HIPPO

• Introduction
• H-bond site detection and representation
  • Complex hydrogen bonding sites
• Covalent bonding sites
• Hydrophobic regions
• Test results
  • The APPA binding site of Trypsin
  • The GDP binding site of ras P21
• References

Introduction

HIPPO is a new module in the SPROUT program suite [1] that automates the selection of interaction sites within a receptor site. These interaction sites are known as target sites and they are used as starting points for structure generation.

The method used in HIPPO is a rule based approach where the rules are derived from experimental data collected from the literature, e.g. [2]. Typical donor and acceptor atoms are located in the protein, intramolecular hydrogen bonds are identified, hydrogen bonding atoms near to the surface of the receptor site are found, and finally hydrogen bonding regions are computed for them with tolerances. Metal ions and residue motifs that tend to form covalent bonds to ligands (e.g. Ser-His-Asp triad) are also identified and the appropriate target sites are generated.

Triad example

The method of defining hydrogen bonding positions is similar to those used in HSITE [3] and LUDI [4]. In HIPPO, however, once the ideal bond has been computed, tolerances are applied to both the distance and the direction so that a target site is defined by a geometric region rather than as a single point or a grid.

H-bond site detection and representation

The input to the program is a PDB (Protein Data Bank) file in which any ligand that is bound in the receptor site has been removed. The hydrogen atoms are also removed and no bonding information is contained in the file.

1. The bonds and bond types are found according to a lookup table that contains the connectivity of the 20 well-known amino acids. The bond types are estimated by distance for unknown atom labels (cofactors, solvent and ligand atoms etc.).
2. The hydrogen positions for donors and lone pair positions for acceptors are then calculated on the basis of hybridisation and the directions of existing connections.
3. Potential intra-molecular hydrogen bonds within the protein are identified using the following criteria:

   Intra H-bond test

4. When residues can exist in different protonation states, e.g. His, Asp, Gln, HIPPO selects the protonation state that allows the maximum number of intra-molecular hydrogen bonds. The orientation of the rotatable terminal -OH and -NH2 groups are optimised for hydrogen bonds (and treated as fixed afterwards). Any solvent molecules are also oriented to form the best possible hydrogen bonds.

   Example for H-bond network

5. The donor hydrogens and acceptor lone pairs that are used in intra-molecular hydrogen bonds are excluded from further calculation.
6. Donor and acceptor atoms that are too far away from the surface of the receptor site are excluded.
7. Complementary acceptor and donor regions are generated within the cavity to correspond to each of the remaining donor hydrogens and acceptor atoms of the receptor, respectively. Regions are also generated to correspond to the solvent molecules. Tolerances are then applied around the optimal directions and distances of the hydrogen bonds to define the geometric regions as illustrated in the following figure.

   H-bond sites

   Rotating the red region about the D-H bond defines the 3D shape of an acceptor region. Similarly, rotating the white and blue regions about the C=O bond defines the 3D shape of a donor and hydrogen region.

In cases when the role of a receptor atom is not certain because it can exist in different protonation states, the most likely site is generated, but it is colored differently (green) to indicate the ambivalency of the site. These sites can be switched to the opposite type by the user. For example, a carboxylic acid is most likely to be negatively charged at normal pH. Therefore, it is primarily an acceptor, but it may also exist in the neutral state with a hydrogen that can be easily donated. In this case, unless an intra-molecular hydrogen bond is detected in which the group already plays donor role, an ambivalent donor site is generated.

Complex hydrogen bonding sites

Complex hydrogen bonding cases, e.g. multicentred and/or bifurcated centres, are identified by the intersections of the single regions (with large tolerances, e.g. with default values of 1.6A-2.3A and 130 degrees or even greater values).

Having generated the single donor and acceptor regions, each pair is tested to see if they intersect. The intersection regions are calculated, and they are tested against the other single regions once more to detect triple intersections.

Covalent bonding sites

The serine hydrolases contain an Asp(Glu)-His-Ser catalytic triad in the active site [5]. The OG_Ser of this triad loses its hydrogen and forms a temporary covalent bond to the ligand in the transition state [6]. Inhibitors can be designed to form a similar covalent bond to that atom.

HIPPO recognises this possibility and suggests a covalent site around the OG_Ser. The geometric representation includes tolerances for the bond length and bond angle, and allows any torsional angle about the CB_Ser - OG_Ser bond that does not cause bad Van der Waals contact between the ligand atom and other receptor atoms (e.g. His).

Covalent site

The 3D shape of the covalent site forms part of a ring.

The metalloproteins constitute one of the four main classes of proteolytic enzymes [7]. The metal ion is at the centre of the active site of the "true" metalloproteins and is involved in the catalytic event. A good inhibitor for these proteins should probably include an atom that can bind to the metal ion.

HIPPO identifies metal ions (Zn, Mg, Cu, Ca, Co, Fe, Ni, Mn) in the receptor PDB file, calculates the most likely direction of the free valency according to the existing connections (to protein or solvent atoms) and generates the appropriate target site.

Metal site

The predicted bond has a default length that is the mean value of the existing connections of the metal ion. The default angle and bond length tolerances are calculated as the maximum deviation among the values of the existing bonds. The 3D shape of a metal target site is shown by example.

Hydrophobic regions

The hydrophobic regions of the cavity are represented on a 3D grid. The cells that are within a given distance limit (the default is 3.9Å) of a hydrophobic residue atom are marked. The number of hydrophobic residues that fall within the limit is registered for each cell so that the program can display multiple regions if desired. The following picture shows the hydrophobic grid cells (yellow cubes) of the APPA binding pocket of Trypsin.

Example

Test results

The program has been tested on a number of protein structures taken from the PDB. Some of them have been crystalised with a bound ligand and others are without a ligand. The predictive power of HIPPO is demonstrated by first removing a bound ligand and then running HIPPO to generate some target sites. The ligand can then be displayed and the actual binding atom positions compared to the predicted regions.

The APPA binding site of Trypsin

The picture below shows APPA (p-amidino-phenylpyruvate) bound to Trypsin. The target regions that are detected by HIPPO and correspond to the binding ligand atoms are displayed with translucent representation.

Detected sites

A movie demonstrates the above scheme in 3D by rotating it about the X axis.

The GDP binding site of ras P21

The picture above shows a possible selection of target sites in the GDP (guanine diphosphate) binding site of ras P21 protein. The picture below shows the solvent accessible surface by a translucent gray representation. The target regions are reduced according to the steric constraints.

Another selection

Click on the picture above to see a movie that demonstrates the scheme in 3D by rotating it about the X axis.

Conclusion

HIPPO successfully predicts hydrogen (and covalent) bonding regions that correspond to the positions of the binding atoms of known ligands. It also finds a number of extra sites that could be used to improve binding by designing ligands that have additional functional groups to satisfy the extra sites.

The geometric hydrogen bonding regions produced by HIPPO initially are quite large, however, they are significantly reduced by removing the portions that violate the steric constraints of the receptor. In addition, the tolerance values and hence the sizes of the regions can be adjusted by the user.

A possible advantage of this method over the grid-based approaches (where the energy is calculated using a probe atom/molecule placed at every possible grid position) is that it does not depend on a discrete grid resolution, but it gives the precise positions for optimal binding. The directionalities of the suggested hydrogen bonds are given precisely as well, and can be used directly for structure generation in SPIDeR.

An important novel feature of the method is that it deals with multicentred and bifurcated hydrogen bonding possibilities. This has been proved to be very useful in finding regions that are occupied by existing ligand atoms with good binding properties. Experimental data shows that about 40% of the known receptor-ligand hydrogen bonding interactions are not simple hydrogen bonds [2].

Further development of the program aims generate target sites from a pharmacophore hypotheses when the receptor is not known (from series of active and passive ligands).

References

1. V. Gillet, A.P. Johnson, P. Mata, S. Sike, P. Williams, J. Comput.-Aided Mol. Design, 7 (1993) 127.
2. G.A. Jeffrey, W. Saenger, "Hydrogen Bonding in Biological Structures", Springer-Verlag 1991.
3. D.J. Danziger, P.M. Dean, Proc. R. Soc. Lond., B236 (1989) 101.
4. H.J. Bohm, J. Comput.-Aided Mol. Design, 6 (1992) 61.
5. Z.S. Derewenda, U. Derewenda, P.M. Kobos, J. Mol. Biol., 241 (1994) 83.
6. G.W. Zhou, J. Guo, W. Huang, R.J. Fletterick, T.S. Scanlan, Science, 265 (1994) 1059.
7. P. Harrison, "Metalloproteins. Part 2: Metal Proteins with Non-Redox Roles", MACMILLAN 1985.