Towards novel herbicide modes of action by inhibiting lysine biosynthesis in plants

Weeds are becoming increasingly resistant to our current herbicides, posing a significant threat to agricultural production. Therefore, new herbicides with novel modes of action are urgently needed. In this study, we exploited a novel herbicide target, dihydrodipicolinate synthase (DHDPS), which catalyses the first and rate-limiting step in lysine biosynthesis. The first class of plant DHDPS inhibitors with micromolar potency against Arabidopsis thaliana DHDPS was identified using a high-throughput chemical screen. We determined that this class of inhibitors binds to a novel and unexplored pocket within DHDPS, which is highly conserved across plant species. The inhibitors also attenuated the germination and growth of A. thaliana seedlings and confirmed their pre-emergence herbicidal activity in soil-grown plants. These results provide proof-of-concept that lysine biosynthesis represents a promising target for the development of herbicides with a novel mode of action to tackle the global rise of herbicide-resistant weeds.


Introduction
Our ability to provide food security for a growing world population is increasingly challenged by the emergence and spread of herbicide-resistant weeds. Resistance has now been observed to the most widely used classes of herbicides. This includes amino acid biosynthesis inhibitors such as chlorsulfuron, glufosinate, and glyphosate, which target enzymes in the biosynthetic pathways leading to the production of branched-chain amino acids, glutamine, and aromatic amino acids, respectively (Gaines et al., 2020;Hall et al., 2020). The impact of herbicide resistance is exacerbated by the lack of new herbicides entering the market in the past 30 years, especially those with new mechanisms of action (Duke, 2012). Nevertheless, the successful commercialisation of such herbicides provides proof-of-concept that targeting the biosynthesis of amino acids offers an excellent strategy for herbicide development (Hall et al., 2020). Amino acids are not only essential building blocks for protein biosynthesis, but they also play important roles in physiological processes that are critical for plant growth and development, such as carbon and nitrogen metabolism, in addition to serving as precursors to a wide range of secondary metabolites (Hildebrandt et al., 2015).
Due to the central role of DHDPS in lysine production in plants, this enzyme has been proposed as a potential target for the development of herbicides (Griffin et al., 2012;Soares da Costa et al., 2018). Indeed, the lysine analogue, S-(2-aminoethyl)-L-cysteine, halts rooting of potato tuber discs at mid-micromolar concentrations (Perl et al., 1993;Ghislain et al., 1995). However, given its poor in vitro potency against plant DHDPS, it is believed that this analogue inhibits plant growth by competing with lysine for incorporation into proteins rather than inhibition of the DHDPS enzyme (Ghislain et al., 1995;Perl et al., 1993). Plants typically have two annotated DHDPS-encoding genes (DHDPS) (Figure 2-figure supplement 1; Craciun et al., 2000;Sarrobert et al., 2000;Vauterin and Jacobs, 1994). In Arabidopsis thaliana, these genes are At3G60880 (DHDPS1) and At2G45440 (DHDPS2), which encode chloroplast-targeted AtDHDPS1 and AtDHDPS2, respectively (Jones-Held et al., 2012). RNA sequencing data have elucidated that both DHDPS-encoding genes are expressed at all stages of A. thaliana development, with the most prominent expression at the seed development stages, in the dry seeds and during germination (Klepikova et al., 2016). Interestingly, maximal DHDPS2 expression is approximately 3-fold greater than that of DHDPS1 (Klepikova et al., 2016). Moreover, upon comparison to the known glyphosate herbicide target, 5enolpyruvyl-shikimate 3-phosphate synthase (At1G48860), expression of DHDPS1 is considerably lower at almost all developmental stages, while expression of DHDPS2 is comparable at all stages, except in the dry seeds (Klepikova et al., 2016). This is a key consideration in target validation as low expressing targets will require less inhibitor to achieve phytotoxicity. Double knockouts of DHDPS1 and DHDPS2 result in non-viable embryos even after exogenous supplementation with lysine, indicating that DHDPS activity is essential (Jones-Held et al., 2012). AtDHDPS enzymes exist as homotetramers (Figure 2A), with the active site located within the (b/a) 8 -barrel ( Figure 2B) and the allosteric cleft in the tight-dimer interface located in the interior of the structure ( Figure 2C; Griffin et al., 2012;Hall et al., 2021).
In this study, we describe a new class of plant DHDPS inhibitors. We show that these compounds display micromolar potency in vitro and in planta against A. thaliana using a combination of enzyme kinetics, seedling, and soil assays, whilst exhibiting no cytotoxic effects in bacterial or human cell lines at equivalent concentrations. Furthermore, we employ X-ray crystallography to show that these compounds target a previously unexplored binding site within DHDPS, which is highly conserved amongst plant species. Thus, this study provides proof-of-concept that lysine biosynthesis represents a promising pathway to target for the development of herbicides with a new mode of action and highlights a novel DHDPS binding pocket to assist in the discovery of herbicide candidates.

Results
High-throughput chemical screen for inhibitor discovery A high-throughput screen of a library of 87,648 compounds was conducted against recombinant DHDPS enzyme by the Walter and Eliza Hall Institute High Throughput Chemical Screening Facility (Melbourne, Australia). The o-aminobenzaldehyde (o-ABA) colourimetric assay was used to estimate DHDPS activity via the formation of a purple chromophore that can be measured at 520-540 nm (Yugari and Gilvarg, 1965). Using a cut-off equal to the mean ± 3Â standard deviation for classification as a hit compound, 435 compounds out of 87,648 were identified as hits at 20 mM (hit rate = 0.50%). The activity of these 435 compounds was confirmed at the same concentration as the primary screen, resulting in 38 compounds demonstrating >40% inhibition of the DHDPS enzymatic reaction (confirmation rate = 8.7%). A counter screen was employed to exclude false-positive compounds i.e., compounds that interacted with o-ABA detection or absorbance quantification. The compounds that displayed confirmed DHDPS inhibition were subsequently progressed to full dose response titration assays using recombinant DHDPS. One promising compound from the screen was (Z)À2-(5-(4-fluorobenzylidene)À2,4-dioxothiazolidin-3-yl)acetic acid (FBDTA). The characterisation of two thiazolidinedione analogues containing methoxy substituents, MBDTA-1 and MBDTA-2, will be discussed here (Figure 3, Supplementary file 1).

Efficacy of inhibitors on recombinant DHDPS
The inhibitory activity of MBDTA-1 and MBDTA-2 against both recombinant A. thaliana DHDPS proteins was quantitated using a DHDPS-DHDPR coupled assay (Atkinson et al., 2013). This was achieved by titrating different concentrations of each compound with substrates fixed at previously determined Michaelis-Menten constant values (Griffin et al., 2012;Hall et al., 2021). The IC 50 values of MBDTA-1 and MBDTA-2 against AtDHDPS1 were 126 ± 6.50 mM and 63.3 ± 1.80 mM, respectively ( Figure 4A). Similarly, the dose-response curves for AtDHDPS2 yielded IC 50 values of 116 ± 5.20 mM and 64.0 ± 1.00 mM for MBDTA-1 and MBDTA-2, respectively ( Figure 4B). As these compounds represent a novel class of inhibitors of plant DHDPS, we set out to assess the mechanism of inhibition by examining the binding of MBDTA-2 to DHDPS using X-ray crystallography.

Molecular basis for inhibitor binding
To probe the molecular determinants for inhibition, recombinant AtDHDPS1 was co-crystallised with MBDTA-2 using the same crystallisation conditions as for the apo enzyme with the addition of inhibitor (Hall et al., 2021). Diffraction data were obtained at a maximal resolution of 2.29 Å , phases solved by molecular replacement and repeating rounds of model building and refinement were performed that allowed us to generate an atomic inhibitor-bound model ( Figure 5A, Table 1). We initially found several MBDTA-2 molecules at the crystal contact formed between protein molecules at chains B and D. Specifically, two parallel MBDTA-2 molecules were bound to H187 (of the symmetry mate) and F210 with complete occupancy ( Figure 5-figure supplement 1). However, given that these molecules were found solely at the crystal interface and were absent in chains A and C, they were assumed to be a result of non-specific binding. representing interspacing residues. Sequences were aligned in BioEdit (v 7.2.5) (Hall, 1999) using the ClustalW algorithm (Thompson et al., 1994). The online version of this article includes the following figure supplement(s) for figure 2: Closer inspection of the crystal structure revealed the presence of four MBDTA-2 molecules bound at the centre of the homotetrameric protein ( Figure 5A), in antiparallel pairs with each molecule, which were stabilised by interactions across three of the monomers ( Figure 5B). Interestingly, this pocket, albeit distinct to the lysine binding site, shares two residues with it, namely W116 and  E147 ( Figure 2C). Specifically, the methoxy group of MBDTA-2 interacts with W116, E147, and H150 from chain B, while the pendant carboxylic acid interacts with S115 from chain A as well as H150 and Q154 from chain C ( Figure 5C). Additionally, we observed that upon binding, MBDTA-2 forces D117 to adopt a different rotamer conformation, which in turn, results in W116 assuming a different conformation. It must be noted that the four MBDTA-2 molecules were present with 50% occupancy. Consequently, each of the two moving residues, D117 and W116, adopt two distinct rotamer conformations, one of the apo-and ligand-bound states of AtDHDPS1. Given that no major rotamer changes or movement of catalytically important residues were noted, the exact mechanism of inhibition remains elusive. Nevertheless, this indicates the presence of a novel DHDPS allosteric pocket that has not been previously exploited for inhibitor discovery. Moreover, an alignment of the primary structure of several DHDPS enzymes from plant species indicates that the residues involved in MBDTA-2 binding are highly conserved across both monocotyledons and dicotyledons ( Figure 5D) and therefore should allow for broad-spectrum inhibition.

Specificity of DHDPS inhibitors
Following determination of the binding site, we examined the specificity of MBDTA-1 and MBDTA-2 to determine whether any future applications would have off-target effects. First, the cytotoxicity of the inhibitors was examined against the human cell lines, HepG2 and HEK293, using the 3-(4,5-dimethylthiazol-2-yl)À2,5-diphenyltetrazolium bromide (MTT) assay ( Figure 6A,B). At the highest concentration assessed (400 mM), treatment with the inhibitors did not affect the viability of either cell line relative to the vehicle control. Second, the effect of the inhibitors on several bacterial species commonly found in the human flora and soil microbiome was assessed by measuring their minimum

Herbicidal efficacy
Given the promising in vitro properties of the inhibitors, we determined their herbicidal efficacy against A. thaliana, initially using seedling agar assays. At high micromolar concentrations of both MBDTA-1 and MBDTA-2, growth was completely attenuated, and most seeds were unable to germinate. Upon quantitation of root lengths, we determined an IC 50 of 98.1 ± 4.34 mM and 47.4 ± 0.450 mM for MBDTA-1 ( Figure 7A) and MBDTA-2 ( Figure 7B), respectively. Based on these results, we  The online version of this article includes the following source data for examined their pre-emergence effect on soil-grown A. thaliana. Specifically, compounds were dissolved in a solution containing a non-ionic organic surfactant (Agral) and seeds were treated immediately after sowing on soil. The vehicle control-treated plants ( Figure 8A) were used as a benchmark to visually assess the effects of inhibitors. The growth of A. thaliana in the presence of MBDTA-1 ( Figure 8B) or MBDTA-2 ( Figure 8C) at 300 mgÁL À1 was severely impeded as evidenced by the  growth area relative to the DMSO control ( Figure 8D), wherein few seeds were able to germinate. This is consistent with the results observed at the highest concentrations of inhibitor on agar. Furthermore, the A. thaliana seeds capable of germinating in the presence of 300 mgÁL À1 MBDTA-2 were halted at the cotyledon stage before the generation of true leaves. As such, our newly discovered MBDTA compounds represent the first DAP pathway inhibitors with soil efficacy against plants.

Discussion
The lack of herbicides with novel modes of action entering the market in the past three decades has led to an over-reliance on our current agrichemicals, which has contributed to the rapid generation of resistance. Although the DAP pathway has gained attention as a way to increase the nutritional content of lysine in crops (Wang et al., 2017), it has remained an unexplored target for the development of herbicides until now. DHDPS catalyses the first step of the DAP pathway and is commonly duplicated in plant species, including A. thaliana. Both DHDPS proteins are localised to the chloroplast and share >85% of primary structure identity, with the majority of differences at the N-terminus. Although DHDPS is essential in A. thaliana (Jones-Held et al., 2012), there have been no published inhibitors of the plant enzymes, with much of the focus on inhibitors of bacterial DHDPS as possible new antibiotics. Several iterations of active site inhibitors against bacterial DHDPS enzymes have been developed, but most of these have, at best, high micromolar potency (Christoff et al., 2019). As the active site of DHDPS is seemingly difficult to target, the allosteric site of the enzyme could represent a more fruitful route for the development of inhibitors. Indeed, lysine is the most potent inhibitor of plant DHDPS orthologues. Furthermore, the most potent bacterial DHDPS inhibitor to date is bislysine, whereby two lysine molecules are linked together via an ethylene bridge (Skovpen et al., 2016). Although herbicides commonly directly target, or bind closely to, the active site of their respective enzyme target, this study shows that allosteric site inhibitors can be employed and afford lethal phytotoxicity.
Our study describes the discovery of the first plant DHDPS inhibitors, with two MBDTA analogues identified and characterised here. The mode of inhibition is via a novel binding pocket adjacent to the lysine binding site, which results in the allosteric inhibition of the enzyme. Lysine has recently been shown to differentially inhibit the AtDHDPS isoforms, with AtDHDPS1 being 10-fold more sensitive to the allosteric inhibitor (Hall et al., 2021). In this study, we demonstrate that the MBDTA compounds have similar inhibitory effects against both enzymes. This further supports crystallography data demonstrating that MBDTA-2 binds in a pocket adjacent to the lysine allosteric site and is likely acting in a different way. However, the exact mechanism of inhibition, much like lysine-mediated allostery, remains elusive. Moreover, this binding pocket is conserved across multiple plant species, including both monocots and dicots. Importantly, our compounds lacked off-target toxicity, whilst resulting in the inhibition of germination and growth of A. thaliana seedlings on solidified media and in soil. However, as expected, plant inhibition was more pronounced on media likely due to the stability, distribution, and persistence of the compounds. Nevertheless, the assays performed on soil demonstrate their potential applicability as pre-emergence treatments. It would also be of interest to investigate the metabolic shifts in plants treated with inhibitors and determine if there is a toxic build-up of other amino acids such as threonine, which has been observed in DHDPS knockout experiments (Sarrobert et al., 2000). Indeed, a common trait of systemic herbicides is that their efficacy is often related to the cascading consequences of inhibiting a key reaction, rather than inhibition of the reaction itself (Hall et al., 2020). Additionally, it would be of interest to determine the metabolic state of the compounds after uptake into the cells. This would elucidate whether the MBDTA compounds act as proherbicides, potentially through the demethylation of their aryl methyl ethers. As such, it would also be of interest to test demethylated analogues in vitro and in planta to assess any changes in potency relative to the methoxy compounds.
Developing enzyme inhibitors into a commercial product is an arduous and costly process. Optimisation of phytotoxicity, water solubility, cell wall penetration, translocation, soil/water persistence, and formulation must all be considered. The MBDTA compounds described here represent an attractive avenue to pursue, and with the elucidation of a novel binding pocket within DHDPS, it may be possible to rationally improve their potency guided by the crystallography data. Specifically, it would be of interest to build compounds that extend into the lysine binding pocket, which could improve potency. Alternatively, novel chemical scaffolds could be explored to target the DHDPS pocket identified. The inhibitors must be able to traverse the chloroplast membrane in order to reach the DHDPS target and be amenable to post-emergence application. It would also be of interest to study inhibitors with increased hydrophobicity and thus, potentially enhanced transport through the epidermis, cell wall, and plastid membrane, to reach the DHDPS target. However, it is important to recognise that there is a fine balance between membrane permeability and aqueous solubility that must be achieved for any new herbicides. Following the synthesis of more potent analogues, it would be of interest to test inhibitors against weed species. To further pursue the compounds described herein and eventually deploy more potent analogues as herbicides, tolerant crops will eventually need to be engineered. As the MBDTA binding pocket is highly conserved among plant species, these inhibitors will likely have broad-spectrum phytotoxicity. Given that the inhibitors do not bind at the active site, it should be possible to engineer tolerant crops with mutant DHDPS enzymes that are not susceptible to inhibition with minimal effects on enzyme function. Importantly, DHDPS inhibitors could also be used in conjunction with other herbicides as part of a combinatorial treatment to yield synergistic responses and circumvent resistance mechanisms to tackle the global rise in herbicide-resistant weeds. High-throughput chemical screen and analogue synthesis

Materials and methods
A high-throughput screen of a library of 87,648 compounds was conducted against recombinant DHDPS enzyme by the Walter and Eliza Hall Institute High Throughput Chemical Screening Facility (Melbourne, Australia). The o-ABA colourimetric assay employed assesses DHDPS activity via the formation of a purple chromophore that can be measured at 520-540 nm (Yugari and Gilvarg, 1965).
The assay was miniaturised, so it could be performed in 384-well plates. For the primary screen, reactions comprised 0.5 mgÁmL À1 DHDPS, 0.5 mM sodium pyruvate, and 0.5 mM ASA. Library compounds were added at final concentrations of 20 mM, with DMSO concentrations kept at 0.4% (v/v). After ASA addition, reactions were incubated at 25˚C for 15 min, before a final concentration of 350 mM HCl was added to stop the reaction. o-ABA was subsequently added to a final concentration of 0.44 mgÁmL À1 , plates incubated at room temperature for 1 hr, and absorbance quantified at 540 nm. Vehicle (DMSO) was used as positive controls, and negative controls lacked ASA. For the secondary screen, 11-point dose-response curves were generated using the same reactions as described above. A counter screen was conducted using the same set-up albeit without the inclusion library compounds before the addition of 350 mM HCl. Library compounds were then added after the reaction was stopped, followed by o-ABA to a final concentration of 0.44 mgÁmL À1 . The plates were subsequently incubated at room temperature for 1 hr, and absorbance was quantified at 540 nm. Analogues were designed and synthesised using the methods described in previous and contemporary work .

Expression and purification of A. thaliana DHDPS enzymes
Both DHDPS isoforms from A. thaliana were expressed and purified as previously described (Hall et al., 2021). Briefly, AtDHDPS isoforms were expressed in Escherichia coli BL21 (DE3) cells, with AtDHDPS2 requiring the GroEL/ES chaperone complex to facilitate correct folding. Purification was performed using immobilised metal affinity chromatography. Lastly, fusion tags were cleaved by human rhinovirus 3C or tobacco etch virus protease for AtDHDPS1 and AtDHDPS2, respectively, whilst simultaneously dialysing into storage buffer (20 mM Tris, 150 mM NaCl, 0.5 mM TCEP, pH 8.0).

Enzyme kinetics
DHDPS enzyme activity was determined using the DHDPS-DHDPR coupled assay as previously described by measuring the oxidation of NADPH (Atkinson et al., 2013;Hall et al., 2021). Assays were carried out in a Cary 4000 UV/Vis spectrophotometer at 30˚C with substrates fixed at the previously determined Michaelis-Menten constant values (Griffin et al., 2012;Hall et al., 2021). Inhibitor was titrated against AtDHDPS enzymes, and reactions were incubated at 30˚C for 12 min before initiation with ASA. Initial velocity data were normalised against a vehicle (DMSO) control and analysed using Equation 1 (log(inhibitor) vs. normalised response -variable slope, GraphPad Prism v 8.3). Dose responses were performed with three technical replicates for each concentration of compound. Dose responses were repeated with three biological replicates, each using a new stock of reagents.
where Y is the normalised rate, logIC 50 is the logarithmic concentration of ligand resulting in 50% activity, X is the concentration of ligand, and Hill Slope is the steepness of the curve.

X-ray crystallography
AtDHDPS1 was co-crystallised as previously described in the presence of MBDTA-2 (Hall et al., 2021).  , 2018). A total of 1800 diffraction images were collected with 0.1˚oscillation using an EIGER 16M detector at a distance of 350 mm, with 20% beam attenuation for a total exposure time of 18 s. X-ray data were integrated using XDS (Kabsch, 2010) and scaled with AIMLESS (Evans and Murshudov, 2013) before phases were determined by molecular replacement through Auto-Rickshaw (Panjikar et al., 2005) with AtDHDPS1 (PDB ID: 6VVI) used as a search model (Hall et al., 2021). Manual building was performed in COOT Emsley et al., 2010 followed by refinement employing REFMAC5 in the CCP4i2 (v7.0) software suite (Emsley et al., 2010;Murshudov et al., 2011;Winn et al., 2011). SMILES string of the inhibitor (MBDTA-2) was processed through AceDRG to generate the coordinate and cif file (Long et al., 2017). Validation was completed using MolProbity (Chen et al., 2010). The structure of MBDTA-2 bound to AtDHDPS1 is deposited in the Protein Data Bank as 7MDS.

Cell lines
Cell lines used were sourced from the American Type Culture Collection and were authenticated using STR DNA profiling, and no mycoplasma contamination was detected.

Toxicity assays
The toxicity of inhibitors against human HepG2 and HEK293 cell lines was assessed using the MTT viability assay as previously described (Soares da Costa et al., 2012). In brief, the cells were suspended in Dulbecco-modified Eagle's medium containing 10% (v/v) fetal bovine serum and then seeded in 96-well tissue culture plates at 5000 cells per well. After 24 hr, cells were treated with 50-400 mM of MBDTA-1 or MBDTA-2, such that the DMSO concentration was consistent at 1% (v/v) in all wells. Alternatively, cells were treated with the cytotoxic defensin protein at 100 mM (Baxter et al., 2017). After treatment for 48 hr, MTT cell proliferation reagent was added to each well and incubated for 3 hr at 37˚C. The percentage viability remaining reported is relative to the vehicle control of 1% (v/v) DMSO. Assays were performed in three biological replicates, using a different batch of reagents and cells.
of the MX2 beamline at the Australian Synchrotron, part of ANSTO and employed the Australian Cancer Research Foundation (ACRF) detector. We acknowledge the CSIRO Collaborative Crystallisation Centre (http://www.csiro/C3; Melbourne, Australia). We also thank the La Trobe University Comprehensive Proteomics Platform for providing infrastructure support.

Additional information
Competing interests Tatiana P Soares da Costa, Belinda M Abbott, Matthew A Perugini: is listed as an inventor on a patent pertaining to inhibitors described in the manuscript. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.