The efficiency of AUG start codon recognition in translation initiation is
modulated by its sequence context. Here me investigated a non-redundant set
of 5914 human genes and show that this context is different in genes locat
ed in different isochores, In particular, of the two main consensus start s
equences, RCCaugR is five-fold more represented than AARaugR in genes from
the GC-rich H3 isochores compared to genes from the GC-poor L isochores, Fu
rthermore, genes located in GC-rich isochores have shorter 5' UTRs and stro
nger avoidance of upstream AUG than genes located in GC-poor isochores, Thi
s suggests that genes requiring highly efficient translation are located in
GC-rich isochores and genes requiring fine modulation of expression are lo
cated in GC-poor isochores, This is in agreement with independent data from
the literature concerning the location of housekeeping and tissue-specific
genes, respectively. (C) 1999 Federation of European Biochemical Societies
.