Results

Slide was previously used to screen for potential ligands to a bacterial aspar-tic protease, the human estrogen receptor, glutathione transferase, and HIV-1 protease [7,48]. Here, we screen for potential ligands to human uracil-DNA glycosylase (coordinates of a complex with 6-amino-uracil provided by C. Mol and J.A. Tainer, The Scripps Research Institute), to the ligand-binding domain of the human progesterone receptor (PDB entry 1a28), and to E. coli dihydrofolate reductase (PDB entry 1ra9). The modeling of inducible complementarity and the control of molecular diversity in the set of potential ligands found by Slide has been described elsewhere [7], and here we include knowledge-based solvation.

We screened two different databases for the three target proteins:

• A subset of 70 113 compounds taken from the Cambridge Crystallo-graphic Database System (CSD, http://www.ccdc.cam.ac.uk). These are all organic compounds with less than 100 atoms and at least three interaction centers that can be mapped onto template points.

• 105 517 compounds from the NCI database (http://dtp.nci.nih.gov), which were taken from the conformers for the open set of this database as they were prepared by the group of J. Gasteiger using CORINA [54].

We used different approaches for designing the binding-site templates. The smallest template, consisting only of six points, was generated for the progesterone receptor. The interaction points were generated based on the centers of four carbon rings and two ketone oxygen atoms in the progesterone bound to the receptor in PDB entry 1a28, resulting in a template consisting of four hy-drophobic and two acceptor points. One water molecule, which interacts with the bound progesterone in this structure was included in the binding site during screening. A search with such a small template is like a pharmacophore-based search, which restricts the set of potential ligands that SLIDE finds to compounds more or less similar to the known ligand, since that ligand and each potential new ligand share at least three interaction centers due to the triangle matching during docking. With this small template for the progesterone receptor, the total screening time for the more than 175 000 compounds was about nine minutes on an Intel Pentium II/450 processor running Solaris 2.7. Figure 1 shows a typical example of the kind of ligand SLIDE found for this screen with the small template. The ligand is 16a,17a-cyclopenteno-progesterone (CSD entry BUBRUJ), which is the known ligand progesterone with a cyclopentene substituent added to its D ring. It obtained a score of 39.9, which ranks it 52th out of the 175 630 screened compounds. The highest-

Figure 1. A cyclopenteno-progesterone (grey) from the CSD (entry BUBPUJ) was identified as a potential ligand and docked by Slide into the ligand-binding site of the human progesterone receptor (PDB 1a28). The template was based on six interaction centers of the progesterone from the crystal structure, which is shown in yellow tubes and is overlaid virtually exactly by Slide's ligand. To fit the additional cyclopentene substituent, four side chains in the receptor underwent minor conformational changes; their native conformation is shown in yellow, and Slide's conformation for these side chains is colored by atom type (green: carbon; red: oxygen). Note that the hydrophobic cyclopentene is in contact with hydrophobic groups in the receptor.

Figure 1. A cyclopenteno-progesterone (grey) from the CSD (entry BUBPUJ) was identified as a potential ligand and docked by Slide into the ligand-binding site of the human progesterone receptor (PDB 1a28). The template was based on six interaction centers of the progesterone from the crystal structure, which is shown in yellow tubes and is overlaid virtually exactly by Slide's ligand. To fit the additional cyclopentene substituent, four side chains in the receptor underwent minor conformational changes; their native conformation is shown in yellow, and Slide's conformation for these side chains is colored by atom type (green: carbon; red: oxygen). Note that the hydrophobic cyclopentene is in contact with hydrophobic groups in the receptor.

ranked progesterone from the CSD received a score of 37.4. A total number of 154 potential ligands were docked into the binding site and obtained a score higher than 35, which is a reasonable cutoff for ligands similar in size and chemistry to the known ligand. Like the progesterone in the crystal structure in PDB 1a28, this ligand makes one water-mediated hydrogen bond. To fit the highly rigid cyclopentene-progesterone into the binding site, adjustments in protein side chains were necessary. The figure shows four side chains in their native conformation together with the final, rotated conformation proposed by Slide. Note that only minor rotations were necessary to accommodate the cyclopentene, which demonstrates favorable hydrophobic complementarity with the neighboring side chains in the progesterone receptor.

In the screening for ligands for the human uracil-DNA glycosylase [55,56], the binding site was taken from a crystal structure of a complex with 6-amino-uracil bound deep in the active-site cleft. Twelve water molecules from this structure were predicted as being conserved by Consolv and included in the binding site during screening. A known inhibitor for this DNA-repair enzyme is a 84-residue protein that mimics DNA but binds irreversibly to the glyc-osylase [57]. We used the positions of five H-bond donors and acceptors in the bound 6-amino-uracil as key points (out of which at least one must be

Figure 2. The structure of 6-amino-uracil, the ligand present in the crystal structure of the human uracil-DNA glycosylase used for screening, is shown together with two highly-ranked molecules suggested by SLIDE as potential ligands for this enzyme.

matched by any potential ligand) in a template consisting otherwise of 150 automatically generated interaction points. The cumulative screening time for both databases was slightly over 17 h. Figure 2 shows the structure of 6-amino-uracil and two of Slide's ligands, one with obvious resemblance to the known ligand. Figures 3 and 4 show these ligands in Slide's binding modes together with key side chains and waters that interact with them. The ligand in Figure 3, CSD entry PICTIE, obtained a score of 32.8 and rank 55 with three water-mediated interactions to the protein, and the ligand in Figure 4, NCI entry 39807 (CAS 6313-89-9), obtained a score of 27.2, which ranked it 384th in the set of 683 potential ligands that were docked by Slide and scored higher than 25.0 out of the data set of over 175 000 compounds that were screened. The latter ligand binds similarly to 6-amino-uracil, but shows better complementarity due to additional water-mediated hydrogen bonds to the protein.

In the screening runs against E. coli dihydrofolate reductase (PDB entry 1ra9), again a hybrid template design was used. To ensure that all ligands docked by Slide interact with the side chains binding pyrimidine in the

Figure 3. This figure shows 3'-deoxysangivamycin (CSD entry PICTIE), which was docked by Slide as a potential ligand into the active site of human uracil-DNA glycosylase, a DNA-repair enzyme. Key side chains and binding-site waters that interact with the ligand are shown, and feasible hydrogen bonds are indicated by dotted lines. Two side chains of the enzyme, a phenylalanine and a glutamine, were rotated by Slide to accommodate the ligand and are also shown in their original conformation (yellow).

Figure 3. This figure shows 3'-deoxysangivamycin (CSD entry PICTIE), which was docked by Slide as a potential ligand into the active site of human uracil-DNA glycosylase, a DNA-repair enzyme. Key side chains and binding-site waters that interact with the ligand are shown, and feasible hydrogen bonds are indicated by dotted lines. Two side chains of the enzyme, a phenylalanine and a glutamine, were rotated by Slide to accommodate the ligand and are also shown in their original conformation (yellow).

Figure 4. A ligand from the NCI database (entry 39807), docked by Slide into the active site of human uracil-DNA glycosylase. It mimics the binding of 6-amino-uracil, the ligand bound in the structure that was used for screening. This ligand shows better complementarity than the original one, due to the additional carboxylate group that interacts with two conserved waters and a histidine side chain.

CSD: FIRNID

nh2

Figure 5. Two CSD compounds that were selected and docked into the active site of dihydrofolate reductase by SLidE and obtained high scores. Both are known DHFR inhibitors.

Figure 5. Two CSD compounds that were selected and docked into the active site of dihydrofolate reductase by SLidE and obtained high scores. Both are known DHFR inhibitors.

known ligands methotrexate and dihydrofolate, two runs of the automatic template generator were done: one to specify 23 key points located in the pyrimidine region of the binding site, and another to fill the remaining part of the binding site with 64 additional template points. Four binding-site waters from the crystal structure of the ligand-free DHFR were predicted to mediate interactions and included during screening. The screening time for the 175 000 compounds was about 14 h. In the set ofpotential ligands identified by SLIDE were at least two known DHFR inhibitors (Figure 5), and their key interactions are shown in Figures 6 (CSD entry JOXTIZ) and 7 (CSD entry FIRNID). Slide's scores for these ligands were 5 1.1 and 5 1.9, which ranked them 205th and 141th, respectively. Both ligands place a pyrimidine group in the same site, and the adamantyl-pyrimidine (FIRNID, Figure 7) binds deeper in the corresponding cleft. In the docking of CSD ligand JOXTIZ (Figure 6) a second water molecule fills the non ligand-occupied space. This water was displaced by an amino group in the docking ofCSD ligand FIRNID (Figure 7), which replaces the hydrogen bonds to the other water and to the aspartic acid side chain of DHFR, so that this displacement was not penalized in Slide's score.

Figure 6. Methylbenzoprim (CSD entry JOXTIZ), a known potent DHFR inhibitor, docked by Slide into the active site of E. coli dihydrofolate reductase. The ligand was selected by Slide out of 175 000 compounds in the screening database. Its pyrimidine group binds in the same cavity as the pyrimidine of the natural ligand, dihydrofolate, which was aided by positioning key template points in that area. The deeper part of this cavity is occupied by two bound water molecules, which were observed in the ligand-free protein structure that was used for screening (PDB 1ra9) and predicted by Consolv as being conserved upon ligand binding. One side chain, a leucine, was rotated by Slide upon ligand docking, and its original conformation is shown in yellow.

Figure 6. Methylbenzoprim (CSD entry JOXTIZ), a known potent DHFR inhibitor, docked by Slide into the active site of E. coli dihydrofolate reductase. The ligand was selected by Slide out of 175 000 compounds in the screening database. Its pyrimidine group binds in the same cavity as the pyrimidine of the natural ligand, dihydrofolate, which was aided by positioning key template points in that area. The deeper part of this cavity is occupied by two bound water molecules, which were observed in the ligand-free protein structure that was used for screening (PDB 1ra9) and predicted by Consolv as being conserved upon ligand binding. One side chain, a leucine, was rotated by Slide upon ligand docking, and its original conformation is shown in yellow.

0 0

Post a comment