Glucose analogue inhibitors of glycogen phosphorylase: the design of potential drugs for diabetes.

The T-state crystal structure of the glucose-phosphorylase b complex has been used as a model for the design of glucose analogue inhibitors that may be effective in the regulation of blood glucose levels. Modeling studies indicated room for additional atoms attached at the C1-beta position of glucose and some scope for additional atoms at the C1-alpha position. Kinetic parameters were determined for alpha-D-glucose: Ki = 1.7 mM, Hill coefficient n = 1.5, and alpha (synergism with caffeine) = 0.2. For beta-D-glucose, Ki = 7.4 mM, n = 1.5, and alpha = 0.4. More than 20 glucose analogues have been synthesized and tested in kinetic experiments. Most were less effective inhibitors than glucose itself and the best inhibitor was alpha-hydroxymethyl-1-deoxy-D-glucose (Ki = 1.5 mM, n = 1.3, alpha = 0.4). The binding of 14 glucose analogues to glycogen phosphorylase b in the crystal has been studied at 2.4-A resolution and the structure have been refined to crystallographic R values of less than 0.20. The kinetic and crystallographic studies have been combined to provide rationalizations for the apparent affinities of glucose and the analogues. The results show the discrimination against beta-D-glucose in favor of alpha-D-glucose is achieved by an additional hydrogen bond made in the alpha-glucose complex through water to a protein group and an unfavorable environment for a polar group in the beta pocket. The compound alpha-hydroxymethyl-1-deoxy-D-glucose has an affinity similar to that of glucose and makes a direct hydrogen bond to a protein group. Comparison of analogues with substituent atoms that have flexible geometry (e.g., 1-hydroxyethyl beta-D-glucoside) with those whose substituent atoms are more rigid (e.g., beta-azidomethyl-1-deoxyglucose or beta-cyanomethyl-1-deoxyglucose) indicates that although all three compounds make similar polar interactions with the enzyme, those with more rigid substituent groups are better inhibitors. In another example, alpha-azidomethyl-1-deoxyglucose was a poor inhibitor. In the crystal structure the compound made several favorable interactions with the enzyme but bound in an unfavorable conformation, thus providing an explanation for its poor inhibition. Attempts to utilize a contact to a buried aspartate group were partially successful for a number of compounds (beta-aminoethyl, beta-mesylate, and beta-azidomethyl analogues). The beta pocket was shown to bind gentiobiose (6-O-beta-D-glucopyranosyl-D-glucose), indicating scope for binding of larger side groups for future studies.