Binding Mode Identification for 7-keto-8-Aminopelargonic Acid Synthase (AtKAPAS) Inhibitors

In this study, we determined the 3D structure of Arabidopsis thaliana KAPAS by homology modeling. We then investigated the binding mode of compounds obtained from the in-house library using computational docking methods. From the flexible docking study, we achieved high dock scores for the active compounds denoted in this study as compound 3 and compound 4. Thus, we highlight the flexibility of specific residues, Lys 312 and Phe 172, when used in active sites. result of several flexible docking processes, this study emphasizes the flexibility of several residues. Noticeably higher


Introduction
Agricultural research efforts for discovery of herbicides acting on new target sites are increasing due to demand from farmers and multinational companies in many countries. However, new modes of action have not been turned to commercial success for the past 10 years. We have recently reported 7-keto-8-aminopelargonic acid synthase (KAPAS, also known as 8amino-7-oxononanoate synthase, ANONS) and have suggested the potential KAPAS inhibitor triphenyltin. Research on herbicides has advanced during the past 50 years to the point that herbicides can now protect crops and improve the quality and quantity of agricultural products. However, the successful development of herbicides has decreased recently owing to new environmental regulations and lack of discovery. To overcome this problem, there is an urgent need for new herbicidal targets and new techniques [1].
While the traditional approach to discover lead compounds heavily depends on serendipity given the poor understanding of biological modes of actions, the structure-based approach utilizes the structure of appropriate target proteins, which have well-known binding sites for possible rational designs. The introduction of new herbicides with either a new mode of action or of a novel chemical class has lingered. However, discovering a chemical structure that could enter the pest, be transported within it, inhibit a key target, get away from detoxification, and also be modified to allow it to fulfill increasing regulatory criteria with respect to environmental compatibility has been required. The structure-based approach uses only appropriate target proteins instead of the entire plant for in vivo testing [2]. In order to perform a structurebased assay, it is necessary to determine a potent target and to have a thorough understanding of the mechanism of action of the target.
Several enzymes in plants are known to be essential enzymes, meaning that they are crucial for the plant's survival. Disrupting a single essential enzyme leads to severe disorder of metabolic processes in the plant, ultimately causing a lethal outcome. The enzyme 7-keto-8aminopelargonic acid synthase from Arabidopsis thaliana plants (AtKAPAS), introduced in this research, and is a new potent herbicide target, which is involved in the early steps of the creation of the biotin biosynthesis pathway. AtKAPAS as a pyridoxal 5'-phosphate-dependent enzyme catalyzes the decarboxylative condensation of L-alanine with pimeloyl-CoA in a stereospecific manner to form KAPA, coenzyme A, and carbon dioxide in the first committed step of biotin biosynthesis. Inhibiting AtKAPAS leads to significant changes in the phenotype, such as growth inhibition, severe growth retardation, and the creation of lethal phenotypes [3].
Although the physiological systems of humans and plants are different in various ways, the misuse of agricultural chemicals can be extremely harmful to humans. Therefore, use of herbicides must follow strict toxicity regulations that are in place to prevent harm to humans and other life. As mentioned above, the novel herbicidal target 7-keto-8-aminopelargonic acid synthase functions in the initial steps of the biosynthetic pathways of biotin (vitamin H) in plants and microorganisms. Because biosynthetic steps of biotin exist only in plants, we expect that the inhibition of the potent target AtKAPAS will not affect the human metabolic system [1,3]. A few publications have also reported beneficial effects of AtKAPAS as a potential herbicidal target. Hwang et al. described the possibility of AtKAPAS as a potential herbicide target enzyme and chemical validation of triphenyltin acetate as a lead compound for the AtKAPAS inhibition in vitro and in vivo [1]. They also suggested AtKAPAS can be a useful target for the rational design of inhibitors in the hope of developing new herbicides.
In this chapter, we aim to obtain potential AtKAPAS inhibitors using the knowledge-based computational informatics method in this research. We described the 3D-structure of AtKAPAS via theoretical method and the binding mode for AtKAPAS and its inhibitors obtained from in vitro assay for in-house compounds.

Building a homology model of AtKAPAS
To apply the structure-based drug design (SBDD) method using current knowledge of protein and drug interactions, a three-dimensional protein structure is necessary [4]. Because the known protein crystal structural information of 7-keto-8-aminopelargonic acid synthase from an experiment was absent, a homology model of AtKAPAS was constructed from its amino acid sequence (Table 1). 10 20 To obtain the 3D-structure of AtKAPAS, a hierarchical protein structure modeling approach was adopted based on secondary-structure enhanced profile-profile threading alignment (PPA) and iterative implementation by the threading assembly refinement (TASSER) program [5]. We obtained five candidate models for the three-dimensional structure of AtKAPAS and then performed molecular dynamics simulations with the use of the CHARMM (Chemistry at HARvard Macromolecular Mechanics) force field (version 27.0) [6] with default parameters interfaced with Accelrys Discovery Studio 3.1. To identify binding sites, we collected the crystal structures of homologous proteins with AtKAPAS as templates from the RCSB Protein Data Bank (PDB) based on three different categories: structural similarity, binding site similarity and functional similarity. Ten different protein crystal structures originating from different species but structurally similar to AtKAPAS were selected. The binding site sequence and conformation of the target protein are particularly important here compared to other sites. Therefore, we found 10 different PDB hits which are similar in terms of their binding site to AtKAPAS. A comparison between each PDB hit and the AtKAPAS biding site sequence was done. The results are denoted using root mean square deviation (RMSD) values ranging from 1.83 to 3.51, as shown in Table 2.  Table 2. Sequence similarity and RMSD between 10 reference PDB proteins. To identify the binding mode of AtKAPAS, 10 different enzymes (PDB code; 1FC4 [7], 1DJE [8], 2WKA [9], 2BWO [10], 3KKI [11], 3DXV [12], 3DXW [12], 2OAT [13], 1GBN [14] and 1MLY [15]) with different functions originating from different species were superimposed (Fig. 1). Consequently, we found that several residues were crucial for protein-ligand interactions. His 210 mainly forms a π-π interaction or a π-cation interaction with the ligands, and all of the reference proteins contain histidine residue at a position homologous to the AtKAPAS. Adjacent to His 210, Phe 172 forms a π-π interaction with compounds that have an aromatic ring moiety as well. 50% of proteins have phenyl residue at a similar position; however, its conformation was considerably different. As shown in Fig. 1, conformation of Phe 172 residue of our homology model, shown in yellow, is uniquely folded toward the active site cavity as opposed to the phenyl residues of the reference PDB. To confirm the effect of Phe 172, we undertook a flexible docking evaluation. Lys 312 also plays an important role as a hydrogen bond donor in the active site. All reference enzymes showed the lysine residue at an analogous location which led to π-cation interaction and hydrogen bonding interaction with its ligand. In addition, due to its comparatively free long aliphatic chain, lysine residue was very flexible. Lysine residues superimposed onto the analogous site verified its flexibility. By superimposing the binding sites of other similar proteins, we reached the conclusion that the residues His 210, Phe 172 and Lys 312 play major roles at the binding site. This research will therefore highlight the ligand binding site residues and its flexibility to search for potent hits.

Rigid docking
As a structure-based drug design, we used an automated docking method [16]. To dock the compounds as shown in Table 3 into the protein active site, we used the rigid docking method implemented in Discovery Studio 3.1 (Accelrys, Inc.), which adopts a Monte-Carlo algorithm to generate ligand conformations and docks the generated ligands into the active site using a shape-based filtering method. The rigid docking process consists of two main steps: defining a binding cavity and docking ligands onto the defined cavity [17]. The AtKAPAS protein as obtained from the homology model needs to be prepared before the docking process is initiated. To prepare the protein, the CHARMM force field was assigned and the docking cavity was defined using the advanced define and edit binding site tool module. Ligands also should be prepared with low-energy conformations so as to be docked onto a 'clean' protein.
The protocol 'Generate Conformations' in Discovery Studio 3.1 was used to obtain threedimensional conformations of each ligand. In-house 17 ligands were generated by the protocol 'Generate Conformation' with the conformation method 'BEST', and the CHARMM force field was applied. For an interaction filter, Lys 312 was selected.

Flexible docking
Various docking methods are utilized by researchers.   Table 3. The structures of the 17 compounds used and their biological activities.

In Vitro assay
Pimeloyl CoA was synthesized according to a previously described method [21]. AtKAPAS activity was determined according to the method described previously [22]

Results and discussion
Interesting results were obtained from the docking processes undertaken in this study. The rigid docking output scores of the in-house compounds are shown in   To obtain a better docking result, we used a flexible docking strategy, as mentioned in the experimental section. In the result for flexible docking 1, which stipulated that the Lys 312 residue of the AtKAPAS model is set to move, most of the compounds formed a stable conformation and formed interactions with the AtKAPAS homology model (Table 5).
According to the docking result, the diverse conformation of Lys 312 directly affects the pose of the ligands and the related activity. The Lys 312 residue forms a hydrogen bond or undergoes the π-cation interaction mostly with the oxygen moiety of the ligand by flexibly moving through the protein cavity (Fig. 2). Interestingly, we obtained relatively high dock scores for compound 1, compound 9, and compound 10. This was unobtainable with the rigid docking process. The correlation coefficient between pIC50 and the dock score for the flexible docking 1 set was 0.72, as shown in Fig. 3   By flexibly moving the residue Phe 172 while the flexible docking 2 process was underway, a better result than the flexible docking 1 process was obtained ( The correlation coefficient between pIC50 and the dock score for the flexible docking 2 set was 0.85, as shown in Fig. 4. The two specific AtKAPAS residues, Lys 132 and Phe 172, can be moved mutually for the flexible docking 3 process. The result shows the clear discrepancy between an active compound and an inactive compound ( Table 7). The flexible docking 3 results showed a higher correlation between pIC50 and the dock score compared to flexible docking 2, whereas it had a slightly lower R 2 value than the flexible docking 2 result (Fig. 5). Because the two residues were set to move at the same time, more diverse results could be obtained. As a result of several flexible docking processes, this study emphasizes the flexibility of several residues. Noticeably higher dock scores were obtained from flexible docking as compared to rigid docking. The rigid Lys 312 residue mostly tends to form π-cation interaction with the aromatic moiety of compounds, allowing a certain amount of space for compounds during the rigid docking process. However, the flexible Lys 312 forms either the π-cation interaction or the hydrogen bonding interaction with hydrogen bond acceptors. The flexibility of the residue, including Lys 312 and Phe 172, allows more space for compounds, thus offering a better docking pose and dock scores.

Summary
In this study, we determined the 3D-structure of AtKAPAS by homology modeling. We then investigated the binding mode of our in-house library using computational docking methods.
In the rigid docking of the in-house compounds as shown in Table 3, the most active compounds, in this case compound 3 and compound 4, obtained high dock scores of 104.21 and 105.47, respectively. However, some active compounds showed rather low dock scores. To obtain a better docking result, we used a flexible docking strategy. From the flexible docking study, we achieved high dock scores and stable binding conformations for the active compounds denoted in this study. Thus, we highlight the flexibility of specific residues, Lys 312 and Phe 172, when used in active sites. Furthermore, we are going to optimize compound 3 and compound 4 using this homology model for AtKAPAS. Following the initial identification of a lead chemical, intensive research and testing needs to be followed to optimize its structure to understand its action and provide data on its environmental compatibility.