Intrinsic transcription terminators are functionally defined as sites that
bring about termination in vitro with purified RNA polymerase alone. Based
on studies in Escherichia coli, intrinsic termination requires a palindromi
c stretch followed by a trail of T (or U) residues in the coding strand. We
have developed a highly efficient algorithm to identify hairpin potential
sequences in bacterial genomes in order to build a general model for intrin
sic transcription termination. The algorithm was applied to analyze the Myc
obacterium tuberculosis genome. We find that hairpin potential sequences ar
e concentrated in the immediate downstream of stop codons. However, most of
these structures either lack the U trail entirely or have a mixed A/U trai
l reflecting an evolutionarily relaxed requirement for the U trail in the m
ycobacterial genome. Predicted atypical structures were shown to work effic
iently as terminators both inside the mycobacterial cell and in vitro with
purified RNA polymerase. The results are discussed in light of the kinetic
competition models for transcription termination. The algorithm identifies
> 90% of experimentally tested terminators in bacteria and is an invaluable
tool in identifying transcription units in whole genomes.