Universal patterns of purifying selection at noncoding positions in bacteria.

To investigate the dependence of the number of regulatory sites per intergenic region on genome size, we developed a new method for detecting purifying selection at noncoding positions in clades of related bacterial genomes. We comprehensively quantified evidence of purifying selection at noncoding positions across bacteria and found several striking universal patterns. Consistent with selection acting at transcriptional regulatory elements near the transcription start, we find a universal positional profile of selection with respect to gene starts and ends, showing most evidence of selection immediately upstream and least immediately downstream from genes. A further set of universal features indicates that selection for translation initiation efficiency is the major determinant of the sequence composition around translation start in all clades. In addition to a peak in selection at ribosomal binding sites, the region immediately around translation start shows a universal pattern of high adenine frequency, significant selection at silent positions, and avoidance of RNA secondary structure. Surprisingly, although the number of transcription factors (TF) increases quadratically with genome size, we present several lines of evidence that small and large genomes have the same average number of regulatory sites per intergenic region. By comparing the sequence diversity of the most and least conserved DNA words in intergenic regions across clades we provide evidence that the structure of transcription regulatory networks changes dramatically with genome size: Small genomes have a small number of TFs with a large number of target sites, whereas large genomes have a large number of TFs with a small number of target sites each.