Neurodegenerative disorders are major consequences of excessive apoptosis caused by a proteolytic enzyme known as caspase-3. Therefore, caspase-3 inhibition has become a validated therapeutic approach for neurodegenerative disorders. We performed pharmacophore modeling on some synthetic derivatives of caspase-3 inhibitors (pyrrolo[3,4-c]quinoline-1,3-diones) using PHASE 3.0. This resulted in the common pharmacophore hypothesis AAHRR.6 which might be responsible for the biological activity: two aromatic rings (R) mainly in the quinoline nucleus, one hydrophobic (H) group (CH3), and two acceptor (A) groups (–C=O). After identifying a valid hypothesis, we also developed an atom-based 3D-QSAR model applying the PLS algorithm. The developed model was statistically robust (
Neurodegenerative disorders like Alzheimer’s disease (AD) and Huntington’s disease [
Caspases, a group of cysteine proteases, are proteolytic in nature and key executioners of apoptosis [
The initiation of caspase cascade reaction can only be regulated by caspase inhibitors. Inhibitors of caspase-3 were described as promising cardioprotectants [
Pharmacophore modeling has been one of the important and successful ligand-based approaches for new drug discovery in the last few years [
In the present study, we have developed an atom-based 3D pharmacophore model using PHASE module, which provides the key structural features required for the biological activity. Furthermore, binding interactions of active molecules were analyzed in binding pocket of caspase-3 by using Glide the in SP mode. Atom-based 3D-QSAR model generates cubes which emphasize the structural features required for caspase-3 inhibition. This piece of information can be useful for further designing of more potent caspase-3 inhibitors.
The computational work was performed on Red Hat Linux Enterprise 3.0 with Intel Pentium Core 2-Duo Processor, 1 GB RAM and 120 GB (hard disk). All of the structures were built on Maestro 8.5, a module of Schrodinger [
A dataset of 82 compounds was selected from two series of caspase-3 inhibitors having similar basic nuclei and activity assays [
The structures were built followed by clean geometry and energy minimization with OPLS_2005 force field method. The activity data for each compound were taken as negative logarithms of IC50 values (molar). All of the structures of compounds with their observed and predicted activities data are given in Tables
Caspase-3 inhibitors (Series I) with the observed and the predicted biological activities.
| |||||
---|---|---|---|---|---|
Compound | R1 | R2 | IC50 (M) | pIC50 observed | pIC50 predicted |
6b | F | H | 0.0000628 | 4.20204 | 4.31 |
6c | Br | H | 0.0000371 | 4.430626 | 4.76 |
6d |
|
H |
|
6.677781 | 6.34 |
6e | H | CH3 |
|
5.196543 | 4.72 |
6f* | Br | CH3 |
|
5.801343 | 5.28 |
6g* |
|
CH3 |
|
7.356547 | 6.91 |
6h | H | CH2CO2CH3 |
|
5.332547 | 5.35 |
6i* | F | CH2CO2CH3 | 0.0000025 | 5.60206 | 5.45 |
6j | Br | CH2CO2CH3 |
|
6.337242 | 5.90 |
6k |
|
CH2CO2CH3 |
|
7.79588 | 7.48 |
6l | H | CH2CH2CO2CH3 | 0.0000233 | 4.632644 | 4.63 |
6m* | F | CH2CH2CO2CH3 | 0.0000055 | 5.259637 | 5.48 |
6n | Br | CH2CH2CO2CH3 |
|
5.966576 | 5.94 |
6o |
|
CH2CH2CO2CH3 |
|
7.431798 | 7.66 |
6p | H |
|
|
5.090979 | 5.33 |
6q* | Br |
|
|
5.595166 | 5.89 |
6r |
|
|
|
7.823909 | 7.46 |
6t | Br |
|
|
6.443697 | 6.71 |
6u* | SO3H |
|
|
7.045757 | 7.16 |
6z |
|
|
|
7.79588 | 8.01 |
8a* |
|
|
|
7.48148 | 7.12 |
8b |
|
|
|
7.69897 | 7.72 |
8c |
|
|
|
7.677781 | 7.90 |
8d |
|
|
|
7.638272 | 7.54 |
8e |
|
|
|
7.259637 | 7.01 |
8f |
|
|
|
8.39794 | 8.37 |
Caspase-3 inhibitors (Series II) with the observed and the predicted biological activities.
| ||||
---|---|---|---|---|
Compound | R | IC50 (M) | pIC50 observed | pIC50 predicted |
3 |
|
|
7.268 | 7.23 |
4 |
|
|
7.638 | 7.70 |
5 |
|
|
7.387 | 7.48 |
6 |
|
|
5.26 | 6.05 |
7 |
|
|
7.699 | 7.89 |
8 |
|
|
7.495 | 7.37 |
9 |
|
|
7.013 | 6.90 |
10 |
|
|
7.252 | 7.01 |
11 |
|
|
7.237 | 7.16 |
12 |
|
|
7.26 | 7.12 |
13* |
|
|
7.276 | 7.02 |
14* |
|
|
7.62 | 7.27 |
15 |
|
|
6.818 | 6.64 |
16 |
|
|
6.708 | 6.30 |
18 |
|
|
7.523 | 7.38 |
19* |
|
|
7.678 | 7.67 |
20 |
|
|
7.071 | 6.96 |
21 |
|
|
7.959 | 8.02 |
22 |
|
|
7.959 | 7.53 |
23 |
|
|
6.18 | 6.21 |
24 |
|
|
7.137 | 7.02 |
25 |
|
|
7.432 | 7.59 |
26 |
|
|
7.721 | 7.58 |
28 |
|
|
7.62 | 7.57 |
29 |
|
|
5.638 | 6.54 |
30 |
|
|
7.62 | 7.64 |
31 |
|
|
8.097 | 8.11 |
32 |
|
|
7.658 | 7.73 |
33 |
|
|
7.699 | 7.91 |
34 |
|
|
8 | 7.92 |
35* |
|
|
8.046 | 7.41 |
36 |
|
|
7.658 | 7.78 |
37* |
|
|
6.391 | 7.26 |
38 |
|
|
7.699 | 7.32 |
39* |
|
|
7.456 | 7.58 |
40* |
|
|
7.77 | 7.45 |
41* |
|
|
7.886 | 7.51 |
42 |
|
|
7.854 | 7.94 |
43 |
|
|
8.097 | 8.40 |
44 |
|
|
8.046 | 8.37 |
45 |
|
|
8.222 | 8.36 |
46 |
|
|
8.222 | 8.16 |
47* |
|
|
8.301 | 8.01 |
48 |
|
|
8.301 | 8.45 |
49 |
|
|
8.523 | 8.42 |
51* |
|
|
8.398 | 8.47 |
54 |
|
|
7.921 | 7.88 |
55* |
|
|
7.745 | 7.91 |
56 |
|
|
7.854 | 8.00 |
57* |
|
|
7.886 | 7.98 |
58 |
|
|
8.523 | 8.19 |
59* |
|
|
7.086 | 8.05 |
60 |
|
|
7.824 | 7.32 |
61 |
|
|
6.394 | 6.21 |
62 |
|
|
7.745 | 7.65 |
63 |
|
|
5.526 | 6.46 |
PHASE, version 3.0, was used for pharmacophore elucidation and QSAR model building. It provides six built-in types of pharmacophore features: hydrogen-bond acceptor (A), hydrogen-bond donor (D), hydrophobe (H), negative ionizable (N), positive ionizable (P), and aromatic ring (R). Ligands were processed with the LigPrep program to assign protonation states appropriate for pH 7.0. All of the molecules were considered to be active for building the hypothesis. Conformer generation for each ligand was carried out with the ConfGen method where the maximum number of conformers was set as 100 per ligand, and total steps per rotatable bond were set as 100 (Table
Ligand preparation.
Compound | Activity | Conformers | Compound | Activity | Conformers |
---|---|---|---|---|---|
6b | 4.202 | 1 | 19 | 7.678 | 29 |
6c | 4.431 | 1 | 20 | 7.071 | 26 |
6d | 6.678 | 4 | 21 | 7.959 | 9 |
6e | 5.197 | 1 | 22 | 7.959 | 13 |
6f | 5.801 | 1 | 23 | 6.18 | 20 |
6g | 7.357 | 4 | 24 | 7.137 | 12 |
6h | 5.333 | 2 | 25 | 7.432 | 18 |
6i | 5.602 | 2 | 26 | 7.721 | 4 |
6j | 6.337 | 2 | 28 | 7.62 | 5 |
6k | 7.796 | 8 | 29 | 5.638 | 14 |
6l | 4.633 | 7 | 30 | 7.62 | 18 |
6m | 5.26 | 7 | 31 | 8.097 | 8 |
6n | 5.967 | 7 | 32 | 7.658 | 15 |
6o | 7.432 | 16 | 33 | 7.699 | 4 |
6p | 5.091 | 2 | 34 | 8 | 4 |
6q | 5.595 | 2 | 35 | 8.046 | 6 |
6r | 7.824 | 7 | 36 | 7.658 | 3 |
6t | 6.444 | 2 | 37 | 6.391 | 9 |
6u | 7.046 | 16 | 38 | 7.699 | 8 |
6z | 7.796 | 2 | 39 | 7.456 | 3 |
8a | 7.481 | 4 | 40 | 7.77 | 4 |
8b | 7.699 | 92 | 41 | 7.886 | 8 |
8c | 7.678 | 35 | 42 | 7.854 | 6 |
8d | 7.638 | 37 | 43 | 8.097 | 8 |
8e | 7.26 | 24 | 44 | 8.046 | 7 |
8f | 8.398 | 8 | 45 | 8.222 | 8 |
3 | 7.268 | 15 | 46 | 8.222 | 7 |
4 | 7.638 | 17 | 47 | 8.301 | 5 |
5 | 7.387 | 32 | 48 | 8.301 | 29 |
6 | 5.26 | 6 | 49 | 8.523 | 8 |
7 | 7.699 | 8 | 51 | 8.398 | 8 |
8 | 7.495 | 17 | 54 | 7.921 | 39 |
9 | 7.013 | 24 | 55 | 7.745 | 5 |
10 | 7.252 | 14 | 56 | 7.854 | 4 |
11 | 7.237 | 27 | 57 | 7.886 | 5 |
12 | 7.26 | 8 | 58 | 8.523 | 4 |
13 | 7.276 | 41 | 59 | 7.086 | 7 |
14 | 7.62 | 35 | 60 | 7.824 | 4 |
15 | 6.818 | 11 | 61 | 6.394 | 6 |
16 | 6.708 | 56 | 62 | 7.745 | 4 |
18 | 7.523 | 16 | 63 | 5.526 | 3 |
Score hypothesis.
ID | Survival | Site | Vector | Volume | Selectivity | Matches | Energy | Activity |
---|---|---|---|---|---|---|---|---|
AAHRR.6 | 3.785 | 0.99 | 1 | 0.792 | 1.711 | 82 | 0 | 8.523 |
AAHRR.7 | 2.493 | 0.44 | 0.595 | 0.454 | 1.728 | 82 | 0.621 | 7.62 |
AAHRR.1 | 2.463 | 0.29 | 0.771 | 0.4 | 1.708 | 82 | 0.936 | 7.387 |
AAHRR.5 | 2.251 | 0.19 | 0.636 | 0.421 | 1.71 | 82 | 6.812 | 7.387 |
Pharmacophore model development was performed after dividing the dataset of 82 compounds into training (62 compounds) and test sets (20 compounds) and after applying PLS factor 4 and grid spacing 1 Å. The training set selection was done on the basis of the information contained in terms of both structural features and biological activity ranges of molecules. The training set included the compounds that cover all ranges of activities (high, moderate, and low active) [
This is done to test the internal stability and the predictive ability of the QSAR models. Developed QSAR models were validated by the following procedures [
Internal validation was carried out using the leave-one-out (
For external validation, the activity of each molecule in the test set was predicted using the model developed by the training set. The pred_
Docking study was performed on Glide 5.0 module of Schrodinger [
We developed an atom-based 3D-QSAR model by using a grid spacing of 1.0 Å and a maximum PLS factor of four. The model was developed using five-point common pharmacophore hypothesis AAHRR.6 (Figure
Statistical data of the 3D-QSAR model for PLS factor four.
Training set correlation | Test set correlation | |||||
---|---|---|---|---|---|---|
|
|
SD |
|
|
RMSE | Pearson’s |
0.925 | 175.9 | 0.298 |
|
0.798 | 0.415 | 0.897 |
Common pharmacophore hypothesis and distances between pharmacophoric sites. Pink spheres with arrows show hydrogen-bond acceptor with lone pairs of electron. Green sphere shows hydrophobe, and yellow rings show ring aromatics.
The plots between the observed and the predicted activities were made for both the training and test sets (Figures
Plot between observed and predicted biological activities of the training set of compounds.
Plot between observed and predicted biological activities of the test set of compounds.
The developed 3D-QSAR model can be visualized as a cluster of cubes, which provided additional information about the structural features required for activity. The blue cubes point out favorable features, and the red cubes point out unfavorable features for activity. A comparative study of these favorable and unfavorable features for the most active (58) and the least active (6b) compounds is shown in Figure
Pictorial representation of cubes for the most active (compound 58) and the least active (compound 6b) compounds where blue and red cubes show favorable and unfavorable regions for activity.
In our previous study [
We have successfully developed a QSAR model that is statistically robust and consistent with earlier observations. It also provided important SAR information that can be used in future drug designing on this target. The developed QSAR model can also serve as a query model for virtual screening of large libraries.
We explored the binding interactions of the two most active compounds (49 and 58) with the receptor. Our aim has been to understand the binding mode of this class of molecules and to cross-check whether the developed pharmacophore model fits properly to the active site. To achieve this, we performed flexible docking using Glide SP mode. Information obtained from binding interactions will be of further help in future design of caspase-3 inhibitors. In addition to analyzing the binding interactions, superposition of most active compounds with respect to common pharmacophore hypothesis was also done. It indicated outstanding superposition of the top two most active compounds. Binding interactions of compound 49 at the active site are depicted in Figure
Binding interactions of compound 49 at the active site where the docked ligand is green in color. Hydrogen bonds are expressed as dotted lines in purple color, and active site residues are demonstrated in orange color.
Binding interactions of compound 58 at the active site where the docked ligand is green in color. Hydrogen bonds are expressed as dotted lines in purple color, and active site residues are demonstrated in orange color.
The interactions of both compounds (49 and 58) with surrounding amino acids were also analyzed by MOE molecular modeling software to provide clear view of
Out of these interactions of the most potent compounds to the active site residues of caspase-3, authors concluded that Ser 209, Ser 251, and Tyr 204 are crucial residues for activity. In addition to these key residues, Ser 205 also has precious contribution to the activity. Since compounds 49 and 58 have similar activities, so they should have almost similar fitness on common pharmacophore, which is clear from Figure
Superposition of compounds 49 and 58 on common pharmacophore hypothesis.
Superposition of pharmacophore hypothesis on docked ligand [compound 58 (X1)] at the binding site where the docked ligand is in red color. Hydrogen bonds are expressed in yellow color, and amino acids are expressed in their standard form (H = histidine, C = cysteine, S = serine, R = arginine, W = tryptophan, Y = tyrosine, D = aspartic acid, F = phenylalanine, and E = glutamic acid).
To cross-check whether the developed pharmacophore model is consistent to the active site residues of caspase-3, we performed the overlay studies. In this study, we aligned the developed pharmacophore on the docked poses of the most potent compound (58) and then manually checked whether the predicted pharmacophoric features were consistent with the protein active site or not. The pharmacophore model predicted the aromatic ring (R) features as an important determinant in biological activity. The docking and the overlay studies revealed a number of favorable aromatic-aromatic and aromatic-aliphatic interactions as shown in Figure S3 provided in the Supporting Information [
The presence of hydrophobe (H) is also consistent with the docking results, and the presence of side chains of many amino acids like ASP253, PHE256, and SER251 makes favorable hydrophobic interactions. The acceptor feature A7 is also somewhat consistent, due to the presence of ARG207 although we did not observe any hydrogen bonding between the ligand and this residue. The feature A6 predicted by the pharmacophore model is inconsistent and redundant. We also feel that the pharmacophore model may have missed the following important features, which are, otherwise, revealed by the docking studies: another aromatic ring feature from the 2nd position of the pyrrolo[3,4-c]quinoline-1,3-diones. hydrophobic substituent on sulfonyl group at the 8th position of pyrrolo[3,4-c]quinoline-1,3-diones,
The identification of intrinsic pharmacophoric features strongly depends on structural diversity and composition of the training set compounds. Missing a few probable pharmacophoric features during a ligand-based model building exercise is obvious. Wherever possible, a developed pharmacophore model must be cross-checked for consistency by docking in the protein active site. Missing these features does not undermine the developed pharmacophore model in any way. Instead, it provides an easier and faster way to screen the larger dataset. Therefore, for future virtual screening of caspase-3, we can adopt a strategy of screening the dataset with the developed pharmacophore and then analyze the top HITs by docking or other structure-based studies.
We were successful in developing an atom-based 3D-QSAR model with high predictive ability. The developed hypothesis consisted of five features with two hydrogen-bond acceptors (A), one hydrophobe (H), and two aromatic rings (R). From the ligand-based study, we can conclude that different electron-withdrawing substituent and with bulkiness at position 8 those heteroaryl substituent at position 2 would be beneficial for designing the new scaffolds of caspase-3 inhibitors. In addition to these suggestions, it is also necessary for a carbonyl group to be electron deficient. These results are also consistent with those of our previous studies. It is further concluded that the developed 3D-QSAR model can also serve as a query model for virtual screening from database to find out new potential caspase-3 inhibitors. Thus, the objective of the present study is to use a computational approach for rapid cost-effective evaluation of caspase-3 inhibitors. These findings can provide direction in order to find out a set of potent caspase-3 inhibitors to be synthesized and examined experimentally for their biological activity.
Author Simant Sharma is thankful to AICTE, New Delhi, India, for providing National Doctoral Fellowship for financial support. Authors are also thankful to Dr. Venkatesan J., Jagannath Behera, Devyani Dube, and Abhishek Jain for software-related technical support and for improving the language of this paper.