Localization of the binding site for modified Gb3 on verotoxin 1 using fluorescence analysis

Verotoxins (VTs) from Escherichia coli elicit human vascular disease as a consequence of specific binding to globotriaosylceramide (Gb3) receptors on endothelial cell surfaces. Molecular models based on the VT1 crystal structure were used previously to investigate the structural basis for receptor recognition by VT1 and other verotoxins. Interestingly, these model-based predictions of glycolipid binding to VT1 differ somewhat from recently published structural data from cocrystals of the VT1 B-subunit (VT1B) and an analogue of the sugar moiety of Gb3. In this study, fluorescence spectroscopy was used to test model-based predictions of the location of Gb3 binding on the B-subunit pentamer of VT1. Resonance energy transfer was used to calculate the distance from a coumarin probe used to replace the acyl tail of Gb3 and the single tryptophan residue (Trp34) present within each VT1B monomer. The observed energy transfer efficiency (greater than 95%) suggests that these two moieties are approximately 13.3 A apart when a single distance is assumed. This distance is consistent with proposed models for the fit of Gb3 within the "cleft site" of the VT1 B-subunit. When the distances from Trp34 to the other coumarinGb3 molecules (bound to each of the four remaining monomers within the VT1B pentamer) are taken into consideration, it appears likely that the coumarin-modified Gb3 analogue used in this study associates with the previously proposed receptor binding site II of VT1. This is consistent with an observed binding preference of VT2c for coumarinGb3. To provide additional information on the association of Gb3 with the VT1 B-subunit, the influence of Gb3 glycolipid binding on the accessibility of Trp34 to different quenching agents in solution was then examined. Taken together, the data suggest that coumarin-labeled Gb3 preferentially binds to site II on VT1 in a position that is consistent with the previously described molecular models.