its infancy. Mutagenesis methods are limited to the common 20 amino acid building blocks, although in a number of cases it has been possible to competitively incorporate close structural analogs of common amino acids throughout the proteome (2, 3). Total synthesis (4) and semisynthetic methodologies (5, 6) have made it possible to synthesize peptides and small proteins containing unnatural amino acids but have limited utility with proteins over 10 kD. Biosynthetic methods that involve chemically acylated orthogonal tRNAs (7, 8) have allowed unnatural amino acids to be incorporated into larger proteins, both in vitro (9) and in microinjected cells (10). However, the stoichiometric nature of chemical acylation severely limits the amount of protein that can be generated. Thus, despite considerable efforts, the properties of proteins, and possibly entire organisms, have been limited throughout evolution by the 20 genetically encoded amino acids [with the rare exceptions of pyrrolysine and selenocysteine (11, 12)].

To overcome this limitation, we recently showed that new components can be added to the protein biosynthetic machinery of the prokaryote Escherichia coli (13), which make it possible to genetically encode unnatural amino acids in vivo. A number of unnatural amino acids with novel chemical, physical, and biological properties (1419) have been incorporated efficiently and selectively into proteins in response to the amber codon TAG. However, because the translational machinery is not well conserved between prokaryotes and eukaryotes is not highly conserved, components of the biosynthetic machinery added to E. coli cannot be used to site-specifically incorporate unnatural amino acids into proteins to study or manipulate cellular processes in eukaryotic cells.

Thus, we set out to create translational components that would allow us to expand the number of genetically encoded amino acids in eukaryotic cells. Saccharomyces cerevisiae was chosen as the initial eukaryotic host organism, because it is a useful model eukaryote, genetic manipulations are facile (20), and its translational machinery is highly homologous to that of higher eukaryotes (21). The addition of new building blocks to the S. cerevisiae genetic code requires a unique codon, tRNA, and aminoacyl–tRNA synthetase (aa RS) that do not cross-react with any components of the yeast translational machinery (2224). One candidate orthogonal pair is the amber suppressor tyrosyl–tRNA synthetase–tRNACUA pair from E. coli (25, 26). E. coli tyrosyl–tRNA synthetase (TyrRS) efficiently aminoacylates E. coli tRNACUA when both are genetically encoded in S. cerevisiae but does not aminoacylate S. cerevisiae cytoplasmic tRNAs (27, 28). In addition, E. coli tyrosyl tRNACUA is a poor substrate for S. cerevisiae aminoacyl–tRNA synthetases (29) but is processed and exported from the nucleus to the cytoplasm (30) and functions efficiently in protein translation in S. cerevisiae (2729). Moreover, E. coli TyrRS does not have an editing mechanism and therefore should not proofread an unnatural amino acid ligated to the tRNA.

To alter the amino acid specificity of the orthogonal TyrRS so that it aminoacylates tRNACUA with a desired unnatural amino acid and none of the endogenous amino acids, we generated a large library of TyrRS mutants and subjected it to a genetic selection (31). On the basis of the crystal structure of the homologous TyrRS from Bacillus stearothermophilus (32), five residues (Fig. 1A) in the active site of E. coli TyrRS that are within 6.5 Å of the para position of the aryl ring of bound tyrosine were randomly mutated. A selection strain of S. cerevisiae [MaV203: pGADGAL4 (2 TAG) (3335)] was transformed with the library to afford 108 independent transformants and grown in the presence of 1 mM unnatural amino acid (Fig. 2C). Suppression of two permissive amber codons in the transcriptional activator GAL4 leads to the production of full-length GAL4 (36) and the transcriptional activation of the GAL4-responsive HIS3, URA3, and lacZ reporter genes (Fig. 2A). Expression of HIS3 and URA3 in media lacking uracil (–ura), or containing 20 mM 3-aminotriazole (37) (3-AT, a competitive inhibitor of HIS3p) and lacking histidine (–his), allows clones expressing active aaRS-tRNACUA pairs to be positively selected. If a mutant TyrRS charges the tRNACUA with an amino acid, then the cell biosynthesizes histidine and uracil and survives. Surviving cells were amplified in the absence of 3-AT and unnatural amino acid to remove full-length GAL4 from cells that selectively incorporate the unnatural amino acid. To remove clones that incorporate endogenous amino acids in response to the amber codon, we grew cells on media containing 0.1% 5-fluorootic acid (5-FOA) but lacking the unnatural amino acid. Those cells expressing URA3, as a result of suppression of the GAL4 amber mutations with natural amino acids, convert 5-FOA to a toxic product, killing the cell (38). Surviving clones were amplified in the presence of unnatural amino acid and reapplied to the positive selection. The lacZ reporter allows active and inactive synthetase-tRNA pairs to be discriminated colorometrically (Fig. 2B).

Fig. 1.

(A) Stereoview of the active site of B. Stearothermophilus tyrosyl–tRNA synthetase with bound tyrosine. The mutated residues are shown in yellow and correspond to residues from E. coli tyrosyl–tRNA synthetase Tyr37 (B. stearothermophilus TyrRS residue Tyr34), Asn126 (Asn123), Asp182 (Asp176), Phe183 (Phe177), and Leu186 (Leu180). (B) Chemical structures of p-acetyl-L-phenylalanine, 1; p-benzoyl-L-phenylalanine, 2; p-azido-L-phenylalanine, 3; O-methyl-L-tyrosine, 4; and p-iodo-L-tyrosine, 5.

Fig. 2.

(A) Plasmids and reporters for the positive and negative selections. (B) Phenotypes of yeast harboring GAL4-responsive HIS3-, URA3-, and lacZ-responsive reporters in response to active (TyrRS) or inactive (A5) aminoacyl–tRNA synthetases on selective media. (C) The scheme used to select mutant synthetases that encode additional amino acids in S. cerevisiae. UAA, unnatural amino acid. (D) Phenotypes of yeast isolated from a selection with p-acetyl-L-phenylalanine.

With the use of this approach, five novel amino acids with distinct steric and electronic properties (Fig. 1B) were independently added to the genetic code of S. cerevisiae. These amino acids include p-acetyl-L-phenylalanine (1), p-benzoyl-L-phenylalanine (2), p-azido-L-phenylalanine (3), O-methyl-L-tyrosine (4), and p-iodo-L-phenylalanine (5). The unique reactivity of the keto functional group of 1 allows selective modification of proteins with an array of hydrazine- or hydroxylamine-containing reagents in vitro and in vivo (39, 40). The heavy atom of 5 may prove useful for phasing x-ray structure data. The benzophenone and phenylazide side chains of 2 and 3 allow efficient in vivo and in vitro photocrosslinking of proteins (1416). The methyl group of 4 can be readily substituted with an isotopically labeled methyl group as a probe of local structure and dynamics with the use of nuclear magnetic resonance and vibrational spectroscopy. After three rounds of selection (positive-negative-positive), several colonies were isolated (Fig. 2D) whose survival on–ura or on 20 mM 3-AT–his media was strictly dependent on the addition of the selected unnatural amino acid (31). The same clones were blue on x-gal only in the presence of 1 mM unnatural amino acid. These experiments demonstrate that the observed phenotypes result from the combination of the evolved aminoacyl–tRNA synthetase–tRNACUA pairs and their cognate amino acids (table S1).

To further demonstrate that the observed phenotypes are due to site-specific incorporation of the unnatural amino acids by the orthogonal mutant TyrRS-tRNA pairs, we generated and characterized mutants of human superoxide dismutase 1 (41) (hSOD) containing each unnatural amino acid (31). Production of hexa-histidine-tagged hSOD from a gene containing an amber codon at position 33 was strictly dependent on p-acetylPheRS-1–tRNACUA (where PheRS is phenylalanine tRNA synthetase) and 1 mM p-acetyl-L-phenylalanine (<0.1% by densitometry, in the absence of either component) (Fig. 3). p-Acetyl-L-phenylalanine containing full-length hSOD was purified with a yield of 50 ng/mL, comparable to that purifed from cells containing E. coli TyrRS-tRNACUA. For comparison, wild-type hSOD could be purified with a yield of 250 ng/mL under identical conditions. The identity of the amino acid incorporated was determined by subjecting a tryptic digest of the mutant protein to liquid chromatography and tandem mass spectrometry (31). The precursor ions corresponding to the singly and doubly charged ions of the peptide Val-Y*-Gly-Ser-Ile-Lys containing the unnatural amino acids (denoted Y*) were separated and fragmented with an ion trap mass spectrometer. The fragment ion masses could be unambiguously assigned, confirming the site-specific incorporation of 1 (fig. S1). No indication of tyrosine or other amino acids in place of 1 was observed (42), and a minimum of 99.8% incorporation purity was obtained from the signal-to-noise ratio of the peptide spectra. Similar fidelity and efficiency in protein expression were observed when p-benzoylPheRS-1, p-azidoPheRS-1, p-iodoPheRS-1, or O-meTyrRS-1 was used to incorporate 2 (43), 3, 4, or 5 into hSOD (Fig. 3 and fig. S1).

Fig. 3.

Protein expression of hSOD (33TAG)HIS in S. cerevisiae genetically encoding unnatural amino acids. (Top) SDS–polyacrylamide gel electrophoresis of hSOD purified from yeast in the presence (+) and absence (–) of the unnatural amino acid indicated (see Fig. 1B) stained with Coomassie. Cells contain the mutant synthetase-tRNA pair selected for the amino acid indicated. (Center) Western blot probed with an antibody against hSOD. (Bottom) Western blot probed with an antibody against the C-terminal His6 tag.

The independent addition of five unnatural amino acids to the genetic code of S. cerevisiae demonstrates the generality of our method and suggests that it should be applicable to other unnatural amino acids including spin-labeled, metal-binding, or photoisomerizable amino acids. This methodology may allow the generation of proteins with new or enhanced properties as well as facilitate studies of protein function in yeast. Moreover, in mammalian cells the E. coli tyrosyl–tRNA synthetase forms an orthogonal pair with the B. stearothermophilus tRNACUA (44). It should therefore be possible to use the aminoacyl–tRNA synthethases that have been evolved in yeast to add unnatural amino acids to the genetic codes of higher eukaryotes.

Supporting Online Material

Materials and Methods

Fig. S1

Table S1

References and Notes

View Abstract

Navigate This Article