The possibilities are endless. A reprogrammed genetic code that goes beyond the natural 20 amino acids and incorporates new chemical building blocks could give rise to newly designed metabolic pathways, new polymers, or even the creation of new organisms. As Jason Chin from the Medical Research Council Laboratory of Molecular Biology in Cambridge points out, a challenge with these aspirational goals is that, in the context of cells, all 64 possible triplets of the genetic code are either assigned the 20 canonical amino acids, or they function as stop codons. Chin has long been interested in engineering the genetic code to encode new building blocks. “If you could reduce the number of codons that encode a particular amino acid,” he reasons, “you could then use the freed-up synonymous codons to encode something else.”

Kaihang Wang, a senior scientist in the Chin lab, decided to pursue this idea and recode some amino acids. Serine, for example, is encoded by six codons, and Wang planned to replace two of these codons with one of the other four and check the effect on viability. Rather than make changes all over the genome, he focused on a cell-division operon to test the recoding schemes. This operon, rich in the targeted codons, carries 12 essential genes expressed over a large dynamic range. It also contains membrane proteins that need to be folded correctly, a process for which codon choice is known to be important. Any detrimental effect caused by the codon swaps could be read out as a decrease in cell viability.

In order to efficiently replace hundreds of kilobases at once, the team combined the established lambda Red recombination strategy with CRISPR–Cas9 in an approach they termed replicon excision for enhanced genome engineering through programmed recombination (REXER). “It uncouples the transformation from the recombination step,” explains Chin. First, cells are transformed with a bacterial artificial chromosome (BAC), which contains the target with the modified codons together with positive and negative selection markers. After selection for stable expression, the cells are transformed with Cas9 and single-guide RNAs that excise the target fragment from the BAC so it is free to recombine into the Escherichia coli genome at the homologous regions.

Wang and the team used REXER to replace two codons each for serine, leucine and alanine in the cell-division operon, and then they sequenced the resulting clones. Some recoding schemes worked very well; for example, the codons TCG and TCA for serine could both be replaced in all instances with AGC. Others failed; for example, TTA and TTG for leucine could not be replaced by CTC. Still others allowed substitutions at most sites but were refractory to them at particular loci. These data yield interesting information about the importance of synonymous codons in certain positions. It will be interesting to further explore why some changes are not tolerated.

This reassignment scheme freed up five codons. In future work, the tRNAs matching these codons will be deleted from the E. coli genome, and new tRNAs aminoacylated with unnatural amino acids will be introduced that recognize these codons.

Chin has calculated that REXER could re-engineer the entire E. coli genome in 14 iterative steps. “We don't understand how to design recoded genomes de novo yet,” he says, “so this sectional approach—which provides feedback on what works—makes sense.” Every successfully designed section would be used as a template for the next replacement step until the entire genome is swapped.

E. coli is not the only target for REXER, as Chin envisions its applications for eukaryotic systems as well.