Analysis of host and pathogen metabolic pathways.

Genome-wide metabolic pathway analysis was performed via the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://www.genome.jp/kegg/) [25], [26]. Metabolic pathways and assigned identification numbers of the pathogen M. ulcerans and the human host were extracted from the KEGG database. A manual comparison was then made, and pathways that did not appear in the host but were present in the pathogen, according to the KEGG database annotations, were selected as unique to M. ulcerans, while the remaining pathways were listed as common. M. ulcerans proteins from common and unique pathways were then identified, and the respective amino acid sequences were obtained from the NCBI genome database.

Identification of non-host and non-host plus essential proteins.

Two-step comparisons were performed between host and pathogen proteomes for the identification of non-host proteins from M. ulcerans. At first, only proteins from pathogen-specific pathways were subjected to NCBI BLASTP analysis [27]. Second, proteins from common pathways were also compared by BLASTP analysis. In each scenario, searching was restricted to proteins from H. sapiens (taxid: 9606), through an option available under the NCBI BLASTP parameters. Proteins without hits below the E-value inclusion threshold of 0.005 and <35% identity, were chosen as non-host bacterial proteins. Once all the non-host proteins were identified, they were further filtered on the basis of essentiality by comparison with protein sequences of the M. ulcerans essential genes for the creation of a non-host essential proteins dataset.

Prioritization parameters for therapeutic targets.

Several molecular and structural criteria that have been proposed to aid in prioritizing suitable therapeutic targets in pathogenic microorganisms [28] were also evaluated for each of the predicted therapeutic targets in M. ulcerans. This involved, calculation of molecular weight (MW) using computational tools and drug target-associated literature available in the Swiss-Prot database [29]. Prediction of biological significance and subcellular localization of therapeutic targets was carried out using MycobacSVM (http://211.83.105.213/server/MycobacSVM/MycobacSVM.htm) [30]. MycobacSVM is a support vector machine based method that uses feature selection techniques to specifically predict subcellular location of mycobacterial proteins at four levels: cytoplasmic, integral membrane, secretory, and membrane attached proteins by a lipid anchor. Transmembrane predictions were made using TMHMM v2.0 (http://www.cbs.dtu.dk/services/TMHMM/) [31] and the TOCONS consensus membrane proteins topology prediction server, which combines predictions from four different algorithms (http://topcons.cbr.su.se/) [32]. Experimentally and computationally solved 3D structures were detected by searching the Protein Data Bank (PDB) (http://www.rcsb.org/pdb) [33] and ModBase (http://salilab.org/modbase) [34] databases, respectively.

Druggability.

Druggability is another important therapeutic prioritization criterion; defined as the likelihood of being able to modulate the activity of the target protein with a small-molecule drug [35], [36]. The druggability potential of each identified drug target was calculated by mining the DrugBank contents (http://www.drugbank.ca/). The DrugBank database is a unique bioinformatics and cheminformatics resource that combines detailed drug (i.e. chemical, pharmacological and pharmaceutical) data with comprehensive drug target (i.e. sequence, structure, and pathway) information. The database contains 6,711 drug entries, including 1,441 FDA-approved small molecule drugs, 134 FDA-approved biotech (protein/peptide) drugs, 84 nutraceuticals, and 5,084 experimental drugs. In addition, 4,231 non-redundant protein (i.e. drug target/enzyme/transporter/carrier) sequences are linked to these drug entries [37]. A BLASTP with default parameters was performed to align the potential drug targets from M. ulcerans against the list of protein targets of compounds found within DrugBank. The selection criteria for filtering BLAST results were as described previously for identification of drug targets in bacterial genomes [23], that is, alignments with e-values less significant than 1×10⁻²⁵ were removed.

Strand-bias among essential genes of M. ulcerans.

It is known that essential genes show strand bias and are preferentially located on leading strands rather than lagging strands. This phenomena was initially studied for E. coli and B. subtilis [15]; and it was postulated that essentiality, rather than expressivity, drives strand-bias. Recently, this phenomena of strand-bias among essential genes has been confirmed in two studies. In one, the authors calculated the biased distribution of essential genes on leading and lagging strands, by analyzing experimentally driven data of essential genes from ten bacterial species [39]. In the other, strand-bias was studied among essential genes from 16 different Mycoplasma species [40]. We analyzed strand-bias in the M. ulcerans genome at two levels, genome-wide and for the predicted set of essential genes. Initially, the replication origins and replication termini were predicted using Ori-Finder [41], based on which the bacterial genes located on the leading and lagging strands were determined. Among the 4,957 genes of M. ulcerans, 2,720 were identified on the leading strand and the 2,237 others on the lagging strand (data available on request). When the same was applied on the 424 essential genes of M. ulcerans, the distribution pattern and strand-bias became significant, as 319 essential genes were found on the leading strand and the remaining 105 were on the lagging strand (Figure 3A and Table S3); thus supporting leading strand-bias among essential genes of M. ulcerans. This and data from previous studies therefore favors that strand-bias can be effectively used as a selection parameter in essential gene prediction algorithms.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Evaluation of M. ulcerans essential genes via essential genes prediction features.

(A) Percentage distribution of M. ulcerans essential genes on leading and lagging stands and among COG functional subcategories. (B) Percentage distribution of M. ulcerans essential genes on leading and lagging strands for six enzyme classes. Strand-bias towards different enzyme classes is also shown.

https://doi.org/10.1371/journal.pone.0043080.g003

Essential genes of M. ulcerans are biased towards COG.

Lin et al. has recently reported that essential genes are biased towards the following ten COG functional subcategories: J (translation, ribosomal structure and biogenesis), K (transcription), L (replication, recombination and repair), D (cell cycle control, cell division, chromosome partitioning), M (cell wall/membrane/envelope biogenesis), O (post-translational modification, protein turnover, chaperones), C (energy production and conversion), G (carbohydrate transport and metabolism), E (amino acid transport and metabolism), and F (nucleotide transport and metabolism) [39]. In parallel to this, Lin et al. also showed that essential genes from the above ten COG categories were preferentially located on leading strands rather than lagging strands. To test if this applies in M. ulcerans, COG functional classes were assigned to 424 essential genes via the NCBI COG database (http://www.ncbi.nlm.nih.gov/COG/). Where one essential gene was assigned to two or more COG functional categories, it was counted in each category, as previously described [39]. The essential genes of M. ulcerans were distributed in 19 COG categories, namely J,K,L,D,M,O,C,G,F,E,H,V,T,I,Q,P,U,R, and N (Figure 3A and Table S3). Most essential genes belonged to COG subcategory E, followed by H, and J. Subcategories D,V (defence mechanisms), and M (cell wall/membrane/envelope biogenesis) had the lowest numbers of essential genes (Figure 3A). None of the essential genes of M. ulcerans were mapped to subcategories B (chromatin structure and dynamics), Y (nuclear structure), Z (cytoskeleton), or W (extracellular structures). These findings are in agreement with the work of Lin et al., with a few exceptions, such as that the M. ulcerans essential genes from COG subcategories D, O, and G did not show significant differences between leading and lagging strands. In addition, essential genes from subcategories P (inorganic ion transport and metabolism), and U (intracellular trafficking, secretion, and vesicular transport) also showed bias towards the leading strand. It is well known that highly-expressed essential genes are preferentially situated on the leading strand in order to avoid head-on collisions between DNA and RNA polymerases [15]. This analysis can be applied to the functional subcategory J, because many genes from this subcategory code for ribosomal proteins. Similarly, the lowest number of essential genes mapped to subcategory D, representing a positive correlation for slow growing bacteria. Furthermore, this shows that COG classifications have the potential to be used as essential gene prediction algorithms, alongside other prediction methods such as strand-bias.

Codon adaptation index of M. ulcerans essential genes.

The preference for synonymous codons among prokaryotes is now accepted to be the result of mutational bias and natural selection acting at the level of translation. Several studies that have focused on essential gene identification by computational methods, have used CAI values as measures of gene essentiality, in combination with other features. The CAI values of genes in a genome range between 0 and 1. A higher CAI value usually suggests that the gene of interest is likely to be highly expressed and therefore essential. Therefore, a CAI >0.5 can be used as a threshold for selection of essential genes that are also highly expressed. We computed the CAI for 424 essential genes of M. ulcerans, using the ACUA program [42], which implements the Sharp and Li method of CAI calculation [43]. Non-synonymous codons and stop codons were excluded during calculation. The CAI values of M. ulcerans essential genes ranged from 0.516 to 0.829. Above the selection threshold of 0.5, all genes are considered highly expressing and likely to be essential (Table S3). However, similarly to strand-bias, CAI alone is a weak selection criterion for the identification of essential genes [44]. It can be best used in combination with other selection features, as we observed that M. ulcerans essential genes located on the leading strand have higher CAI values than those located on the lagging strand (Table S3). Similarly, CAI values can also be used in combination with COG classes, as we observed that essential genes associated with J, E, and H classes have higher CAI values than those associated with other classes (Table S3). We propose therefore that CAI, COG, and strand-bias can be used in combination, for the development of an effective algorithm for computational identification of essential genes.

Enzyme enrichment in M. ulcerans essential genes.

It has recently been shown that enzymes are enriched in essential bacterial genes by overrepresented ligases and underrepresented oxidoreductases when compared to non-essential genes of same bacteria [45]. We did not perform a cross-comparison of enzyme types among essential and non-essential genes of M. ulcerans. Instead, we were interested to determine the specific enzyme class to which maximum number of essential enzymes belonged. To perform this analysis, the enzyme commission (EC) code annotations available as GBK files for M. ulcerans were retrieved from the NCBI FTP server (ftp://ftp.ncbi.nih.gov/genomes/Bacteria). In addition, information about EC numbers was also taken from NCBI, Swiss-Prot, and KEGG databases to cover the maximum possible number of enzymes. Among the 424 essential genes, EC numbers were available for 336 essential enzymes. Among these 336 essential enzymes, 128 were transferases, 65 oxidoreductases, 43 ligases, 43 hydrolases, 40 lyases, and 17 isomerases (Figure 3B and Table S3).

While performing this analysis, we became interested in determining whether enzymes from different classes also show strand-bias. Unlike for previous comparisons among two or more essential gene prediction features, such as COG categories and strand-bias [39], to the best of our knowledge, strand-bias among essential enzymes is not known. To determine any potential relationship between two these features, enzymes from each class were evaluated for their location on either strand. Interestingly, a significant level of bias towards the leading strand was observed in each class (Figure 3B and Table S3). We roughly propose that transferases and isomerases show maximum strand-bias. As in case of M. ulcerans, although only 17 isomerases were identified as essential, 15 of them were located on leading strand. As stated earlier, there were equal number of essential ligases (43/336) and hydrolases (43/336), the pattern of strand-bias exhibited by enzymes from these classes was also found to be similar i.e., there were 35 ligases and 35 hydrolases on leading strand and 8 ligases and 8 hydrolases on lagging strand. As strand-bias and enzyme classes distribution are now known to exist among essential genes, it would be worthwhile to further validate this relationship in those bacteria whose essential genes have already been identified experimentally (Table S1). Identification and further validation of this feature can provide an excellent opportunity to add additional parameter in essential genes identification algorithms.

It is known that some non-coding regions are also essential if they contain DNA sequences responsible for important biological functions. Such sequences include the chromosomal origins of replication, promoters, tRNAs, rRNAs, and small RNAs. It is likely that some essential genes were missed due to limitations in available experimental data about essential genes from other bacterial species. However, until large scale experimental gene essentiality studies are conducted for M. ulcerans, essential genes data from our study can serve as a reference. Furthermore, the post identification evaluation of essential genes, based on other genomics features, also supports the likelihood that the identified genes are essential for M. ulcerans.

Bacterial secretion systems.

Bacterial protein secretion pathways play a key role in modulating biotic associations as well as pathogenicity, via secretion of virulence factors. In Gram-negative bacteria, six general classes of secretion systems have been identified which show considerable diversity, and facilitate the entry of secreted proteins into host cells, modification of host physiology, and colonization. In Gram-positive bacteria, secreted proteins are commonly translocated across the single membrane by general secretion (Sec), and twin-arginine translocation (Tat) [54]. Although Gram-positive bacteria share some of the same secretion systems as Gram-negative bacteria, others such as M. tuberculosis and M. marium, which have a hydrophobic, nearly impermeable cell wall called the mycomembrane, have developed a specialized secretion system. This is known as the type VII secretion system and is responsible for the virulence and translocation of proteins across both the membrane and the cell wall [55], [56]. Although M. ulcerans also belongs to the genus Mycobacteria and is thought to have evolved from M. marium, it does not have a type VII secretion system; this is believed to be an immune evasion strategy of the pathogen [57]. The importance of bacterial secretion systems is well established in terms of both bacterial viability and pathogenicity and they have therefore been widely suggested as targets for new drugs, vaccines, and diagnostic markers. Several proteins from secretion pathways have been proposed as drug and vaccine targets using both computational and experimental approaches for different Gram-positive and Gram-negative bacteria, including B. pseudomallei [20], S. aureus [21], and M. tuberculosis [58]. No such information is yet available for M. ulcerans. We therefore evaluated the therapeutic potential of M. ulcerans secretion pathways via comparative genomics analysis. Initially, in accordance with the KEGG annotations, we identified 12 proteins from the M. ulcerans genome that mapped to protein export pathway (Table 1). Among these, eight were further identified as non-host (Table S2) and seven as essential (Table S3). Our focus was to select proteins that were non-host plus essential; six proteins were selected: preprotein translocase subunit (SecY), preprotein translocase subunit (SecE), inner membrane protein translocase component (YidC), preprotein translocase subunit (SecA), lipoprotein signal peptidase (LspA), and sec-independent protein translocase transmembrane protein (TatC) (Table S4).

SecE, SecY, SecA, and YidC are part of the M. ulcerans Sec pathway and TatC belongs to the Tat export pathway. In Mycobacteria including M. ulcerans, both these secretory pathways are functional and essential or indispensable. SecY, SecE, and SecG form an essential heterotrimeric protein complex that is central to the Sec pathway. This SecYEG complex serves as a transport channel for the movement of protein synthesized in the cytoplasm to the extracytoplasmic environment. This transport and movement of protein through the SecYEG channel is regulated by cytoplasmic ATPase SecA, via repeated cycles of ATP-binding and hydrolysis. SecG is expendable, although it increases the efficiency of protein export. It has been observed that disruption of SecYEG-SecA leads to aggregation of unfolded bacterial proteins in the cytoplasm and triggers a cellular stress response. LspA, also identified as a potential drug target (Table S4), is responsible for the cleavage of signal peptides from lipoproteins, thereby leading to the folding of proteins into mature conformations. These functional lipoproteins are known to be involved in the virulence of mycobacteria. Therefore, following the biological significance and increasing evidence supporting the therapeutic potential of protein export machinery, we suggest that the development of inhibitors against these proteins holds great therapeutic potential for the treatment of M. ulcerans infections.

Peptidoglycan biosynthesis.

Cell walls are important and integral components helping bacteria to maintain their morphology as well as to withstand unfavorable conditions. The disruption of bacterial cell walls leads to cell lysis and hence cell death. Several antibiotics, such as penicillin, bacitracin, and vancomycin, kill bacteria by interfering with the biosynthesis of their cell walls [59]. Cell wall biosynthesis therefore remains a valid target for novel antibiotic development, especially for those drugs that can specifically inhibit any one of the series of essential enzymatic functions involved in the assembly of peptidoglycan (PG). To evaluate the therapeutic potential of the M. ulcerans PG biosynthesis pathway, we initially performed genome-wide metabolic pathway analysis, and identified 14 proteins from the M. ulcerans genome associated with PG biosynthesis. Of these, 12 (MurA, MurB, MurC, MurD, MurE, MurF, MurG, DacC, Ddl, MraY, UppP, and PbpB) were further identified as essential, mapped to COG subcategory M (cell wall biogenesis), and preferentially located on the leading strand, except for MurB. Following subcellular localization analysis, MurA–MurF and Ddl were identified as cytoplasmic, and MraY, UppP, PbpB, and DacC as membrane proteins. As this pathway is absent in humans, none of the 11 enzymes showed homology with human proteins, and they were therefore classified as non-host plus essential (Table S4). PG, which forms more than 70% of the weight of the cell wall, is a large molecule responsible for maintaining morphology and balance via osmotic pressure [60]. Biosynthesis of PG is a complex process of assembly and polymerization. Briefly, at the first assembly step, MurA and MurB catalyze the formation of UDP-GlcNAc-enolpyruvate and UDP-MurNAc, respectively, followed by the successive additions of L-alanine, D-glutamic acid, meso-diaminopimelic acid or L-lysine, and D-alanyl-D-alanine, to form UDP-MurNAc-L-Ala-γ-D-Glu-meso-A₂pm-D-Ala-D-Ala. These steps are catalyzed by specific peptide ligases, designated MurC, MurD, MurE, and MurF. The second stage of PG biosynthesis involves transglycosylation and transpeptidation reactions of the disaccharide pentapeptide monomers, and takes place in the periplasmic space catalyzed by several membrane and periplasmic enzymes [61]. The therapeutic potential of targeting the above enzymes as drug targets is evident from several experimental studies conducted on different bacteria, including M. tuberculosis. We therefore propose that these enzymes can effectively be targeted against M. ulcerans. MurA, for instance, which catalyzes the condensation of phosphoenolpyruvate and UDP-N-acetylglucosamine, is a valid target of the antibiotic, fosfomycin [62]. However, one potential drawback of targeting MurA is the presence of two separate genes, murA1 and murA2, that encode proteins with the same enzymatic activity in Gram-positive pathogens such as S. aureus and Streptococcus pneumoniae. In addition, differences in the active sites also make it difficult to develop MurA-specific antibiotics that could effectively inhibit both enzymes because the two homologs of the murA gene share less than 60% identity among Gram-positive species. Mutagenic studies have shown that disruption of either the murA1 or the murA2 gene had no significant effect on cell growth, but cells were unable to survive when both genes were removed [63]. To check whether M. ulcerans also contains two copies of murA, we performed a genome-wide search and found that M. ulcerans contains only one copy of murA gene (data not shown); meaning that the MurA of M. ulcerans is an attractive drug target for already available antibiotics. Another attractive drug target is D-Alanine:D-alanine ligase (Ddl); this catalyzes the ATP-driven ligation of two D-alanine molecules to form the D-alanyl:D-alanine dipeptide. This molecule is a key building block in PG biosynthesis and inhibition of this step leads to extensively weaker cell walls and cell death. D-cycloserine is a competitive inhibitor of Ddl, and is used as a second line of defense in the treatment of tuberculosis. We therefore suggest that fosfomycin and D-cycloserine should be tested as drugs of choice against the MurA and Ddl of M. ulcerans, respectively. Although not tested experimentally for M. ulcerans, based on genome-wide analysis, we suggest that antibiotics such as bacitracin and others of similar mechanism may not be good choices against M. ulcerans due to the presence of the uppP gene, known for encoding a protein responsible for conferring resistance to bacitracin.

Polyketide sugar unit biosynthesis.

Polyketides are secondary metabolites that are formed as a result of the polymerization of acetyl and propionyl subunits. Polyketides play important roles in intercellular communication, maintenance of cell wall viability and defense mechanisms in different bacteria, including M. ulcerans, and potentially help through periods of desiccation [64]. Hence, the therapeutic potential of this pathway has been proposed in several studies [65], [66]. No such information is yet available for M. ulcerans. We identified four enzymes encoded by rml genes: Alpha-D-glucose-1-phosphate thymidylyl-transferase (RmlA), dTDP-glucose-4,6-dehydratase (RmlB), DTDP-4-dehydrorhamnose 3,5-epimerase (RmlC), and dTDP-6-deoxy-Llyxo-4-hexulose reductase (RmlD), associated with the polyketide sugar unit biosynthesis pathway. These four enzymes catalyze the steps of biosynthesis of dTDP-rhamnose from dTDP and glucose-1-phosphate. Following our comparative genomics analysis, each enzyme was also identified as essential, thereby highlighting the importance of targeting these enzymes as drug targets (Table S3). However, RmlA, RmlB, and RmlD showed more than 45% identity with human proteins. In contrast, RmlC showed no homology with human proteins. Furthermore, a search across DrugBank showed that M. ulcerans RmlC shares significant levels of identity with the binding partner (M. tuberculosis RmlC) of a small molecule experimental drug S,S-(2-hydroxyethyl) thiocysteine (Table S5). RmlC catalyzes a key step in dTDP-rhamnose synthesis by converting glucose-1-phosphate to dTDP-L-rhamnose via epimerizing the C3′ and C5′ positions of dTDP-6-deoxy-D-xylo-4 hexulose, making dTDP-6-deoxy-L lyxo-4-hexulose [67]. The essentiality of RmlC in mycobacterial survival is also evident from previous gene knock-out studies of M. tuberculosis and M. smegmatis [65]. We propose therefore that RmlC can serve as an effective drug target for disruption of M. ulcerans cell wall synthesis, for its essential and non-host nature, involvement in cell wall synthesis and, in comparison with other Rml enzymes, high substrate-specificity and cofactor-independent activity [65].

Thiamine biosynthesis.

We identified five drug targets from the thiamin biosynthesis pathway of M. ulcerans. These included the thiamine biosynthesis protein (ThiC), phosphomethylpyrimidine kinase (ThiD), thiamine biosynthesis oxidoreductase (ThiO), thiamine monophosphate kinase (ThiL) and thiazole synthase (ThiG) (Table S5). Thiamin (Vitamin B1) is an essential cofactor and is indispensable for the activity of the carbohydrate and branched-chain amino acid metabolic enzymes. The active form of thiamin is thiamin diphosphate (ThDP). Synthesis of bacterial ThDP is a two-step process involving the formation of the thiazole moiety, 4-methyl-5-β-hydroxyethyl thiazole phosphate (THZ-P) and the pyrimidine moiety, 4-amino-5-hydroxymethyl-2-methyl-pyrimidine pyrophosphate (HMP-PP) [68]. During the first step, THZ-P is derived from an oxidative condensation of tyrosine or glycine, cysteine and 1-deoxy-D-xylulose 5-phosphate via seven different enzymes including ThiG and ThiO. Parallel with this, formation of HMP-PP is accomplished in two steps regulated by ThiC and ThiD. In the second and final step, THZ-P and HMP-PP are coupled into the active form of thiamin (ThDP) through a final phosphorylation step mediated by ThiL [69]. thiL and other genes from the thiamin metabolism pathways of Plasmodium falciparum [70] and M. tuberculosis [71] have also been identified as essential for survival and as potential drug targets. Designing inhibitors of identified enzymes to block the biosynthesis of thiamin therefore represents an attractive strategy with potential to damage the growth and survival of M. ulcerans.

Aminoacyl-tRNA biosynthesis.

Aminoacyl-tRNA synthetases (AaRS) are the group of enzymes that catalyze the acylation of amino acids to tRNA molecules in the translation stage of protein biosynthesis. Being identified as essential for the survival of the pathogen and playing a crucial role in protein biosynthesis, AaRS enzymes have received much attention for antibacterial drug discovery by several pharmaceutical companies and academic research groups [72]. In addition to essentiality, there are several features that favor targeting AaRS for drug discovery. These include: (i) the presence of considerable evolutionary divergence between prokaryotic and eukaryotic enzymes thus making them ideal candidates whose inhibition will not likely affect human enzymes [73], [74]; (ii) development of potent and broad spectrum antibiotics, as AaRs are highly conserved among pathogenic bacteria [73], [74]; (iii) the full complement of 20 synthetases is found in most bacterial pathogens and may represent 20 independent antibacterial targets [73], [74]; (iv) these enzymes are soluble, stable, and easy to purify in large quantities from recombinant expression systems, and can be assayed by one or more conventional methods amenable to high-throughput screening [75], [76]; and (v) the x-ray crystal structures for most of the synthetases are known from several bacteria, and provide a platform for rational drug design [73], [77].

We have highlighted three M. ulcerans AaRS as drug targets: arginyl-tRNA synthetase (ArgS), histidyl-tRNA synthetase (HisS), and phenylalanyl-tRNA synthetase (PheS) (Table S5). Although 18 AaRS were identified as essential for M. ulcerans (Table S3), 15 of these showed a considerable level of homology with human AaRS, and were therefore not included in further analysis. Currently, only one AaRS inhibitor, mupirocin, which inhibits bacterial IleS, is marketed as an antibacterial agent [78]. However, the IleS of M. ulcerans showed 48% identity with human IleS, and resistance to this agent is also widely reported; it may therefore not be an ideal choice. Recently, phenyl-thiazolylurea-sulfonamides have been identified as successful inhibitors of PheS in Escherichia coli, Haemophilus influenzae, S. pneumoniae, and S. aureus, showing high potency, broad-spectrum activity and selectivity for bacterial PheS versus the corresponding mammalian cytoplasmic and human mitochondrial enzymes [79]. It has also been observed that inhibition of these enzymes leads to disruption of protein biosynthesis, in turn resulting in the attenuation of bacterial growth under both in vitro and infectious conditions [80]. We propose therefore, that inhibition of ArgS, HisS, and PheS can lead to disruption of M. ulcerans protein synthesis with no side effects for its human host. It would be worthwhile to evaluate the antibacterial activity of phenyl-thiazolylurea-sulfonamides against the PheS of M. ulcerans.

Folate biosynthesis.

Folate biosynthesis is an important biochemical pathway whose enzymes have been targeted since 1930 as key for antimicrobial therapy. Inhibition of two enzymes in this pathway, dihydropteroate synthase (FolP) and dihydrofolate reductase (DfrA), has been widely used for treatment of infections caused by bacteria such as Pneumocystis carinii [81], T. gondii [82], and protozoan parasite P. falciparum [83]. Recently, we have also reported that the DfrA of M. genitalium can serve as a potential therapeutic target [13]. The therapeutic potential of targeting the folate biosynthesis pathway of M. ulcerans is not well elucidated. The folate biosynthesis pathway of M. ulcerans comprises of 11 proteins. Among these, we identified four enzymes; DfrA, FolP, dihydroneopterin aldolase (FolB), and 6-pyruvoyl tetrahydrobiopterin synthase (PTPS) as non-host plus essential (Table S3). DfrA is a ubiquitous enzyme that is responsible for the reduction of dihydrofolate to tetrahydrofolate, an important co-factor in the biosynthesis of thymine. Inhibition of DfrA leads to cell death through lack of thymine as the cells have no alternative [84]. FolP catalyzes a condensation reaction yielding dihydropteroate, an intermediary metabolite subsequently converted to tetrahydrofolic acid, and essential for the syntheses of purine, thymidylate, glycine, methionine, pantothenic acid, and N-formylmethionyl-tRNA. In addition, it is an identified drug target in P. falciparum [83] and T. gondii [82]. As folate biosynthesis is a common pathway, significant homology between host and pathogen proteins can occur. However, none of the identified drug targets had human homolog. We suggest therefore that the DfrA and FolP of M. ulcerans hold strong therapeutic potential worthy of experimental follow-up.

Histidine biosynthesis.

12 proteins from the M. ulcerans genome mapped to the histidine metabolism pathway, and out of these, eight were identified as non-host plus essential (Table S4). Druggability analysis also revealed several drug-like compounds that can be tested as inhibitors of the ATP phosphoribosyl transferase (HisG) and Histidinol dehydrogenase (HisD) of M. ulcerans (Table S5). HisG catalyzes the first committed step in histidine biosynthesis, condensation of ATP with phosphoribosyl pyrophosphate to produce phosphoribosyl ATP and inorganic pyrophosphate, leading to intermediates that play a role in purine biosynthesis. HisD regulates the final oxidation step of histidine synthesis. The therapeutic potential of HisG, HisD, and other enzymes from this pathway has been reported previously from comparative genomics in M. tuberculosis [85], M. leprae [19], P. aeruginosa [11], and S. aureus [21]. These predictions have recently been confirmed in a study in which nitrobenzothiazole-containing compounds were identified as successful inhibitors of M. tuberculosis HisG [86]. Similarly, in another recent study, several inhibitors of S. aureus histidine biosynthesis were computationally predicted and then confirmed experimentally [87]. These findings collectively highlights the therapeutic potential of the histidine metabolism pathway. We thus propose that M. ulcerans histidine metabolism pathway enzymes can serve as novel drug targets.

In addition to the drug targets discussed above, we identified several other novel targets from the common and unique pathways of M. ulcerans (Table S4). Homoserine dehydrogenase (ThrA) and homoserine kinase (ThrB), for example, are two enzymes involved in the glycine, homoserine, threonine, cysteine, and lysine metabolism pathways of M. ulcerans (Table S4). The therapeutic potential of both these enzymes has been confirmed in a recent study of M. tuberculosis [85]. In M. ulcerans, we identified both these enzymes as cytoplasmic, and non-host plus essential, suggesting that their inhibition holds therapeutic potential, as their successful disruption has been shown to deprive the pathogen of essential nutrients and synthesis of necessary components such as cell wall PG.