Triazolopyrimidine herbicides are potent inhibitors of Aspergillus fumigatus acetohydroxyacid synthase and potential antifungal drug leads

Aspergillus fumigatus is a fungal pathogen whose effects can be debilitating and potentially fatal in immunocompromised patients. Current drug treatment options for this infectious disease are limited to just a few choices (e.g. voriconazole and amphotericin B) and these themselves have limitations due to potentially adverse side effects. Furthermore, the likelihood of the development of resistance to these current drugs is ever present. Thus, new treatment options are needed for this infection. A new potential antifungal drug target is acetohydroxyacid synthase (AHAS; EC 2.2.1.6), the first enzyme in the branched chain amino acid biosynthesis pathway, and a target for many commercial herbicides. In this study, we have expressed, purified and characterised the catalytic subunit of AHAS from A. fumigatus and determined the inhibition constants for several known herbicides. The most potent of these, penoxsulam and metosulam, have Ki values of 1.8 ± 0.9 nM and 1.4 ± 0.2 nM, respectively. Molecular modelling shows that these compounds are likely to bind into the herbicide binding pocket in a mode similar to Candida albicans AHAS. We have also shown that these two compounds inhibit A. fumigatus growth at a concentration of 25 µg/mL. Thus, AHAS inhibitors are promising leads for the development of new anti-aspergillosis therapeutics.

. Chemical structures of the commercial herbicides evaluated here. A general feature of the sulfonylureas, triazolopyrimidines, and sulfonylamino-carbonyl-triazolinones is the presence of both an aromatic and a heterocyclic ring. The structural images were created using CHEMDRAW 20.0.

K i values of commercial herbicides for AfuAHAS. The herbicide binding stoichiometry of AfuAHAS
was investigated by measuring the activity of increasing concentrations AfuAHAS in the absence and presence of 0.5 µM of the sulfonylurea, chlorimuron ethyl (CE; Fig. 1). The results show that 0.5 µM CE can fully inhibit 1.06 µM of AfuAHAS (Fig. 3a), giving a 1:2 inhibitor to enzyme ratio, which has also been observed in other AHAS studies 10 . Thus, once one active site is blocked, communication is halted and the second active site is no longer capable of substrate turnover 20 .
Eighteen commercial herbicides from five chemical families ( Fig. 1) were tested for their ability to inhibit AfuAHAS. In order to accurately determine the K i values of herbicides for AfuAHAS, a partial anaerobic environment was established through nitrogen bubbling of the assay buffer in the presence of 2-mercaptoethanol (see "Materials and methods", Supplementary Fig. 2) 12 .
The K i values of the herbicides for AfuAHAS under these conditions are listed in Table 1. Among the five chemical families, the triazolopyrimidines generally have the highest affinity. PS and MT have the lowest K i values of 1.8 ± 0.92 nM and 1.4 ± 0.18 nM, respectively (Fig. 3b). For the triazolopyrimidine family there is a strong correlation in K i values for AfuAHAS and CaAHAS (Table 1), suggesting the herbicides of this family have very similar binding modes in the two enzymes (discussed below).
Amongst the sulfonylureas, CE has the lowest K i (23.4 ± 1.3 nM), while the other sulfonylureas studied here have K i values in the range from 220 to 3100 nM (Table 1). Relative to CaAHAS, the sulfonylureas bind to AfuAHAS with an increase in K i of ~ 4-10-fold. For the sulfonylamino-carbonyl-triazolinones and pyrimidinylbenzoates, the herbicides have K i values that are generally higher than those of the triazolopyrimidine and sulfonylurea families. Furthermore, the imidazolinones tested here show no inhibition of AfuAHAS at concentrations up to 100 µM, whilst they weakly inhibit CaAHAS. Overall, however, there is a good correlation of the K i values between families and between the two enzymes (Fig. 3c).

Accumulative inhibition by the commercial herbicides on AfuAHAS. Accumulative inhibition of
AfuAHAS was evaluated through the determination of the apparent first-order rate constants of enzyme inactivation (k iapp ) and enzyme recovery (k 3 ) 11 , and the efficiency of accumulative inhibition, k iapp /k 3 13 (Table 1). Among all the herbicides, the members of the triazolopyrimidine family were generally shown to be most efficient accumulative inhibitors of AfuAHAS (Table 1). In particular, PS and MT have the most potent time-dependent inhibition (k iapp /k 3 values of 1426 and 1750, respectively), reflected by the high k iapp of accumulative inhibition and the relatively low rate of enzyme recovery (k 3 ) (Table 1). PS has a ~ four-fold higher rate of enzyme inactivation compared to MT. However, due to its ~ five-fold lower enzyme recovery rate, stronger accumulative inhibition is achieved by MT over time (Fig. 3d)  and AtAHAS catalytic subunit sequences shows that most of the herbicide binding site residues are highly conserved amongst these AHASs (Fig. 4). The exceptions are S254 and S259 in AfuAHAS, which are hydrophobic residues in the other fungal and plant AHASs, as well as A719, which is a glycine or serine in ScAHAS and AtAHAS, respectively (Fig. 4). The conservation of herbicide binding site residues in AfuAHAS and CaAHAS is consistent with the correlation found in the K i values (Fig. 3c). Attempts to crystallize AfuAHAS, either in the presence or absence of inhibitors were not successful, so molecular modelling and docking was used to provide structural explanations for the inhibition results. A homology model of AfuA-HAS was generated using SWISS-MODEL 21 and the crystal structures of CaAHAS with bound herbicides were used to identify herbicide interactions in AfuAHAS (Fig. 5). The homology model has an rmsd of 0.368 Å for 1196 out of 1231 Cα atoms after superimposition with CaAHAS. In total, 95.7% of the amino acid residues have favourable Ramachandran dihedral angles and there are 0.4% outliers. The modelling shows that the herbicide binding site structures in CaAHAS and the AfuAHAS model are similar (Fig. 5a-c). The most significant differences are A191 and A196 in CaAHAS which are replaced by serine (i.e. S254 and S259; Figs. 4 and 5). Based on the published CaAHAS structures, A191 and A196 form hydrophobic interactions with the aromatic ring ( Fig. 1) of the herbicides (Fig. 5) 13 . Inhibitor docking with the AfuAHAS model shows little change in the distances of S254 and S259 to the bound herbicide compared to A191 and A196 in CaAHAS (Fig. 5). The proximity of the serine residues to the bound herbicides remain sufficient for hydrophobic contacts to occur. The change in K i values (Table 1) observed when sulfonylureas and sulfonylamino-carbonyl-triazolinones bind could be the result of the subtle changes to the herbicide binding site in AfuAHAS, e.g. the serine residues might promote binding of an ordered water molecule, thereby modifying the binding mode of the inhibitors. However, for PS and MT these sequence changes have no apparent effect on binding.
The presence of S254 and S259, not observed in the other pathogenic fungi and A. thaliana, could provide a target for the design of specific A. fumigatus inhibitors. Introducing hydrophilicity into the aromatic ring structure of the herbicide to form hydrogen bonds with the serine residues may potentially improve inhibitor binding affinity.   Table 2). PS and MT also inhibit A. nidulans growth at 25 µg/mL. None of the other compounds tested showed activity at < 100 µg/mL, and the growth of A. niger and A. flavus is not impaired by any of the herbicides listed in Table 2.
The difference in susceptibility of the four Aspergillus species to the same herbicide (Table 2) may be attributed to a number of factors. However, comparison of the herbicide binding site residues between the different Aspergillus species (Fig. 6) show few differences, suggesting binding to the enzyme is not one of these. The lack of inhibitory activity on A. niger and A. flavus growth may be due to the production of proteins that manage oxidative stress, or differentially metabolize these herbicides (e.g. P 450 s) 24 .
Richie et al. showed that A. fumigatus can scavenge BCAA from sheep blood, raising concerns on the efficacy of inhibiting the BCAA biosynthesis pathway for this fungus 15 . However, the primary route of A. fumigatus infection is through the lungs, where a porcine model has provided evidence that BCAA availability is limited in pulmonary secretions 25 . A. fumigatus is thus unlikely to be able to scavenge sufficient BCAAs from the pulmonary Table 1. Apparent first-order rate of inhibition (k iapp ), first-order rate of enzyme recovery (k 3 ), and inhibition constants (K i ) of the commercial herbicides for AfuAHAS. *K i values for CaAHAS are shown for comparison were obtained from Garcia et al. 13 . NAI No reversible accumulative inhibition observed, ND Not determined. See Fig. 1 for chemical structures of compounds.    www.nature.com/scientificreports/

Conclusion
Evaluation of the five chemical families of commercial herbicides show that the trizaolopyrimidines have potent inhibitory properties against AfuAHAS. In particular, the herbicides PS and MT are tight binding inhibitors and also have strong accumulative inhibitory properties. These two herbicides are therefore good starting points for the design of novel antifungal compounds that target AfuAHAS. However, the bioavailability of the herbicide in Aspergillus cells needs to be improved for the design of more effective compounds that can prevent growth of this fungal pathogen.

Materials and methods
Preparation of AfuAHAS gene construct. The amino acid sequence for the catalytic subunit of AfuA-HAS was obtained from the National Centre for Biotechnology Information (NCBI) (NCBI reference sequence: XP_754588.1). The construct was modified to remove the DNA coding for the first 111 amino acids belonging to the mitochondrial transit peptide sequence. An AUG start codon was added at the N-terminus of the peptide sequence. A Tobacco Etch Virus protease restriction site and a hexahistidine tag were added at the C-terminus of the peptide sequence. The gene sequence was synthesised in a pUC57 plasmid by Biomatik (Cambridge, Canada). The coding sequence for AfuAHAS was excised from the plasmid by HindIII/NdeI (New England Biolabs) digestion. The gene was then ligated with the pET30A(+) vector cut using the same restriction sites. Escherichia coli BL21 (DE3) cells were then transformed using the resultant pET30A(+)-AfuAHAS plasmid. www.nature.com/scientificreports/ Protein expression and purification. All reagents were obtained from Sigma-Aldrich (St Louis, MO, USA) and were of analytical grade, unless otherwise stated. Protein expression and purification were performed as described previously for CaAHAS, but with some differences 13 . The gel filtration buffer used for size exclusion chromatography contains 10 mM potassium phosphate buffer pH 7.2, 300 mM NaCl, 10 µM FAD and 1 mM DTT. The fractions containing the folded enzyme were pooled, and the potassium phosphate concentration was increased to 200 mM and 10 mM of MgCl 2 was added before concentration and storage at − 80 °C.

Determination of the kinetic parameters (K cat and K M ) of AfuAHAS reaction.
The continuous spectrophotometric method (see above) was used to determine the K M of the substrate, pyruvate. The FAD of AfuAHAS was fully reduced to remove the lag phase prior to measuring the rates 19 . 100 µL of assay mixture was incubated at 30 °C at varying pyruvate concentrations (0.03-200 mM) containing 6.6 µM AfuAHAS. The data were fitted to the Michaelis-Menten equation to obtain the K M and the k cat .
Preparation of partial anoxic buffer. Partial anoxic assay buffer was made by bubbling nitrogen gas through the standard assay buffer for 30 min to purge dissolved oxygen and by including 14.2 µM 2-mercaptoethanol 12 .
The combined effect of 2-mercaptoethanol and nitrogen bubbling on accumulative inhibition generated by PS is shown in Supplementary Fig. 2.
K i determination by the colorimetric single point method. 49.37 nM AfuAHAS was first incubated at 30 °C in partial anoxic assay buffer for 20 min. Next, 90 µL of the assay mixture containing the enzyme was added to 8 µL of nitrogen-gas-treated water containing 2 uL of varying concentrations of inhibitors dissolved in dimethyl sulfoxide (DMSO), including DMSO alone for the control. The mixture was then incubated at 30 °C for 15 min, before stopping the reaction with 10% H 2 SO 4 and quantifying the amount of 2-acetolactate (see colorimetric method above). The experiments were repeated three times. Equation (1) 28 was used to fit the data for tight binding inhibitors (CE and all triazolopyrimidines apart from FT) for which the enzyme concentration has to be taken into account. Equation (2) 29 was used to fit the data for all other herbicides (medium binding inhibitors). V 0 and V 1 are the uninhibited rate and inhibited rate respectively, [I] is the total inhibitor concentration and [E] is the total enzyme concentration.
Accumulative inhibition assay. The continuous spectrophotometric method (see above) was used where 2.6 µM AfuAHAS was incubated with the standard assay buffer at 30 °C for 20 min to reach maximum enzyme activity before addition of inhibitor. 25 nM of PS, FS, and MT, and 100 nM of the other herbicides were used in the assay. The accumulative inhibition induced by these herbicides was observed for 45 min. All experiments were repeated three times. The data were fitted to Eq. (3) 10,11 to calculate the kinetic rate constants k iapp and k 3 (Table 1).
where F represents the ratio of free enzyme/enzyme-inhibitor complex. (2) www.nature.com/scientificreports/ to an enzyme, which has to be taken into account in the calculation of the F value. The effective concentration of enzyme-inhibitor complex was calculated using the inhibition constant formula [Eq. (4)].
where [E] represents the uninhibited enzyme concentration, [I] the unbound inhibitor concentration, [EI] the enzyme-inhibitor complex concentration.
Stoichiometry of enzyme inhibition. AHAS activity was monitored using the continuous method. For this study, CE was chosen as the inhibitor as it binds tightly to the enzyme and without accumulative inhibition in the presence of 2-mercaptpethanol. Increasing concentrations (2-6 µM) of AfuAHAS were incubated in the standard assay buffer with the addition 14.2 µM 2-mercaptoethanol for 15 min until maximum activity was reached. 0.5 µM of CE was then added and the enzyme activity for the first 200 s of inhibition was recorded to determine the initial inhibition rates without accumulative inhibition. A control without inhibitor was included and the measurements were performed in triplicates.
Fitting of the data. Fitting of data was performed using GraphPad Prism 7.01 (GraphPad Software, San Diego California USA, www. graph pad. com).
Antifungal susceptibility assays. The antifungal susceptibility assays were conducted according to Clinical and Laboratory Standards Institute guidelines for broth microdilution M32-A2 30 , except the testing media contained Yeast Nitrogen Base, without amino acids or ammonium sulfate, buffered with 50 mM 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid (HEPES). 2% glucose and 10 mM ammonium sulfate were added to the solution as the carbon and nitrogen sources, respectively. The medium was adjusted to pH 7.0 using 4 M sodium hydroxide and then stored at room temperature. 10 mg/mL stocks of each herbicide were prepared in DMSO. 10 serial twofold dilutions were conducted, giving a dilution range of 0.097656-100 µg/mL for each antifungal agent. A. fumigatus strain ATCC MYAA 3626 was used in this assay. The 96-well plates were incubated at 35 °C for 48 h, and the OD 530 of each well was recorded using the Molecular Devices Spectramax 250 microplate reader (Marshall Scientific). Each experiment was performed in triplicate.