Comparative structural and mechanistic studies of 4-hydroxy-tetrahydrodipicolinate reductases from Mycobacterium tuberculosis and Vibrio vulnificus

Abstract

Background

The products of the lysine biosynthesis pathway, meso-diaminopimelate and lysine, are essential for bacterial survival. This paper focuses on the structural and mechanistic characterization of 4-hydroxy-tetrahydrodipicolinate reductase (DapB), which is one of enzymes from the lysine biosynthesis pathway. DapB catalyzes the conversion of (2S, 4S)-4-hydroxy-2,3,4,5-tetrahydrodipicolinate (HTPA) to 2,3,4,5-tetrahydrodipicolinate in an NADH/NADPH dependent reaction. Genes coding for DapBs were identified as essential for many pathogenic bacteria, and therefore DapB is an interesting new target for the development of antibiotics.

Methods

We have combined experimental and computational approaches to provide novel insights into mechanism of the DapB catalyzed reaction.

Results

Structures of DapBs originating from Mycobacterium tuberculosis and Vibrio vulnificus in complexes with NAD⁺, NADP⁺, as well as with inhibitors, were determined and described. The structures determined by us, as well as currently available structures of DapBs from other bacterial species, were compared and used to elucidate a mechanism of reaction catalyzed by this group of enzymes. Several different computational methods were used to provide a detailed description of a plausible reaction mechanism.

Conclusions

This is the first report presenting the detailed mechanism of reaction catalyzes by DapB.

General Significance

Structural data in combination with information on the reaction mechanism provide a background for development of DapB inhibitors, including transition-state analogues.

Introduction

Studies on the antibiotic-pathogen interactions at the molecular level show that the majority of antibiotics target cell wall synthesis, protein synthesis, and DNA replication and repair [1]. Enzymes containing trans peptidase and trans glycosylase domains, ribosomal machinery and DNA gyrase act as sites of action for antibiotics within these specific pathways [2]. Unfortunately, pathogenic bacteria have developed many defensive mechanisms to evade the antibiotics, including production of β-lactamases, enzymatic modification of drug targets, and development of drug efflux systems [1]. Researchers have identified several approaches to combat rising antibiotic resistance. The most direct involve the development of new classes of antibiotics and compounds that disrupt drug resistance mechanisms [2]. In parallel, several new bacterial components are also being studied as potential novel drug targets. These include bacterial proteases [3], two component signal transduction systems [4], riboswitches [5] and several enzymes involved in fatty acid, nucleotide and amino acid biosynthesis [6]. Disrupting enzymes involved in central biosynthetic pathways will greatly reduce the survival and pathogenicity of bacteria due to the lack of alternative biosynthetic cycles. These will also have an additional advantage of not having homologous enzymes in humans [6]. One such pathway in bacteria is the amino acid lysine or meso-diaminopimelate biosynthetic pathway.

The products of the pathway, meso-diaminopimelate and lysine, are essential for bacterial survival. The meso-diaminopimelate and lysine (for some bacterial species) are essential for bacterial cell shape and rigidity due their role in the covalent linkage of peptidoglycan in the cell wall [7]. Lysine is also an essential amino acid for protein synthesis. Several enzymes from this pathway are currently under study as potential drug targets including aspartate semi aldehyde dehydrogenase (ASADH) [8], 4-hydroxy-tetrahydrodipicolinate synthase (DapA) [9], 4-hydroxy-tetrahydrodipicolinate reductase (DapB) [10], and succinyl diaminopimilate desuccinylase (DapE) [11]. According to the Database of Essential Genes, the genes that code for these enzymes, including DapB, are essential in many pathogenic bacteria (Table S1) [12]. This paper focuses on structural and mechanistic studies of DapB. Functionally, recent work demonstrates that this enzyme catalyzes conversion of (2S, 4S)-4-hydroxy-2,3,4,5-tetrahydrodipicolinate (HTPA) to 2,3,4,5-tetrahydrodipicolinate in an NADH/NADPH dependent reaction (Figure 1) and dihydrodipicolinate (DHDP) is not the substrate as previously thought [13].

Figure 1.

Reaction catalyzed by DapB.

The structure for DapB has been determined from various bacterial species including Escherichia coli [14], Bartonella henselae [15], Staphylococcus aureus [16], Neisseria gonorrhoeae [10], Thermotoga maritima [17], Coxiella burnetii, and Pseudomonas aeruginosa among others. All of these structures indicate that bacterial DapB is a homotetrameric enzyme. Each monomer can be divided into two functional domains. The first domain is responsible for oligomerization and the second is the NADH/NADPH binding domain (Figure 2). An interesting feature observed in DapB enzymes that drives the enzyme reaction is the change in conformation upon ligand binding. Binding of pyridine nucleotide or the cofactor induces a change primarily in nucleotide binding domain resulting into partial closure of each monomer (Figure S1). This change is usually less than 10°. However, binding of both the cofactor and substrate/substrate analogs induces a major conformational change (>30°) resulting into complete closure of the monomer [18]. Even though most DapB homologs can bind both NADH and NADPH, they usually prefer one of the nucleotides. Currently, the reasons for this nucleotide preference remain elusive. In addition, the details of the mechanism for the DapB catalyzed reaction are not known. This manuscript focuses on DapBs homologs from Mycobacterium tuberculosis and Vibrio vulnificus. M. tuberculosis is the causative agent of tuberculosis. It primarily infects the lungs and has been described in a recent report by the World Health Organization as a leading cause of death by a single infectious agent. The incidence of drug resistance reported in this pathogen is extremely high, with over 160 thousand cases documented globally in 2017 [19]. Drug resistant M. tuberculosis has been classified as a serious threat by the Centers for Disease Control and Prevention (USA) [20]. V. vulnificus on the other hand is an opportunistic human pathogen. The natural habitat of this bacterium is coastal or estuarine environments and it is known to colonize shrimp, fish, oysters, and clams [21]. It enters the human body via the consumption of raw seafood. V. vulnificus infection is characterized by gastroenteritis, wound infections, septicemia, and has a very high mortality rate amongst infected patients [22]. This high mortality rate may be attributed to the drug resistance for many antibiotics including penicillin, ampicillin, and tetracycline seen in the bacterium [23]. Given the high drug resistance observed in both bacteria, research needs to be directed towards identifying new drug targets and inhibitors in order to curb these infections. The manuscript describes the first crystal structures of DapB from V. vulnificus (VvDapB) and several new crystal structures of M. tuberculosis DapB (MtDapB) in apo and ligand bound forms. We have combined the structural analysis and various computational approaches to provide novel insights into mechanism of the DapB catalyzed reaction mechanism.

Figure 2.

(A) Tetrameric assembly of DapB from V. vulnificus (PDB ID: 5TEN). One of the protein chains in the tetramer is in partially open conformation. (B) Close-up of the chain in in the partially open conformation. NAD⁺ is shown in orange and 2,5-furan dicarboxylic acid in magenta.

Results and Discussion

Screening and identification of potential inhibitors

Differential Scanning Fluorimetry (DSF) assays were performed to identify potential inhibitory compounds for DapB enzymes. The effect of ligands structurally similar to 2,6-pyridine dicarboxylic acid (2,6-PDC), a known inhibitor of DapB enzymes, was evaluated. All the experiments were done using VvDapB as a representative DapB enzyme. Since DapBs share a very high sequence similarity, the results obtained through these experiments may be applicable for other DapB homologs as well [24]. The ligands studied and the experimental results are shown in Figure S2. The melting temperature (Tm) of VvDapB was found to be 60°C. Addition of NADH caused a slight increase of 2°C in the Tm. As expected, addition of 2,6-PDC increased the thermal stability of VvDapB with a rise in Tm to 76°C with NADH and 70°C without NADH. Another ligand, 2,5-furan dicarboxylic acid (2,5-FDC) was able to increase the melting point to 68°C with NADH but no increase was seen without NADH. The crystal structure of VvDapB bound to NADH and 2,5-FDC has been described below. Chelidonic acid and nitro-5-furoic acid were also found to increase Tm for VvDapB.

Overall structure of enzymes

Several different structures were obtained for each recombinant protein. The biological assembly for both MtDapB and VvDapB was determined to be tetrameric (Figure 2), which is characteristic of all bacterial DapBs [10, 14]. Each monomer or chain of VvDapB is composed of 269 residues and can be divided into the coenzyme or pyridine nucleotide binding domain (residues 1–123 and 236–269) and oligomerization domain (residues 124–235). VvDapB was crystallized in two crystal forms with NAD⁺ and 2,6-PDC, a substrate analog and inhibitor. Here we would like to note, that despite the fact that NADH was used for crystallization, in all cases we observed NAD⁺ in the crystal structures. In the first crystal form (PDB ID: 5US6), three chains in the tetramer were bound to NAD⁺ and 2,6-PDC (closed conformation) and one chain was in open conformation whereas in the second form (PDB ID: 5TEM), all the chains were in closed conformation. We also obtained two crystal structures of VvDapB with NAD⁺ and a new substrate analog and potential inhibitor 2,5-FDC. In one structure, three chains were in the closed conformation with both the ligands bound, and one chain was in the partially closed conformation (PDB ID: 5TEJ), as it had only NAD⁺ bound. The second structure with 2,5-FDC (PDB ID: 5TEN) has two tetramers in the asymmetric unit. Interestingly, one tetramer is in completely closed conformation, but the second tetramer has 2,5-FDC and NAD⁺ in three chains, while the fourth chain has NAD⁺ bound only. There are crystal structures of other DapB homologs with ligands partly bound. For example, the E. coli DapB structure (PDB ID: 1ARZ) has three chains of the tetramer in the completely closed conformation and one chain in the open conformation, which is similar to one of the VvDapB structures. On the other hand, B. henselae DapB was crystallized with cofactor bound to two chains of the tetramer (PDB ID: 3IJP) [15].

The MtDapB enzyme displays only ~30% sequence identity to VvDapB (Supplementary Figure S3) and has a shorter protein chain with 245 amino acids present in a single peptide chain. Residues 1–108 and 216–245 form the coenzyme binding domain whereas residues 107–215 form the oligomerization domain. Crystallographic studies have been done previously on MtDapB [25, 26]. We obtained MtDapB in new crystal forms with better resolution. Here we report a 1.5 Å resolution structure of MtDapB co-crystallized with NADP⁺ and 2,6-PDC with all chains in closed conformation (PDB ID: 5TJZ). Currently, this is the highest resolution structure reported amongst DapB enzymes. Along with this structure, we also obtained better resolution structure of the apo enzyme (PDB ID: 5TEK) at 2.0 Å. Comparison of experimental data from previously described structures with ours reveals that our recombinant protein had an additional N-terminal hexa-histidine tag. Details on crystallization conditions, data collection, processing and refinement statistics for all the structures are shown in Supplementary Information (Tables S2–S4).

Interactions with pyridine nucleotides in solution

We have used Isothermal Titration Calorimetry (ITC) to study interactions of MtDapB or VvDapB with NADH in solution (Figure S4). The ITC studies indicated that MtDapB binds NADH with a K_d of 3.6 ± 0.4 μM whereas VvDapB has a K_d of 3.9 ± 1.2 μM for NADH binding. Both the binding interactions are energetically favorable with ΔG value of −30.6 ± 0.8 kJ/mol (ΔH = −40 ± 3 kJ/mol, ΔS= −30.6 ± 12.9 J/mol·K) for VvDapB and ΔG value of −31.0 kJ/mol (ΔH= −46.3 ± 2.5 kJ/mol, ΔS= −51.2 ± 9.1 J/mol·K) for MtDapB. Similar to other bacterial DapB enzymes, NADH binds with a stoichiometry of one. Binding affinities of DapB enzymes from different bacterial species with NADH are listed in Table 1. Various experiments were also conducted to study the binding of DapB enzymes with NADPH. All the experiments indicated that NADPH binds to both MtDapB and VvDapB (data not shown). Despite several modifications in the experimental design, we could not achieve complete saturation under the conditions tested in order to calculate the thermodynamic parameters for NADPH binding.

Table 1:

Cofactor binding affinities for DapB enzymes. Dissociation constants (K_d) for NADH complexes with DapBs from different bacteria determined by using Isothermal Titration Calorimetry (ITC).

Organism	NADH (μM)	Reference
V. vulnificus (VvDapB)	3.9 ± 1.2	This work
M. tuberculosis (MtDapB)	3.6 ± 0.4	This work
N. gonorrhoeae	2.5 ± 0.5	[10]
E. coli	0.46 ± 0.04	[59]
S. aureus	9.2 ± 0.2	[16]
Methicillin resistant S. aureus	29 ± 2	[60]
Paenisporosarcina sp.	No binding	[18]

Active site

The substrate and inhibitor (e.g. 2,6-PDC) binding site is located near a highly conserved region in the oligomerization domain of the enzyme (Figure 3) [14]. This region is composed of amino acids ¹⁵³EAHHRHKVDAPSGTA¹⁶⁷ in VvDapB and ¹³⁰ELHHPHKADAPSGTA¹⁴⁴ in MtDapB (residues shown in bold are conserved among all bacterial DapBs that have their structures determined; Supplementary Information Figure S3). Apart from the conserved residues, the inhibitor site in VvDapB is also surrounded by Val213 and His215 from the oligomerization domain and several residues from the nucleotide binding domain (Pro123, Asn124, Thr100, Arg236 and Phe239). Similarly, the MtDapB substrate binding pocket also has additional residues from the nucleotide binding domain (Thr77, Pro103, Asn104, Phe217) and oligomerization domain (Ala192). We determined several structures of VvDapB (PDB IDs: 5US6, 5TEM) and MtDapB (PDB IDs: 5TJY, 5TJZ) bound to 2,6-PDC, which is a known substrate analog and inhibitor of DapB. The positioning of 2,6-PDC in all the MtDapB structures is very similar, and the 2,6-PDC forms several hydrogen bonding interactions with His133, Lys136, Gly142 and Thr143 (Figure 4). Analysis of VvDapB structures bound to 2,6-PDC reveals that the same amino acids His156, Lys159, Gly165 and Thr166 interact with the ligand. In E. coli DapB, an additional interaction with Ala127 has been reported [14].

Figure 3.

Structure of MtDapB single chain in closed (A) and open conformation (B). Residues that are conserved in DapBs that have their structures determined are shown as red sticks. Residue numbering in black corresponds to MtDapB and in green to VvDapB. 2,6-PDC is shown in magenta. In the case of NADP⁺ carbon atoms are shown in orange, oxygen atoms in red and nitrogen atoms in blue. The high-resolution structure of MtDapB (PDB ID: 5TJZ) was used to prepare figure (A), and structure of apo-MtDapB (PDB ID: 1YL5) was used to prepare figure (B).

Figure 4.

Binding of 2,6-PDC (magenta) and NADP⁺ in the active site of MtDapB (PDB ID: 5TJZ). Residue numbering in black corresponds to MtDapB and in green to VvDapB. Only residues that are conserved in DapBs are shown. Hydrogen bonds are marked using dashed lines.

Our DSF and structural studies identified a novel small compound, 2,5-furan dicarboxylic acid (2,5-FDC) that binds to DapB enzymes. Further structure details show that 2,5-FDC binds in the active site of the enzyme. Moreover, VvDapB interacts very similarly to 2,6-PDC with this ligand, indicating that 2,5-FDC is a substrate analog and potential inhibitor (Figure 5).

Figure 5.

Binding of 2,6-PDC (A) and 2,5-FDC (B) in the active site of VvDapB. Residue numbering in black corresponds to MtDapB and in green to VvDapB. Only residues that are forming hydrogen bonds with the inhibitors are shown. Hydrogen bonds are marked using dashed lines. Distances are reported in Ångstroms.

Conformational changes in enzymes

Superposition of apo and ligand bound structures of DapBs revealed that these enzymes undergo major conformational changes upon ligand binding (Figure 3). The conformational changes in VvDapB and MtDapB were studied using the DynDom server [27, 28]. Details of the angles of rotation, as well as exemplary figure showing domains, hinge regions and the rotation axis are provided in the Supplementary Information section (Tables S5–S7 and Figure S1). Analysis shows that binding of both cofactor and substrate/inhibitor causes a significant conformation change (> 30°, as defined by DynDom) in the protein however binding of just the cofactor induces a relatively minor conformational change (<10°). Chains in the DapB tetramer can be divided into two domains, namely the fixed domain and the moving domain. When a chain undergoes a conformational change, the moving domain rotates by a specific angle through the bending region (Figure S1). DynDom analysis of apo MtDapB (PDB ID: 5TEK chain A) and NAD⁺ bound MtDapB (PDB ID: 5UGV chain A) structures reveal that the fixed domain is composed of residues Arg3-Pro104 and Phe218-Leu243, with residues Asp105-Ser217 forming the moving domain. The bending region is formed by residues Pro104-Phe106 and Thr216-Val219. On the other hand, analysis of apo-VvDapB (PDB ID: 5US6 chain A) and ligand bound VvDapB (PDB ID: 5US6 chain D) chains indicate that the bending region is composed of Asn124-Gly128 and Thr234-Thr238. Arg3-Tyr125 and Met237-Gly265 form the fixed domain whereas Ser126-Arg236 form the moving domain. Out of all the bending region residues, Asn104 (124 in VvDapB) is highly conserved through bacterial DapB homologs (Supplementary Figure S3, Figure 3). Through a hydrogen bond, Asn104 interacts with Thr143 that is directly binding to one of the carboxyl groups of HTPA or 2,6-PDC (Figure 4). Moreover, the identified bending regions are in vicinity of highly conserved glycine residues Gly108 and Gly220 of MtDapB or Gly128 and Gly242 of VvDapB. We suggest that these glycine residues are highly important for MtDapB and VvDapB conformational flexibility and therefore for catalytic activity. In contrast, the conserved Gly7, Gly10 and Gly142 (Gly8, 11 and 165 in VvDapB) are directly involved in binding of substrates (Figures 3–5). The analysis of various DapB conformations also highlights importance of Asp138 (Asp161 in VvDapB), as this amino acid not only positions Lys136 (Lys159 in VvDapB) for substrate binding, but after the closure forms a hydrogen bond with Thr77 (Thr100), which is critical for proper positioning of NADH/NADPH (Figure 4). There are several structures of DapBs with just the cofactor bound to the enzyme. However, insufficient structural data is available identifying binding of substrate or substrate analogs in absence of the cofactor. There are studies in literature that show that the substrate and substrate analogs can bind the enzyme independently of the cofactor [29]. In support of that we also recently characterized the thermodynamics of binding of 2,6-PDC with DapB from Neisseria gonorrhoeae [10]. The conformational changes described here may have major implications in targeting DapB in a drug discovery process, as one may target with inhibitor various conformational states of the enzyme, which has been shown for other proteins [30, 31].

Catalytic Mechanism for DapB

Binding orientation of substrates in the enzyme active site: MM-PBSA calculation

The structural studies strongly suggest that closing of DapB is necessary to bring both substrates in vicinity and initiate the chemical change. Therefore, for all computational calculations we used our VvDapB structure as a model representing the closed conformation of the protein, hence the numbering of amino acids in the following paragraph will correspond to this enzyme. Since both the substrate HTPA and intermediate DHDP have asymmetric structures, which is in contrast with 2,6-PDC and 2,5-FDC that we co-crystallized with VvDapB, the first step towards understanding the catalytic mechanism of DapB is to probe the binding orientation of substrates in the enzyme active site. First, we generated the enzyme-substrate complexes DHDP_M1, DHDP_M2, HTPA_M1 and HTPA_M2 based on two binding orientations (M1 and M2 (Figure S5)) of substrate in the enzyme active site. The substrates were placed in the active site at the position that was occupied by the inhibitor according to the determined by us VvDapB crystal structure. After addition of hydrogen atoms and charge neutralization, all the complexes were subjected to explicit water (TIP3P) molecular dynamics (MD) simulations using AMBER16, after prior minimization to remove close contacts between solute and solvent, to get enough sampling of stable enzyme-substrate complex for further calculation of binding free energies (generation of the enzyme-substrate complexes and the MD procedure are described in the Methods section). The binding free energies (ΔG) of enzyme-substrate complexes DHDP_M1, DHDP_M2, HTPA_M1 and HTPA_M2 (Figure S5), calculated by MM-GBSA (std. deviation) are: −39.5 (5.5), −33.9 (6.3), −33.9 (6.0) and −32.6 (6.7) kcal/mol and by MM-PBSA (std. deviation) are: −14.6(8.4), −12.7 (12.3), −22.1 (10.1) and −11 (9.6), respectively. According to ΔG values, M1 orientation is comparably more stable than M2 for both the substrate HTPA and intermediate DHDP. In the case of product THDP, it was also observed that ΔG for the M1 orientation is more favorable, i.e. −67.2 kcal/mol while for M2 it is −41.4 kcal/mol.

His155 protonation: hydrogen bond analysis and MM-PBSA calculation

The MM-PBSA analysis results presented above show that for both HTPA and DHDP enzyme-substrate complexes the M1 orientation is energetically more stable compared to the M2 orientation. Thus, next we proceeded to check protonation state of active site residue His155 needed for enzyme catalysis. Structural studies clearly indicate that in the closed DapB conformation this residue is in vicinity of inhibitors (2,6-PDC or 2,5-FDC; Figure 4), but in contrast with His156 does not form a H-bond with the inhibitor/substrate (Figure 5). Again, here we used the same procedure of MD simulation and MM-PBSA analysis as we did for determination of substrate binding orientation in the active site. According to protonation state of the active site residue His155, i.e. doubly protonated positively-charged HIP, and singly protonated (on δ-N) charge-neutral HID, we generated the enzyme-substrate complexes DHDP_HIP, DHDP_HID, HTPA_HIP, and HTPA_HID. After MD simulations of all these enzyme-substrate complexes, analyses of hydrogen bond interactions between enzyme and substrate were performed and compared with MD simulated DapB-2,6-PDC (Table S8). Also, binding free energies were calculated for all the complexes (Table S9). In the case of DapB-2,6-PDC complex, 2,6-PDC makes interactions with active site residues His156, Lys159, Gly165, and Thr166 both in the crystal structure (Figure 5) as well as during MD simulation. All these interactions, except interaction with Gly165, are also observed in all the enzyme-substrate complexes during MD simulations. Due to double protonation of His155 in DHDP_HIP and HTPA_HIP complexes, this residue interacts with carboxyl groups of DHDP and HTPA. In the case of DHDP_HIP, DHDP is stabilized by additional hydrogen bonds with active site residues Ser164 and His216, while these interactions were not observed in DapB-2,6-PDC as well as DHDP_HID, HTPA_HIP, and HTPA_HID. It is worth to note, that Ser164 is conserved in all DapBs that have their structures determined (Figure S3). His216 is not completely conserved, and in MtDapB is replaced with Gln194. However, it is likely that the amide group of MtDapB Gln194 may still form a hydrogen bond with the substrate. According to binding free energies, DapB-2,6-PDC, DHDP_HIP, and HTPA_HIP are more stable than DHDP_HID and HTPA_HID. Differences in the energetic stability of HIP and HID complexes agree with the differences in number of protein-substrate interactions as well as their resident time though out the MD simulation (Table S8). DHDP_HID and HTPA_HID have a comparatively smaller number of enzyme-substrate interactions as well as their resident time is shorter, so they are comparatively less stable.

Reduction of DHDP to THDP

Since calculations suggest the enzyme-substrate complex with doubly protonated His155 (HIP155), DHDP_HIP is energetically more stable compared to DHDP_HID, it was considered for further study of enzymatic reduction (Figure 6). Thus, to calculate the energy barrier for the reduction of DHDP to THDP, we considered a reaction scheme-1 (Figure 7) in which a carbanion intermediate, derived from DHDP after accepting H⁻ from NADH, undergoes resonance stabilization. Then subsequent transfer of H⁺ from Nε2 atom of His155 (doubly protonated) to C5 of the intermediate completes C=C bond reduction to C-C. For the first step of the reduction reaction, energy barrier calculated by DFT method is 12.9 kcal/mol (Figure 7C) and for TS1a the distance between C4 of DHDP and hydrogen atom attached to C4 of NADH (to be transferred to DHDP as H⁻) decreased from 2.61 Å in the optimized E-S structure (Figure 6A) to 1.37 Å. Resulting intermediate (Int1) has energy of −24.5 kcal/mol (Figure 7C). It was observed that this intermediate is stabilized by accepting at the imine nitrogen H⁺ derived from Lys159 and delivered through the sugar moiety of NADH. Lysine residue at position 159 (numbering according to VvDapB) is completely conserved among all the bacterial DapBs (Figure S3) [32] and the role of this residue in enzyme catalysis has already been suggested by mutational analysis of E. coli DapB [14]. Further, energy barrier for transfer of H⁺ from His155 to C5 of Int1, connected with TS2, was calculated to be 16.1 kcal/mol (Figure 7C) and the distance between hydrogen of His155 and C5 of DHDP reduced to 1.4 Å from 4.49 Å in the optimized E-S structure (Figure 6A). Finally, energy for the resultant E-P (product is protonated, H⁺ taken from Lys159 for stabilization of intermediate state) structure was calculated to be −22.0 kcal/mol (Figure 7C).

Figure 6.

Optimized active site model of A) DHDP_HIP, E-S complex, and B) HTPA_HIP, E-S complex. Active site key residues are shown in grey. C-alpha atoms fixed during geometry optimization are in green color. The carbon atoms of DHDP and HTPA are in magenta and of NADH are in orange. Distances between C4 of DHDP and hydrogen attached to C4 of NADH, hydrogen attached to Nε2 of His155 and C5 of DHDP, hydrogen attached to C5 of HTPA and Nε2 of His155, and oxygen of HTPA OH group and Nε2 of His155, respectively are shown in black dashed line in the optimized active site model. Water molecules involved in the HTPA dehydration reaction are labeled as ‘W’.

Figure 7.

DHDP to THDP reduction reaction. A) Scheme-1, B) Scheme-2, and C) Energy profile (both Schemes). Marvin was used for drawing the chemical structure, Marvin 5.3.4, 2010, ChemAxon (http://www.chemaxon.com)

Since within the Scheme-1, Int1 is stabilized by Lys159, we also calculated the energy barrier for an alternative pathway, i.e. a Scheme-2 (Figure 7B). Here, the order of H⁻ and H⁺ transfer to DHDP by NADH and Lys159 are reversed. According to the Scheme-2, there will be formation of carbocation after H⁺ transfer to DHDP from Lys159, and the energy barrier for the H⁺ transfer (TS1b) was calculated to be only 2.87 kcal/mol (Figure 7C). The resultant carbocation intermediate (Intb), which has the energy of 1.58 kcal/mol (Figure 7C), accepts H⁻ from NADH and gives the intermediate Int1, which is common for both the Scheme-1 & 2 (Figures 7A and B). The energy barrier for H⁻ transfer by NADH to carbocation (TS2b) was calculated to be 2.45 kcal/mol and the distance between C4 of carbocation and hydrogen atom attached to C4 of NADH decreased from 2.45 Å to 1.57 Å. The energy difference between Int1 structures obtained within the Scheme-1 and Scheme-2 is only 0.17 kcal/mol, which is negligible.

Comparison of the reaction energy profiles for both of the schemes reveals that the alternative pathway, i.e. the Scheme-2, involves a much lower energy barrier for H⁻ transfer from NADH to DHDP, and the role of Lys159 is not only to stabilize the reaction intermediate, but also to trigger the reaction by donating H⁺ to substrate. Thus, we consider the Scheme-2 to be more plausible.

Dehydration of HTPA to DHDP

Since HTPA_HIP enzyme-substrate complex is structurally as well as energetically more stable compared to HTPA_HID complex, thus we considered it for further density functional theory (DFT) calculations. Starting from the HTPA_HIP complex, the Scheme-1 (Figure 8A), first, positively charged His155 will donate H⁺ to OH group of HTPA resulting in the release of an H₂O molecule and formation of carbocation. Based on the high-resolution structure of MtDapB that we have determined (PDB ID: 5TJZ), we speculate that this water molecule may be trapped by the protein. In the high-resolution structure of the MtDapB complex with 2,6-PDC there are four water molecules trapped “behind” the inhibitor. This small water filled cavity is delineated by residues that are conserved in both MtDapB and VvDapb. Moreover, the water molecules are in contact with previously mentioned His216 (Q194 in MtDapB).

Figure 8.

HTPA to DHDP dehydration reaction. A) Scheme-1, B) Scheme-2, C) Scheme-3, and D) Energy profile (Scheme-2 and Scheme-3). Marvin was used for drawing the chemical structure, Marvin 5.3.4, 2010, ChemAxon (http://www.chemaxon.com).

In the next step, by transferring its H⁺ from C5 to neutral His155, the carbocation will be finally converted to DHDP. Despite many different attempts, in our DFT calculation, no saddle point for transfer of H⁺ and release of H₂O molecule was obtained. In all scans the energy of the system increased to prohibitively high values (above 50 kcal/mol). Thus, we abandoned this pathway and considered a mechanism of dehydration reaction starting from the HTPA_HID complex (Scheme-2).

In the first step of this scheme (Figure 8B), HTPA transfers its H⁺ from C5 to Nε2 of His155 and turns to a carbanion reaction intermediate. The energy barrier for the first step of dehydration reaction, i.e. TS1a, was calculated to be 21.1 kcal/mol (Figure 8D) and the distance between hydrogen atom attached to C5 (to be transferred as H⁺) and Nε2 of His155 decreased from 3.4 to 1.7 Å. The resulting intermediate (Int1a) displayed an energy of 22.1 kcal/mol (Figure 8D), which is slightly higher in comparison to the energy of TS1a (Int1a is slightly more stable than TS1a at the level used for geometry optimization, but including zero-point energy (ZPE) and basis set corrections reverses their energy order). In the reduction reaction of DHDP to HTPA, it was observed that the carbanionic reaction intermediate (Int1 formed after accepting H⁻ from NADH by DHDP) was stabilized by a proton transfer from Lys159. We considered that this active site residue could also be involved in stabilizing the Int1a of dehydration reaction. Therefore, we calculated the energy barrier for transfer of H⁺ from Lys159 to imine nitrogen, proceeding through TS2a, which is −1.2 kcal/mol (again, the negative activation energy is an artifact of the computational procedure, as at the level of theory used to optimize the geometry the barrier is slightly positive). The water molecule present between Lys159 and substrate mediates the transfer of H⁺ from Lys to substrate‟s nitrogen (Figure 8B). The subsequent intermediate (Int2a) has energy of 0.4 kcal/mol (Figure 8D). Further, the energy barrier for transfer of H⁺ from His155 to OH group of Int2a coupled with release of water molecule, i.e. TS3a, was calculated to have a relative energy of 10.5 kcal/mol. The resultant intermediate (Int3a) with energy of 2.1 kcal/mol is still protonated (H⁺ taken from Lys159 for Int1a stabilization). The energy barrier (TS4a) associated with transfer of H⁺ back to Lys159 was calculated as 1.7 kcal/mol. Finally, the energy of the resultant E-Pa complex was calculated to be −1.2 kcal/mol (Figure 8D).

Since in the case of the reduction reaction reverting the order of H⁻ and H⁺ transfer to DHDP decreases the rate-limiting barrier, which is connected with H⁻ transfer, we considered a similar alternative for the dehydration reaction, i.e. Scheme-3. In this scheme, we proceeded with the same complex HTPA_HID, which was already taken for Scheme-2 (Figure 8C). In the first step of the dehydration reaction, i.e. transfer of H⁺ from Lys159 to HTPA, the energy barrier (TS1b) was calculated to be 4.9 kcal/mol and the resulting carbocation (Int1b) is 4.8 kcal/mol higher in energy relative to E-S (Figure 8D). In the second step of the reaction, C5 of carbocation transfers its H⁺ to His155 with energy barrier (TS2b) of 11 kcal/mol and forms the second reaction intermediate (Int2b) with energy of 6.9 kcal/mol. Taking into account Scheme-2, although there is 5.3 kcal/mol decrease in energy barrier for H⁺ transfer from HTPA to His155 in Scheme-3, the reaction species Int2a and Int2b formed in Scheme-2 and Scheme-3, respectively, have an energy difference of 6.5 kcal/mol. Int2a is energetically more stable than Int2b. Structure superimposition of Int2a and Int2b reveals that H-bond between His155 (doubly protonated) and the reaction intermediate is present in Int2a, but absent in Int2b. Moreover, Int2a and Int2b differ substantial in conformations of Glu153 and Asn124, which participate in extended H-bonding networks. These networks are also very well visible in crystal structures (Figure 4). The distances between hydrogen (to be transferred to the intermediate for release of water) of His155 and the OH group of the intermediate are 1.9 and 2.7 Å, respectively, in Int2a and Int2b. We attempted to calculate the transition state between Int2b and Int2a but did not succeed because of nontrivial differences in H-bond networks of these two species. So, we proceeded from Int2b to calculate the energy barrier (TS3b) for release of water molecule by transfer of H⁺ from His155 to the OH group of Int2b, which is calculated to be 6.9 kcal/mol. Finally, to get the resultant product, DHDP, H⁺ from protonated intermediate Int3b (having energy of 4.1 kcal/mol) was transferred to Lys159. The activation energy for this step was calculated to be 1.4 kcal/mol and the product (E-Pb) has energy of 1.2 kcal/mol (Figure 8D).

So, despite the fact that doubly protonated HTPA_HIP complex is structurally and energetically more stable than singly protonated complex HTPA_HID, it is unable to trigger the dehydration reaction. Moreover, comparison of the calculated energies of Scheme-2 & 3 suggests that the change in the order of H⁺ transfer to HTPA, from His155 and Lys159, respectively, decreases the energy barrier, as we observed for the reduction reaction. Similarly, Lys159 has a role in triggering the dehydration reaction in addition to stabilizing the reaction intermediate. Thus, Scheme-3 is more relevant for dehydration of HTPA into DHDP than Scheme-2.

Conclusions

This manuscript describes several different structures of VvDapB and MtDapB. The sequences of these enzymes are highly conserved and there are almost no changes in the amino acid composition of these enzymes in various strains that were sequenced. We have not noticed any mutation in MtDapB, and there is one mutation in VvDapB. Namely the Val26 present in the enzyme from CMCP6 and SC9794 strains is replaced with Ala in YJ016. It was demonstrated the dapB gene mutation significantly impaired the growth rate of M. tuberculosis [33, 34]. Gram-positive and Gram-negative bacteria use lysine or meso-diaminopimelate to covalently link the peptidoglycan monomers in the bacterial cell wall. As such, we hypothesize that blocking the activity of DapB will induce defects in the bacterial cell wall that are similar to those caused by β-lactam antibiotics. Thus, DapB is a very attractive drug target with unexploited antimicrobial potential. This claim is additionally strengthened by the fact that the M. smegmatis mutant containing an insertion in the dapB gene was β-lactam antibiotic-hypersusceptible [35]. Therefore, DapB inhibitors may also be used in combination with other antibiotics to achieve a synergistic therapeutic effect, broadening their appeal, and providing increased longevity to other compounds for which resistance has been gained.

Biological assembly for all the structures was found to be tetrameric, which is consistent with all the other known structures of DapB homologs. MtDapB was crystallized in its apo form as well as ligand bound form. The structure of MtDapB co-crystallized with NADPH and 2,6-PDC was solved at 1.5 Å resolution. This is the highest resolution structure for any DapB in Protein Data Bank. VvDapB was co-crystallized in different forms with NADH and 2,6-PDC, and NADH and 2,5-FDC. Analysis of these structures revealed that different chains in the tetramer can exist in different conformations ranging from open to fully closed. Analysis of chains of DapB enzymes in different conformational states indicates that these enzymes undergo a minor change upon cofactor binding. Binding of both the cofactor and substrate/substrate analogs on the other hand induces a major conformational change in the protein. This information will be helpful in elucidating the dynamics between the tetrameric chains during enzyme reaction.

The thermodynamic parameters of binding of VvDapB and MtDapB with co-factor NADH was studied using ITC. The experiments indicate that binding of NADH to both the enzymes is thermodynamically favorable with a K_d of 3.6 ± 0.4 μM for MtDapB and K_d of 3.9 ± 1.2 μM for VvDapB. DSF experiments were conducted to screen for novel potential inhibitors of DapB enzymes. Several compounds including 2,5-FDC, chelidoninc acid and nitro-5-furoic acid were identified as potential inhibitory compounds. Crystal structure of VvDapB bound to NAD⁺ and 2,5-FDC has been described. This is the first crystal structure of DapB enzyme bound to a ligand other than 2,6-PDC and its cofactor. Moreover, the binding of 2,5-FDC is very similar to that of 2,6-PDC highlighting its potential as a novel inhibitor.

Structural studies indicate that the enzyme has to adopt the closed conformation in order for the reaction to proceed. Therefore, the crystal structure of VvDapB bound to NAD⁺ and 2,6-PDC was used as a starting point for computational studies on E-S structures and reactions mechanisms. Molecular dynamics simulations coupled with MM-PBSA energy analysis indicate that the reaction substrate binds within the active site in the same fashion as the inhibitor, yet, due to its lower symmetry compared to 2,6-PDC, two orientations of the substrate were considered. The orientation with the sp³-hybridized C2 located in the vicinity of Gly165-Thr166 is energetically preferred. Reaction energy profiles, obtained with the DFT method applied to a large active site model, revealed that His155 and Lys159 play crucial roles during the dehydration and reduction stages of the reaction. Lys159 provides a proton to the imino nitrogen of the substrate/reaction intermediate and it this way activates it for dehydration/reduction. During the dehydration stage His155 first acts as a general base accepting proton from the substrate and then as a general acid donating proton to the leaving OH group. During the reduction stage His155 acts as a general acid providing proton required to complete the reduction of the double bond initiated by H⁻ transfer from NADH. Our computational studies confirm the importance of Lys159 and His155 for the DapB catalyzed reaction. These two residues were experimentally confirmed to be critical for the activity of DapB from E. coli [14], and mutation of each of these amino acids to alanine or glutamine resulted in almost 200-fold reduction of enzyme activity.

Relatively little research has been conducted on the inhibition of DapB [14, 29, 36–38]. The main focus of what has been performed is related to 2,6-PDC, which closely resembles the substrate and has been co-crystallized with DapB homologs. The inhibitory properties were first explored by Scapin et. al. using EcDapB, and it was shown that 2,6-PDC was a competitive inhibitor with K_i of 55 μM [14]. Another effort at inhibitor development for DapB focused on both molecular modeling and conventional high throughput screening to generate novel inhibitors of MtDapB [36]. This led to a new group of inhibitors that are based on a scaffold with benzyl sulfonamide core structure. Several of these compounds were tested using the DapA/DapB dual enzyme assay and all of the inhibitors were found to be competitive and gave K_i values in low micromolar range. A high throughput screening also identified a number of novel compounds that also showed micromolar K_i values [37,38]. One of these compounds that included a 2-[(5Z)-5-[(3,4-dihydroxy-phenyl)methylidene]-2,4-dioxo-1,3-thiazolidin-3-yl]acetic acid fragment was combined with 2,6-PDC to obtain a DapB inhibitor with K_i equal 100nM. However, no further optimization of the identified inhibitors is reported in the literature. It is possible that this was caused by a poor antimicrobial activity of these compounds. Therefore, in the design of the new class of DapB inhibitors we not only want to use compounds that are transition state analog, but we would like to derivatize them to improve their ability to penetrate bacterial cells.

Methods

Cloning

4-hydroxy-tetrahydrodipicolinate reductase (DapB) genes from Vibrio vulnificus strain CMCP6 (Uniprot: Q8DEM0) (VvDapB) and Mycobacterium tuberculosis HKBS1 (Uniprot: W6GQ56) (MtDapB) were cloned into pJExpress411 plasmid vector by ATUM, Inc. (Newark, CA). Both the genes were codon optimized for expression in E. coli and had an additional N-terminal hexa-histidine tag followed by a TEV protease cut site. pJExpress411 plasmid has a selectable kanamycin resistance marker and isopropyl β-D-1-thiogalactopyranoside (IPTG) inducible T7 promotor. For protein expression the plasmids were transformed in E. coli BL21 (DE3) cells using heat shock [39].

Expression and purification

For expression of proteins, E. coli BL21 (DE3) cells containing respective plasmids were grown at 37°C to an OD₆₀₀ of 0.6–0.8 in Luria-Bertani (LB) broth. This was followed by induction of protein expression by the addition of 0.5 mM IPTG and growth at 16°C overnight. The proteins were purified as described previously [24]. Briefly, bacterial cells were collected and resuspended in lysis buffer (50 mM Tris, 500 mM NaCl, and 10 mM imidazole, pH 8.0) supplemented with a protease inhibitor cocktail (Pierce, Rockford, IL). The cells were lysed by sonication and the supernatants containing protein were separated from cell debris by centrifugation (16,000 rpm for 30 min at 4°C). The supernatants were then loaded onto Ni-NTA agarose bead columns (Qiagen or Pierce). The columns were washed with 50 mM Tris, 500 mM NaCl, and 30 mM imidazole, pH 8.0. The proteins were eluted with 50 mM Tris, 500 mM NaCl, and 300 mM imidazole, pH 8.0 and dialyzed overnight in 10 mM Tris, 150 mM NaCl, pH 7.5. Dialyzed proteins were further concentrated and purified by size exclusion chromatography using a GE Healthcare AKTA-Pure FPLC and HiLoad Superdex 200 column.

Crystallization data collection and structure determination

All crystallization experiments were done by vapor diffusion technique using 96 well crystallization plates (Hampton Research, Aliso Viejo, CA). Experiments were set by mixing 1 μL of protein solution (3–5 mg/mL and 10–15 mg/mL of both the proteins) with 1 μL of crystallization solution. A number of well diffracting crystals were obtained for both the proteins in apo and ligand bound forms from different crystallization conditions from the Index (Hampton Research, Aliso Viejo, CA) and Wizard crystallization screens (Emerald Biosystems, Bainbridge Island, WA). The crystallization condition for all the structures are listed in Table S2 (Supplementary Information). All the crystals were cryo-cooled in liquid nitrogen for X-ray diffraction experiments.

X-ray diffraction experiments were done using Life Sciences Collaborative Access Team (LS-CAT), the Structural Biology Center (SBC), and the South East Regional Collaborative Access Team (SER-CAT) beamlines at Advanced Photon Source (APS), Argonne National Laboratory (Lemont, IL). Data were processed using HKL-2000 [40] and structures were solved by molecular replacement using MOLREP [41] integrated with HKL-3000 [40]. E. coli DapB (PDB ID: 1ARZ) and M. tuberculosis DapB (PDB ID: 1C3V) were used as search models. Refinement was done using HKL-3000 integrated with REFMAC5 [42] and various components of the CCP4 package [43]. Manual building was done using COOT [44]. MOLPROBITY [45] and COOT were used as validation tools. The final coordinate files along with structure factors were deposited to the Protein Data Bank with accession codes 5UGV, 5TJY, 5TJZ, 5TEK, 5US6, 5TEJ, 5TEM, and 5TEN.

Isothermal Titration Calorimetry

Thermodynamics of interaction between pyridine nucleotide NADH and MtDapB or VvDapB was studied using Isothermal Titration Calorimetry (ITC). For ITC experiments, His-tag was cleaved from both the proteins and they were stored in 25 mM HEPES, pH 7.5 buffer. MtDapB and VvDapB were used in 100 μM and 52 μM concentrations respectively in the sample cell. NADH solution was made in 25 mM HEPES. 2 mM solution was used for titration with MtDapB whereas 1mM solution was used for VvDapB. A total of 20 injections of 2 μL each were used to study binding of both the proteins with NADH. The syringe was rotated with a speed of 100 rpm and an interval of 250 s was maintained between injections. All the experiments were done at 25°C. Experiments were carried out using Affinity ITC instrument and data analysis was done in Nano Analyze software. (TA Instruments, New Castle, DE)

Catalytic mechanism for DapB

Protein-ligand complexes

The closed conformation structure of VvDapB (PDB ID: 5TEM) bound with inhibitor (2,6-PDC) and NAD⁺ was used for calculations. Structures of the enzyme with substrate HTPA or intermediate DHDP, as well as product THDP, are not available. Thus, employing program Quatfit [46] substrate molecules were superimposed onto the 2,6-PDC molecule bound in the enzyme active site, the latter was subsequently removed from the model and hence two enzyme-substrate complexes were generated with DHDP and HTPA, respectively, bound to enzyme active site and NADH bound to the cofactor binding site. Since both substrates have asymmetric structure, to determine the binding orientation of DHDP and HTPA, enzyme-substrate complexes with two binding orientations M1 and M2 (Figure S5A & B) were generated for further computations: DHDP_M1, HTPA_M1, DHDP_M2, and HTPA_M2. Similarly, an enzyme-product complex of DapB with THDP and NAD⁺ was generated for two binding orientations M1 (THDP_M1) and M2 (THDP_M2).

In all the above enzyme-substrate and enzyme-product complexes, the protonation state of the active site residue His155 was charge neutral, i.e. HID. This residue is not only important for substrate binding, but even more important for catalysis [14]. Thus, to decipher the role of His155 in the enzymatic reaction, MD simulations were performed for enzyme-substrate complexes for two protonation states of His155: doubly protonated - positively charged HIP155, and singly protonated – charge neutral HID155, respectively, for both DHDP (DHDP_HIP and DHDP_HID) and HTPA (HTPA_HIP and HTPA_HID). Since 2,6-PDC is a substrate analog for DapB, ligand-protein interactions in modeled enzyme-substrate complexes were confirmed by comparison of these interactions in MD simulated enzyme-substrate and enzyme-inhibitor (DapB-2,6-PDC) complexes.

Molecular Dynamics Simulation

Explicit water (TIP3P) MD simulations of all the protein-ligand complexes were carried out with AMBER16 [47] employing ff03 force field [48, 49] and GAFF [50] for protein and organic ligands, respectively. Leap module of AMBER16 was used for setting up all the initial structures. To prevent any steric clashes between solute and solvent during MD simulation, all the solvated systems were initially minimized in two steps. First, minimization of solvent and ions was performed by applying 50 kcal/mol/Ų positional restraint on all the atoms of protein and ligand, followed by the second minimization of the whole system without any positional restraint. Before the production run, each system was heated using Langevin dynamics from 10 to 300K at NVT ensemble with the positional restraint of 5 kcal/mol/Ų on the protein’s backbone heavy atoms and ligand’s heavy atoms. Positional restraint was released gradually in the next two steps i.e. 3 kcal/mol/Ų in 1^st step and then 1 kcal/mol/Ų in 2^nd step. Further, system was equilibrated for 100 ps at NVT followed by 2400 ps at NPT ensemble. Finally, the production runs (DHDP_M1 = 100 ns, DHDP_M2 = 100 ns, HTPA_M1 = 100 ns, HTPA_M2 = 100 ns, DHDP_HIP and DHDP_HID = 75 ns, HTPA_HIP = 75 ns, HTPA_HID = 50 ns, THDP_M1 and THDP_M2 = 75 ns, DapB-2,6-PDC = 50 ns) were carried out at NPT ensemble by integrating the Newtonian equation of motion at every 2 fs. Trajectories were analyzed by cpptraj module of AMBER16. Hydrogen bonds were calculated for Donor-Acceptor distance cutoff 3.5 Å and the Donor-Hydrogen-Acceptor angle cutoff 135°.

MM-PBSA analysis

For the calculation of ligand binding free energy, MMPBSA.py of AMBER16 was used [51]. Since no conformational changes were observed in protein during MD simulations, a single trajectory method was used for binding free energy calculation. Whole production run trajectories of complexes DHDP_M1, DHDP_M2, HTPA_M1, and HTPA_M2 were taken for MM-PBSA analysis. For enzyme-substrate complexes, DHDP_HIP and DHDP_HID, last 40 ns of total 75 ns production run trajectories, respectively, were considered. In the cases of enzyme-substrate complexes with HTPA, product complexes THDP_M1, and THDP_M2, last 50 ns of total 75 ns, respectively, and for the enzyme-inhibitor complex DapB-2,6-PDC, last 30 ns of total 50 ns production run trajectories were considered for calculation of binding free energy. Gas phase (electrostatic, vdw and internal) energies were calculated by the sander program within AMBER. Polar solvation (electrostatic) and nonpolar solvation (hydrophobic) free energies were calculated using the Poisson Boltzmann (PB) equation/Generalized Born (GB) method and Linear Combination of Pairwise Overlaps (LCPO) method implemented within sander, respectively.

Active Site Model

To understand the NADH dependent reduction (DHDP to THDP) and NADH independent dehydration (HTPA to DHDP) reactions catalyzed by DapB, DFT calculations were carried out. Cluster representative structures of whole production run MD trajectories were taken to generate active site models for quantum calculations. The active site model for reduction of DHDP to THDP (Figure 6A) consists of key fragments of active site residues: 1) making hydrogen bond with DHDP: His155, His156, Lys159, Ser164, Thr166, His216, and Arg236, 2) making hydrogen bonds with co-factor NADH: Pro123 and Tyr125, 3) making hydrophobic core of the active site: Met13, Phe75, and Phe239 and 4) making hydrogen bond with another active site residue: Asn124 (interacting with Thr166), and 5) connecting residue between other active site residues: Gly165 (between Ser164 and Thr166). The active site model for dehydration reaction of substrate HTPA to DHDP (Figure 6B) consists of key fragments of active site residues: 1) making direct hydrogen bond with substrate HTPA: Thr100, Asn124, His156, Lys159, and Gly165, 2) making water-mediated interaction with HTPA: His155, Lys159, Thr166, and Arg236, 3) making hydrogen bond with co-factor NADH: Pro123 and Tyr125, 4) making hydrophobic core of the active site: Val213 and 5) making hydrogen bond with other active site residues: Glu153 (interacting with HiS155 and Arg208), Ser164 (interacting with Glu153), Arg208 (interacting with Glu153), and Ile212 (interacting with His156), and 6) four water molecules. One nucleotide of NADH consisting nicotinamide and ribose sugar is present in surrounding of HTPA, and making π stacking to DHDP, in case of HTPA and DHDP complexes, respectively. Another nucleotide, consisting of adenine and ribose sugar, is far away from the active site. Thus, NADH is truncated up to the nucleotide consisting of nicotinamide and sugar.

DFT calculation

In the present study, all DFT calculations were carried out with Gaussian 16 [52] employing a hybrid density functional method B3LYP and D3 Grimme‟s dispersion correction with Becke-Johnson damping scheme [53]. Geometry optimization was performed using split valence polarized basis set def2-SVP, the effect of solvation and rest of the enzyme (which is not included in the model) were calculated by self consistent reaction field (SCRF) method using CPCM (conductor like polarizable continuum model) with radius probe of 1.4 Å and dielectric constant 4.0. During geometry optimization Cα atoms were fixed in order to preserve the geometry of the active site. To calculate zero-point energy (ZPE) correction, frequency calculation was performed at the same level of theory and SCRF method as used for geometry optimization. The accurate single point energies were calculated employing triple zeta valence polarized basis set def2-TZVP [54] and SCRF method [55]. These energies corrected for ZPE are presented and discussed beneath.

Screening and identification of potential inhibitors

Differential Scanning Fluorimetry (DSF) was used as a screening tool to identify potential inhibitors for DapB enzymes. This assay uses the melting temperature of proteins as a measure of protein stability. The assay was designed as described by Booth et al., 2018 [24]. The stability of VvDapB was evaluated in the presence of different ligands. Ligands were selected on the basis of their similarity to 2,6-PDC which is known to bind DapB enzymes and acts also as an inhibitor. Different ligands that were tested are listed in Figure 9. The effect of these ligands was evaluated in presence and absence of co-factor NADH. All the ligands were used at a final concentration of 10 mM. For all experiments, Sypro orange was diluted 1:1000 in the protein solution. A step temperature gradient was setup from 30–90 °C in 2 °C increments, with incubation at each temperature followed by fluorescence measurement. A Biorad CFX96 RT-PCR set to the FRET channel was used for screening. The melting temperature (T_m) was calculated using the proprietary Biorad software with the minimum of the first derivative plot as the melting point of the protein.

Various computational calculations

The SSM algorithm in COOT was used for the superposition of structures [56]. PyMOL [57] and Espript [58] were used for preparation of figures. DynDom server was used to compare various chains in the structures [27, 28].

Highlights.

• Lysine biosynthesis pathway is a target for development of new antibiotics.

• Several crystal structures of DapB were determined.

• Detailed computational analysis of possible reaction mechanisms was performed.

Abbreviations

2,5-FDC

2,5-furan dicarboxylic acid

2,6-PDC

2,6-pyridine dicarboxylic acid

DapB

4-hydroxy-tetrahydrodipicolinate reductase

DFT

Density Functional Theory

DHDP

dihydrodipicolinate

DSF

Differential Scanning Fluorimetry

ITC

Isothermal Titration Calorimetry

HTPA

(2S, 4S)-4-hydroxy-2,3,4,5-tetrahydrodipicolinate

MD

molecular dynamics

MtDapB

DapB from Mycobacterium tuberculosis

PDB

Protein Data Bank

VvDapB

DapB from Vibrio vulnificus