The nature and variation of synonymous codon usage in 47 open reading
frames from Kluyveromyces lactis have been investigated. Using multiva
riate statistical analysis, a single major trend among K. lactis genes
was identified that differentiates among genes by expression level: h
ighly expressed genes have high codon usage bias, while genes of low e
xpression level have low bias. A relatively minor secondary trend diff
erentiates among genes according to G+C content at silent sites. In th
ese respects, K. lactis is similar to both Saccharomyces cerevisiae an
d Candida albicans, and the same 'optimal' codons appear to be selecte
d in highly expressed genes in all three species. In addition, silent
sites in K. lactis and S. cerevisiae have similar G+C contents, but in
C albicans genes they are more A+T-rich. Thus, in all essential featu
res, codon usage in K. lactis is very similar to that in S. cerevisiae
, even though silent sites in genes compared between these two species
have undergone sufficient mutation to be saturated with changes. We c
onclude that the factors influencing overall codon usage, namely mutat
ional biases and the abundances of particular tRNAs, have not diverged
between the two species. Nevertheless, in a few cases, codon usage di
ffers between homologous genes from K. lactis and S. cerevisae. The st
rength of codon usage bias in cytochrome c genes differs considerably,
presumably because of different expression patterns in the two specie
s. Two other, linked, genes have very different G+C content at silent
sites in the two species, which may be a reflection of their chromosom
al locations. Correspondence analysis was used to identify two open re
ading frames with highly atypical codon usage that are probably not ge
nes.