Project description:While the “universal” genetic code is now known not to be universal, and stop codons can have multiple meanings, one regularity remains, namely that for a given sense codon there is a unique translation. Examining CUG usage in yeasts that have transferred CUG away from leucine, we here report the first example of dual coding: Ascoidea asiatica stochastically encodes CUG as both serine and leucine. This is deleterious as evidenced by CUG codons being rare, never at conserved serine/leucine residues and predominantly in lowly expressed genes. Related yeasts solve the problem by loss-of-function of one of the two tRNAs. This dual-coding is consistent with the tRNA-loss driven codon reassignment hypothesis and provides a unique example of a proteome that cannot be deterministically predicted.