Botulinum neurotoxins are the most potent toxins known to date. They are zinc-metalloproteases able to cleave selectively an essential component of neurotransmitter exocytosis, causing the syndrome of botulism characterized by a flaccid paralysis. There is a great interest in designing antagonists of the action of these toxins. One way is to inhibit their catalytic activity. In this study, we report the design of such inhibitors directed toward BoNT/B. A study of the S(1) subsite specificity, using several beta-amino thiols, has shown that this subsite prefers a p-carboxybenzyl moiety. The specificity of the S(1)' and S(2)' subsites was studied using two libraries of pseudotripeptides containing the S(1) synthon derived from the best beta-amino thiol tested. Finally, a selection of various non natural amino acids for the recognition of the "prime" domain led to the most potent inhibitor of BoNT/B described to date with a K(i) value of 20 nM.