Small-Molecule Inhibitors of the Pseudaminic Acid Biosynthetic Pathway: Targeting Motility as a Key Bacterial Virulence Factor

Abstract

Helicobacter pylori is motile by means of polar flagella, and this motility has been shown to play a critical role in pathogenicity. The major structural flagellin proteins have been shown to be glycosylated with the nonulosonate sugar, pseudaminic acid (Pse). This glycan is unique to microorganisms, and the process of flagellin glycosylation is required for H. pylori flagellar assembly and consequent motility. As such, the Pse biosynthetic pathway offers considerable potential as an antivirulence drug target, especially since motility is required for H. pylori colonization and persistence in the host. This report describes screening the five Pse biosynthetic enzymes for small-molecule inhibitors using both high-throughput screening (HTS) and in silico (virtual screening [VS]) approaches. Using a 100,000-compound library, 1,773 hits that exhibited a 40% threshold inhibition at a 10 μM concentration were identified by HTS. In addition, VS efforts using a 1.6-million compound library directed at two pathway enzymes identified 80 hits, 4 of which exhibited reasonable inhibition at a 10 μM concentration in vitro. Further secondary screening which identified 320 unique molecular structures or validated hits was performed. Following kinetic studies and structure-activity relationship (SAR) analysis of selected inhibitors from our refined list of 320 compounds, we demonstrated that three inhibitors with 50% inhibitory concentrations (IC₅₀s) of approximately 14 μM, which belonged to a distinct chemical cluster, were able to penetrate the Gram-negative cell membrane and prevent formation of flagella.

INTRODUCTION

Infections caused by bacteria continue to represent major challenges to health care in both the hospital environment and the community setting. Microbial resistance to antibiotics has increased to the point where the current arsenal of antibacterial drugs is inadequate, and on many occasions, bacterial resistance to these drugs can lead to life-threatening infection (1, 2). The development of new, effective antimicrobials is clearly needed, and targeting bacterial virulence has gained considerable attention recently as an alternative approach to identify novel antibacterial therapeutic agents (3–6). Depriving pathogenic bacteria of their virulence functions might prevent the establishment of infection and allow the host immune system sufficient time to facilitate clearance of the organism. As virulence-targeted drugs are organism specific, they are less likely to impact host commensal flora. As such, the risk of opportunistic infections due to alterations in the host microbiome would be greatly reduced. New antivirulence drugs might also be used in combination with existing antibiotics to improve efficacy in current treatment strategies.

Helicobacter pylori is a significant gastrointestinal pathogen responsible for chronic active gastritis, peptic ulcers, and related gastric cancers (7). The current established treatments for H. pylori infection are numerous and include triple and quadruple therapy, both of which utilize two antibiotics (metronidazole, amoxicillin, tetracycline, or clarithromycin) in addition to either a proton pump inhibitor (PPI) (triple therapy) or a PPI and bismuth (quadruple therapy). The efficacies of these treatment strategies have been severely hampered in recent years due to the rise in antibiotic resistance of H. pylori isolates and are now at the point where the current rate of eradication has dropped below 70% in many countries (8). Thus, there is a clear need to develop alternative therapeutic strategies for the management of H. pylori-related disease.

It has been clearly established that motility is a critical virulence factor for H. pylori infections (9–13). This motility, observed under conditions of elevated viscosity (as found in the gastric lumen), is due to a unipolar bundle of sheathed flagella, the structural filaments of which are composed of two flagellin protein species, FlaA and FlaB. To infect the stomach, the bacteria must first transit the mucus layer from the gastric lumen, with the final destination being the epithelial surface, which is the site of infection. The directed motility of cells is essential to this process as H. pylori colonizes the interface of separate mucosa (antral and fundic) in the stomach, and the organism must continually seek out this niche as conditions vary between fasting and feeding (14). Importantly, in addition to being required for initial colonization of the stomach, motility has also been shown to be required for robust, long-term, persistent infections (11, 12, 15).

In previous studies, we demonstrated that the structural flagellin proteins from H. pylori and Campylobacter jejuni are glycosylated with the novel “sialic acid-like” nonulosonate sugar, pseudaminic acid (Pse). Targeted gene disruption of the Pse biosynthetic pathway genes showed that this glycosylation is essential for flagellar filament assembly and consequent motility (9, 16). The H. pylori Pse pathway isogenic mutant strains were unable to colonize the stomach in a mouse model of infection, and C. jejuni Pse isogenic mutant strains were attenuated in the ferret diarrheal disease model (9, 17). Pseudaminic acid derivatives are also found in a number of other bacterial species as components of cell surface glycans such as lipopolysaccharide (LPS) O antigens, capsular polysaccharides, and pili, and in many examples, these surface glycans are essential for bacterial virulence (18–21).

With Pse being a key virulence factor as well as a unique product made by microorganisms, the Pse biosynthetic pathway offers potential as a novel therapeutic target. The Pse biosynthetic pathways from H. pylori and C. jejuni have been elucidated, and the function of each of the pathway's five biosynthetic enzymes has been determined following recombinant production and purification of each biosynthetic enzyme (22–25). In addition, it has been demonstrated that all five Pse pathway enzymes can be combined in a single one-pot reaction for the synthesis of Pse using UDP-GlcNAc as an initial substrate (22). Structural studies of three of the biosynthetic enzymes have also been completed (26–28).

The observation that glycosylation of the flagellin structural proteins is required for flagellar assembly and subsequent motility, in addition to the extensive body of work characterizing the novel bacterial pseudaminic acid biosynthetic pathway, has set the groundwork for small-molecule inhibitor screening of this key H. pylori/C. jejuni virulence factor. In this study, we have identified small-molecule hits from high-throughput screening (HTS) and virtual screening (VS) campaigns. We disclose a subset of chemically related small-molecule lead compounds that inhibit H. pylori and C. jejuni Pse biosynthetic pathway enzymes and prevent formation of flagella in cell-based assays with C. jejuni.

MATERIALS AND METHODS

HTS library.

The screened HTS library consists of 96,116 pure chemical compounds acquired from the commercial HTS libraries of ChemDiv Inc. (San Diego, CA) and Tripos International (St. Louis, MO), which were selected based on their chemical diversity and drug-like properties.

Pse pathway enzymes.

Recombinant production and purification of H. pylori and C. jejuni Pse biosynthetic enzymes were as previously described (22, 24), and recombinant plasmids are listed in Table S1 in the supplemental material. Prior to assays, purified proteins were dialyzed against 20 mM HEPES [pH 7.2] and 50 mM NaCl.

Phosphate-based primary screening assay.

For HTS in 384-well plates, the reaction volume was 10 μl per well. A substrate master mix of 7.26 μl (containing 0.5 mM UDP-GlcNAc, 0.5 mM pyridoxal phosphate [PLP], 7 mM l-Glu, 0.5 mM acetyl-coenzyme A [CoA], and 0.5 mM phosphoenolpyruvate [PEP]) was combined with 2.74 μl of enzyme mix at the concentrations indicated in Table S2 in the supplemental material. Of note, the PseB cofactor NADP⁺ does not have to be added exogenously as it remains tightly bound to the enzyme during purification (24). Each well contained 10 to 20 μM library compound, and the reaction mixture was incubated for 60 min at 37°C with 95% relative humidity (RH). P_i was detected by addition of 40 μl of ALS reagent (P_iColorLock assay kit; Innova Biosciences) and incubation for 5 min, and then the reaction was stopped by addition of 4 μl of stabilizer. The final optical density at 595 nm (OD₅₉₅) reading was done on a PerkinElmer Envision plate reader.

RapidFire HTMS secondary screening.

For secondary screening, each of the substrates (PseC, PseH, PseG, and PseI) was prepared enzymatically, as these reagents are not available commercially. Typically, reactions were performed in a mixture containing 20 mM HEPES [pH 7.2], 50 mM NaCl, 15 mM UDP-GlcNAc, 1 mM PLP, 75 mM l-Glu, and 17 mM acetyl-CoA (as appropriate), using approximately 5 mg of each enzyme (as appropriate) with 10- to 15-ml volumes. Reaction mixtures were incubated at 37°C and upon completion were filtered through an Amicon Ultra-15 (10,000 molecular weight cutoff) filter membrane to remove enzymes. Each product was quantitated using the molar extinction coefficient of UDP (ε₂₆₀ = 10,000) to measure total UDP-containing species and capillary electrophoresis to measure the relative abundances of specific UDP species. In addition, PseH and PseG substrates were also purified using methods described for metabolite purification (29). Pse enzyme products were typically in the range of 7 to 13 mM final concentrations, where these stock substrate solutions were used for screening individual Pse enzymes (i.e., secondary screening). High-throughput mass spectroscopy (HTMS) reactions were carried out in 384-well plates with final reaction volumes of 10 μl per well. The reaction conditions for each assay are provided in Table S3 in the supplemental material. Reactions were stopped by addition of 30 μl of stop solution (acetonitrile-water-formic acid, 20:80:0.2), and then 25 μl of the stopped reaction mixture was transferred from the 384-well assay plates onto 4 × 96-well plates containing 175 μl of stop solution. HTMS plates were sealed, centrifuged, and kept at 4°C until processed by the RapidFire 200 (Agilent) coupled to an Agilent 6410B triple-quadrupole (QQQ) mass spectrometer. Samples (10 μl) were delivered directly from 96-well plates to the Graphitic Carbon cartridge to replace the nonvolatile buffer with H₂O in a 2.5-s wash cycle at a flow rate of 1.5 ml/min. The substrate and product were coeluted to the mass spectrometer in 4 to 8 s using acetonitrile-water-ammonium hydroxide (70:30:0.02) at a flow rate of 0.9 ml/min. The chromatograph system produced baseline-resolved peaks. The eluted sample passed directly into the mass spectrometer ion source under the negative ion mode for quantitative analysis. The other parameters for each analyte are given in Table S4 in the supplemental material. The concentration of each inhibitor in the HTMS screening reactions was 30 μM.

Chemical clustering.

Chemical clustering of compounds was carried out in SYBYL 8.1.1 (Tripos International) by hierarchical clustering. Two-dimensional (2D) fingerprints, which are binary variables for the presence (1) or absence (0) of specific fragments, and ATOM_PAIR_FP, which are fingerprints describing the minimum path lengths between atoms in molecules, were used as clustering descriptors of molecular structure at 2D level. Hierarchical clustering was complete, that is, taking the intercluster distance to be the greatest separation between their elements, thus producing dendrograms with multiple, compact root clusters and minimizing the generation of singletons.

PseB kinetic assay.

Reactions were carried out in 384-well plates in 20 mM HEPES, 20 mM NaCl [pH 7.2], and 1% dimethyl sulfoxide (DMSO), with a final reaction volume of 50 μl per well. PseB at a concentration of 2 μM was preincubated with library compounds at 100, 66.7, 44.4, 29.6, 19.8, 13.2, 8.8, 5.9, 3.9, 2.6, 1.7, or 1.2 μM for 10 min at 25°C. The reaction was started by addition of 200 μM UDP-GlcNAc. After incubation for 30 min, the reactions were stopped by transferring 4 μl of each reaction mixture into 195 μl of stop solution (acetonitrile-water-formic acid, 20:80:0.2) in 96-well plates. The samples were than analyzed by the RapidFire 200 coupled to the mass spectrometer.

Cell-based assays.

C. jejuni 81-176 and C. jejuni 81-176 pseB::Cm were grown overnight on Mueller-Hinton (MH) agar under microaerophilic conditions at 37°C. Cells were harvested from an agar plate into Mueller-Hinton broth, and this suspension was used to inoculate a well with 1 ml of Mueller-Hinton broth containing 0.01% DMSO in a Falcon multiwell 6-well plate to an OD₆₀₀ of 0.1. Inhibitors (in DMSO) were added to wells to final concentrations as specified. The multiwell plates were incubated with shaking (200 rpm) at 37°C under microaerophilic conditions for 7 h, the OD₆₀₀ was measured, and cells were harvested by centrifugation. Cells were fixed overnight in 3% formalin in phosphate-buffered saline (PBS) and washed in PBS, and the OD₆₀₀ was adjusted to 0.08 for coating on 96-well plates for enzyme-linked immunosorbent assays (ELISAs). C. jejuni strains 81-176 and the 81-176 pseB::Cm grown in multiwell plates in identical fashions to inhibitor test samples were used as positive and negative controls for ELISAs. Assays were completed on two independent occasions.

ELISAs.

Nunc MaxiSorp plates were coated with 100 μl of formalin-fixed cells overnight at 37°C. Plates were blocked (1% bovine serum albumin [BSA] in PBS) and then washed 3 times with PBS-0.05% Tween (PBS-T). A His-tagged single-domain antibody (sdAb) specific for 81-176 flagellin (a gift from M. Arbabi, National Research Council) (30) was then added (1:1,000 in PBS-BSA, 0.85 mg/ml), and the plates were incubated for 2 h at room temperature (RT). The plates were washed 3 times (PBS-T) and then incubated with rabbit anti-His horseradish peroxidase (HRP)-conjugated antibody (1:10,000 in PBS-BSA, 1 mg/ml) for 1 h at RT. Following washing, the antibody was detected with 3,3′,5,5′-tetramethylbenzidine (TMB) for 10 min, and the reaction was stopped with 1 M H₃PO₄. The samples were analyzed in triplicate, and the absorbance was measured at 450 nm.

Inhibitor docking and virtual screening.

The 1.85-Å resolution crystal structure of PseG in complex with UDP (Protein Data Bank [PDB] code 3HBN) and the 1.9-Å resolution crystal structure of PseB in complex with UDP-GlcNAc (PDB code 2GN4) were used for virtual screening and inhibitor docking after removal of water, substrate/product molecules, and cosolvent/buffer molecules and addition of hydrogen atoms according to standard ionization states. In the case of PseB, the NADP⁺ cofactor and two subunits of the hexamer were retained from the crystal structure as they are essential for shaping the PseB substrate-binding cleft.

Virtual screening was carried out on a library of 1.6 million commercially available drug-like compounds from the ZINC database (31). We used a high-throughput VS docking-scoring pipeline (32, 33). The exhaustive docking program Wilma (32) within the VS pipeline was used with default increment parameters and the WilmaScore1 energy function (34). The ligand conformations were generated by Omega (OpenEye, Inc., Santa Fe, NM) and controlled by setting the internal energy cutoff to 20 kcal/mol and adjusting the pose clustering parameter to produce at most 5,000 conformations. Wilma-generated poses were refined by constrained energy minimization as described previously (32–34) prior to binding affinity scoring with the solvated interaction energy (SIE) function (35) using default parameters for the electrostatic and nonpolar contributions to interaction and desolvation energies (34, 36). Pose selection for a given inhibitor against a given enzyme variant was based on the lowest SIE score. These SIE scores were then used for binding affinity ranking among various inhibitor-enzyme complexes. Top-scored 200 complexes in the case of PseG and 357 in the case of PseB were subjected to SIE averaging on molecular dynamic trajectories using SIETRAJ as described previously (34, 36, 37) to obtain the final binding affinity scores.

RESULTS

Development of an optimized HTS assay for Pse biosynthetic enzymes.

The Pse biosynthetic pathway is outlined in Fig. 1. Previous work had demonstrated that the synthesis of Pse might be accomplished by combining all five enzymes of the pathway in a single reaction with UDP-GlcNAc and necessary cofactors. Since the last enzymatic step results in the release of inorganic phosphate, the entire pathway can be screened for inhibitors simultaneously using a phosphate detection-based assay. As such, we first developed a 384-well plate assay using the 5 enzymes of the Pse pathway and measuring the release of free phosphate with the P_iColorLock ALS reagents (Innova Biosciences) according to the manufacturer's instructions. The assay conditions (relative enzyme concentrations, stoichiometry, reagent concentrations, incubation times, etc.) were optimized to allow identification of hits that target any one of the five Pse enzymes (see Materials and Methods). The final concentrations of each H. pylori enzyme are provided in Table S2 in the supplemental material. In a preliminary screening of >20,000 compounds, these conditions led to a Z′ factor of 0.88, indicating that the assay was robust and suitable for HTS (see Fig. S1 in the supplemental material).

FIG 1.

The pseudaminic acid biosynthetic pathway in H. pylori and C. jejuni. Biosynthetic intermediates are UDP-GlcNAc (I), UDP-2-acetamido-2,6-dideoxy-β-l-arabino-hexos-4-ulose (II), UDP-4-amino-4,6-dideoxy-β-l-AltNAc (III), UDP-2,4-diacetamido-2,4,6-trideoxy-β-l-altropyranose (IV), 2,4-diacetamido-2,4,6-trideoxy-l-altropyranose (V), and pseudaminic acid (VI).

Using this assay, we have completed in duplicate an HTS of a library comprising approximately 100,000 small-molecule compounds in order to identify inhibitors for the 5 enzymes of the pathway (Fig. 2). This screening was carried out at the compound concentration of 10 μM. The quality of the screen allowed us to establish 40% inhibition as a threshold for the identification of hits. This led to a hit list of 1,773 unique compounds from two independent primary screens (within the red rectangle in Fig. 2). Chemical similarity-based clustering of these hits revealed that a certain fraction of them comprises derivatives of a small number of chemical scaffolds. For example, even at a deep level of clustering corresponding to 500 clusters, there were 28 clusters, each containing at least 10 close analogs, with 7 clusters represented by at least 20 congeners.

FIG 2.

HTS screening using a 5-enzyme primary assay. The HTS library of 96,116 compounds was screened twice against a mixture of H. pylori Pse pathway enzymes with concentrations and assay conditions as shown in Table S1 in the supplemental material. The area enclosed in the red rectangle focuses on 1,847 hits (corresponding to 1,773 unique compounds) showing >40% inhibition in both screens. Note that a result of 3 standard deviations from mean inhibition corresponds to inhibition thresholds of at least 41% and 42% in screens 1 and 2, respectively.

Secondary screening hit validation.

It was important to confirm the primary hits obtained with a secondary assay using a different technique. To this end, we next used label-free high-throughput mass spectroscopy (HTMS) screening with RapidFire technology both to confirm the primary hits and to identify which of the 5 enzymes was the respective target of each of those primary hits. The RapidFire HTMS incorporates automated sample handling, an in-line solid-phase extraction (SPE) cartridge system for sample cleanup and analyte concentration, and an injection system coupled to a triple-quadrupole mass spectrometer. The typical throughput of the instrument is approximately 500 samples per hour. The secondary HTMS screens were carried out against each of the five enzymes from the Pse pathway, separately or in combination.

RapidFire HTMS methods were established for the simultaneous detection of multiple reaction monitoring (MRM) signals of substrate and product for individual enzyme, giving no signal cross talk between the channels, thus allowing the percentage of conversion to be calculated. Since all of the substrates and products in the Pse pathway are hydrophilic (Fig. 1), solid-phase cartridges of hydrophilic interaction liquid chromatography (HILIC) (type Z) and graphitic carbon (type D) were evaluated, and the graphitic carbon cartridge gave the better performance and was chosen for the analysis. The sampling and running parameters were optimized so as to give the best throughput without compromising the sensitivity or reproducibility of detection. The mass spectrometric parameters for each compound and the linear ranges of all substrates and products are provided in Table S4 in the supplemental material.

Enzymatic characterization of the recombinant purified enzymes PseC, PseH, and PseG was performed to develop optimal assay conditions for inhibitors. For the enzymes PseB and PseI, the conditions developed in the HTS assay were used. The reaction conditions are given in Table S3 in the supplemental material.

The HTMS assay was used for screening of 1,773 unique hits identified from the primary screening. Single-concentration screening was performed at a 30 μM compound concentration. The compounds were screened in a total of 155 96-well plates for each of the 5 enzymes. The Z′ scores for individual plates were calculated using the formula (38)

$${\text{Z}}^{\prime }=1-\left[\frac{3\times ({\text{SD}}_{\text{PosCtrl}}+{\text{SD}}_{\text{NegCtrl}})}{{\text{Avg}}_{\text{PosCtrl}}-{\text{Avg}}_{\text{NegCtrl}}}\right]$$

where Avg and SD are the averages and standard deviations, respectively, of the inhibition signals produced by the positive and negative controls (PosCtrl and NegCtrl, respectively) on each plate. Thus, Z′ is reflective of both the assay signal dynamic range and the data variation associated with the signal measurements and is used for comparison and evaluation of the quality of assays. A high-throughput assay with a 1 > Z′ > 0.5 score is considered an excellent assay (28). Therefore, results from the plates with Z′ >0.5 were analyzed. The hit selection for each enzyme was based on the Z′ scores obtained for that particular enzyme. Therefore, hit selection was not necessarily based on the level of inhibition by a particular compound but on how many standard deviations (SD) that inhibition level deviated from the mean inhibition across all tested hits (Fig. 3). The following thresholds for hit selection were established: >1 SD for PseB and coupled PseB/PseC assays, >2 SD for PseG, PseH, and PseI, and more than 3 SD for PseC. This led to 169 hits for PseB, 25 hits for PseC, 89 hits for the coupled PseB/PseC assay, 118 hits for PseG, 89 hits for PseH, and 100 hits for PseI. Because some of these hits were found to inhibit more than one enzyme, the validated set of Pse pathway inhibitors thus comprises a total of 320 compounds that inhibit at least one of the five enzymes in the pathway.

FIG 3.

HTMS-based enzyme-specific secondary screening. The level of inhibition of each H. pylori Pse-pathway enzyme is plotted for the 1,847 hits (corresponding to 1,773 unique compounds) from the primary screens and color coded according to the number of standard deviations (SD) from the mean for each screened enzyme (see legend on the right-hand side).

We repeated chemical clustering for these 320 unique molecular structures and obtained 146 singletons and 174 compounds clustered into 22 clusters with at least 2 compounds, the most populated cluster reaching 40 compounds (Fig. 4). The fraction of compounds that were found to inhibit more than one enzyme in the pathway varies with the chemical nature of the scaffolds representing the clusters. Overall, 53.4% of clustered compounds inhibit more than one enzyme, which is similar to the fraction of 52.7% for multienzyme inhibitors found among the remaining unclustered compounds (singletons). Most of these Pse enzyme inhibitors, however, act on only two or three Pse enzymes, with only 12 compounds exhibiting statistically significant inhibition of all 5 enzymes. This helps to address some concern that multienzyme inhibitors might act generally as nonspecific promiscuous ligands or aggregators (39, 40). In order to further address this issue, we have taken 106 hits that act on at least 2 Pse enzymes and queried the historical data acquired from HTS campaigns that we have carried out with the same compound library against various targets unrelated to Pse enzymes (9 enzyme and 4 nonenzyme targets). We found that the vast majority of these multi-Pse enzyme inhibitors were not hits against any of these unrelated targets, and only 4 compounds were among the hits for 2 of these 13 unrelated targets (see Table S5 in the supplemental material).

FIG 4.

Clustering of validated hits from secondary screening. The 320 compounds identified after HTMS-based enzyme-specific screening were clustered by 2D chemical similarity. With two exceptions, these clusters contain at least a fraction of the compounds that inhibit more than one enzyme in the pathway.

Structure-based virtual screening.

In parallel with the HTS screening of an actual compound library, we took advantage of some of the structural information available for enzymes in the Pse biosynthetic pathway. We carried out VS of a virtual library of 1.6 million compounds against the crystal structures of PseB and PseG, using a proprietary VS pipeline that demonstrated some of the better performances in blind tests (32, 33). We acquired 38 virtual hits from the PseG screening and 42 virtual hits from the PseB screening and tested them in the 5-enzyme assay described earlier. Four of the PseB virtual hits showed reasonable inhibition of the pathway at a 10 μM inhibitor concentration (see Table S6 in the supplemental material). The PseB inhibition was further confirmed in the secondary HTMS assay, with degrees of inhibition above the noise level established from the HTMS-based secondary screening experiments. Interestingly, three of these compounds were also found to significantly inhibit PseG, possibly suggesting certain mimicry by these compounds of the similar UDP-sugar substrates of PseB and PseG. The comparisons of the docked poses of these inhibitors with the actual binding mode of UDP-GlcNAc in the PseB active site appear to support this hypothesis (Fig. 5; see also Fig. S2 and S4 in the supplemental material). Given that the in silico screening did not utilize structural information about the substrate chemical nature and its binding mode and relied solely on the structure of the binding site of the target enzyme, it is a measure of success for the computational method that binding modes of confirmed VS hits display elements seen in the experimentally determined binding modes of known ligands or substrates. The four confirmed VS hits display inhibition potencies (see Table S6 in the supplemental material) which are somewhat weaker than those for the best confirmed HTS hits that were progressed (Fig. 6 and 7) but nevertheless within the range of confirmed HTS hits, ranking 79, 103, 121, and 145 compared against the set of 169 HTMS-confirmed PseB hits.

FIG 5.

Example of confirmed virtual screening hit. The chemical identification is given as the identification code in the ZINC database (http://zinc.docking.org). In vitro data are from 5-enzyme pathway and HTMS enzyme-specific assays. The image on the right depicts a docked pose of the inhibitor (stick model with C atoms in white) overlaid with the actual binding mode of the UDP-GlcNAc substrate (stick model with C atoms in purple) in the PseB active site located at the interface of the two enzyme molecules (shown as ribbons of different colors; PDB code 1GN4).

FIG 6.

Structure-activity data for select PseB inhibitors. The chemical structures and IC₅₀ values for H. pylori PseB inhibition are shown in the left column. The corresponding dose dependencies are presented in the right column.

FIG 7.

Correlation between inhibition of the PseB enzymes from H. pylori and C. jejuni. Data shown are for the 5 inhibitors (30 μM) selected for kinetic analysis and molecular docking (Fig. 6). Squares denote cluster 5 inhibitors and circles are for cluster 6 analogs.

Kinetic studies and SAR of select PseB inhibitors.

We next focused our attention on inhibitors of PseB, the first enzyme in the pathway. The reasons to focus on PseB were that (i) it is the first enzyme in the pathway and hence it may lead to more effective inhibition, (ii) it has a solved crystal structure, and (iii) it may act as a kinetic bottleneck based on the optimized conditions for the pot assay (see Table S2 in the supplemental material). Hence, the remaining hits (non-PseB inhibitors) were deemed less attractive at the moment. From the PseB hits, we selected 20 hits from the top 50 hits that had at least 3 SD above the average inhibition in the screen and that were also present in the 6 most populated chemical clusters (Fig. 4). The rationale was to allow retrospective structure-activity relationships (SAR) for progressed compounds that will inform not only on the importance of various substituents but also on compound selectivity and possible promiscuity. This is not to say that other hits are not worth pursuing in the future; these are still under evaluation. Kinetic studies were completed with the PseB enzyme for each of the 20 inhibitors using the RapidFire HTMS PseB assay (see above) where substrate and product were quantified. Of the 20 compounds tested, 5 showed good dose-dependent inhibition behaviors, with 50% inhibitory concentration (IC₅₀) values ranging from 12 μM to 72 μM (Fig. 6). These compounds have drug-like physicochemical properties for oral bioavailability, albeit three of them violate slightly Lipinski's rule of five (LR5) (41) in terms of molecular weight and hydrophobicity. Nonetheless, these five compounds are well within the drug-likeliness ranges for the number of rotatable bonds and the polar surface area, which together have been found to be better predictors of oral bioavailability than the LR5 rule of thumb (42) (see Table S7 in the supplemental material).

Following chemical clustering analysis of the top 320 confirmed hits against PseB (Fig. 4), we found that these five PseB inhibitors belonged to two distinct chemical clusters but nevertheless shared a related core substructure (Fig. 6). Inhibitors CD09463 and CD23703 belong to cluster 6 and are both substituted N-phenyl-2,5-dimethylpyrroles, differing only in the position of the hydroxyl and carboxylate substituents of the phenyl ring. Inhibitors CD26389, CD24868, and CD36508 belong to cluster 5 and are almost double in size relative to the sizes of the two cluster 6 inhibitors. One can easily notice that these larger inhibitors bear the core substructure N-phenyl-2-pyrrolidone which is structurally similar to the N-phenyl-2,5-dimethylpyrrole structure of the smaller inhibitors. The substitution pattern on the phenyl ring of the core structure is also similar, with CD23703 from cluster 6 matching CD24868 and CD26389 from cluster 5 and with CD09463 from cluster 6 matching CD36508 from cluster 5. Hence, one can expect that the core substructure affords most of the binding affinity to PseB, while the larger substituents of the cluster 5 compounds (at positions 3 and 5 of the unsaturated 2-pyrrolidone ring) have a smaller, albeit favorable, contribution to the binding affinity and hence enzyme inhibition. Molecular docking of these inhibitors in the PseB substrate-binding cleft supports these structure-activity relationship data (Fig. 6; also see Fig. S3 in the supplemental material). Additional analog analysis based on secondary screening of 10 congeneric compounds from cluster 5 and 9 congeneric compounds from cluster 6 helped better define the Pse target specificity (see Tables S8 and S9 in the supplemental material). It appears that cluster 5 analogs are more target specific and afford statistically significant inhibition levels only for PseB and/or PseG. In contrast, some of cluster 6 analogs are pan-Pse enzyme inhibitors, and inhibition levels afforded by cluster 6 compounds also tend to correlate across various Pse enzymes. The SAR analysis helps to highlight the importance of various substituents around the common core structure of the progressed hits (a more detailed analysis is provided in the supplemental material).

Inhibition of Campylobacter jejuni PseB enzyme.

As H. pylori flagella produced at the cell surface are covered with a membranous sheath, detection of the assembled flagellin filaments at the cell surface presents a challenge. In contrast, the assembled flagellin filaments on Campylobacter cells are not covered with this membranous sheath, which permits the detection of assembled filaments at the cell surface with flagellin-specific immunological reagents. To allow us to examine the top five PseB compounds identified above in a cell-based assay, it was necessary to confirm that the inhibitors selected were active for the PseB enzyme from the related organism C. jejuni. Four of the compounds indeed inhibited C. jejuni PseB, albeit at levels slightly lower than the inhibition of H. pylori PseB (Fig. 7). However, one of the cluster 5 compounds (CD36508) appeared not to inhibit C. jejuni PseB in vitro at the concentration tested, which indicates weaker inhibitory potency at best.

Cell-based assay.

To demonstrate the efficacy of Pse inhibitors which were identified by secondary screening, we next developed a cell-based assay (ELISA) to measure flagellin production on bacterial cells when grown in the presence of an inhibitor. We focused these assays on compounds identified as PseB inhibitors. As indicated above, the assay uses C. jejuni 81-176, which has been shown to also glycosylate the flagellin structural proteins with pseudaminic acid derivatives (16). The flagellin proteins are detected on the cell surface using a single-domain antibody (sdAb) specific for the C. jejuni 81-176 flagellin protein (30). The three compounds of the same structural cluster 5 (Fig. 6) were found to inhibit flagellin production in a dose-dependent manner. Inhibition was observed at compound concentrations of ≥100 μM (Fig. 8). Growth of C. jejuni was monitored by measuring OD₆₀₀ values, and in each case was unaffected by the presence of inhibitors (data not shown). Cell-based confirmation of activity in future studies might be carried out either by overexpression of the target enzyme and titration of cellular effects or by metabolomics-based analysis of cytosolic extracts by capillary electrophoresis-mass spectrometry (CE-MS) with precursor ion scanning and screening for CMP-Pse sugar as described in our earlier work (43).

FIG 8.

Inhibition of C. jejuni flagellin production. Whole-cell ELISAs using the flagellin-specific antibody FlagV1. ■, C. jejuni 81-176 (81176) and C. jejuni 81-176 pseB::Cm (1293) controls grown in MH broth with DMSO; □, C. jejuni 81-176 grown in MH broth with variable concentrations of inhibitors as indicated. (A) Inhibitor concentrations of 500 μM, 50 μM, 500 nM, 50 nM, and 5 nM; (B) inhibitor concentrations of 500 μM, 250 μM, 100 μM, and 50 μM.

DISCUSSION

Due to significant drug resistance, it is becoming increasingly difficult to eradicate H. pylori, and as such, the need for new, effective therapeutic regimes to treat H. pylori infections is globally recognized (44). Here, we present the results of a combined HTS and in silico approach to identify inhibitors of the pseudaminic acid biosynthetic pathway enzymes. Development of a pathway screening assay offered a considerable advantage over parallel screenings of individual enzymes. The efficiency of screening multiple enzymes simultaneously reduced the total number of assays required in the HTS and also permitted the use of a commercially available substrate (UDP-GlcNAc) rather than the more unique sugar pathway intermediates which are not available commercially and would require more costly enzymatic or chemical synthesis strategies. The optimized HTS 5-enzyme assay was used to identify 1,773 unique compounds, and secondary HTMS was used to confirm the primary hits and to identify the respective inhibitor enzyme targets. From these analyses, a validated set of 320 compounds have been identified as inhibitors of at least one Pse biosynthetic enzyme.

Analysis of these validated hits indicated the presence of several chemical clusters with close congeners, which can be used to generate structure-activity relationship (SAR) data valuable for further optimization efforts. Interestingly, with two exceptions (clusters 14 and 20) (Fig. 4), these clusters include analogs that inhibit more than one enzyme in the pathway. Inhibiting the biosynthetic pathway simultaneously at several points can be beneficial not only for increased efficacy and specificity but also for circumventing eventual bacterial resistance. One may speculate that the discovery of multienzyme inhibitors is a reflection of the substrate similarities shared by the component enzymes in the pathway. It is also possible that multienzyme inhibition is a direct consequence of the HTS strategy developed here, as it originates in the 5-enzyme one-pot assay used during the primary screening.

The kinetic analysis combined with molecular docking of five validated PseB inhibitors provided valuable SAR data (Fig. 6, see also Fig. S3 in the supplemental material). The common core substructure of these analogs (two smaller cluster 6 inhibitors and three larger cluster 5 inhibitors) interacts tightly and complementarily into the deep pocket of the binding site that accommodates the pyrophosphate and the sugar moieties of the sugar-UDP substrate. However, the large substituent at position 3 of the unsaturated 2-pyrrolidone ring of the larger cluster 5 analogs is solvent exposed and makes minimal contact with the enzyme and is not predicted to bind in the enzyme pocket used for UDP binding. Interestingly, the size of the substituent at position 5 of the pyrrol(e/idine) ring is predicted to override the substitution pattern on the N-phenyl ring. Consequently, the predicted binding modes of the inhibitors from the two clusters correspond to flipped orientations of the N-phenyl ring and its substituents within the binding pocket. In the case of cluster 6 inhibitors, the smaller 5-methyl substituent allows the pyrrole ring to interact snugly with the enzyme subpocket used for binding the ribose ring of the UDP-sugar substrate. In contrast, the cluster 5 inhibitors are predicted to have the larger 5-phenyl substituent interacting in this subpocket. This would displace the pyrrolidone ring toward a more central position in the binding cleft.

The subsequent finding that three out of these five analogs can inhibit flagellin production in cellular assays is also informative from several viewpoints. First, one of the three analogs that inhibited flagellin production in C. jejuni was found not to inhibit in vitro the PseB enzyme at the tested concentration, which is indicative at best of weaker inhibition, although all three inhibited PseB from H. pylori similarly (Fig. 7). However, we noted that all three compounds are also in vitro inhibitors of PseG in H. pylori (52 to 70% inhibition at the concentration in the HTMS assay). Although we have not tested their activity against PseG from C. jejuni, one can speculate that the multienzyme inhibitory characteristic of these and other compounds discovered in this study may provide a mechanism for achieving robust efficacy in the biological environment. Second, we found that the two smaller cluster 6 analogs do not possess cellular activity despite showing inhibition of both PseB and PseG enzymes (in both H. pylori and C. jejuni). Given that these analogs are negatively charged with a carboxylate moiety, their membrane permeability is expected to be quite poor. However, the derivatization of the larger analogs from cluster 5 leads to a significant increase in their hydrophobicity relative to that of the smaller inactive analogs from cluster 6 (see Table S7 in the supplemental material). Hence, the increased hydrophobicity and membrane permeability of these pathway inhibitors are other key properties required for achieving cellular efficacy.

The identification of three inhibitors which show activity in bacterial cell-based assays provides a robust starting point for further hit-to-lead evaluation. Further characterization of the binding modes of these inhibitors will provide a guide for structure-based design of more potent and selective inhibitors toward the Pse biosynthetic enzymes which can then be tested as novel antivirulence therapeutic agents targeting H. pylori. Furthermore, the strategies utilized here may be useful for identifying inhibitors of other related nonulosonate pathways, such as legionaminic acid from Legionella pneumophila.