Construction of models of P450-substrate complexes by a combination of NMR and homology modelling

In modelling human P450s on the basis of the structures of the bacterial enzymes, we are faced with a relatively low sequence similarity, coupled with the fact that the part of the structure in which we are most interested, the active site, is generally the least similar in the different enzymes. In order to help overcome this problern, we have introduced additional experimental data. in tho form of distances from the haem iron to atoms of a bound substrate, measured by NMR relaxation methods (1). This is most easily done by using a homology modelling procedure which is based upon a constraint satisfaction approach; we have used MODKLLKR (2). The presence of the substrate. and the expérimentai constraints on its position relative to the haem, in turn introduce valuable constraints on the positions of amino-arid side-chains in the active site. The method used will be outlined and its advantages and limitations will be discussed; it has been applied to a number of complexes of P450 2D6 and 3A4, and the resulting models have been successfully used to design mutants with altered specificity. (1) Modi et al. (I996) Biochemistry. ;i5, 4540-4550; (1997) Biochemistry, in press. (2) Sali and Blundell (1993) .1. Mol. Biol., 234, 779-815.