Abstract
The target-specific cytotoxicity for gram-negative bacteria and the endotoxin-neutralizing activity of the 55-kDa bactericidal/Permeability-increasing protein (BPI) and its bioactive 23-kDa NH2-terminal fragment depend on the strong attraction of BPI for the lipid A region of lipopolysaccharides (LPS). We have shown before that smooth gram-negative bacteria with long-chain LPS are more resistant to BPI (especially holo-BPI) than are rough strains. It has been suggested that the high BPI resistance of some gram-negative bacteria, such as Proteus mirabilis, might also reflect the structural diversity of lipid A. To explore this possibility, we compared the antibacterial activity and binding of natural and recombinant holo-BPI and a recombinant NH2-terminal fragment (rBPI-23) to an isogenic rough (Re-LPS chemotype) and a smooth (S-LPS chemotype) strain of P. mirabilis and to LPS isolated from the two strains. Holo-BPI and rBPI-23 were both potently active against the Re strain of P. mirabilis (90% lethal dose, 20 nM). In contrast, the smooth strain was > or = 100 times more resistant to holo-BPI but only 10 times more resistant to rBPI-23. rBPI-23 was also more potent against several Escherichia coli strains from clinical bacteremia isolates. Differences in the antibacterial potency of BPI toward the Re and S strains of P. mirabilis correlated with differences in the binding of holo-BPI and rBPI-23 to these bacteria. In contrast, the binding of biosynthetically (in vitro transcribed and translated) 35S-labeled holo-BPI and NH2-terminal fragment to isolated Re- and S-LPS from P. mirabilis in solution was similar. Moreover, in the Limulus amebocyte lysate assay, holo-BPI and rBPI-23 potently neutralized both forms of LPS with equal effectiveness. Together, these results strongly suggest that BPI recognizes Proteus lipid A and that the relative resistance of (smooth) P. mirabilis to holo-BPI is due to the inhibitory effect of long polysaccharide chains of tightly packed LPS in the envelope.