Background: Many pharmacologically important peptides are synthesized nonri
bosomally by multimodular peptide synthetases (NRPSs). These enzyme templat
es:consist of iterated modules that, in their number and organization, dete
rmine the primary structure of the corresponding peptide products. At the c
ore of-each module is an adenylation domain that recognizes the cognate sub
strate and activates it as its aminoacyl adenylate. Recently, the,crystal :
structure of the phenylalanine-activating adenylation domain PheA was solv
ed with phenylalanine and AMP, illustrating the structural basis for substr
ate recognition,
Results: By comparing the residues that line the phenylalanine-binding pock
et in PheA with the corresponding moieties in other adenylation domains, ge
neral rules for deducing substrate specificity were developed. We tested th
ese in silico 'rules' by mutating specificity-conferring residues within Ph
eA, The substrate specificity of most mutants was altered or relaxed. Gener
alization of the selectivity determinants also allowed the targeted specifi
city switch of an aspartate activating adenylation domain, the crystal stru
cture of which has not yet been solved, by introducing a single mutation.
Conclusions: In silico studies and structure-function mutagenesis have defi
ned general rules for the structural basis of substrate recognition in aden
ylation domains of NRPSs, These rules can be used to rationally alter the s
pecificity of adenylation domains and to predict from the primary sequence
the specificity of biochemically uncharacterized adenylation domains. Such
efforts Could enhance the structural diversity of peptide antibiotics such
as penicillins, cyclosporins and vancomycins by allowing synthesis of 'unna
tural' natural products.