Virtual screening, identification and in vitro validation of small molecule GDP-mannose dehydrogenase inhibitors

Upon undergoing mucoid conversion within the lungs of cystic fibrosis patients, the pathogenic bacterium Pseudomonas aeruginosa synthesises copious quantities of the virulence factor and exopolysaccharide alginate. The enzyme guanosine diphosphate mannose dehydrogenase (GMD) catalyses the rate-limiting step and irreversible formation of the alginate sugar nucleotide building block, guanosine diphosphate mannuronic acid. Since there is no corresponding enzyme in humans, strategies that could prevent its mechanism of action could open a pathway for new and selective inhibitors to disrupt bacterial alginate production. Using virtual screening, a library of 1447 compounds within the Known Drug Space parameters were evaluated against the GMD active site using the Glide, FRED and GOLD algorithms. Compound hit evaluation with recombinant GMD refined the panel of 40 potential hits to 6 compounds which reduced NADH production in a time-dependent manner; of which, an usnic acid derivative demonstrated inhibition six-fold stronger than a previously established sugar nucleotide inhibitor, with an IC50 value of 17 μM. Further analysis by covalent docking and mass spectrometry confirm a single site of GMD alkylation.


Virtual screening methodology using GOLD
The GOLD 8,9 version 5.6.2software suite was also used as the docking engine to predict for biologically active hits from the same compound library.The same protein set up and centre of binding coordinates were used with 10 Å radius.The basic amino acids lysine and arginine were defined as protonated.Furthermore, aspartic and glutamic acids were assumed to be deprotonated.Three scoring functions were implemented to validate for the predicted binding modes: Goldscore (GS), 10 Chemscore (CS), 11 Chem Piecewise Linear Potential (ChemPLP), 12 and Astex Statistical Potential (ASP). 13ovalent docking and MM-GBSA calculations Covalent docking was preformed using CovDock 14 panel of Maestro (2019-4, Schrödinger LLC, New York, NY) with the same centroid coordinates above.For modelling of the Michael addition of lysine, a custom cdock file was generated manually.A maximum 200 initial poses were kept for refinement and 50 final poses were generated. MMGBSA was performed using Prime panel of Maestro (2019-4, Schrödinger LLC, New York, NY) with VSGB solvation model according to the OPLS3e force field parameters.Residues at 4 Å away from 13 were kept flexible and the optimized complexes were sampled using force field minimization.

1.1: Glide & FRED
Virtual Screening results using Glide and FRED: The results from Glide and FRED are listed according to the docking scores, the ligands with docking score < -9.0 kcal/mol from Glide (top 12 compounds) and < -12.5 kcal/mol from FRED (top 20 compounds) were visually evaluated regarding integrations with the Cys268, the water molecule, and the residues interacting with GDP-mannuronic acid and those interacting with less than two of these residues were discarded.Table S1.Top-scoring compounds according to Glide, their scores and the key GMD residues in electrostatic interactions (H bond, -, -cation, salt bridge, halogen bond).Virtual screening results using GOLD A 3D library of 1447 small molecules from the in-house library were docked into the catalytic site and virtually screened (see methodology) at 30% efficiency at 20 GA runs.Efficiency indicates the thoroughness in which the algorithm searches for a thermodynamically favourable binding pose in the enzyme active site.All compounds with a hydrogen bonding score of less than 1.0 were removed and the top scoring 200 compounds from each of GS, CS and ChemPLP scoring functions were kept resulting in 263 compounds.The resulting 263 compounds were virtually screened using the same protocol at 100% efficiency at 50 GA runs.All compounds with hydrogen bonding scores of less than 1.0 were removed and the top scoring 100 compounds from each of the three scoring functions were selected for visual inspection leaving a pool of 179 compounds.Most of the compounds at this stage formed hydrogen bond interactions with some of the residues involved in stabilising the cocrystallised GDP-mannuronic acid.Considering numerous hydrogen bonding residues were formed with the co-crystallised ligand, the hydrogen bond scores and networks were prioritised, i.e., compounds predicted to form numerous hydrogen bonds with important residues.A total of 8 compounds were chosen (Table 3).
Table S3.The docking scores and hydrogen bonding residues of the co-crystallised ligand and the 8 predicted inhibitors.

1.4: Library Compounds & Purity
All compounds were used within this study were received as gifts from the lab of Konstantin P. Volcho (N.N.Vorozhtsov Novosibirsk Institute of Organic Chemistry, Russia) & used without additional purification.][17][18][19][20] Following identification of compound 13 as a potential inhibitor of GMD, additional characterisation was performed by NMR to ensure purity.1H NMR data matched the previously reported literature data.

Section 2: GMD expression
The recombinant plasmid (pET-3a) containing the algD gene encoding for GDP-mannose dehydrogenase (GMD) from P. aeruginosa was kindly donated by P. Tipton.The plasmid was transformed into E. coli soluBL21(DE3) chemically competent cells and the transformant grown according to the literature. 4,21riefly, 1 L of the transformant in LB medium containing the appropriate antibiotic (carbenicillin, 100 µg/mL) was incubated at 37 °C with gentle shaking in baffled flasks until an OD 600 of 0.6-0.8 was reached.Heterologous protein expression was induced by adding isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.4 mM, followed by incubation at 37 °C for 4 hours at 180 rpm.Afterwards the cells were harvested by centrifugation (4000 x g, 4 °C, 20 mins) and stored at -80 °C until use.Frozen cells were thawed in 20 mM HEPES (pH 7.5), 150 mM NaCl supplemented with DNase A (10 µg/ml, Sigma) and proteinase inhibitor cocktail (Roche), then lysed by sonication on ice.The supernatant was recovered by centrifugation (20,000 x g, 4 °C, 20 min) and nucleic acid precipitated through the addition of protamine sulfate (5 mg per gram wet cell pellet) and incubated on ice for 30 mins.Precipitated nucleic acid removed by centrifugation (20,000 x g, 4 °C, 20 min), the crude protein solution was fractionated with ammonium sulfate, with GMD precipitating between 45 and 60% saturation.Protein pellets were redissolved in 20 mM HEPES (pH 7.5), 150 mM NaCl and purified using an ÄKTA pure FPLC system (GE Healthcare) by gel filtration chromatography using a Superdex S200 16/600 column (GE Healthcare).Proteins were eluted with 20 mM HEPES (pH 7.5) and 150 mM NaCl at the flow rate of 1 ml/min.GMD comprising fractions were combined and concentrated to ~4.5 mg/mL (concentration determined by Pierce™ BCA assay, ThermoFisher or Bradfords Assay, Sigma).Concentrated GMD was then divided into aliquots and stored at -80°C until required.

3.1: Initial screening of all 40 compounds with no preincubation
The assay was performed in 96-well flat bottomed, non-binding, polystyrene microtiter plates (Grenier 655906).NAD + (200 µM), inhibitor (50 µM) and GMD (50 µg/mL) were prepared in 50 mM sodium phosphate (pH 7.4) containing 0.5 mM MgCl 2 and 1 mM DTT.A solution of GDP-Man (final: 50 µM) was added to the plate and the fluorescence was measured at 25 °C for 30 minutes using a BMG labtech FLUOStar Omega microplate reader (excitation 355 nm; emission 460 nm).The limits of detection were analysed by control samples as followed: positive control contained no inhibitor; negative control contained no inhibitor or GDP-Man.

Data Processing
Initial rate of fluorescence increase was calculated over the first 20 minutes.The fluorescence was converted to % NADH production by comparison with the maximum and minimum values obtained from the positive and negative controls.

3.2: Screening of 21 compounds following 1-hour preincubation with GMD
The assay was performed in 96-well flat bottomed, non-binding, polystyrene microtiter plates (Grenier 655906).NAD + (200 µM), inhibitor (50 µM) and GMD (50 µg/mL) were prepared in 50 mM sodium phosphate (pH 7.4) containing 0.5 mM MgCl 2 and 1 mM DTT.The samples were incubated at 4 °C for 1 hour before a solution of GDP-Man (final: 50 µM) was added to the plate and the fluorescence was measured at 25 °C for 60 minutes using a BMG labtech FLUOStar Omega microplate reader (excitation 355 nm; emission 460 nm).The limits of detection were analysed by control samples as followed: positive control contained no inhibitor; negative control contained no inhibitor or GDP-Man.

Figure S2 .
Figure S2.Bar chart comparing percentage NADH production in the presence of the 40 potential inhibitors hits.Percentage NADH production determined relative to the positive control containing no inhibitor.Inhibitors 7 & 35 demonstrate increased NADH production compared to the positive control.

Table S4 .
The structure of virtual screening hits and LogP values.
1.2: Structures of Virtual Screening Hits

Table S5 .
Definition of lead-like, drug-like and Known Drug Space (KDS) in terms of molecular descriptors.The values given are the maxima for each descriptor for the volumes of chemical space used (Zhu, F.; Logan, G.; Reynisson, J. Wine Compounds as a Source for HTS Screening Collections.A Feasibility Study.Mol.Inf.2012, 31, 847 -855.).