Lipopolysaccharide (LPS), a major component of Gram-negative bacteria, signals bacterial invasion and triggers defensive host responses. However, excessive responses also lead to the serious pathophysiological consequence of septic shock. To develop Gram-negative selective compounds that can inhibit the effects of LPS-induced sepsis, we have designed constrained cyclic antimicrobial peptides based on a cystine-stabilized beta-stranded framework mimicking the putative LPS-binding sites of the LPS-binding protein family. Our prototype termed R4A, c(PACRCRAG-PARCRCAG), consists of an eight amino acid degenerated repeat constrained by a head-to-tail cyclic peptide backbone and two cross-bracing disulfides. NMR study of K4A, an R4A analogue with four Arg --> Lys replacements, confirmed the amphipathic design elements with four Lys on one face of the antiparallel beta-strand and two hydrophobic cystine pairs plus two Ala on the opposite face. K4A and R4A displayed moderate microbicidal potency and Gram-negative selectivity. However, R4A analogues with single or multiple replacements of Ala and Gly with Arg or bulky hydrophobic amino acids displayed increased potency and selectivity in both low- and high-salt conditions. Analogues R5L and R6Y containing additional cationic and bulky hydrophobic amino acids proved the best mimics of the amphipathic topology of the "active-site" beta-strands of LPS-binding proteins. They displayed potent activity against Gram-negative E. coli with a minimal inhibitory concentration of 20 nM and a >200-fold selectivity over Gram-positive S. aureus. Our results suggest that an LPS-targeted design may present an effective approach for preparing selective peptide antibiotics.