Research Article

Dynamic control of plant water use using designed ABA receptor agonists

See allHide authors and affiliations

Science  25 Oct 2019:
Vol. 366, Issue 6464, eaaw8848
DOI: 10.1126/science.aaw8848

Plant thirst quenched without water

Drought causes many billions of dollars of annual losses to farmers worldwide. Central to a plant's water use efficiency are signaling pathways regulated by the hormone abscisic acid and its receptors. Vaidya et al. screened a pool of candidate small molecules and used structure-guided design to optimize the function of an abscisic acid receptor agonist (see the Perspective by Phillips and Sussman). Application of the agonist protected Arabidopsis, wheat, and tomato from underwatering.

Science, this issue p. eaaw8848; see also p. 416

Structured Abstract


Climate extremes create a need to mitigate the effects of drought on agriculture. The contributions of water to crop yield vary over a growing season but peak during reproductive development. Water banking strategies, which save soil water early in a growing season, reserve water for flowering and can increase yield under modest drought. Antitranspirants based on mimics of the phytohormone abscisic acid (ABA), which controls stomatal aperture, are sought as next-generation agrochemicals for controlling water use and increasing yield during drought.


Information on the structure and function of ABA receptors has created opportunities for agrochemical development. Current lead molecules have low and short-lived bioactivity in some relevant crop species, including wheat, the world’s most widely grown staple crop. This liability is likely a consequence of incomplete activation of different ABA receptor subclasses. We reasoned that the idiosyncratic activity of these molecules was due, in part, to a lack of interaction between the agonist and a conserved lysine in ABA receptors that forms a salt bridge to ABA’s carboxylate. We performed virtual screening to identify candidate agonists that interact with this lysine.


Two ABA receptor structures were used to screen against the ZINC database, a collection of commercially available ligands, using Glide docking protocols, requiring that hits interact with the conserved lysine. Candidate agonists were obtained and tested for receptor activation using in vitro assays. A family of substituted phenyl acetamido-cyclohexane carboxylic acid ABA receptor panagonists was identified. Scaffold merging was used to improve binding: We grafted an optimized headgroup of an existing sulfonamide onto our phenyl acetamido-cyclohexane carboxylic acid scaffold to yield a chimeric ligand (3CB) that displayed an improvement toward target sites of up to three orders of magnitude. Analysis of a 3CB-PYL10 crystal structure suggested that addition of appropriately situated hydrophobes to 3CB might improve activity. A 3CB derivative was synthesized, yielding an agonist that we have named opabactin (OP) for overpowered ABA receptor activation. Thermodynamic characterization of 3CB or OP receptor binding reactions indicates that the newly generated scaffold’s improvements are enthalpically driven relative to sulfonamides, consistent with the salt bridge observed in our crystal structure. Biological studies show that OP is ~10-fold more active in inhibiting seed germination (a response driven by ABA) than ABA itself. Experiments in wheat, tomato, barley, Arabidopsis, and Commelina demonstrate bioactivity in vegetative tissues across diverse species. Time course experiments in wheat and tomato using thermal imaging show that OP induces a more sustained antitranspirant response than ABA, and they document poor activity of sulfonamide agonists in those two species. To understand which receptors are necessary for OP action, we used Arabidopsis mutant strains to show that OP requires the subfamily III receptors PYR1, PYL1, and PYL2 for maximal activity. Thus, our virtual screening experiments yielded an ABA receptor agonist that functions as an antitranspirant.


A newly generated ABA agonist scaffold was identified and optimized through a structure-guided approach. The chemical biology of ABA receptor agonists can be broadened by designing ligands that engage a conserved lysine residue in the binding pocket. This pharmacophoric feature results in a favorable enthalpic binding profile and lower dissociation constants than existing sulfonamide-based ligands. Our ABA agonist, opabactin, has activity in multiple monocots and eudicots and addresses the limitations of existing sulfonamide molecules being explored as tools for mitigating the effects of drought on crop yields.

Water savings and drought protection activity in crops.

Virtual screening yielded a scaffold 3B4 whose potency was improved by ~1600-fold by scaffold merging, yielding 3CB. Structure-based optimization gave opabactin (OP). Thermodynamic profiling shows that OP binding to its targets PYR1 and HAB1 is enthalpically driven with about one-tenth the dissociation constant Kd of ABA. OP exhibits more potent and longer-lasting antitranspirant effects than existing sulfonamide-based ligands.


Drought causes crop losses worldwide, and its impact is expected to increase as the world warms. This has motivated the development of small-molecule tools for mitigating the effects of drought on agriculture. We show here that current leads are limited by poor bioactivity in wheat, a widely grown staple crop, and in tomato. To address this limitation, we combined virtual screening, x-ray crystallography, and structure-guided design to develop opabactin (OP), an abscisic acid (ABA) mimic with up to an approximately sevenfold increase in receptor affinity relative to ABA and up to 10-fold greater activity in vivo. Studies in Arabidopsis thaliana reveal a role of the type III receptor PYRABACTIN RESISTANCE-LIKE 2 for the antitranspirant efficacy of OP. Thus, virtual screening and structure-guided optimization yielded newly discovered agonists for manipulating crop abiotic stress tolerance and water use.

Water flux from soil to atmosphere occurs primarily by its movement through plants (1). Water enters the plant body through roots and exits as vapor from leaf surface stomata, small pores formed by neighboring guard cells that open and close to regulate gas exchange. The accumulation of biomass by photosynthesis requires CO2 to access inner leaf mesophyll cells through stomata, but this comes at the cost of H2O loss, which is driven by the difference in water vapor pressure between the inner leaf and atmosphere. Plants, therefore, face a trade-off between water use and growth. One consequence of this trade-off is that attempts to improve water use efficiency often come at the cost of reduced growth and yield (2). Conversely, studies of high-yielding wheat varieties have shown that they transpire more water than lower-yielding ones, and reduced water use efficiency has been proposed to be a consequence of selection for high grain yields in wheat (3, 4).

It is, therefore, intrinsically challenging to create crops with both high yield and high water use efficiency. Small molecules can, in principle, address this dilemma by coaxing high-yielding plant varieties with low water use efficiency into more water-efficient states on demand (5). This strategy would also enable data streams from precision agriculture to be directed toward conserving limited water resources and responding to emergent adverse events. Moreover, given that ~11% of Earth’s surface is dedicated to crop production (6) (mostly nonirrigated), better tools for managing crop water use could have global effects on water usage and help to address the ~$29 billion U.S. dollars in agricultural losses attributed to drought annually (7).

The scope and global importance of water use efficiency is driving the development of next-generation agrochemicals for managing responses to drought (820). These efforts have focused primarily on small-molecule mimics of the phytohormone abscisic acid (ABA), which coordinates plant physiology, growth, and stress responses with water availability. ABA responses are mediated by a signaling module that involves soluble Pyrabactin Resistance 1/PYR1-Like/Regulatory Component of ABA Receptor (PYL) ABA receptors, clade A type 2C protein phosphatases (PP2Cs), and subfamily three SNF1-related kinases (SnRK2s) (21, 22). When ABA binds to its soluble receptors, PYLs, a conformational change enables the receptors to bind to PP2C active sites and inhibit phosphatase activity. This, in turn, leads to the accumulation of activated SnRK2 kinases, which phosphorylate downstream factors and trigger physiological responses such as guard cell closure. This system is thus a target for both genetic and agrochemical control of drought tolerance. Overexpression of ABA receptors can increase water use and photosynthetic efficiency in both Arabidopsis and wheat with negligible effects on growth in controlled environment studies (23, 24), such genetic interventions can complement chemical strategies because the two can be combined for improved effects (25).

The most potent synthetic ABA receptor agonists to date are ABA analogs or molecules that share a core sulfonamide linkage related to quinabactin (18, 20, 2630). Flowering plant ABA receptors cluster into three phylogenetically distinct subfamilies (31, 32). Genetic analyses in Arabidopsis have demonstrated that quinabactin and other sulfonamide agonists require subfamily IIIA receptors (PYR1 and PYL1) to exert their antitranspirant effects (26, 33), but it is unclear whether activation of other ABA receptors, which in many angiosperms exceed a dozen per genome, would enhance antitranspirant effects. Although quinabactin activates PYR1 with nanomolar potency, it shows reduced or negligible activity on other receptors (26), including PYL4, one of the key receptors expressed in Arabidopsis guard cell microarray experiments (34, 35), a liability that could potentially limit efficacy. Upon investigating the effects of quinabactin and other sulfonamides in wheat and tomato, we observed low and short-lived bioactivity, which we speculated was the result of weak affinity for necessary targets. A structure-guided ligand discovery and optimization campaign led us to opabactin (OP), a potent agonist with activity in wheat and tomato. Genetic analyses show that PYR1, PYL1, and PYL2 are key target sites necessary for OP’s potent antitranspirant activity.

Discovery of a nonsulfonamide ABA pan agonist scaffold by virtual screening

In ABA/receptor cocrystal structures, ABA’s carboxylate forms a salt bridge with a conserved lysine (K59 in PYR1) that current sulfonamide agonists are unable to form (33, 3639) (Fig. 1A). We reasoned that the idiosyncratic activation of different ABA receptor types was due, in part, to lack of K59 engagement. We therefore used Glide docking protocols (40) to screen a collection of available ligands (41) for agonists that interact with K59. A subset of the predicted binders were tested for activity in vitro, using a pool of diverse recombinant ABA receptors (Arabidopsis PYL1, PYL2, PYL4, and PYL8, and maize PYL8). This process yielded 22 hits (half maximal inhibitory concentration, IC50 ≤ 25 μM) (Fig. 1B, fig. S1, and table S1), most of which could be clustered into three scaffolds: a group of sulfonamides reminiscent of pyrabactin and quinabactin, a family of substituted aminopropanediols, and a set of amino acid amide conjugates (Fig. 1B, fig. S1, and table S1). We focused optimization efforts on the amide scaffold because of its relatively high activity across the eleven Arabidopsis receptors tested (tables S1 and S2).

Fig. 1 Virtual screening and scaffold merging enabled the design of overpowered ABA receptor agonists.

(A) Structures of existing ABA receptor agonists; the gray box highlights the 4-cyano-3-cyclopropylphenyl head group used in our scaffold merging strategy. tABA, tetralone ABA; CB, cyanabactin; QB, quinabactin; R, radical. (B) Structures of representative hits (IC50 ≤ 25 μM) belonging to the different scaffolds identified; the yellow box highlights the amide scaffold that was merged with the cyanabactin headgroup. The complete list of hits identified is provided in fig. S1 and table S1. (C) Binding modes of ABA and hit 3B4, as originally predicted by docking to PYR1; the figure illustrates that 3B4 lacks a hydrogen bond acceptor for the Trp-lock water with thin red lines representing hydrogen bonds from 2.7 to 3.1 Å. (D) Hit discovery and optimization. The ZINC collection of 18 million purchasable drug-like compounds was screened against PYR1 and PYL9 to identify candidate binders that interact with K59. Of the top 10,000 hits, 1700 were obtained and tested for activity in vitro. The best hit obtained, 3B4, lacked an H-bond acceptor for the Trp lock. We therefore installed the cyanabactin head group on the 3B4 amide scaffold. A subsequent structure of 3CB-PYL10 suggested that bioactivity could be improved through the addition of an additional hydrophobe to interact with the receptors’ 3′ tunnels, which yielded the agonist opabactin (OP). IC50 values shown on the right (in red) are for the compounds tested against subfamily III receptor PYL2 (see Fig. 2 and fig. S2 for complete IC50 data). Synthetic schemes for the synthesis of 3CB and OP are provided in fig. S2.

Scaffold merging yields a potent panreceptor agonist

The most active amide identified [the chemical in screening plate 3, well B4 (3B4)] displayed submicromolar IC50 values and apparent panreceptor agonism (table S2). Our virtual screening data predicted that its carboxylate could form a salt bridge to K59 but that it lacked a hydrogen bond acceptor needed to engage the tryptophan (Trp)–lock water, which interacts with ABA’s ring ketone and coordinates an H-bond network that stabilizes activated receptors (Fig. 1C) (4244). In previous work, we showed that this function could be served by a nitrile in the designed ligand cyanabactin (Fig. 1A) (33). We hypothesized that a chimeric ligand possessing cyanabactin’s 4-cyano-3-cyclopropylphenyl headgroup grafted onto the 3B4 amide scaffold would improve binding affinity by providing simultaneous interactions with K59 and the Trp lock, essentially adopting a “scaffold merging” approach that has been exploited in the design of heat shock protein 90 inhibitors (45) and strigolactone mimics (46). To test this, we synthesized and characterized the chimeric 3B4-cyanabactin hybrid compound 3CB (scheme shown in fig. S2 and the materials and methods). We also synthesized AMF4 (ABA mimic 1 tetrafluoro derivative), a quinabactin derivative and sulfonamide agonist with improved potency (25) and tetralone ABA (29, 30), an ABA derivative with increased potency, so that we could compare these other potent agonists to our chimeric ligand.

Our synthesized chimeric ligand (3CB) displayed an increase of up to three orders of magnitude in affinity for subfamily III receptors and low nanomolar pan agonist activity (Fig. 2A, fig. S3A, and table S2), validating the scaffold-merging strategy used. To establish whether the agonist was active in monocots, which diverged from the eudicot lineage ~200 million years ago and include the world’s major grain crops, we tested the sensitivity of wheat ABA receptors to 3CB using a panel of subgenome D ABA receptors and a native wheat PP2C (24). These data show that 3CB is a similarly potent pan agonist of wheat receptors, unlike sulfonamide agonists (Fig. 2B, fig. S3B, and table S3). Moreover, we observe that the chimeric ligand is approximately twofold more active than ABA in Arabidopsis seed germination assays, indicating adequate bioavailability (Fig. 2C and fig. S4). Collectively, these experiments demonstrate that the combined use of virtual screening and scaffold merging enabled us to discover and optimize an agonist that binds multiple ABA receptors and is active in both eudicots and monocots.

Fig. 2 3CB and OP are best-in-class ABA receptor agonists.

(A) Agonist potency against different Arabidopsis receptors, as measured using agonist/receptor-mediated inhibition of ∆N-HAB1 phosphatase activity (n = 3 replicates). A dash indicates that PP2C activity was >75% control at 50 μM test chemical (the highest concentration tested). The colored boxes indicate the subfamily membership of the receptors assayed. (B) Agonist potency against different wheat receptors, using indicated receptors and TaPP2C. A dash indicates that PP2C activity was >75% control at 10 μM test chemical (the highest concentration tested). We note that TaPYL9 is a pseudoreceptor with an atypical binding pocket. (C) OP shows an ~10-fold increase in in vivo potency relative to ABA. The potency of different agonists on Arabidopsis seed germination and their corresponding IC50 values (concentrations required to inhibit germination by 50%). Photographs were taken 4 days post stratification (n = 6 replicates). The full-dose response curves used to infer IC50 values for all three figure panels are shown in figs. S3 and S4 and tables S2 and S3.

Structure-guided optimization of an overpowered agonist

To understand the atomic basis of 3CB’s high activity, we cocrystallized it with the Arabidopsis subfamily I receptor PYL10, which was chosen because it expresses in Escherichia coli at high levels relative to other subfamily I receptors, and solved an x-ray crystal structure of a PYL1025−183:3CB complex using molecular replacement at a resolution of 2.4 Å (table S4). The PYL1025−183:3CB complex crystallized in the space group P3121 and contained a single protomer in the asymmetric unit. Several rounds of structural refinement were carried out before modeling 3CB into the ligand-binding pocket’s unbiased electron density (fig. S5). A real-space correlation coefficient of 0.95 calculated between the unbiased electron density and 3CB indicates agreement between the model and observed electron density. These data position 3CB in the PYL10 ligand-binding pocket with its carboxylate forming a salt bridge to K56 (homologous to PYR1’s K59), and its aryl ring oriented toward the gate and latch loops, mimicking ABA’s cyclohexenone ring and gem-dimethyl group (Fig. 3A). Direct polar contact between N163’s amide NH and 3CB’s amide carbonyl, and two water-mediated contacts to E90 and F98 (Fig. 3A), additionally stabilize 3CB-PYL10 interactions. 3CB’s cyclohexyl ring packs against L159, V156, Y116, and I106, which line the C6 cleft, a small hydrophobic pocket that normally interacts with ABA’s C6 methyl and is a ligand-binding hotspot (47).

Fig. 3 Structure of 3CB bound to PYL10.

(A) Cocrystal structure of 3CB bound to PYL10, dashes representing hydrogen bonds from 2.6 to 3.3 Å. (B) Differences in orientation of the cyclopropyl group present in the 3CB:PYL10 structure versus the CB:PYR1 structure (PDB ID 5UR5). (C) Positioning of H111 and L159 in PYL10 relative to 3CB’s 3-cyclopropyl-4-cyanobenzyl headgroup. This illustrates the site of predicted steric clash that would occur between L159 and OP’s C5 cyclopropyl substituent when bound to Arabidopsis subfamily I receptors. (D) Sequence variation at the selectivity-determining residue (homologous to PYL10 L159), which is the only residue lining the 3′ tunnel that is variable between ABA receptors). (E) Genetic evidence that sequence variation at the selectivity-determining residue (L159 in PYL10) determines OP selectivity. V163 in PYR1 is homologous to L159 in PYL10. To test the role of this residue in OP selectivity, wild-type PYR1, V163I, and V163L were tested for interactions with HAB1 using an established yeast two-hybrid assay (chemicals tested at 25 μM); the bulkier leucine substitution corresponding to that present in Arabidopsis subfamily I receptors blocks OP action. Single-letter abbreviations for the amino acid residues are as follows: E, Glu; F, Phe; H, His; I, Ile; K, Lys; L, Leu; N, Asn; V, Val.

On the basis of our prior crystallographic analyses of a PYR1-cyanabactin complex (33), we anticipated that 3CB’s cyclopropyl substituent might extend into the hydrophobic 3′ tunnel, which contacts ABA’s C7′ methyl (Fig. 3B). Instead, our structure shows 3CB’s cyclopropyl group oriented away from the 3′ tunnel, packing against the gate and latch loops (Fig. 3A). The vacancy of the 3′ tunnel, therefore, suggested that we might increase 3CB’s potency by adding hydrophobic substituents to interact with this pocket, a strategy used to improve other agonists (33, 48). To test this idea, we synthesized a dicyclopropyl 3CB analog, which we have named opabactin (OP) for overpowered ABA receptor activation (synthetic details are provided in fig. S2). We tested OP activity in vitro using a receptor-agonist N-terminal deletion of HAB1 (∆N-HAB1) inhibition assay and observed that it activates Arabidopsis subfamily III receptors with potencies that are up to ~10-fold greater than ABA’s; in vivo, OP is ~10-fold more bioactive, as measured using Arabidopsis seed germination assays (Fig. 2C), and induces an approximately fivefold stronger effect than ABA on guard cells, as measured using a Commelina guard cell closure assay (fig. S6). Although activity against subfamily III receptors improved, activity against Arabidopsis subfamily I receptors diminished (Fig. 2A). In contrast, we observed that OP activates the wheat subfamily I receptor TaPYL8 and is thus a pan agonist of the wheat ABA receptors tested (Fig. 2B). Inspection of sequence alignments reveals a single amino acid difference between Arabidopsis and wheat subfamily I ligand-binding pockets (a bulky leucine is replaced by a smaller valine in TaPYL8); moreover, this residue packs tightly against the location of the second cyclopropyl in OP and would cause steric clash in Arabidopsis receptors (Fig. 3, C and D). Yeast two-hybrid assays confirm that the leucine-to-valine pocket difference determines subfamily I sensitivity to OP (Fig. 3E). Thus, the 3CB-PYL10 cocrystal structure enabled us to increase its activity against subfamily III targets and yielded an agonist that is ~10-fold more potent than ABA in seeds.

OP-PYR1 binding is enthalpically driven

The 3CB carboxylate-K56 salt bridge is likely to improve binding affinity via an enthalpic contribution. To examine this, we conducted isothermal titration calorimetry (ITC) experiments comparing 3CB- and OP-PYR1 binding isotherms (Fig. 4A). Prior analyses (47) revealed an apparent dissociation constant (Kd) of 4.8 μM for quinabactin-PYR1 binding and indicated that quinabactin binding is entropy driven (change in enthalpy, ∆H = −2.8; change in entropy, −TS = −4.7 kcal/mol) (Fig. 4A). In contrast, 3CB- and OP-PYR1 binding reactions are more enthalpic and of higher affinity relative to quinabactin and cyanabactin (OP Kd = 1.9 μM; quinabactin Kd = 4.8 μM) (Fig. 4A). The presence of the second cyclopropyl substituent in OP relative to 3CB provides an additional −1.7 kcal/mol ∆H for PYR1 binding; this may indicate protein movement in response to ligand binding and/or be a consequence of van der Waals contacts, which in some systems are sufficient to create exothermic binding isotherms without conformational changes in protein structure (4951). We additionally examined ligand–PYR1-HAB1 binding reactions, comparing ABA, quinabactin, and OP (Fig. 4B). These experiments show that OP’s Kd for PYR1-HAB1 binding is approximately one-seventh that of ABA (28 ± 5 nM versus 201 ± 14 nM). This increase in binding affinity translates into improved potency of OP in Arabidopsis seed germination assays described above (Fig. 2C). Collectively, our ITC data show that our newly identified amide scaffold provides more enthalpy-driven, higher-affinity binding than previous sulfonamides, a feature of many best-in-class ligands (52, 53).

Fig. 4 The high binding affinities of 3CB and OP are enthalpy driven, in contrast to the entropy-driven binding of sulfonamides.

(A) Measured thermodynamic parameters of binding of agonists to PYR1 in the presence or absence of HAB1; the top two rows are previously published observations, and all the rest were determined in this study. The bar graphs summarize the ITC data and illustrate that sulfonamide agonist binding is driven primarily by entropy, whereas the 3CB and OP binding are more enthalpically driven. (B) Representative thermogram and binding isotherm. ITC experiments were conducted by repeated injections of 2.5 μl of OP (60 μM) into a 1:1 mixture of PYR1:∆N-HAB1 (10 μM). Binding isotherms were fitted to an independent one-site binding model using NanoAnalyze software (TA Instruments, USA); n = 3 replicate experiments, standard deviations shown. The red trace is a control of buffer injections (rather than buffer + ligand) into PYR1–∆N-HAB1. The data are consistent with high-affinity binding of OP to the receptor complex; however, precise Kd values in the low nanomolar range are challenging to deduce.

OP has antitranspirant activity and activates ABA signaling in vegetative tissues

Our initial tests indicated that both 3CB and OP activate multiple ABA responses in Arabidopsis vegetative tissues (fig. S7) and demonstrate antitranspirant activity in the grain crop barley (fig. S8). To investigate OP’s effects in other crop species, we treated tomato and wheat plants with different concentrations of ABA, quinabactin, AMF4, or OP and used thermal imaging to measure their effects on transpiration for up to 5 days after application (relative to mock-treated controls). In both tomato and wheat, ABA’s effects on leaf temperature dissipate after 2 to 3 days, and sulfonamide agonist effects are no longer evident after 48 hours (Fig. 5). In contrast, OP’s effects remained evident at 5 days (the latest time point for which we collected data) in both species (Fig. 5). This increased bioactivity is likely the result of a combination of improved persistence and target site potency; however, metabolic studies will be required to resolve this. Whatever the case, these data show that OP has longer-lasting effects relative to ABA and other agonists in wheat, a globally important staple crop.

Fig. 5 OP has bioactivity and long-lasting effects in tomato and wheat, whereas sulfonamide agonists do not.

(A) Infrared images of 6-week-old tomato seedlings (UC82) treated with 100 μM of different chemical solutions composed of 0.5% dimethyl sulfoxide (DMSO) and 0.05% Silwet-77. Graphs depict the variation in difference in temperature levels of treated plants relative to mock-treated plants up to 5 days after application (n = 18 replicates for mocks and n = 6 for chemical treatment). (B) Infrared images of 4-week-old wheat seedlings (WB-9229) sprayed with 11 μM of different chemicals solution composed of 0.5% DMSO and 0.05% Silwet-77. Graphs depict the variation in difference in temperature levels of treated plants, relative to mock-treated plants up to 5 days after application. Statistical analyses were conducted in R using a one-way analysis of variance (ANOVA) and post-hoc Dunnett tests to obtain multiplicity adjusted P values for treatment effects relative to mock-treated controls (n = 10 replicates for mocks and n = 5 for chemical treatment). 95% confidence intervals are shown.

We next examined OP’s effects on both ABA-induced transcriptional responses and drought tolerance in wheat. Using a progressive water stress regime, we observed that wheat seedlings treated with ABA, 3CB, and OP delayed wilting under water stress and retained more chlorophyll than did mock-treated plants, indicating less stress-induced senescence in treated plants relative to mock-treated controls (Fig. 6, A and B). In contrast, quinabactin-treated plants were only modestly protected under the same conditions, consistent with the weak activity of sulfonamide agonists described above (Fig. 6) and results from direct receptor-based assays (Fig. 2B). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) experiments of treated seedlings for three ABA-responsive genes also demonstrate that OP shows increased activity relative to ABA and quinabactin (Fig. 6, C to E). Collectively, these data show that OP’s high binding affinity for ABA receptors (relative to ABA and sulfonamides) translates into bioactivities that exceed both ABA and sulfonamide agonists in planta.

Fig. 6 OP enhances drought tolerance and hyperactivates ABA transcriptional responses in wheat.

(A) Effects of ABA receptor agonists in wheat. Two-week-old plants were sprayed with chemical solution containing 0.05% Tween-20 and 50 μM compounds. After 3 days of water deprivation, plants were rewatered, and stress recovery images were acquired after 12 hours. (B) Chlorophyll content of aerial parts of plants after water stress treatments was measured as described in the methods. (C to E) Comparison of transcript levels for TaPP2C6, TaLea, and TaAOS induced by individual ligands in 7-day-old wheat seedlings after a treatment of 50 μM for 12 hours. Statistics were done using a one-way ANOVA and post-hoc Dunnett tests to obtain multiplicity adjusted P values for treatment effects relative to mock-treated controls, with n = 5 replicates for chlorophyll measurements and n = 4 replicates for qRT-PCR studies. Error bars indicate standard error of mean, individual P values are shown on the graph. n.s., nonsignificant.

Subfamily III receptors are crucial for mediating the effects of OP in planta

To examine the receptor dependence of OP’s action, we examined agonist-induced responses in wild-type or mutant strains lacking the major members of the different receptor subtypes. The effects of quinabactin and cyanabactin on Arabidopsis transpiration, germination, and gene expression are mediated primarily by the subfamily III receptors PYR1 and PYL1, which form a subclade within family III (33). To examine if this subclade similarly executes OP’s effects, we examined the sensitivity of wild type; pyr1,pyl1 (referred to as 0;1); pyr1,pyl1,pyl2 (0;1;2, subfamily III); pyl4,pyl5 (4;5, subfamily II); and pyl8,pyl9 (8;9, subfamily I) mutant strains to OP and other agonists. Both 3CB and OP inhibit Arabidopsis seed germination and seedling establishment, but the effects of OP are largely absent in the 0;1;2 triple mutant strain, while those of 3CB are not (Fig. 7A). These genetic observations therefore suggest that OP’s action on seeds is mediated by subfamily III receptors because OP-treated seeds germinate in the 0;1;2 mutant. As expected, quinabactin and AMF4 effects on seeds are absent in the 0;1 double mutant; however, OP retains bioactivity in the 0;1 mutant, and its effects are only greatly reduced in the 0;1;2 triple mutant (Fig. 7A). This demonstrates that, in contrast to previously characterized sulfonamide agonists, PYL2 is additionally necessary for full OP action in seeds. Collectively these results further confirm the importance of subfamily III receptors for sulfonamide agonist action and illuminate an added role for PYL2 in OP’s action.

Fig. 7 OP acts primarily through subfamily III receptors and demonstrates that PYL2 contributes to OP’s antitranspirant activity.

(A) Effects of different ABA receptors agonists on seed germination in wild type and different receptor mutant lines. Chemical effects were tested at concentrations approximately two times their IC50 values (Fig. 2C) to normalize for differences in agonist potency (ABA, 1200 nM; QB, 1800 nM; AMF4, 1600 nM; 3CB, 440 nM; and OP, 120 nM). Photographs were taken after 4-day incubation under constant illumination. (B) Adult Col-0 and different mutant lines were sprayed with 50 μM compounds and imaged by thermography 24 hours after treatment and (C) average leaf temperatures for each treatment were quantified. Statistics done using a one-way ANOVA and post-hoc Dunnett tests to obtain multiplicity adjusted P values for treatment effects relative to mock-treated controls (n = 8 replicates). Error bars indicate standard error of mean. Individual P values for comparisons to mock-treated controls are shown on the graph.

To determine whether the involvement of PYL2 in OP’s action also applies to vegetative responses, we examined OP’s effects on leaf temperature in wild type, 0;1, 0;1;2, and pyl2 mutant seedlings. These experiments show that OP’s activity in leaf tissues can be greatly reduced in the 0;1;2 triple mutant but not the 0;1 double mutant strain, paralleling our observations in seeds (Fig. 7, B and C). Furthermore, both ABA and OP inhibit primary root growth and greening in Arabidopsis seedlings in a dose-dependent manner, as expected, and OP’s effects in these growth assays can be greatly reduced in the 0;1;2 triple mutant (fig. S9). These genetic analyses demonstrate that PYL2, in addition to PYR1 and PYL1, is a cellular target that can be activated by synthetic agonists to manipulate plant transpiration and ABA responses.


Synthetic ABA mimics are being investigated as tools for mitigating the impact of drought stress on crop yields. We used virtual screening to identify ABA receptor agonists in distinct chemical scaffolds, one of which we modified using scaffold merging to create 3CB and subsequently opabactin, a molecule with high bioactivity and receptor binding affinity. Our cocrystal structure of 3CB:PYL10 reveals a salt bridge to K56 and direct hydrogen bond of its amide to N163, features that likely contribute to the enthalpic binding isotherms observed, and the approximately sevenfold improvement in Kd relative to ABA. Prior sulfonamide ABA mimics do not benefit from the lysine salt bridge, and our work suggests that contacts to this residue are essential for high-affinity binding and pan agonism and should be considered in future ligand design and optimization. Because the sulfonamides in quinabactin, pyrabactin, CB, and AMF4 are positioned proximally to this lysine, it should be possible to generate contacts by lowering the pKa (where Ka is the acid dissociation constant) of their sulfonamide NH by the installation of adjacent electron withdrawing groups, a strategy previously used to improve sulfonamide dihydrofolate reductase and carbonic anhydrase inhibitors (54, 55). Thus, there are likely many routes to developing more-potent ABA receptor agonists, but our development of a new amide-based scaffold broadens the current chemical space available for manipulating ABA receptor function.

OP’s activity in Arabidopsis demonstrates that PYL2 contributes to OP’s effects on transpiration, in addition to PYR1 and PYL1. In contrast, the effects of sulfonamide agonists require primarily PYR1 and PYL1 (26, 33). We conclude, therefore, that PYL2 is a useful antitranspirant target site in addition to PYR1 and PYL1; this further demonstrates the relevance of subfamily III receptors as targets for the development of synthetic antitranspirants but does not preclude the activation of other receptor types for antitranspirant activity. The development of agonists selective for different receptor subfamilies could help untangle the relative importance of each subfamily to transpiration. Additionally, because increases in both water use and photosynthetic efficiency have been observed by overexpressing subfamily II receptors (23, 24), selective agonists may provide access to the control of distinct water use–related traits not achievable by only activating subfamily III receptors.

The triple mutant pyr1,pyl1,pyl2 has a relatively modest phenotype. If subfamily III receptors are such critical mediators of OP and sulfonamide effects on transpiration, why is their multilocus loss-of-function phenotype so mild? We do not yet have a complete answer to this question, but one possibility is that other receptors may compensate for the loss of subfamily III receptors, as has been observed by systematic analyses of yeast gene families, which show compensation by paralogs (56). Alternatively, it may be that the major effects of OP and sulfonamides require both subfamily II and subfamily III receptors to be present, i.e., different subfamily II receptors than we are currently able to test genetically (pyl4 and pyl5). Given that PYL4 and PYL5 are the most highly expressed subfamily II ABA receptors, this seems unlikely, but future analyses of genetic strains lacking all members of each subfamily should help resolve this. Nonetheless, weighting a protein’s relevance as a target for manipulating a biological process on the basis of only genetic studies may be misleading because of compensation. Thus, chemical interrogation of wild-type systems provides a more direct route to assessing the relevance of a protein as a target.

ABA has many effects on plant growth and development in addition to its roles in water relations, seed dormancy, and gene regulation. Although OP mimics ABA’s effects, further characterization will be required to establish which of ABA’s many responses are activated by OP. It will also be necessary to conduct toxicological assays and regulatory procedures before OP (or any agonist) is suited for agricultural purposes. ABA receptor overexpression can enable increases in water use efficiency without growth penalties in both Arabidopsis and wheat in controlled-environment studies (23, 24). The application of ABA—which, like OP, inhibits growth at high concentrations—to crops in the field improves yield under modest drought conditions (5759), so the growth inhibitory effects of ABA are not an intrinsic limitation to their other benefits. The value of chemical approaches is that they can be deployed on demand, whereas crop genetics cannot be changed dynamically in a growing season. It is likely that both genetics and chemistry will work together for maximal benefit and flexibility (25). OP is a new chemical tool for dynamically manipulating plant water use.

Materials and methods summary

Virtual screening of the ZINC collection, a database of commercially available chemical compounds, was carried out using Protein Data Bank (PDB) IDs for PYR1 (3K3K) and PYL9 (3W9R) using Schrodinger’s Glide package. Top scoring 0.1% hits were redocked against PYR1 to eliminate ligands not predicted to contact PYR1-K59. 1724 predicted agonists were purchased from Enamine (Ukraine) and tested at 25 μM in a direct receptor activation assay using a pool of recombinant ABA receptors from the three major clades (PYL1, PYL2, PYL4, PYL8, and ZmPYL8). The pooled receptor assay measures agonist dependent receptor mediated inhibition of the protein phosphatase ∆N-HAB1 using a fluorogenic substrate (4-methylumbelliferyl phosphate). This process yielded 25 hits (IC50 ≤ 25 μM), one of which (3B4) was optimized by chemical synthesis to yield 3CB, a pan agonist. After optimizing one of our hits, we obtained a cocrystal structure of the lead molecule 3CB with the subfamily I ABA receptor PYL10. PYL1025-183:3CB crystals were grown in 33% tacsimate (pH 7.0) and flash frozen using perfluoropolyether oil. Diffraction data was processed in HKL2000, refined in Phenix.refine and validated using MolProbity; the final model was deposited in PDB (ID 6NWC). We next synthesized OP, a derivative of 3CB. 3CB and OP were synthesized from methyl (4-cyanophenyl)acetate over five steps involving (i) palladium catalyzed ortho-halogenation of the nitrile, (ii) suzuki coupling with cyclopropyl boronic acid, (iii) base-catalyzed hydrolysis of the ester, (iv) EDCI/DMAP coupling to methyl 1-aminocyclohexanoate, and (v) subsequent hydrolysis to yield OP (34 to 58% yield). The OP that we synthesized for this paper can be purchased from Kerafast (USA), a distributor of academic research reagents. The AMF4 and tetralone ABA used in our studies were synthesized according to literature protocols, using chiral chromatography to purify the bioactive (+) isomer of tetralone ABA, as previously described. To directly measure agonist-receptor interactions, we conducted isothermal titration calorimetry (ITC) experiments using low-volume Nano ITC (TA Instruments, USA). To determine which residues in PYR1 mediate OP selectivity, we used an established yeast two-hybrid assay to analyze agonist responses in wild-type PYR1, PYR1-V163, and PYR1-V163L mutants. To measure bioactivity of the agonists in adult plants, chemicals were applied to plants as aerosols in a solution supplemented with the wetting agent Silwet-77; leaf temperatures were measured by thermal imaging using a FLIR camera. To profile Arabidopsis mutant responses to agonists, we used previously described receptor mutants and a newly constructed pyr1,pyl1,pyl2 strain that we made by crossing a pyr1,pyl1,pyl2,pyl4 quadruple mutant strain to wild type (Col-0) and isolating the desired genotype amongst F3 plants.

Supplementary Materials

Materials and Methods

Figs. S1 to S9

Tables S1 to S6

NMR Spectra

References (6075)

References and Notes

Acknowledgments: Funding: NSF IOS (grant 1656890) and Syngenta Crop Protection AG to S.R.C. JST PRESTO (JPMJPR15Q5), KAKENHI (17H05009), and the Joint Research Program of Arid Land Research Center, Tottori University (30C2007) to M.O. Author contributions: A.S.V. designed and synthesized 3CB and OP, with technical assistance from B.K. and S.B., and performed Arabidopsis PP2C assays, seed germination assays, luciferase imaging, and thermal imaging experiments. J.D.M.H. performed virtual screening and hit characterization. F.C.P. performed crystallographic studies in coordination with B.F.V. D.E. performed isothermal titration calorimetry studies. R.M. performed plasmid vector construction for PP2C assay. W.D. performed SEM studies, SAI, root growth, and greening measurements. A.K. and J.D.M.H. conducted small molecule screens of virtual screening hits. S.-Y.P. performed yeast two-hybrid assays. Z.X. and S.-Y.P. purified and characterized Arabidopsis PYR/PYL proteins. M.O. identified, cloned, and expressed wheat ABA receptors and performed wheat PP2C assays, qRT-PCR, and drought stress assays. J.T. and Y.T. synthesized tetralone ABA for comparison in various assays. S.R.C. conceived of the research, and A.S.V. and S.R.C. wrote the paper with contributions from all authors. Competing interests: S.R.C., J.D.M.H., and A.S.V. are inventors on a UC-owned patent application (62/691,534) covering OP and related structures. Data and materials availability: All data are presented in the main text and supplementary materials, the OP synthesized for this study can be obtained from Kerafast, and other biological materials may be obtained by contacting S.R.C. (Arabidopsis) or M.O. (wheat). The atomic coordinates and structure factor files for the PYL103CB structure have been deposited in Protein Data Bank (ID 6NWC).
View Abstract

Stay Connected to Science

Navigate This Article