Making sense out of nonsense.

T is considerable interest among protein scientists in methods that permit unnatural amino acids to be incorporated at specific sites in proteins. Such methods would facilitate studies of protein structure and function and would allow for the synthesis of proteins having novel properties. Current methods use nonsense-suppressing tRNAs (1) to incorporate an unnatural amino acid at the desired site in a protein and chemical (rather than enzymatic) means to attach the unnatural amino acid to the nonsensesuppressing tRNA (2). However, protein yields are typically modest because the suppressor tRNA participates in only one round of translation. Protein expression could be increased by the development of in vivo systems, but this poses significant challenges. It requires a synthetase that activates an unnatural amino acid and only attaches it to a designated tRNA. In addition, it requires that the designated tRNA is not aminoacylated by any other synthetase in the cell and that it efficiently translates a nonsense codon. A ‘‘21st cognate pair’’ is essentially created if all of these criteria are met (Fig. 1). The paper by Kowal et al. (3) in this issue of PNAS makes an important contribution by describing a novel approach for creating tRNA-synthetase pairs that efficiently incorporate natural amino acids at specific sites in proteins in vivo. To incorporate a desired amino acid at a specific codon, the translation machinery must be manipulated without disrupting its overall accuracy and efficiency. Each amino acid is normally activated by a single aminoacyl-tRNA synthetase, which subsequently attaches the amino acid to only the corresponding tRNA. The charged tRNA is then bound by an elongation factor that transports it to the ribosome. At the ribosome, complementary base-pairing between tRNA anticodons and mRNA codons dictates the order of addition of amino acids to the growing polypeptide chain. Ordinarily, there are no tRNAs that read the three nonsense (stop) codons. However, tRNAs competent to insert an amino acid in response to nonsense codons can be engineered by altering the tRNA anticodon (4–6). Coexpression of a suppressor and a gene having an introduced nonsense codon gives a full-length protein. Thus, this method permits a spectrum of desired amino acids to be inserted at a single site in a protein. Engineering an endogenous tRNA with the requisite characteristics is problematic. Abundant structural and functional data (7) make it relatively easy to identify nucleotide changes in a tRNA that discourage aminoacylation by the cognate synthetase. The difficulty lies in identifying changes that do not promote recognition by another synthetase. All of the synthetases and the 20 types of tRNA are coexpressed in the same cellular compartment, resulting in a complex network of potential interactions among these macromolecules. The subset of correct interactions that gives rise to accurate amino acid incorporation thus depends on a proper balance among the concentrations of tRNAs and synthetases (8, 9) as well as on positive elements that promote productive interactions with the cognate synthetase and negative elements that discourage interactions with the 19 noncognate synthetases (10, 11). The sequence space available for creating a tRNA that is not recognized by the endogenous synthetases but that is recognized by a ‘‘21st’’ synthetase is limited (11). tRNAs are comprised of only about 74–90 nt; and typically, nucleotides in only a few regions of the tRNA are specifically contacted by synthetases (7, 12–15). Consequently, changes that discourage aminoacylation by the cognate synthetase often concomitantly encourage aminoacylation by a noncognate synthetase(s). The problem is particularly acute for nonsense suppressors (6). By definition, these tRNA variants have an altered anticodon trimer. Because most synthetases recognize at least one of the three anticodon nucleotides (15), a change in the anticodon trimer can abolish recognition by the cognate synthetase but can concomitantly promote recognition by another synthetase in the cell (6, 15, 16). Fortunately, there are loopholes in the rules governing tRNA recognition. In particular, the identity of nucleotides that comprise the complex network of positive and negative interactions among the tRNAs and synthetases differs from organism to organism. Thus, when a tRNA is expressed in a heterologous system, it often accepts an amino acid but does so with a reduced efficiency (7, 17–19). The introduction of mutations into these debilitated tRNAs can render them unrecognizable by any endogenous synthetase in the cell, giving them one of the characteristics required of tRNA for the sitespecific incorporation of amino acids.