Abstract
Fungal pathogens, including Candida albicans, Candida glabrata and Aspergillus fumigatus, pose serious health risks for the immune suppressed or immune compromised i.e. cancer and transplant patients, AIDS patients, diabetics, newborns and the elderly. Antifungal drugs currently used in the clinic have significant practical limitations. Antifungal resistance is also an escalating problem. This includes the natural resistance of species such as A. fumigatus and Candida krusei to the widely-used azole antifungal drugs and the emergence of a significant proportion of C. glabrata strains that are intrinsically resistant. There is a growing unmet need for new classes of well-tolerated, low-cost, broad-spectrum antifungal drugs that circumvent drug resistance. A bioinformatics analysis indicated the riboflavin biosynthetic pathway, found in fungi but not in humans, could be a useful target for antifungal intervention. The aim of this thesis was to validate two enzymes from the riboflavin biosynthetic pathway, Dihydroxy-butanone-phosphate (DHBP) synthase and lumazine synthase (LS), as antifungal drug targets.
LS is the penultimate enzyme of the riboflavin synthesis pathway. It converts 3,4-dihydroxy-2-butanone-4-phosphate (DHBP), and 5-amino-6-ribitylamino pyrimidinedione to 6,7-dimethyllumazine. The gene RIB4 encoding C. glabrata LS (CgLS) was cloned in the pQE30 vector and expressed as a soluble and active N-terminal 6×histidine tagged construct using the M15 strain of Escherichia coli. CgLS was purified by Ni-NTA chromatography and size-exclusion chromatography. It was recovered mainly as a pentamer with a small proportion of the enzyme in the form of a decamer. Crystals of purified CgLS obtained using the hanging drop vapour diffusion method and diffraction data were phased using molecular replacement and refined to 2.24 Å. Molecular replacement using the S. cerevisiae structure (1EJB) revealed two pentamers in the asymmetric unit. The N-termini of all subunits showed high flexibility based on high B factors or lack of density. The current structure differs from the decamer from Brucella spp. as the pentamer interface is formed via a tail-to-tail interaction. Pentamers are connected via water-mediated hydrogen bonds. After several rounds of refinement, density was seen in several active sites. Glycerol was initially modelled into these positions. However, additional contiguous density made it clear that there was another molecule present. One molecule of the catalytic product 6,7-dimethyl-8-(D-ribityl)lumazine was modelled into each pentamer. In addition, attempts were made to cocrystallize C. glabrata LS with previously published inhibitors of Saccharomyces pombe and Mycobacterium tuberculosis LS. In one of these structures, the product molecule fitted the density maps within the active site better than the inhibitor.
In silico screening for new compounds was carried out in our laboratory by Ms Ann Walter. The imide moiety of the substrate and known inhibitors of LS are known to make important hydrogen bonding interactions with the enzyme. Using this cyclic imide as a pharmacophore several search queries were generated resulting in ~14,000 hits from an initial database of ~5.3 million compounds. Four sets of hits were subsequently docked separately into S. cerevisiae LS (pdb1EJB) using GOLD5.0 flexible ligand docking software. Hydrogen bonding constraints were used to maintain the imide-protein interaction and the top 500 hits from GOLD- and Chem-score were kept. A total of 36 molecules were selected, out of which 17 compounds were purchased for subsequent testing.
One of the compounds (compound10) showed pH-dependent antifungal activity against Saccharomyces cerevisiae cells with apparent specificity for LS in knockdown experiments. As the pH was reduced below pH 7, the compound showed an increase in activity. By pH 4.5 this compound also showed ~50% activity compared to AmpB on a molar basis (positive control). We infer that because the compound is less ionised at lower pH this property will aid in its passage to its intracellular target. The in vitro and in vivo activity of the compound against recombinant CgLS (IC50 of 2.4 µM) confirmed the compound’s specificity for LS. Compound 10 (6-(5-chloro-2-hydroxybenzoyl)-1-(2-methylpropyl)pyrido[2,3-d]pyrimidine-2,4(1H,3H)-dione obeys Lipinski’s rule of 5 and is a potential druggable candidate. These results validate LS as a potential drug target. It was found that the enzyme assay is not ideal as one of the substrates is unstable and degrades rapidly when exposed to oxygen. This required, for each experiment, preparation of the substrate from a more stable precursor.
DHBP synthase converts ribulose 5-phosphate (Ru5P) to 3,4-dihydroxy-2-butanone-4-phosphate (DHBP) which is a substrate for LS in the riboflavin biosynthetic pathway. Recombinant DHBP synthase from C. glabrata and A. fumigatus were cloned, expressed and purified using an E. coli host. The cloning of C. glabrata DHBP synthase proved problematic due to the presence of rare codons. Thus, the DHBP synthase genes (RIB3) were codon optimised/ harmonised and purchased commercially. Both enzymes were successfully expressed as soluble, active, constructs that were purified by Ni-NTA affinity and size exclusion chromatography (SEC) and extensively characterized. DHBP synthase from both the enzymes eluted as dimers in SEC. The sequence and molecular mass of each protein were confirmed by MS analysis. Parameters including pH, temperature and metal ion concentration were optimised for enzyme activity of the recombinant DHBP synthase from both the organisms. The optimal pH for recombinant DHBP synthase for C. glabrata and A. fumigatus was found to be 7 and 7.5, respectively. DHBP synthase requires metal ion (Mg2+) and concentrations for the optimal enzyme activity for CgDHBP synthase and AfDHBP synthase were found to be 1 mM and 5 mM, respectively. DHBP synthase from C. glabrata showed optimal activity at 30°C and A. fumigatus at 40°C. Putative substrate analogue inhibitors of DHBP synthase were identified via similarity searches of commercially available molecules. Docking was carried out to test if these substrate-analogues fit the active site in the same way manner as the substrate and maintained the correct stereochemistry. The carbonyl group of the natural substrate Ru5P forms coordinate bonds with metal ions which is essential for the enzyme activity. The proposed inhibitors should be able to interact with both metal ions in a similar way to have an effect on the protein. The compounds tested were not active against DHBP synthase. The substrate analogues used in this study had the same stereochemistry as the substrate itself but subtle differences including the planarity of the carbonyl in the substrate-analogue phosphoglycerate renders limited flexibility leading to different conformational accessibility. This could be the reason for the compounds having no effect on the enzyme. These features will need to be considered in the future design of inhibitors. Future experiments should also check the activity of substrate analogue inhibitors against human enzymes, ribulose 5-phosphate-3-epimerase and ribulose-5-phosphate isomerase.
This thesis provides partially and fully characterized enzyme assays for recombinant fungal LS and DHBP synthase, respectively. I have solved the structure of product complex of LS from C. glabrata. I have identified the first inhibitor of LS that is active against fungal cells. This inhibitor compound also follows Lipinski’s rule of five. Thus, in this thesis I show that LS is a valid target and the inhibitor will serve as a starting point for future antifungal drug discovery program.