Reply to Simmonds et al.: Codon pair and dinucleotide bias have not been functionally distinguished.

Simmonds et al. (1) have criticized our paper (2). Two disagreements stand out. First, the authors suggest that our viruses are attenuated by increased “dinucleotide frequencies” (1, 3), but “codon pair deoptimization” they dismiss as “artefact” (3). We believe this is just semantic word juggling. When a protein is recoded, UpA/CpG dinucleotides available for change are largely at codon–codon junctions, where they define codon pairs. In mammals, many disfavored codon pairs have UpA or CpG at their junctions (4). Therefore, altering UpA/CpG dinucleotides simultaneously alters codon pairs, and vice versa. Accordingly, all of the viral constructs in question (1, 2) coordinately alter dinucleotides and codon pairs (see below) and so do not distinguish the phenomena. No mechanism is known for either phenomenon. Nothing about our results, conclusions, or the future uses of these attenuated viruses would change if we called them “dinucleotide deoptimized.”