Designing novel inhibitors against Mycobacterium tuberculosis FadA5 (acetyl-CoA acetyltransferase) by virtual screening of known anti-tuberculosis (bioactive) compounds

By-products of fatty acid degradation are extensively utilized by Mycobacterium tuberculosis (Mtb) for lipid synthesis and energy production during the infection phase. Cholesterol from host is scavenged by Mtb to fulfill its metabolic requirements, evade host immunity and invade macrophages. Blocking cholesterol catabolic pathways leads to bacteriostasis. FadA5 (Acetyl-CoA acetyltransferase), a thiolase encoded by fadA5 (Rv3546) gene in Mtb, plays a crucial role in cholesterol aliphatic chain degradation. Hence, FadA5 is a potential target for designing antitubercular inhibitors. In this study, 60,284 anti-tuberculosis (bioactive) compounds from ChEMBL database and analogous library from ZINC database of commercially available compounds have been screened against FadA5 active site to identify compounds having inhibitory potential against both the apo (state I) and the intermediate (state II) states of FadA5. Altogether, this study reports 7 potential inhibitors against two functional states of FadA5, which can be further taken for invitro studies.


Background:
Hypercholesterolemia is known to impair immunity against TB and contributes towards the infection development [1]. Host lipids have been shown to play an important role in Mtb survival against the host immunity [2]. Mtb primarily uses fatty acids as their main source of carbon during infection [3] and requires cholesterol for macrophage invasion [4]. Mtb does not synthesize cholesterol instead utilizes host cholesterol to accomplish its metabolism [5]. They penetrate the macrophage membrane where cholesterol-rich microdomains are present [6]. By-products of cholesterol catabolism are used by the bacterium for lipid synthesis and energy production [7]. Deletion of mce4 transporter in Mtb blocked cholesterol import thus resulting in reduced infection both in the activated and the mouse model [7]. Blocking cholesterol catabolic pathways at certain steps have been shown to cause bacteriostasis and cell deaths in Mtb [8]. These literature evidences justify that cholesterol and cholesterol catabolic pathway is crucial for Mtb survival in macrophages and can be targeted to develop new antitubercular drugs.
Cholesterol catabolism in Mtb starts with the degradation of aliphatic side chains of cholesterol followed by the sequential degradation of rings [5]. Many enzymes are involved in this process. FadA5 is one such enzyme, encoded by fadA5 (Rv3546) that catalyzes the thiolysis of keto CoA-esters formed during beta-oxidation of the cholesterol side chain [10] and produces androsterone metabolites, which contribute towards Mtb persistence [9]. The complete degradation process follows a pingpong mechanism where an acylated-cysteine intermediate is formed. This active site cysteine is crucial for catalysis as its mutation leads to attenuation of infection in the mouse model  In this study, we have screened known antituberculosis bioactive chemical library (ChEMBL database) and an analogous library (ZINC database) to identify novel active inhibitors against FadA5. The intermediate form of FadA5 (state II) has acetylated serine (OAS93), which is larger in size as compared to C93 present in FadA5 state I. This is why, firstly, CoA bound intermediate form (OAS93-CoA; acetylated-serine-CoA) was used to identify top hits which were then cross-docked against apo-form to identify potential inhibitors against both the functional states of the enzyme. We report 7 compounds, which have good predicted inhibitory potential towards state I and state II of FadA5. CID: Compound identification number used in this paper. HB: hydrogen bond forming residues. NBI: Non-bonded interactions.

Methodology:
The workflow used in this work is shown in supplementary figure 1.

Target protein structure
FadA5 belongs to thiolase family of enzyme, involved in the catabolism of fatty acids. The crystal structure of the enzyme was solved with acetyl-CoA and CoA. The active site of the enzyme consists of amino acid residues C93, R221, H347, A242, G243, Q177, S246, T224 and G227. In addition to active site residues, CoA, which is a substrate to the enzyme, also makes direct interactions with Q151, T223, and S246 and water-mediated interaction with Q247 [10]. In this study, two structures, one representing the apo form (PDB ID: 4UBW) and one representing the modified C93S variant of the 3-ketoacyl-CoA thiolase FadA5 (PDB ID: 4UBU) were selected as target, which will be referred as FadA5 state I and FadA5 state II, respectively in rest of the manuscript. FadA5 state II was obtained by modifying C93to OAS93 (acetylated-serine). Both the states of the protein were prepared using protein prep wizard of Schrödinger suite [11].
While preparation of the receptor, all water molecules were deleted and missing hydrogen atoms were added. Missing side chain residues were modeled using prime tool available in Schrödinger suite [12]. It was followed by restrained energy minimization by fixing the main chain to remove steric clashes between side chains. While preparation of the receptor of FadA5 state II, CoA was retained to define the active site of the enzyme. The receptor grid was generated using Glide module [13, 14] of Schrödinger suite. The receptor grid was generated using the center of CoA as grid center, and the grid boundary was defined in such a way that the minimum distance between any atom of ligand and grid boundary is at least 5Å. The active site of the apo form of FadA5 (state I) was defined by the amino acid residues, which are in close proximity of CoA in state II of FadA5, which was later, used for receptor grid generation. were added in calculation of scores in each case. In the last stage, screening of analogs of the top-ranked ChEMBL compounds from ZINC database was performed. The analogous library was screened directly in XP mode. The binding affinities of the obtained best binding pose of top ranking compounds were predicted using X-score (v1.2.1). X-score estimates the binding affinity of a given binding mode of a ligand within the binding pocket using empirical scoring function. Its scoring function estimates the binding affinity on the basis of four energy terms viz Van der Waals interaction energy, hydrogen bonding energy, deformation penalty and hydrophobic effect.

Screening of top ranked compounds of FadA5 state II against FadA5 state I:
The top-ranked compounds obtained after screening of ChEMBL and analogues library against FadA5 state II were re-docked against FadA5 state I in XP mode to identify compounds that bind to both the states of the enzyme. The compounds were then scored and evaluated.

Results and Discussion: Screening of anti-tuberculosis (bioactive) compounds against FadA5 State II:
After screening ChEMBL compounds against FadA5 state II, the compounds were ranked on the basis of G-score. The G-Score of the top 10 compounds ranged from -12.252 kcal/mol (C1) to -10 34 kcal/mol (C10) (supplementary table 1). The predicted binding affinity of each compound using X-score lie in the range of -10.67 kcal/mol (C8) to -8.46 kcal/mol (C3) (supplementary table 1). To understand the interactions of top-ranked compounds, the 2D interaction profiles were generated using LigPlus, and different types of interactions between ligands and protein were analyzed.
As mentioned in supplementary table 1, compound C8 possess the highest binding affinity towards FadA5 state II among all the top 10 compounds when examined in terms of X-score. Compounds C1, C2, C4, C5, C6, C7 and C10 were found to have binding affinities close to -9 kcal/mol calculated using X-score. Compound C3 although raked better than C4, C5, C6, C7, C8, C9 and C10in terms of G-score; the calculated binding affinity using X-score suggests that it has lower binding affinity compared to other compounds enlisted in top 10 list. All the top 10 listed compounds have more or less similar binding affinity towardsFadA5 state II as they share similar structure scaffold.
Combining all these properties, we can say that the compound C1 (2D interaction profile of C1 is shown in supplementary figure 2) and C8 (2D interaction profile is shown in supplementary figure 3) have the highest inhibitory potential in terms of X-score, G-score, and the number of hydrogen bonds they share with binding pocket residues (supplementary figure  2).

Molecular docking of analogues chemical library against FadA5 State II:
The docking studies for analogues compound library obtained from ZINC database was performed and the compounds were ranked on the basis of G-score (supplementary table 1). The Gscore and X-score of the top ten compounds (Z1-Z10) vary between -12.626 Kcal/mol (Z1) to -11.582 kcal/mol (Z10) and -9.81 kcal/mol (Z2 and Z4) to -8.17 kcal/mol (Z10) (supplementary table 1), respectively. Compounds Z2 and Z4 were found to have the highest affinity, i.e., -9.81 kcal/mol; and compound Z10 was found to have the least affinity, i.e., -8.17 kcal/mol towards FadA5 state II as per X-score. 2D interaction profile of the complexes reveal that compounds Z2 and Z5 form the highest number of hydrogen bond interaction which involves amino acid residues Lys16, Arg 17, Gln151, Gln177, Thr223, Ser246 Ile248, Ala317, and Ile343, out of which the hydrogen bonds with all residues except Ile248 and Ser246 are conserved.
The obtained results imply that compounds Z2 (supplementary figure 4) and Z5 (supplementary figure 3) have good affinity towards FadA5 state II. In case of FadA5 state I, the G-score ranges between -11.065 kcal/mol to -9.906 kcal/mol being lowest for Z1-state I and highest for C1-state I. The binding affinities predicted using Xscore ranges from -9.58 kcal/mol to -8.28 kcal/mol ( Table 1). Z1state I is shown to have the highest affinity towards FadA5 state I. The 2D interaction profile of Z1-state I with both the states of FadA5 is shown in Figure 1. For other compounds, the ranking order has changed, but have almost similar type of binding affinities with the FadA5 state I as have been observed for FadA5 state II. Number of hydrogen bond interaction ranges between 4 to 6, numbers of lipophilic interactions varies between 7-13 and number of non-bonded interactions varies between 42 and 63 ( Table 1). These results imply that the compounds enlisted in

Conclusions:
Docking and post-docking analysis suggest that the top-ranked compounds reported in this study have similar type of interaction profile and affinity towards both the states of FadA5, except compound C3 and C9 which have relatively less affinity towards the enzyme. Altogether, this study reports 7 potential inhibitors against both the functional states of FadA5, which can be taken further for in-vitro studies.  CID: Compound identification number used in this paper.