QSAR Studies of Glycogen Phosphorelase Inhibitors

C687 (Spring Semester, 1999)

A series of compounds, as potential inhibitors of Glycogen Phosphorelase (GP), have been discovered. The inhibition constants have been analyzed by enzyme studies against GP. The Quantitative Structure Activity Relationship ( QSAR) of these compounds and their relative activity, are studied using InsightII, a molecular visualization software.

Glycogen phosphorylase (GP) plays a very important role in the regulation of muscle and hepatic glycogen metabolism. The inhibition study of GP has a meaningful influence to the treatment of the Non-insulin-dependent Diabetes Mellitus, which is a very common type of diabetes disease.

There are some known inhibitors of GB, namely compounds I-VI (Fig. 1).^[1] The inhibition constants against GB are also shown with the structures in Figure 1.

Figure 1: Some known GP inhibitors.

Recently, several new molecules (Fig. 2) were synthesized and tested for their GP inhibition ability. Among them, VII is a competitive inhibitor with a K_i of 8.2(± 0.4) x 10^-3 M and VIII is a weak competitive inhibitor with a K_i of 2.2(± 0.3) x 10^-1 M. Compounds IX and X show neither inhibition nor activation in the same conditions.^[2]

Fig. 2: Recently reported GP inhibitors.

This project studies the QSAR of these GP inhibitors. Relations between the molecular properties, including dipole moment, solvent accessible area, solvation energy and electrostatistic surface, and the inhibition constant K_I are analyzed.

The compounds in both figure 1 and 2 are used as input and these substances can be divided into six different structural subsets:

1. Compounds with the common structural moiety of glucose (I - VI)
2. Compounds with five-membered sugar ring (VII – X)
3. Compounds with five-membered spiro-heterocycles (IV, V, VII - X)
4. Compounds with six-membered spiro-heterocycles (III)
5. Compounds with spiro-fused imidazolidione ring (II - V, VII - X)
6. Compounds without spiro-fused imidazolidione ring (I, II, VI)

By the QSAR studies, there are several things need to be clarified:

1. The effect of the sugar size to the activity (comparing subsets 1 to 2)
2. The effect of the heterocycle size to the activity (comparing subsets 3 to 4)
3. The effect of the spiro-fused system to the activity (comparing subsets 5 to 6)
4. The effect of the mode of fusion (R/S) at the spiro ring junction to the activity (comparing compounds inside subset 5)

Several modules of InsightII were used in this research, including Biopolymer, Solvation, Delphi and Apex-3D. (For details about the modeling technique, please see Appendix I.) All of them are 95.0 version. Biopolymer is used to build and modify molecules. Delphi is a software package th at calculates the electrostatic potential in and around macromolecules. Solvation is an implemented module in order to simplify and speed up calculating the solvation energy using Delphi. Apex-3d (also called Activity_Prediction) is a very powerful system for biological activity classification and prediction in designing novel drugs. Unfortunately, some part of the system did not work, so the activity prediction could not be studied as planed.

(1). Generation of 3D models and geometry optimization:

The models of the molecules I – X were built in the module of Biopolymer with the help of the Fragment Libraries of InsightII. After creating a new molecule, the potentials and charges may still be incorrect. The builder/forcefield/ potentials menu was used to fix the potentials, charges, and formal charges.

The generated molecular structures are geometry optimized to the individual energy minimum state by using Optimize menu in Apex-3D module. The forcefield I used is "cvff".

(2) Creating multiple Conformations

Molecules are not rigid, they have different conformations. The conformers were generated by using Conformers/Generate menu in Discover module.

(3) Calculation

Calculation includes two parts:

(a) Calculation of molecule properties, such as heat formation, dipole moment, ion potential, etc.

(b) Calculation of solvent accessible area and solvation energy.

The first part was performed in the Similarity menu of Apex-3D module and the second part was performed in the Solvation module. Properties of every conformer of each molecule was calculated and then the average value for each molec ule was calculated using Microsoft Excel.

(4)Creation of the electrostatic surface.

A 3D grid of points, centered on one molecule with all its conformers, was created using Setup menu in Delphi module. The electostatic potential felt by each grid point from the partial charges of each atom was then calculated. Finally, the surface of the protein was created and colored based upon the electrostatic values. Because each molecule has different number of conformers, the minimum and maximum value of the spectrum was modified to make the surface color comparable with each ot her.

Among the molecule properties, dipole moment is the most important for the interaction between a drug and a receptor. Because, at the beginning of the interaction, the distance between the drug and receptor is large and electrost atic field has the largest effect at that situation. After data analysis, it was found that with the increase of dipole moment the inhibition constant is decreased (Fig. 1). So that means, a good inhibitor of GP should have a small dipole moment.

Fig. 1 Dipole Moment vs Ki

After drug entering into human body, it will be existing either in an aqueous environment or a lipid environment. The solvent accessible area has an important effect for drug’s moving and functioning. Here, it is found that the molecule s with larger solvent (water is used as the model in calculation) accessible area, are more active.

Fig. 2 Solvent Accessible Area vs Ki

The free energy transfer of a molecule from vacuum to water is called its solvation free energy D G_sol. Delphi-Based (DB) model is the model type to calculate the solvation energy from vacuum to water. From the data analysis, it is found that with the decrease of solvation energy, the inhibition activity also decreases. So that means, good drug candidates are not like to be surrounded by water.

Fig. 3 Solvation Energy(DB) vs Ki

Octanol is often used as an approximation to the environment felt by an amino acid or a small molecule in a protein interior. The partition coefficients of a small molecules between water and octanol are frequently used as a way of char acterizing the behavior of drug candidates in drug design. The Eisenberg McLachlin (EM) is the model type to calculate the solvation energy from octanol to water. Opposite to fig. 3, Fig.4 shows that with the decrease of solvation energy, the inhibition a ctivity increases. That means, compared with organic environment, active molecules more like the aqueous environment.

Fig. 4 Solvation Energy(EM) vs Ki

The solvation energy from vacuum to octanol was calculated by deduct the EM part from the DB value. In fig. 5, it shows that good drug candidates are more difficult to dissolve in octanol.

Fig. 5 Solvation Energy(DB - EM) vs Ki

The calculation of solvent (water) accessible area is compared with the solvation energy. The conclutions are:

• Large molecules have higher DB solvation energy. So compared with small molecules, they are more difficult to dissolve in water. (Fig. 6)
• Small molecules have higher EM solvation energy. So compared with large molecules, they are more like to be in lipid environment. (Fig. 7) Fig. 8 further improves that small molecules more like to be in lipid environment.

Fig. 6 Solvation Energy(DB) vs Solvent Accessible Area

Fig. 7 Solvation Energy(EM) vs Solvent Accessible Area

Fig. 8 Solvation Energy(DB - EM) vs Solvent Accessible Area

The initial step in the formation of a drug-receptor complex is a recognition event. This process occurs at rather large distances, so the 3D electrostatic field surrounding each molecule therefore plays a crucial role. The molec ular electrostatic potential (MEP) is determined by systematic calculation and sampling of interaction energies using a chemical probe. The result is displayed by making an electrostatic surface of each molecule (Fig. 9, 10, 11).

Fig. 9 Electrostatic surface of molecule I and II

Fig. 10 Electrostatic surface of molecule III - V (top: III and IV, bottom: V and VI)

Fig. 11 Electrostatic surface of molecule VII - X (top: VII and VIII, bottom IX and X)

As the most active inhibitor, molecule IV is used as the base of surface analysis. It is divided into 11 areas, named A to H (Fig. 12). By comparing with other molecules, it is foound that

1. The structure of the right side, from E to G is very important to the activity. All of the inhibitors have E, G, I and J as bumps, and F and H as dents, with E and F negative and others positive. By losing this shape, molecules IX a nd X also lose the activity.
2. The shape of the left side, from A to D is not important to the activity. For example, with a very big bump, Vi is still active.
3. The fusion of -amide with the sugar ring seems effect the activity. With big blue area at B and C, which results from "S" fusion, molecules are more active.
4. The hydroxy groups on the sugar ring are important. For example, the red F area is one of the hydroxy groups on the sugar ring. By losing this hydroxy group, molecules IX and X have a white F. Their activities are also lost.
5. The size of heterocycle does not effect the structures a lot. Molecule III and IV look very similar.
6. The size of sugar ring effects the structures a little bit. Molecule III - IV and VI - X look different. Since III - IV are more active than VI - X, six membered ring should be better than five-membered ring.

Fig. 12 Electrostatic surface of molecule IV

Generally, I will give a high evaluation to the computing methods. They are very powerful and fast in calculations. But due to short of data (there are totally just ten molecules), it is not very easy to interpret the modelin g results. For example, in the data analysis by Excel, it is always difficult to see an obvious trend for the whole serious data. So certain kind of selection is necessary.

Because activity prediction function of Apex-3D did not work, I could not identify the biophore with a good model. If there is chance in the fututure, I want to finish this part.

1. (a) Johnson, L. N.; Barford, D. Glycogen phosphorylase: the structural basis of the allosteric response and comparison with other allosteric proteins, J. Biol. Chem. 1990, 265(5), 2409-2412. (b) M artin, J. L.; Veluraja, K., Ross, K.; Johnson, L. N., Fleet, G. W. J.; Ramsden, N. J.; Bruce, I.; Orchard, M. G.; Oikonomakos, N. G.; Papageorgiou, A. C.; Leonidas, D. D.; Tsitoura, H. S. Glucose analogue inhibitors of glycogen phosphorylase: the design of potential drugs for diabetes, Biochemistry 1991, 30, 10101-10116. (c) Krulle, T. M.; Watson, K. A.; Gregoriou, M.; Johnson, L. N. et al. Specific inhibition of glycogen by a spirodiketopiperazine at the anomeric position of glucopyranose, Tetrahydron Letters 1995, 36(44), 8291-8294. (d) Board M.; Hadwen M.; Johnson, L. N. Effects of novel analogues of D-glucose on glycogen phosphorylase activities in Crude extracts of liver and skeletal muscle. E ur. J. Biochem. 1995, 228, 753-761, (e) Krulle, T. M.; Watson, K. A.; Gregorious, M.; Johnson L. N., Crook S.; Watkin, D. J.; Griffiths R. C.; Nash R. J.; Tsitsanou, K. E.; Zographos S. E.; Oikonomakos, N. G.; Fleet G. W. Specific inhib ition of glycogen phosphorylase by a spirodiketopiperazine at the anomeric position of glucopyranose, Tetrahedron Letters 1995, 36, 8291-8294
2. Agasimundin, Y. S.; Mumper, M. W. and Hosmane, R. S., Inhibition of glycogen phosphorylase B: synthesis, biochemical screening, and molecular modeling studies of novel analogues of hydantocidin, 1998
3. Cambridge Structural Database, Dr. Olga Kennard, F.R.S., Cambridge Crystallographic Data Center, 12 Union Road, Cambridge CB2 1EZ, U.K.