Introduction

Herbicides have been a mainstay in agriculture to protect crops against weed infestations for more than 70 years. Currently, agricultural practices require agrochemicals to be highly efficient and selective and environmentally and ecologically safe1. Branched-chain amino acids (BCAAs; valine, leucine, and isoleucine) are fundamental building blocks of proteins that are synthesized in bacteria, plants, and fungi but not in animals. Therefore, inhibitors targeting the BCAA synthesis pathway exhibit high selectivity among animals, plants and microorganisms. Therefore, enzymes integral to the BCAA biosynthetic pathway are considered prime targets for the development of selective herbicides2,3,4. The initial steps in the BCAA biosynthetic pathway involve acetolactate synthase (AHAS)5, acetohydroxy acid isomeroreductase (KARI)6, and dihydroxyacid dehydratase (DHAD) (Fig. 1a)7. To date, more than 50 herbicides targeting AHAS have been developed and widely used8. While these classes of herbicides have high efficacy and established safety profiles, the emergence of weed resistance presents a significant problem for their long-term usage. Not only are AHAS-inhibiting herbicides encountering the development of resistance, but most other herbicides with alternative targets have also succumb to resistance issues. This underscores the urgent need to discover active herbicides with alternative modes of action.

Fig. 1: Structure and function of DHAD and its catalytic mechanism.
figure 1

a Branched-chain amino acid biosynthetic pathway. b Amino acid sequence alignment of DHADs from Arabidopsis thaliana (AtDHAD, 5ZE4), Spinacia oleracea (SoDHAD, XP_021849033.1), Synechocystis sp. PCC 6803 (SnDHAD, 6NTE), Mycobacterium tuberculosis (MtbDHAD, 6OVT) and Escherichia coli (EcDHAD, ACX41805.1). The truncated sequences primarily display key amino acid residues within the catalytic active site of the protein. c Superposition of the three-dimensional structures of AtDHAD and MtbDHAD and residues in the active pocket. d Superposition of the three-dimensional structures of AtDHAD–AA (9JPI) and AtDHAD (5ZE4). Conformational changes in amino acid residues in the active pocket.

Full size image

DHAD, a member of the IlvD/EDD protein family, is encoded by a single gene in plants9. In contrast to the well-researched enzyme AHAS, DHAD has not been as extensively explored as a target for the development of herbicides. Dennis H. Flint pioneered the synthesis of a series of derivatives that aimed to inhibit DHAD with inhibition constants (Ki values) in the submicromolar range10. However, despite their effectiveness as DHAD inhibitors, these compounds do not exhibit herbicidal activity10. Leveraging a resistance gene-directed approach, the natural product aspterric acid (AA) was found to target DHAD and exhibits herbicidal activity against glufosinate-resistant Arabidopsis thaliana at a concentration of 250 µM2. Subsequently, the Zhou group reported the crystal structure of AA with A. thaliana DHAD (AtDHAD), which shows that hydroxycarboxylic acid moiety binds with Mg2+ and that hydrogen bonding networks and hydrophobic interactions stabilize AA in the enzyme active site11. The well-understood interactions between DHAD and this inhibitor represent a significant breakthrough in the rational design of DHAD inhibitors. Recently, DHAD has gained attention not only as a target for herbicidal action but also as a promising candidate for the precise growth control of microbes2,3,7. In 2019, the Ding group reported that tartronic acid (TA) inhibited the growth of cyanobacteria by targeting DHAD3. Despite our comprehensive understanding of the structure and function of DHAD, designing inhibitor molecules for this novel target remains a significant challenge. However, innovative approaches, including structure-based iterative learning12, peptidomimetics13,14, and virtual synthon hierarchical enumeration screening (V-SYNTHES)15,16, offer promising pathways forward. Nevertheless, when the mechanism of action of an inhibitor on its target is not completely understood, the reliability of applying existing data models to swiftly and precisely identify target inhibitors remains uncertain.

In this study, leveraging a model of substrates bound to DHAD, we adopt a substrate-based strategy to design a series of benzoxazinone derivatives incorporating an α-hydroxycarboxylic acid fragment and develop an efficient method to synthesize target compounds. Microscale thermophoresis (MST) assays reveal that compounds I-6c, I-6e, I-6g, I-6j, and I-6i have stronger binding affinities for DHAD than the natural substrate does. In addition, the cocrystal structure of AtDHAD in complex with I-6e is determined at 2.19 Å resolution, revealing how the inhibitor interacts within the active site pocket. These results are instrumental in guiding the future design of targeted DHAD inhibitors with significant potential for agricultural use.

Results

Structure of DHAD and its active site

All DHADs from different species contain [2Fe-2S] or [4Fe-4S] clusters, which are essential for the catalytic function of these enzymes17. The [4Fe-4S] cluster was initially observed in DHAD from Escherichia coli, while the [2Fe-2S] cluster was first reported in DHAD from spinach17,18. Although the [2Fe-2S] cluster is more stable in the presence of oxygen than the [4Fe-4S] cluster, both DHADs can be inactivated under aerobic conditions3,7. The oxygen sensitivity of DHAD proteins does indeed present a challenge for experimental studies, particularly when aiming to elucidate their catalytic mechanism and interactions with inhibitors.

Recently, crystal structures of DHADs from A. thaliana2, Mycobacterium tuberculosis19, and Synechocystis sp. PCC 68033 have been obtained, which provide insights into how DHAD substrates interact with this enzyme. Amino acid sequence alignment of revealed that the sequence similarity between A. thaliana and spinach DHADs exceeded 83% (Supplementary Table 1), whereas greater sequence divergence was observed between plant DHADs and those from fungi and bacteria (Supplementary Fig. 1). Despite structural variations in holo-DHADs across species, the key amino acid residues that interact with iron‒sulfur clusters and Mg2+ ions remain highly conserved (Fig. 1b)2,3,19. Comparative analysis of a single subunit from the crystal structures of AtDHAD (PDB ID: 5ZE4) and M. tuberculosis DHAD (MtbDHAD, PDB ID: 6OVT) (Fig. 1c) revealed that the configurations of the three cysteine residues coordinating with the 2Fe-2S cluster were essentially identical. However, a notable shift was observed in the positions of the Asp140 and Glu463 residues associated with the binding of Mg2+. Additionally, the spatial arrangements of several key residues within the active site involved in substrate interactions (Phe181, Ser489, and Tyr215) were altered. These differences can be exploited to develop species-specific inhibitors that selectively target the DHAD enzyme in one organism without affecting the same enzyme in another, which is particularly valuable in the development of selective herbicides. The cocrystal structure of AtDHAD–AA has been reported (PDB ID: 9JPI)11. When AA binds to the active site of AtDHAD, Tyr215, Phe181 and Asp140 exhibit significant conformational changes, as shown in the superimposed crystal structure (Fig. 1d). Furthermore, the mutations I177F and V178F in AtDHAD have been shown to decrease the stability of AA binding to DHAD. On the other hand, during the enzymatic reaction catalyzed by DHAD, the coordination interactions between the substrate and the metal ions (iron‒sulfur clusters and Mg2+) and the hydrogen bonding interactions with neighboring amino acids are crucial10,19. These interactions provide a theoretical foundation for designing the molecular framework of targeted DHAD inhibitors.

Substrate-based molecular design

In the model of AtDHAD bound with its natural substrate, 2,3-dihydroxy-3-methylbutanoic acid (Fig. 2a), the α-hydroxycarboxylic acid moiety not only coordinates with the iron‒sulfur clusters but also engages with Mg2+ and forms hydrogen bonds with Glu463. These findings suggest that the α-hydroxycarboxylic acid fragment is a critical component and should be included in the design of improved DHAD inhibitors. In addition to coordination interactions, the aromatic amino acid residues Tyr215 and Phe181 establish hydrophobic contacts with the substrate. This observation indicates that hydrophobic interactions are another fundamental aspect of the inhibitor structure. These key binding residues are also observed in the crystal structure of the AtDHAD–AA complex. These findings imply that incorporating aromatic groups into the inhibitor skeleton may allow the formation of π‒π interactions with neighboring residues Tyr215 or Phe181, as well as hydrophobic interactions with Ile177. Additionally, Ser489 interacts with the substrate via hydrogen bonds, which increases the stability of a small molecule in the active cavity and the binding affinity of an inhibitor to the enzyme.

Fig. 2: Design and synthesis of α-hydroxycarboxylic acid derivatives.
figure 2

a Design strategy for compounds I-6 and the model of the substrate (2,3-dihydroxy-3-methylbutanoic acid) and I-6 docked into AtDHAD. Light blue indicates the substrate with AtDHAD, and the light coffee color indicates I-6 with AtDHAD. b Kd value of compound I-6 binding to AtDHAD and the inhibitory effect of 25 and 50 μg/mL I-6 on A. thaliana germination. The data are presented as the mean ± SD. c The general synthetic scheme for the title compounds. d Cs2CO3-initiated α-C (sp3)-H hydroxylation of carbonyl derivatives I-5a − I-5m and target compounds I-6a − I-6n. The yields shown are for the isolated products. The crystal structures of compounds I-5b and I-6c are shown. Source data are provided as a Source Data file.

Full size image

Additionally, a benzoxazinone skeleton from the pesticide fragment bank attracted our attention because of its extensive biological activity in the pharmaceutical and agrochemical fields20,21. This core structure is particularly prominent in agrochemicals and serves as a key functional group in several commercial herbicides, including flumioxazin, thidiazimin, and trifludimoxazin. Therefore, we chose two molecular fragments to form a DHAD-targeted inhibitor scaffold (I-6). Molecular docking studies demonstrated that the inhibitors could stably bind within the DHAD active site, utilizing the α-hydroxycarboxylic acid fragment for metal ion coordination and the benzoxazinone skeleton for forming hydrophobic interactions with aromatic amino acids. The binding model of compound I-6 with DHAD is similar to that of the substrate (Fig. 2a). This finding was validated by MST assays, which demonstrated the strong binding affinity (Kd = 2.79 µM) of I-6 for DHAD in vitro. Moreover, 50 μg/mL I-6 completely inhibited the germination of A. thaliana seeds (Fig. 2b). These findings not only confirm the viability of our rational design strategy but also provide a foundation for further studies on the interaction mechanisms between these inhibitors and DHAD.

Synthesis of the α-hydroxycarboxylic acid derivatives

A comprehensive synthesis scheme is depicted in Fig. 2c, with the detailed methods provided in Supplementary Note 1. The most challenging aspect of the synthesis route was determining a convenient and efficient method for the synthesis of α-hydroxycarboxylic acid fragments. Although several methods for C − H hydroxylation have been established22,23,24, recent research has not yet fully realized green and environmentally benign synthetic approaches that avoid the use of expensive metal catalysts, hazardous stoichiometric oxidants, and reductants. In this study, we achieved α-C(sp3)-H hydroxylation of carbonyls by employing cost-effective atmospheric oxygen (O2) as an oxidant and cesium carbonate as a catalyst at ambient temperature.

According to previous studies, keto–enol tautomerism is among the core steps for the α-C(sp3)-H hydroxylation of carbonyl groups22,23. Experiments conducted without the addition of alkali revealed no detectable products other than I-4g (Supplementary Table 2, entry 1). However, the introduction of various stoichiometries of inorganic alkali into the reaction mixture significantly enhanced the reaction efficiency. Notably, the use of Cs2CO3 as a base achieved impressive yields exceeding 95% (entries 2-5). A recent study by Li’s group focused on the use of the organic base TBD for α-C − H hydroxylation25. However, in our experiments, organic bases such as TBD, DBU, t-BuOK, and Et3N delivered yields of less than 70% (entries 6-9). In contrast, our results demonstrate that inorganic alkalies outperform their organic counterparts in catalyzing this specific reaction. Among all the inorganic bases tested, Cs2CO3 was the most effective and was considered the optimal reagent for achieving the α-C(sp3)-H hydroxylation of carbonyl groups in high yields.

We subsequently explored the effects of varying stoichiometries of Cs2CO3 on the reaction (entries 10-13). When the quantity of Cs2CO3 exceeded 1.0 equivalents, the starting material I-4g was almost completely converted to the desired product I-5g. In contrast, the yield of I-5g significantly decreased in the presence of less than 1 equivalent of Cs2CO3. However, extending the reaction duration to 2 h or 3 h gave I-5g in excellent yield (greater than 90%) (entries 14-15). Additionally, we determined that light exposure had no significant effect on the reaction, as the yield remained consistent regardless of whether the reaction was conducted in the dark or light (entries 16 and 2). We then assessed the effect of the solvent on C − H hydroxylation (entries 17-24). Alternative solvents such as toluene, DMSO, DCM, THF, 1,4-dioxane, n-butanol, MeOH, and acetone were tested in lieu of DMF. However, none of these solvents outperformed DMF. Notably, alcohol solvents, specifically n-butanol and MeOH, promoted the hydrolysis of ester groups under mildly alkaline conditions, which resulted in a higher conversion of I-4g but yielded only a minimal quantity of the target product I-5g (entries 22, 23). Furthermore, following the methodology of the Gnanaprakasam group26, we attempted C − H hydroxylation using t-BuOK or TBD in DMSO (entries 25, 26), but these conditions produced only modest amounts of I-5g (yield < 30%). These experiments conclusively demonstrated that the synergistic combination of Cs2CO3 and DMF constituted the most favorable reaction conditions for the C(sp3)-H hydroxylation of the benzoxazinone ring. We subsequently successfully converted the hydroxylated intermediates into the final compounds through ester bond hydrolysis. Moreover, we obtained single crystals of compounds I-5b and I-6c directly from the NMR solvent CDCl3 that were suitable for characterization using single-crystal X-ray diffraction analysis. The detailed crystal structure data are provided in Supplementary Table 3 and 4. The structural characterization data for all the compounds are presented in Supplementary Note 1. The 1H NMR and 13C NMR spectra of compounds I-5a to I-5m and I-6a to I-6n are shown in Supplementary Fig. 2–55.

Possible mechanism of the C(sp3)-H α-hydroxylation of benzoxazinone

To date, the reaction mechanism for C(sp3)-H hydroxylation has not been conclusively determined. Under nitrogen, the reaction failed even after 24 h. Introducing radical inhibitors such as TEMPO or BHT did not affect the yield (Supplementary Fig. 56a). These results revealed that atmospheric oxygen is the oxygen source and that the reaction follows a nonradical pathway. Moreover, the NH-substituted compound I-3a did not react (Supplementary Fig. 56a), whereas the N-benzyl-substituted compound I-4g reacted normally. This discrepancy could be due to compound I-3a potentially forming two transition states, with one being incapable of C(sp3)-H α-hydroxylation. After comparing the Gibbs free energy (∆G) of the two transition states, ∆G2 for the C = N transition state was found to be lower than ∆G1 for the C = C transition state (14.02 kcal/mol vs. 19.11 kcal/mol), as calculated using the Gaussian 09 software (Supplementary Fig. 57). These results support the conclusion that the hydroxylation reaction is highly dependent on the keto-enol tautomerism of the C = C double bond rather than on the C = N double bond in the benzoxazinone ring.

The potential transition state was further elucidated through real-time NMR monitoring (Supplementary Fig. 56b). The C(sp3)-H (Ha) signal (δ = 5.79 ppm) nearly vanished within 30 min, whereas the signals for Hb and Hc gradually diminished. Concurrently, the signals for Hb’ and Hc’ appeared and then weakened over time, disappearing completely after 60 min. Analysis of the changes in chemical shifts of the hydrogen proton signals over time revealed the initial occurrence of α-C(sp3)-H carbonyl keto-enol tautomerism (I-4g-1), which subsequently leads to the generation of an intermediate that is stable in the reaction medium. The reaction mechanism of oxygen insertion into a C − H bond to form a C − O − O − H intermediate is widely recognized23,25,27. In our experiments, the chemical shifts of the other three hydrogen signals on the benzoxazinone ring shifted from 7.09 − 6.83 ppm to 6.68 − 6.41 ppm. This shift was attributed to the O − O − H group decreasing the charge density on the benzoxazinone ring, leading us to postulate that the intermediate structure is most likely I-4g-2 (Supplementary Fig. 56b). Unfortunately, despite its apparent stability in the reaction system, this intermediate could not be isolated. Additionally, the introduction of a small quantity of trifluoroacetic acid (CF3COOH, TFA) to the reaction system resulted in NMR spectra that matched those of the control compound I-5g with the addition of TFA. Therefore, the most plausible reaction mechanism is that under alkaline conditions, α-C (sp3)-H carbonyl keto-enol tautomerism allows the generation of the reactive intermediate I-4g-2 in the presence of atmospheric oxygen. Subsequent acid catalysis facilitates the cleavage of the peroxyl ether bond, leading to the formation of the final product I-5g.

Optimizing inhibitor activity on the basis of the target structure

Structure-based molecular optimization integrates detailed knowledge of the biological target into the drug design process, leading to more efficient and potentially more successful drug discovery and development28. The surface model of the compound I-6–DHAD structure (Supplementary Fig. 58) shows that the active pocket of AtDHAD is a large internal cavity that is occupied by N-substituent groups and the benzoxazinone ring. These observations imply that the introduction of hydrophobic or bulky groups at these positions may increase the binding affinity of the inhibitor for DHAD (Supplementary Table 5). Furthermore, halogen bonding can be exploited for additional protein‒ligand interactions29. Leveraging data from theoretical predictive models, we introduced a variety of substituents, such as halogens, hydrogen atoms, alkyl groups, alkoxypropargyl, and various substituted phenoxy groups, on either the benzene ring or at the N-position (Fig. 2d).

Inhibition of A. thaliana seed germination and revealing structure-activity relationship

We tested the inhibitory effects of all the compounds on A. thaliana seed germination (Supplementary Table 6). These compounds generally had a more pronounced inhibitory effect on root development than on hypocotyl growth. Notably, compound I-6g, with a N-benzyl substituent, inhibited both root and hypocotyl growth by more than 99%, outperforming the other N-substituted analogs (methyl, ethyl, and n-propyl) when tested at a concentration of 100 μg/mL. Among the compounds with unsaturated alkyl substituents, compound I-6e (propargyl) showed greater inhibitory activity than I-6d (allyl) and I-6f (isopentenyl) did, yet none were as effective as I-6g (benzyl). As a result, we maintained the benzyl group at the N-position and redirected our efforts to investigate how varying the substituent at the 7-position of the benzoxazinone nucleus affects inhibitory activity. Replacing the fluorine atom with other substituents, such as hydrogen (I-6h), chlorine (I-6i), bromine (I-6j), a methyl group (I-6k), and a methoxy group (I-6l), resulted in decreased inhibitory activity in comparison with that of compound I-6g. Compound I-6m, which contains a substituent at this position with large steric hindrance, demonstrated potent inhibition of root growth (~90%) but lower inhibition of hypocotyl growth at the same concentration.

Given the potent inhibitory effects of these compounds on A. thaliana seed germination, compounds I-6c, I-6e, I-6g, and AA were applied for postemergence treatment of two-week-old A. thaliana plants, which resulted in the cessation of new leaf production and the onset of leaf wilting (Fig. 3a). After 20 days of treatment with 200 μg/mL compound, the fresh weight inhibition by compound I-6g reached 76.01%, which was the highest among all the tested compounds and was significantly greater than that of AA, which exhibited 55.96% inhibition (Fig. 3b). Additionally, confocal laser scanning microscopy (CLSM) was used to observe the localization of DHAD in Nicotiana benthamiana leaves 24, 48, and 72 h after treatment with water or compound I-6e. The results revealed that DHAD was predominantly localized in chloroplasts (Supplementary Fig. 59). Our findings suggest that compounds with substantial steric bulk at the N-position are more effective DHAD inhibitors and that halogen F is a favorable substituent at the 7-position of the benzoxazinone ring. These experimental observations align well with our computational predictions.

Fig. 3: Effects of the α-hydroxylated acid derivatives on Arabidopsis thaliana growth and validation of targeting AtDHAD.
figure 3

a Col-0 seedlings were grown in soil in pots and treated with AA, I-6c, I-6e, or I-6g at a dose of 200 μg/mL by foliar spray. b Fresh weight inhibition rates of A. thaliana plants 20 days after treatment with AA, I-6c, I-6e, or I-6g. The data are presented as the mean ± SD. One-way analysis of variance was used, and the data for the test compounds were compared with those of AA. Three independent replicates were performed. c Wild-type A. thaliana (Col-0) seedlings were grown on vertically oriented 1/2 MS media alone or media supplemented with 25 μg/mL BCAAs and treated with either 50 μg/mL AA or I-6g. Control seedlings were treated with 50 μg/mL DMSO. All the seedlings were grown under etiolation conditions for 7 days. d KEGG enrichment analysis of differentially expressed genes in the leaves of A. thaliana after treatment with I-6g. The vertical axis displays the pathway name, and the horizontal axis presents the Rich factor (sample number/background number). The greater the Rich factor is, the greater the degree of enrichment. Generally, the FDR ranges from 0–1, and the closer it is to zero, the more significant the enrichment. Kd values of I-6e (e), AA (f) and TA (g) binding to AtDHAD. Each set of data comprises three replicates. R2 is the correlation coefficient, and R2 values closer to 1 represent a better fit. The data are presented as the mean ± SD. Source data are provided as a Source Data file.

Full size image

Inhibition of DHAD by the α-hydroxycarboxylic acid derivatives

To investigate whether compound I-6g exerts its inhibitory effects on A. thaliana by specifically targeting the BCAA synthesis pathway, we conducted amino acid supplementation experiments. We attempted to grow A. thaliana Col-0 seeds on 1/2 MS media supplemented with 50 µg/mL I-6g, but germination was completely inhibited (Fig. 3c). Notably, the addition of 25 µg/mL BCAAs to the I-6g-supplemented medium restored germination and normal seedling development, with root and hypocotyl lengths similar to those observed in the DMSO-treated control (Supplementary Fig. 60a and 60b). The same effect was noted in the AA-treated positive control group, both on standard 1/2 MS medium and that with added BCAAs. These results implicate α-hydroxycarboxylic acid derivatives as inhibitors of BCAA biosynthesis.

We subsequently applied compound I-6g to A. thaliana plants via stem and leaf spraying for transcriptomic analysis. The results revealed that the top 10 pathways impacted by the treatment, as identified by KEGG database analysis, included both BCAA biosynthesis and degradation (Fig. 3d). As the duration of exposure to the compound increased, a broader spectrum of metabolic pathways was influenced, encompassing pathways such as photosynthesis–antenna proteins, starch and sucrose metabolism, and cysteine and methionine metabolism (Supplementary Fig. 60c). We further expressed and purified AtDHAD proteins and measured the binding affinity (Kd value) of the agent for the protein using MST assays30,31. The MST data are shown in Fig. 3e, g. The Kd values for compounds I-6c, I-6e, I-6g, I-6j, and I-6k were all less than 3 µM, indicating their stronger affinities for the enzyme than that of the positive controls TA (Kd = 506.60 µM) and AA (Kd = 3.41 µM, Supplementary Table 7). These results support that α-hydroxycarboxylic acid derivatives target DHAD.

Cocrystal structure of AtDHAD-I-6e

Among all the compounds, I-6e exhibited the highest affinity for AtDHAD, with a Kd value of 1.00 µM. Consequently, I-6e was selected as a ligand for cocrystallization with DHAD. The best crystals diffracted to a resolution of 2.19 Å (Supplementary Table 8). Despite the use of racemic compounds in our studies, only one enantiomer was observed to bind to DHAD.

The cocrystal structure of AtDHAD with I-6e (Fig. 4a) reveals a clear electron density map that highlights the presence of ligand I-6e in the active pocket. The electron density contours (2Fo-Fc maps, contoured at 1.0 σ) for the ligand I-6e in different orientations are also shown in Fig. 4a. Additionally, the Fo-Fc omit map of I-6e (contoured at 1.0 σ) is presented in Supplementary Fig. 61. The [2Fe-2S] cluster, anchored at the base of the binding pocket, forms coordination bonds with three conserved cysteine residues, Cys211, Cys139, and Cys66. This leaves an open coordination site on one of the iron atoms, which is available to interact with ligand I-6e, aligning with previous studies4,18,19. Furthermore, Mg2+ ion in the pocket coordinates with three acidic residues, Asp98, Asp140, and Glu463, as well as with the hydroxyl group of I-6e, securing the small molecule in the active site. Notably, the carboxylic acid group of I-6e establishes a hydrogen bond with Ser489, and the fluorine atom interacts with the nearby residues Ser287 and Thr288 via a water-mediated hydrogen bond, further enhancing the binding stability of I-6e to DHAD (Fig. 4b, c).

Fig. 4: Mechanism of the interaction between I-6e and DHAD and the pharmacophore of the targeted DHAD inhibitor.
figure 4

a Cocrystal structure of AtDHAD with I-6e and electron density contours (2Fo-Fc maps contoured at 1.0 σ) for the ligand in different orientations. b Interaction mode of compound I-6e in the active pocket of AtDHAD (9L8R). The blue dashed lines represent hydrogen bonding interactions. c Surface model of the active pocket of the AtDHAD–I-6e complex. The images were generated with PyMOL software. d Superimposition of the protein crystal structures of AtDHAD with and without the inhibitor bound. e MST dose–response curve showing the interaction between I-6n and AtDHAD. The data are presented as the mean ± SD. f Col-0 seedlings were grown on 1/2 MS medium and treated with AA, I-6g, or I-6n at a dose of 50 μg/mL. Source data are provided as a Source Data file.

Full size image

Moreover, superimposing the three-dimensional DHAD structures (Fig. 4d) revealed distinct similarities in the binding interactions of I-6e and AA within the enzyme’s active pocket. Notably, the π‒π stacking interactions with Tyr215 and Phe181 were less prominent for I-6e. Correspondingly, the conformations of these aromatic residues remained akin to those observed in the absence of a bound inhibitor. Our comparative crystallographic analysis revealed additional interactions within the DHAD enzyme that are critical for inhibitor binding. These findings facilitate the design of a diverse array of inhibitor molecules with various structures that can specifically target DHAD.

α-hydroxycarboxylic acid is a crucial pharmacophore of targeted DHAD inhibitors

Analysis of the cocrystal structure of AtDHAD with I-6e underscores the significant role of the α-hydroxycarboxylic acid moiety in enhancing the interaction between the inhibitor and its target. Thus, this raises the question of whether the hydroxyl group is essential for the efficacy of this series of DHAD inhibitors. Therefore, we synthesized a derivative, compound I-6n, which lacks this hydroxyl group. Subsequent in vitro binding assays revealed that I-6n displayed significantly reduced affinity, with a Kd of 1355.11 µM (Fig. 4e), which is nearly three orders of magnitude lower than that of I-6g (Kd = 1.43 µM). In line with these findings, the inhibitory effect of I-6n on seed germination was also markedly decreased compared with that of I-6g, even at a high concentration of 100 μg/mL (Fig. 4f). The combined in vitro and in vivo experimental results highlight the critical importance of the α-hydroxycarboxylic acid group for the activity of DHAD these inhibitors, confirming its status as a key pharmacophore.

Using structure of I-6e to design targeted DHAD inhibitors

The glasshouse tests to examine the preemergence and postemergence herbicidal activities of compounds I-6a to I-6m provided important insights into their efficacy and selectivity. In these assays, water with 5% DMSO was used as a blank control, and penoxsulam, which targets AHAS, was used as a positive control. The ability of the synthetic compounds to inhibit weed or crop growth at a dose of 150 g ai/ha was evaluated. The results, as summarized in Table 1 and Supplementary Fig. 62, revealed that compounds I-6c, I-6e, I-6g, I-6i, I-6j, and I-6m inhibited the growth of the six weeds tested by more than 50%, as determined by weight. Notably, Avena fatua was the most effectively controlled by these compounds. In addition to their herbicidal activity, the safety of these compounds to crops was assessed upon preemergence treatment. Most of the compounds were safe for rice, with a fresh weight inhibition rate of less than 5%. Some compounds with certain crops, such as I-6j for soybean, I-6m for corn, and I-6b for wheat, had no significant harmful effects. Notably, compound I-6e demonstrated over 70% efficacy against broadleaf weeds such as Abutilon theophrasti Medicus and Amaranthus retroflexus and 100% efficacy against gramineous weeds such as Setaria viridis P. Beauv. and A. fatua and was less harmful to rice. Thus, compound I-6e is a particularly promising candidate as a lead structure for targeted DHAD inhibitors because of their potent herbicidal activity and crop safety.

Table 1 Average fresh weight inhibition rates (%) after preemergence and postemergence treatment
Full size table

Discussion

DHAD plays a critical role in the biosynthesis of BCAAs, which are essential for the growth and development of plants. This enzyme is responsible for dehydrating dihydroxy-acid substrates, yielding α-keto acids that serve as precursors of leucine, isoleucine, and valine. Although DHAD has been revealed to be a potential herbicidal target, no lead compounds with broad-spectrum herbicidal activity that target this enzyme have been identified2,3. Protein structure superposition revealed that the three cysteine residues involved in iron‒sulfur coordination exhibit consistent conformations (Fig. 1c). However, the binding of the inhibitor to the active site induces a conformational shift in Phe181 and Tyr215, and the I177F mutation alters the stability of the binding of the inhibitor AA to DHAD. Also considering the binding model of the substrate to DHAD (Fig. 2a), coordination interactions with metal ions, hydrophobic interactions with Tyr215, Phe181, and Ile177, and hydrogen bonding interactions with Ser489 not only are crucial for the natural substrate’s binding with DHAD but also constitute the fundamental forces enabling the inhibitor molecules to stably bind to the active site of DHAD.

Adopting a substrate-based molecular design approach, we utilized the α-hydroxycarboxylic acid moiety as a foundation unit and introduced an aromatic fragment to form hydrophobic interactions. The benzoxazinone ring is a prevalent scaffold in both medicinal chemistry and agrochemical design. Therefore, we combined the benzoxazinone moiety with α-hydroxycarboxylic acid units to design an inhibitor skeleton. Molecular docking simulations indicated that these derivatives can engage in coordination interactions with the 2Fe-2S cluster and Mg2+, in addition to forming hydrophobic contacts with Phe181 and hydrogen bonding interactions with Ser489. Our design strategy offers a different and effective approach for developing more potent and selective herbicidal compounds that target different sites of action.

Furthermore, by utilizing cost-effective atmospheric oxygen (O2) as an oxidant and cesium carbonate as a catalyst at ambient temperature, we developed an efficient and eco-friendly method for synthesizing α-hydroxycarboxylic acids without the need for expensive metal catalysts. In contrast to previous studies25,26,32, we discovered that an inorganic base (Cs2CO3) outperforms an organic base in our reaction and demonstrates compatibility with a broad spectrum of substrates. Unfortunately, this method does not allow control of the chirality of the compound. Additionally, during chiral resolution, the compound underwent spontaneous racemization in the solvent. This phenomenon was attributed to the formation of hydrogen bonds33,34, as supported by the detection of such bonds in the crystal structure of compound I-5b (Fig. 2c). As a result, the synthesized target compounds, which were employed for bioactivity evaluations, are racemic mixtures.

Furthermore, our research revealed that α-hydroxycarboxylic acid derivatives act as inhibitors of the BCAA synthesis pathway by targeting DHAD. BCAA deficiency leads to distinct phenotypes in plants, including stunted growth, underdeveloped root systems, and reduced root lengths9. In our study, treatment with these derivatives or AA induced identical phenotypic symptoms, which were reversed upon supplementation of BCAAs in the medium (Fig. 3c). Transcriptomic analysis further highlighted significant changes in BCAA biosynthesis and metabolism at both 24 h (Fig. 3d) and 72 h (Supplementary Fig. 60c), confirming the influence of hydroxycarboxylic acid derivatives on BCAA production. The in vitro assays revealed that the Kd value of I-6e with DHAD was 1.00 μM, indicating its stronger affinity for DHAD than that of dihydroxyisovalerate (Kd = 5.39 μM). Additionally, we successfully determined the crystal structure of the AtDHAD–I-6e complex at a resolution of 2.19 Å and observed that the binding patterns of I-6e and AA within the active pocket were markedly different. In addition to the coordination interaction, the presence of hydrogen bonds between the inhibitor and Ser489, as well as the interactions of the halogen atom of the inhibitor with Thr288 through a water-mediated hydrogen bond (Fig. 4b), contributed to the superior binding stability of I-6e to DHAD compared with that of AA. The I-6e binding site of DHAD paves the way for the design of structurally diverse targeted DHAD inhibitors. Furthermore, our findings suggest that the α-hydroxycarboxylic acid functional group is crucial for enhancing inhibitor binding stability and retaining in vitro herbicidal activity, indicating that it is an essential pharmacophore of targeted DHAD inhibitors. Notably, compound I-6e displayed a broad spectrum of herbicidal effects, most notably completely inhibiting the growth of S. viridis and A. fatua while causing little harm to rice at a dosage of 150 g ai/ha (Table 1).

In summary, we present a class of inhibitor molecules that target DHAD designed via a substrate-based design strategy. Within this class, I-6e has emerged as a particularly promising DHAD inhibitor because of its potent herbicidal efficacy and favorable safety profile in crops. Crystallographic analysis revealed that the binding pattern of I-6e with DHAD is distinct from that of the AA-DHAD interaction, highlighting a different site of action that can be exploited when designing inhibitory molecules. This discovery opens potential pathways for the development of more selective and efficient DHAD-targeted herbicides and provides a theoretical foundation for the development of targeted DHAD inhibitors, which hold great promise for agricultural applications.

Methods

Chemical synthesis and analysis

Reagents were purchased from Energy Chemical, Aladdin, JK Chemical, Macklin and other commercial suppliers. AA was purchased from GlpBio Technology. Biological reagents and consumables were purchased from Vazyme, Yeasen and other commercial suppliers. Primer synthesis was conducted by GenScript Biotechnology (Nanjing, China). Reactions were monitored by thin layer chromatography (TLC) on silica. Silica gel (200–300 mesh) for column chromatography was purchased from Qingdao Marine Chemical Inc., China. 1H NMR and 13C NMR spectra were collected with a JNM-ECZ500R instrument (500 MHz, JEOL, Japan) in CDCl3 or DMSO-d6 and with tetramethylsilane (TMS) as the internal standard. The NMR data were analyzed using MestReNova 14.0.0 software. High-resolution electrospray ionization mass spectrometry (HR-ESI–MS) data were obtained with an Agilent 6210 TOF LC‒MS spectrometer (Agilent, USA). High-performance liquid chromatography (HPLC) was conducted using an Eclipse XDB-C18 column (5 μm, 4.6 × 250 mm; AkzoNobel, China) on an Agilent 1260 liquid chromatograph (Agilent Technologies, USA). HPLC gradient elution was performed with a linear gradient of 60–95% MeOH–H2O over 15 min followed by 95% MeOH for 3 min and then 60% MeOH–H2O for 3 min; the flow rate was 1 mL/min. Compound X-ray single-crystal diffraction data were collected on a SMART APEX DUO CCD area detector diffractometer (Bruker AXS, Madison, WI) using Mo or Ga Kα radiation at 100 (2) K. The diffraction profiles were integrated, and structural analysis was performed using SAINT Plus software (Bruker AXS, Madison, WI) and the SHELXS97 program (University of Gottingen, Gottingen, Germany), respectively. The detailed crystal structure data are provided in Supplementary Table 3 and 4.

Real-time NMR monitoring

Real-time NMR was performed to monitor reaction process35,36. Because the monitored reaction proceeds slower in DMSO than in DMF, DMSO-d6 was used as the solvent instead of DMF-d7, and an excess of Cs2CO3 (1.5 eq) was added to ensure that the reaction proceeded completely. Compound I-4g (50 mg, 0.15 mmol) was dissolved in 3 mL of DMSO-d6, and the time after the addition of Cs2CO3 (0.225 mmol) and stirring at room temperature was measured. The mixture (volume of 600 μL) was quickly transferred to an NMR tube after 10 min, 20 min, 30 min, and 60 min, and NMR data were collected. The last sample (60 min) was analyzed again after the addition of TFA (30 μL) to the tube. As a control, compound I-5g (10 mg, 0.03 mmol) was mixed with 600 μL of DMSO-d6 and 0.045 mmol of Cs2CO3 and stirred at room temperature for 60 min. Then, 30 μL of TFA was added to the reaction mixture, and NMR data were collected. All NMR data were collected at 25 °C and analyzed using MestReNova 14.0.0 software.

Molecular docking

The crystal structure of AtDHAD (PDB ID: 5ZE4) was downloaded from the RCSB Protein Data Bank. The 3D structures of our compounds were prepared using SYBYL-X 2.0. The [2Fe-2S] cluster was chosen as the center of the binding pocket. The energies of the ligand and receptor were optimized on the SYBYL platform, and the other parameters were set to the defaults according to the Sybyl user guide. A total of 200 runs were performed, after which we selected the top-ranked conformation as the most suitable binding conformation. The total score was used as a reference for binding strength comparison37,38.

Plant materials and growth conditions

Wild-type A. thaliana Col-0 seeds were provided by L R. Wei. N. benthamiana seeds were provided by M F. Feng. A. thaliana seeds were incubated at 4 °C for 3 days before sowing. Afterward, A. thaliana was germinated in the cultivation substrate, transplanted into a small pot, and grown in a light incubator with a light/dark cycle of 8 h (22 °C)/16 h (22 °C) for 3‒4 weeks. N. benthamiana plants were grown in a growth chamber with a light/dark cycle of 16 h (25 °C)/8 h (25 °C). Weed and crop seeds were planted in soil and grown in a glasshouse under natural conditions at temperatures ranging from 20 °C to 32 °C.

Germination of A. thaliana seeds

A. thaliana seeds (Col-0) were surface sterilized with chlorine gas for 3 h and vernalized at 4 °C for 3 days before use. Afterward, the A. thaliana seeds were sown in Petri dishes (10 cm × 10 cm) containing 1/2 MS media (2.2 g/L MS salts, 10.0 g/L sucrose, and 8 g/L agar) supplemented with different concentrations of various agents (compounds I-6 to I-6n or AA at 3.13 µg/mL, 6.25 µg/mL, 12.5 µg/mL, 25 µg/mL, 50 µg/mL, or 100 µg/mL). Separately, BCAAs (mLeu:mIle:mVal = 1:1:1, total final concentration of 25 μg/mL) was added to 1/2 MS medium. ~12–15 surface-sterilized Col-0 seeds were sown on the 1/2 MS medium and cultivated at 22 °C in the dark for 7 days. Then, ten germinated seeds were randomly selected from each Petri dish, and the lengths of the hypocotyls and roots were measured. The root or hypocotyl inhibition rate was calculated as equation 1:

$${{{\rm{Inhibition}}}}\; {{{\rm{rate}}}}= ({{{\rm{Length}}}}\; {{{\rm{in}}}}\; {{{\rm{control}}}}\; {{{\rm{group}}}}\\ {{{-}}}{{{\rm{Length}}}}\; {{{\rm{in}}}}\; {{{\rm{treatment}}}}\; {{{\rm{group}}}})\\ /{{{\rm{Length}}}}\; {{{\rm{in}}}}\; {{{\rm{control}}}}\; {{{\rm{group}}}} * 100$$
(1)

All the experiments were independently repeated three times.

Postemergence treatment of A. thaliana

Compounds I-6e, I-6c, I-6g, and AA were dissolved in DMSO and diluted with 0.5% aqueous Tween 80 to a final concentration of 200 μg/mL, with a final concentration of DMSO of less than 5%. The prepared solutions (1 mL) were sprayed evenly on the stems and leaves of 2-week-old A. thaliana plants exhibiting good growth, and the plants were grown in a light incubator for 20 days.

Protein expression, purification, and cocrystal structure determination

The plasmid AtDHAD-pET was transformed into E. coli BL21 (DE3). After IPTG induction of protein expression, the bacterial precipitate was collected by ultracentrifugation. Next, lysozyme and benzonase nuclease were added and crushed using an ultrasonic crusher. After ultracentrifugation, the supernatant was collected and separated by purification using Ni-NTA affinity column in a Vigor anaerobic chamber. During E. coli BL21 (DE3) culture, FeCl3 and L-Ara were added to stabilize the [2Fe-2S] cluster in the protein. We obtained a high-purity AtDHAD protein by optimizing the expression and purification conditions. We obtained a high-purity AtDHAD protein by optimizing the expression and purification conditions2. The protein was diluted with Buffer C (150 mM NaCl, 25 mM Tris-HCl, pH 8.0) to a final protein concentration of 10 mg/mL. The mixed solution of protein and I-6e was incubated on ice for 1–2 h in an anaerobic box, and then crystallization was carried out by the sitting-drop vapor diffusion method at 16 °C. Diffraction of the tetragonal orange cocrystals was performed in 1.0 M ammonium sulfate and 0.1 M ammonium acetate (pH 5.24) in the absence of oxygen. All the crystals were flash-cooled in liquid nitrogen after being cryoprotected with glycerol. The data were collected at 100 K at Beam Line 19U1 in the Shanghai Synchrotron Radiation Facility (SSRF). All datasets were indexed, integrated, and scaled using the HKL3000 package39. Finally, the crystal structure data (Supplementary Table 8) were obtained with an R-free value of 0.2311, R-work value of 0.1814 and resolution of 2.19 Å.

MST assay

With purified AtDHAD in hand, buffer replacement and lysine residue labeling of the AtDHAD protein in solution (10 µM) were performed according to standard operating procedures using the Protein Labeling Kit RED-NHS 2nd Generation (MO-L011, NanoTemper). Ligand solutions at various concentrations (ranging from 10−10 M to 10−1 M) were incubated with 50 nM dye-labeled tagged protein at 4 °C for 30 min in assay buffer (3.2 mM Na2HPO4, 0.5 mM KH2PO4, 1.3 mM KCl, 135 mM NaCl, 0.05% Tween 20, pH 7.4). The samples were subsequently loaded into label-free standard capillaries (MO-K022, NanoTemper) for data collection on a Monolith NT.115 (MO. Control). The excitation power was 100% or 40%, and the MST power was medium. The MST data were analyzed by MO.Affinity Analysis, and GraphPad Prism 7.00 were used for plot construction.

Determination of subcellular localization via CLSM

The full-length coding sequence of AtDHAD was cloned and inserted into the p2300-YFP vector to generate the p2300-YFP-AtDHAD construct. The constructed plasmid p2300-YFP-AtDHAD was subsequently transformed into Agrobacterium tumefaciens strain GV3101 cells. After incubation in YEB liquid medium at 200 × g for 3 h, the bacteria were collected by centrifugation at 4000 × g for 4 min, and after discarding the supernatant, the cells were resuspended in infiltration buffer (10 mM MES, 10 mM MgCl2, pH 5.6, and 100 μM acetosyringone) to an OD600 of 0.6. Fully expanded leaves of N. benthamiana plants at the 6–7-leaf stage were infiltrated with Agrobacterium cultures using 1 mL needleless syringes and grown in a growth chamber at 25 °C with a 16 h light/8 h dark photoperiod. After 2 days of low-light culture, the plants were observed by confocal microscopy. Confocal images were captured with an inverted Zeiss LSM 900 Airyscan 2 and a 20× water immersion objective lenses. YFP was excited at a wavelength of 488 nm, and emission in the range of 497–520 nm was captured. To visualize chloroplasts, chlorophylls were excited at a wavelength of 488 nm, and emission in the range of 660–720 nm was captured. Images were processed using a Zeiss LSM 900 Airyscan 2. Signal overlap was observed by merging the YFP-AtDHAD and chlorophyll fluorescence signals.

Transcriptomic analysis

The stems and leaves of A. thaliana plants were sprayed with compound I-6g, and samples without roots were collected 24 h and 72 h later and frozen with liquid nitrogen. Then, RNA extraction and data analysis were performed according to standard methods (Shanghai Personalbio Technology Co., Ltd., China). DESeq was used to analyze differential gene expressions, and genes were considered differentially expressed when they met the criteria of |log2FoldChange | å 1 and P value < 0.05. KEGG enrichment analysis (KAAS, bidirectional best hit) of the differentially expressed genes considered the enrichment degree determined by the Rich factor, the false discovery rate (FDR) and the number of genes enriched in each pathway. The Rich factor is the ratio of the number of differentially enriched genes and the number of annotated genes in a pathway. The greater the rich factor is, the greater the degree of enrichment. Generally, FDR ranges from 0-1. The closer the value is to zero, the more significant the enrichment. The top 10 KEGG pathways with the smallest FDR values (the pathways enriched in the greatest number of differentially expressed genes), were selected for display (http://www.kegg.jp/). Three replicates were used in the experiment. The raw transcriptome data were uploaded to the NCBI database (PRJNA1229804).

Preemergence herbicidal activity

Weed and crop seeds with an emergence rate greater than 95% were evenly planted in small pots, provided sufficient water, and allowed to grow in a glass greenhouse for 12 h. Then, solutions of the test compounds were evenly sprayed on the surface of the soil at a dose of 150 g ai/ha. Each treatment contained three replicates. After 21 days, the weeds above the soil surface were removed, and the fresh weight was measured with a balance with a precision of 1/10,000 (BSA2245, BSISL). All experiments were independently repeated three times. The average fresh weight inhibition rates were calculated as eq. 2 and presented in Table 1.

$${{{\rm{Fresh}}}}\; {{{\rm{weight}}}}\; {{{\rm{inhibition}}}}\; {{{\rm{rate}}}}= ({{{\rm{Weight}}}}\; {{{\rm{in}}}}\; {{{\rm{control}}}}\; {{{\rm{group}}}}\\ -{{{\rm{Weight}}}}\; {{{\rm{in}}}}\,{{{\rm{treatment}}}}\,{{{\rm{group}}}})\\ /{{{\rm{Weight}}}}\; {{{\rm{in}}}}\,{{{\rm{control}}}}\,{{{\rm{group}}}} * 100$$
(2)

Postemergence herbicidal activity

Weeds and crops were grown to the 2–3-leaf stage in a greenhouse, with 10 plants in each pot. The test compound solutions were sprayed on the plants with a WPSH-500D bioassay spray tower (Nanjing Institute of Agricultural Mechanization) at a dose of 150 g ai/ha. The instrument was operated with a disk diameter of 50 cm, a nozzle aperture of 0.3 mm, a spray pressure of 0.3 MPa, a dropper diameter of 100 µm, and a nozzle flow rate of 90 mL/min. Each treatment consisted of three replicates. After 21 days of treatment, 10 weeds above the soil surface were removed, the fresh weight was measured, and the fresh weight inhibition rate was calculated. The average values obtained for the fresh weight inhibition rates are shown in Table 1.

Mapping and data processing

All the data were summarized in Microsoft Excel 2016. Histogram analysis and two-way analysis of variance were performed with GraphPad Prism 7.00. Image processing was performed with Adobe Photoshop 2021. The structures of the compounds were constructed with ChemBioDraw 14.0.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.