3D QSAR and docking studies on benzoylsulfonohydrazides as histone acetyltransferase KAT6A inhibitors

Sixty-one analogs of benzoylsulfonohydrazides were subjected to 3D QSAR studies using CoMFA and CoMSIA techniques followed by docking studies to develop a correlation of the structure with their respective activities. The generated model had shown good predictability and the contour analysis followed by docking study has provided an insight to develop new inhibitors. The cross-validation values corresponding to CoMFA and COMSIA were observed to be within the acceptable criterion (q2 > 0.5). The docking analysis of the best active compound shown was −41.81 kcal/mol. From the obtained analysis results of CoMFA as well as CoMSIA, the data can be useful to develop more potent histone acetyltransferase inhibitors.


INTRODUCTION
Histone acetyltransferases has a crucial role in hematogenesis and are one among the chromatin modifying enzymes that are responsible for the post-translational modifications of the histone in the nucleosome of a cell which include methylation, acetylation, ADP-ribosylation, phosphorylation, and ubiquitination (Allfrey et al., 1964;Des Jarlais and Tummino, 2016;Lawrence et al., 2016;Luger and Richmond, 1998;Sterner and Berger, 2000;Sadakierska and Filip, 2015) Among these, acetylation is known to be the earliest modification which is related to the gene activation as it is linked functionally with transcription activation by adding acetyl group (-COCH3) to the s-amino group present in lysine residue resulting in loosening of the nucleosome structure (Bannister and Miska, 2000;Roth et al., 2001;Parthun, 2012). This acetylation occurs at the N-terminal of the basic amino acid (lysine) dense region of the histone core; as a result, the acetyl-CoA gets transferred to -NH+ of lysine neutralizing the +ve charge (Loidl, 1994). There are three different families of histone acetyltransferase (HAT) which are p300, , and MYST (MOZ, Ybf2, Sas2, and Tip60 as founding members) (Voss and Thomas, 2018). Monocytic leukemia zinc finger protein (MOZ HAT) is an oncogene of MYST family which is involved directly in the process of hematopoiesis as it forms HAT complex that acetylates H2 (A and B), H3 and H4 corresponding to the up-regulation of gene there by activating the oncogene resulting in acute myeloid leukemia (AML) (Borrow et al., 1996;Champagne et al., 2001;Dohner et al., 2015). Granulocytic leukemia (synonym for AML) is a cancer that is characterized by the over production of white blood cells that are immature (myeloblasts) and functionally causes affect on blood as well as bone marrow. These cells prevent leukopoiesis of normal blood cells that act as the defense system in the body resulting in poor immune system and also cause anemia, bruising, and easy bleeding (Camos et al., 2006;Ullah et al., 2008). KAT6A is one among the five subfamilies of MYST HAT that is responsible for the hostile form of acute myeloid leukemia due to rearrangement in the KAT6A gene (Lowenberg et al., 1999). Hence, inhibiting HAT KAT6A would help preventing the continuous growth of tumors and their metastasis in case of AML.
Literature survey reveals only two compounds that were discovered which include WM-8014 and WM-1119 of which latter was found to be the most active compound with IC50 value of 0.25 pM showing less protein-binding than the former with IC50 value of 2.3 pM and high protein-binding (Baell et al., 2018). Based on the study of WM-1119, the same group further discovered benzoylsulfonohydrazides as the potent inhibitors of HAT KAT6A (Leaver et al., 2019). Therefore, in the present in silico study, we endeavored to develop a 3D QSAR model adopting CoMFA and CoMSIA techniques on 61 benzoylsulfonohydrazide analogs from which the contour maps of the most active compound could give an insight in developing inhibitors with enhanced activity against HAT KAT6A.

Data set preparation
Sixty-one benzoylsulfonohydrazides as potent inhibitor of histone acetyltransferase KAT6A were taken from the literature (Leaver et al., 2019). These molecules were reported to inhibit in micro-molar range (IC50) and were converted to pIC50; which is the negative logarithm of IC50 (i.e., pIC50 = −log IC50). The molecules were constructed in SYBYL X and minimized by Gasteiger-Huckel charges using distance dependent dielectric and Powell-conjugate gradient algorithm with 0.05 kcal/mol convergence. All the default parameters were adopted during the minimization of the molecules.

Alignment
The alignment of the molecules determines the accuracy of the model. The molecules in the present study were aligned upon the most active compound by selecting a basic skeleton of benzoylsulfonohydrazide. Figure 1 represents the common structure used to sketch the molecules, whereas alignment of all 61 optimized molecules on the basic skeleton are presented in Figure 2.

3D QSAR Model construction
Model of 3D QSAR of benzoylsulfonohydrazides in the present study was constructed on SYBYL X in which CoMFA and CoMSIA methods were adopted to determine the relation between the bioactivity with their corresponding 3D structure of molecules. CoMFA model describes the steric fields and the electrostatic fields of the molecules under study, whereas CoMSIA model describes the hydrophobic, HBD and HBA along with steric as well as electrostatic fields.

Statistical validation
Partial least square method (PLS) is a standard statistical regression tool used to predict the 3D QSAR model. It was adopted for the present study as it can analyze the data in a realistic way and interpret the contribution of the molecular structure with the biological activity. For CoMFA, all the parameters which include cross-validation, correlation coefficient, standard error of estimate, f-value etc. were obtained taking number of components of 5 and 6, whereas column filtering of 2 and 1 for CoMFA and CoMSIA, respectively. The study was conducted dividing the molecules in 1:3 ratio of test and training molecules, respectively. Therefore, 15 molecules were selected randomly which were grouped as test set and the remaining 46 were grouped as training set. With this, leave one out (LOO) was used in order to establish the reliability of the generated model for CoMFA as well as CoMSIA. All other parameters were recorded for no-validation, cross-validation, and bootstrapping. The activity was predicted for the test as well as training set and correlated with the experimental pIC50 values.

Molecular docking
It is an important technique to find the interaction of the ligand with a specific protein of interest. The study was done using AutoDockTools software (Morris et al., 2009). The protein structure (6CT2) of 2.128A resolution in PDB format was downloaded from database of protein (https://www.rcsb.org) and the ligand structures that were constructed and their energy was minimized in sybyl X and were further prepared within the docking software. The active ligand interacting site of the protein was noted from PDBSum (http://www.ebi.ac.uk) as SER 690 (A). The x, y, z coordinates were taken from Spdbv protein viewer tool and entered in the grid to generate active site grid box. The crystal water was removed prior to docking simulation followed by docking of the protein with highest and lowest active ligands. Compound PS-97 was the most active while the least active compound was PS-9.

Statistical results
A model for CoMFA and CoMSIA was produced using 61 KAT6A inhibitors using benzoylsulfonohydrazide as the  skeleton for alignment. The image of alignment of the inhibitors can be visualized in Figure 2.
The model was developed by randomly dividing the molecules in 1:3 ratio of test and training set and performing PLS analysis to determine the prediction power of the model. The pIC50 values were predicted for both series. Table 1 shows the predicted as well as residual values of these analogues and their correlation is shown in Figures 3 and 4, respectively.

Statistical analysis
The statistical results obtained after running PLS regression for both the variables, i.e., CoMFA and CoMSIA, which includes LOO, no validation, cross validation, and also Bootstrapping, were recorded to evaluate reliability of the developed model. In case of CoMFA, the q 2 obtained was 0.678; r 2 was 0.948, whereas F-value and SEE was 144.505 and 0.226, respectively. Steric field contribution was 47.8%, while

Contour analysis
Contour maps analyze the characteristics of the fields around the molecules. These are used to find the basic structural requirement for the bioactivity which facilitate for the development of inhibitors with high potency. Best active compound (P5-97) of training set was considered to analyze contour maps by setting contribution values of favored as well as disfavored region to 80% and 20%, respectively. The steric contributions of the best active compound can be visualized in Figures 5 (CoMFA) and 6 (CoMSIA), whereas electrostatic contribution can be observed in Figures 7 and 8, respectively. Other parameters of CoMSIA which include hydrophobicity, HBD and HBA are shown in Figures 9-11, respectively.

Molecular docking analysis
Docking study reveals the interaction of the selected ligand with the protein of interest which enables us to understand the model of 3D QSAR. The docking study was performed using          AutoDock tools software and the results are outlined in Table 3. From the docked images of the potent compound (P5-97) as viewed in Figures 12 and 13, it was observed that all the three oxygen atoms were responsible for binding with the amino acids GLY657, ARG660, and ARG655 of the active region. The calculated binding energy was found to be −11.81 kcal/mol. Moreover, from the docking (Figs. 14 and 15) of the least active compound, P5-9 shows that the oxygen atoms, one from sulfono group, the other from benzoyl group, and the R2 methoxy group were responsible for binding the drug with GLY657, GLY659, and LYS763 amino acids in the active region of the protein. The binding energy corresponding to P5-9 was −9.68 kcal/mol.

DISCUSSION
From steric contour maps corresponding to CoMFA and CoMSIA represented in Figures 5 and 6, respectively, the result suggests that, by attaching substituent at green colored region would enhance the activity while yellow contour indicate the decreased activity. Therefore, attachment of various groups at R3 position of the benzoyl group would increase the potency of     compound P5-97. Electrostatic contour maps of respective CoMFA and CoMSIA, red color represent the negative favorable region where as blue color represents the positive favorable regions. Increasing the positive nature at the benzoyl ring and increasing negative charge at R2 position could enhance the activity of the compound. Hydrophobic contour maps from Figure 9 with white and yellow region indicate attachment of hydrophilic and hydrophobic groups, respectively, would result in compounds with increase in the activity. The substitution with hydrophobic group at R3 position and hydrophilic groups at the R5 position of benzoyl and the phenyl ring at the sulfo terminal would significantly enhance the inhibitory action.
From Figure 10 of H-bond donor contour map, substitution with electron withdrawing groups to both the nitrogens of the hydrazine may favor to increase the activity of the molecule. Furthermore, it is clear from the docking study that the key components responsible for binding the ligand (P5-97) within the active region are the oxygen atoms of benzoylsulfonohydrazide. Moreover, from the hydrophobic contour map (Fig. 9) obtained by 3D QSAR study, presence of hydrophilic groups on both the phenyl rings of benzoylsulfonohydrazides would show more binding interactions such that it fits better in the active region of protein. From the obtained results of the study, it provides significant proposition to develop further new compounds retaining the benzoylsulfonohydrazides as the key components.

CONCLUSION
The in silico 3D QSAR study of 61 benzoylsulfonohydrazide analogs of histone acetyltransferase KAT6A inhibitors were carried out. Partial least square analysis was done to evaluate the model developed for CoMFA and CoMSIA. The cross validation (q 2 ) and no validation (r 2 ) for CoMFA were 0.678 and 0.948, while for CoMSIA were 0.719 and 0.953, respectively. Therefore, the obtained results were convincing as they were within the acceptable statistical criterion (q 2 > 0.5). From the cross-validation results, the model of CoMFA and CoMSIA are nearly similar, however, CoMSIA shows to have a better predictive ability. The contour maps that were obtained from study of CoMFA and CoM-SIA of the compound (P5-97), it provides significant insight to design molecules with better inhibitory activity.