Wide ranging studies of the readthrough of translational stop codons w
ithin the last 25 years have suggested that the stop codon might be on
ly part of the molecular signature for recognition of the termination
signal. Such studies do not distinguish between effects on suppression
and effects on termination, and so we have used a number of different
approaches to deduce whether the stop signal is a codon with a contex
t or an extended factor recognition element. A data base of natural te
rmination sites from a wide range of organisms (148 organisms, similar
to 40000 sequences) shows a very marked bias in the bases surrounding
the stop codon in the genes for all organisms examined, with the most
dramatic bias in the base following the codon (+4). The nature of thi
s base determines the efficiency of the stop signal in vivo, and in Es
cherichia coli this is reinforced by overexpressing the stimulatory fa
ctor, release factor-3. Strong signals, defined by their high relative
rates of selecting the decoding release factors, are enhanced wheras
weak signals respond relatively poorly. Site-directed cross-linking fr
om the +1, and bases up to +6 but not beyond make close contact with t
he bacterial release factor-2. The translational stop signal is deduce
d to be an extended factor recognition sequence with a core element, r
ather than simply a factor recognition triplet codon influenced by con
text.