Analysis of drug binding pockets and repurposing opportunities for twelve essential enzymes of ESKAPE pathogens

Selection of pathogens

Nine different bacterial pathogens were selected including six ESKAPE pathogens, namely Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa and Enterobacter aurogenes. Mycobacterium tuberculosis and two model non-pathogenic bacterial species Mycobacterium smegmatis and Escherichia coli were also included in our binding pocket comparative analysis.

Selection of essential enzyme targets

Twelve essential and mostly conserved metabolic enzymes were selected as priority targets for our study from a larger set of high-ranked metabolic drug targets based on previous comparative and functional genomics analyses (Osterman & Begley, 2007) (Table 1). These critical enzymes catalyze key steps in various metabolic pathways including folate biosynthesis, fatty acid biosynthesis, isoprenoid biosynthesis, peptidoglycan biosynthesis, nicotinamide adenine dinucleotide (NAD) and NADH biosynthesis, and riboflavin metabolism. More importantly, our target selection was driven by the availability of experimentally determined 3D structures and complexes of the respective protein target, independent of the species. In addition to other criteria, the majority of these selected enzymes lack human orthologs (or close homologs), which is important for antibiotic discovery to minimize side effects via inhibition of respective host enzymes (Table 1).

Uniprot sequences and their alignment

The amino acid sequences of the selected enzymes for each of the respective bacterial pathogens were obtained from the Uniprot database (The UniProt Consortium, 2017). The sequences for each enzyme were then aligned using ICM v3.8-4a or above (Molsoft L.L.C., La Jolla, CA) (Abagyan & Totrov, 1994). The selected enzyme targets were matched against the Pocketome encyclopedia (Kufareva, Ilatovskiy & Abagyan, 2011), which contains a set of collected and annotated binding pocket structures derived from the Protein Data Bank (PDB), allowing accurate, experimentally-derived definitions of their respective binding pockets used for this analysis (Table 1).

Generation of distance-based maps for binding pocket sequence identity comparison

Based on the full protein sequence, the sequence identity between each of the target enzymes across each species was calculated; sequence identity was measured as identical residue/total number of aligned residues, transformed by the Dayhoff correction and distance-based maps were generated in ICM. The binding pocket of each enzyme was determined by the residue contacts made by small molecule inhibitors co-crystallized with an enzyme from one of the selected pathogens. A subselection of the full sequence alignment was extracted by propagating the determined binding pocket residues over orthologous protein sequences of the selected pathogens. This followed binding pocket identity calculations and regeneration of sequence maps based on the binding site as described above.

The labels used for various microorganisms were as followed: ENTFC for Enterococcus faecium, STAAU for Staphylococcus aureus, KLEPN for Klebsiella pneumoniae, ACIBT for Acinetobacter baumannii, PSEAE for Pseudomonas aeruginosa, ENTAE for Enterobacter aurogenes, MYCTU for Mycobacterium tuberculosis, MYCSM for Mycobacterium smegmatis and ECOLI for Escherichia coli.

Generation of homology model

A homology model was generated for the enzyme 1-deoxy-D-xylulose 5-phosphate reductoisomerase (Dxr) from P. aeruginosa for predictive binding of known inhibitor fosmidomycin into the binding pocket. The crystal structure of Dxr from Mycobacterium tuberculosis (PDB ID: 4AIC) was used as the template (45% sequence identity). This model was built by using the ICM homology modeling module.

Drug binding pocket conservation of enzymes involved in fatty acid biosynthesis

Acetyl-Coenzyme A carboxylase (AccC) plays a key role in synthesis and degradation of fatty acids. Targeting bacterial acetyl-CoA carboxylase can be an effective approach for discovery of novel antibiotics with low side effects. Of the selected pathogens, acetyl-CoA carboxylase (AccC) from E. coli has been crystallized (PDB: 2V58). The comparison of the putative binding pockets based on the AccC E. coli crystal structure revealed that binding pocket of AccC from E. coli (PDB: 2v58) (Miller et al., 2009) had 100% identity with its ortholog in E. aurogenes and K. pneumoniae and up to 97% identity in P. aeruginosa and A. baumannii. In comparison with the full-length sequence, the binding pocket of this enzyme is highly conserved and interestingly, the level of conservation of drug binding pocket is high among Gram-negative species (97–100%) as compared to Gram-positive bacterial species like S. aureus and E. faecium (up to 74%) as depicted in Figs. 2A and 3A.

Beta-ketoacyl-acyl carrier protein synthase III (FabH) is another key enzyme involve in fatty acid biosynthetic pathway. Using the FabH crystal structure from E. coli (PDB: 1MSZ) (Daines et al., 2003), the putative drug binding pocket of K. pneumoniae shared 89% identity with the co-crystallized structure whilst the drug binding pockets of FabH among all species in the study have high identity (77–100%) as suggested by the maps (Fig. 2B). Previously, 4,5-dichloro-1,2-dithiole-3-one was reported as potent inhibitor of FabH from S. aureus but also in E. coli, further supporting our hypothesis that inhibitors effective in one species have potential off target activity against other species (He et al., 2004).

The binding pocket of enoyl-acyl carrier protein reductase (FabI) is highly conserved among various species as compared to the overall sequence. By using the E. coli FabI crystal structure (PDB: 1QSG) (Stewart et al., 1999), we found that the respective FabI ortholog putative binding pockets in E. aurogenes, K. pneumoniae, P. aeruginosa, A. baumannii were 100% identitical to one another, while sharing greater than 80% sequence identity compared to the putative FabI pocket E. faecium and M. tuberculosis (Fig. 2C).

Drug binding pocket conservation of enzymes involved in folate synthesis

The sequence of drug binding pockets of key enzyme dihydrofolate reductase (folA) of folate biosynthesis were found to be highly conserved (81–100%) among various ESKAPE pathogens by using PDB: 4NX6 from E.coli for comparison and the same trend was observed in case of 6-Hydroxymethyl-7,8-dihydropterin pyrophosphokinase (folK) except S. aureus with reference PDB: 4M5H (Figs. 2D and 2E).

Drug binding pocket conservation of enzymes involved in isoprenoid biosynthesis

The species S. aureus and E. faecium involve the mevalonate pathway for isoprenoid biosynthesis so they lack the Dxs, Dxr and IspH enzymes. The amino acid residues involved in ligand interactions of 1-deoxy-D-xylulose 5-phosphate synthase (Dxs) from E.coli (PDB: 2EGH) (Yajima et al., 2007) showed sequence identity values below 70% among various bacterial species in current comparison. The other two enzymes of non-mevalonate pathway like IspH and Dxr showed 98–100% conservation of amino acid sequence among all species in comparison based on the crystal structures PDB: 3ZGL and PDB: 3ANL (Deng et al., 2010) from E.coli respectively (Figs. 2G–2I).

Drug binding pocket conservation of enzymes involved in peptidoglycan and CoA biosynthesis

D-alanyl-D-alanine ligase (ddlB) is a significant target for antibiotic discovery as it is essential for bacterial growth. Propagating the drug binding pocket of ddlB from E.coli (PDB: 1IOW) (Fan et al., 1997) onto all the other species, it revealed that they all shared 96–100% sequence identity amongst one another as shown in Fig. 2F.

Phosphopantetheine adenylyltransferase (CoaD) enzyme involved in coenzyme A biosynthesis is another attractive target for antibiotic discovery. Similarly, the drug binding pocket of Phosphopantetheine adenylyltransferase based on PDB: 1B6T (Izard & Geerlof, 1999) was found to be highly conserved amongst the ESKAPE, Mycobacterium tuberculosis as well as other bacterial species in the range of 85–100% as compared to full sequence (Figs. 1 and 2K).

Drug binding pocket conservation of NADD and 6,7-dimethyl-8 -ribityllumazine synthase enzyme

The drug binding pocket of co-crystal structure of NADD from E. coli (PDB: 1K4M) (Zhang et al., 2002) showed 100% identity with all the species in this comparison, except A. baumannii, while the inhibitor binding pocket of the enzyme ribH, which catalyzes a key step in metabolism of Riboflavin resulting in synthesis of flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD) cofactors, was conserved (90–100%) across the various species as compared to reference binding pocket (PDB: 2VI5) (Zhang et al., 2008) (Figs. 2J and 2L).

Comparison of fosmidomycin interaction with Dxr binding pockets across various species

From the ChEMBL dataset of active compounds for the enzyme 1-deoxy-D-xylulose 5-phosphate reductoisomerase (Dxr), we found that Fosmidomycin, a natural antibiotic, is an inhibitor for Dxr in both E.coli and M. tuberculosis. Crystal structures of E.coli and M. tuberculosis Dxr have been solved with Fosmidomycin (PDB ID: 1ONP and 4AIC respectively) and they display a similar shape and orientation in three-dimensional space (Henriksson et al., 2007; Steinbacher et al., 2003) (Figs. 4A and 4B). In addition, the inhibitor showed similar, if not identical, interactions with binding pocket residues (Fig. 4C); they display 41% sequence identity across the full length sequence, which further strengthens our hypothesis that inhibitors active in one species may have activity for other enzyme orthologs. As Fosmidomycin has shown activity against Dxr of M. tuberculosis and E. coli, this will help the design of new drugs for tuberculosis by taking advantage of the fosmidomycin chemical scaffold. Fosmidomycin and its derivatives has also been reported as potent inhibitors of Dxr of P. falciparum, the causative agent of malaria. To predict its interaction-binding mode, we generated homology model of Dxr from P. aeruginosa using PDB: 4AIC as a template and docked Fosmidomycin into the putative binding pocket. The highest ranked docking pose predicted a similar interaction pattern compared to the crystal structures (Fig. 4C). In another example, oxacillin, a β-lactam antibiotic of the penicillin group has been co-crystallized with the Class D β-lactamases oxacillinase OXA-1 from E. coli and OXA-24/40 from A. baumannii (which shares 99% full length sequence identity with OXA-1) (PDB ID: 4MLL and 4F94, respectively) (June et al., 2014). The comparison of the ligand binding pockets display similar orientation (with slight rotation of an R side chain group between the structures), binding interactions, as well as shape (June et al., 2014) as depicted in Figs. 4D and 4E. The examples of fosmidomycin and oxacillin further demonstrates the idea of conserved binding pockets and cross-species activity.