Transcription initiation by RNA polymerase (RNP) carrying the house-keeping sigma subunit, sigma70 (Esigma70), is repressed by H-NS at a number of promoters including hdeABp in Escherichia coli, while initiation with RNP carrying the stationary phase sigma, sigma38 (Esigma38), is not. We investigated the molecular mechanism of selective repression by H-NS to identify the differences in transcription initiation by the two forms of RNPs, which show indistinguishable promoter selectivities in vitro. Using hdeABp as a model promoter, we observed with purified components that H-NS, acting at a sequence centered at -118, selectively repressed transcription by Esigma70. This selective repression is attributed to the differences in the interactions between hdeABp and the two forms of RNPs, since no other factor is required for the repression. We observed that the two forms of RNPs could form an open initiation complex (RP(O)) at hdeABp, but that Esigma70 failed to initiate transcription in the presence of H-NS. Interestingly, KMnO4 assays and high-resolution atomic force microscopy (AFM) revealed that hdeABp DNA wrapped around Esigma70 more tightly than around Esigma38, resulting in the potential crossing over of the DNA arms that project out of Esigma70 . RP(O) but not out of Esigma38 . RP(O). Based on these observations, we postulated that H-NS bound at -118 laterally extends by the cooperative recruitment of H-NS molecules to the promoter-downstream sequence joined by wrapping of the DNA around Esigma70 . RP(O), resulting in effective sealing of the DNA loop and trapping of Esigma70. Such a ternary complex of H-NS . Esigma70 hdeABp was demonstrated by AFM. In this case, therefore, Esigma70 acts as a cofactor for DNA looping. Expression of this class of genes by Esigma38 in the stationary phase is not due to its promoter specificity but to the architecture of the promoter . Esigma38 complex.