Modification of auxinic phenoxyalkanoic acid herbicides by the acyl acid amido synthetase GH3.15 from Arabidopsis

Herbicide-resistance traits are the most widely used agriculture biotechnology products. Yet, to maintain their effectiveness and to mitigate selection of herbicide-resistant weeds, the discovery of new resistance traits that use different chemical modes of action is essential. In plants, the Gretchen Hagen 3 (GH3) acyl acid amido synthetases catalyze the conjugation of amino acids to jasmonate and auxin phytohormones. This reaction chemistry has not been explored as a possible approach for herbicide modification and inactivation. Here, we examined a set of Arabidopsis GH3 proteins that use the auxins indole-3-acetic acid (IAA) and indole-3-butyric acid (IBA) as substrates along with the corresponding auxinic phenoxyalkanoic acid herbicides 2,4-dichlorophenoxylacetic acid (2,4-D) and 4-(2,4-dichlorophenoxy)butyric acid (2,4-DB). The IBA-specific AtGH3.15 protein displayed high catalytic activity with 2,4-DB, which was comparable to its activity with IBA. Screening of phenoxyalkanoic and phenylalkyl acids indicated that side-chain length of alkanoic and alkyl acids is a key feature of AtGH3.15's substrate preference. The X-ray crystal structure of the AtGH3.15·2,4-DB complex revealed how the herbicide binds in the active site. In root elongation assays, Arabidopsis AtGH3.15-knockout and -overexpression lines grown in the presence of 2,4-DB exhibited hypersensitivity and tolerance, respectively, indicating that the AtGH3.15-catalyzed modification inactivates 2,4-DB. These findings suggest a potential use for AtGH3.15, and perhaps other GH3 proteins, as herbicide-modifying enzymes that employ a mode of action different from those of currently available herbicide-resistance traits.

Herbicide-resistance traits accounted for 47% of genetically engineered soybean, maize, canola, cotton, sugar beet, and alfalfa plantings worldwide in 2017 (1). Since the introduction of the first crops with a glyphosate-resistant trait, the use of herbicides with distinct modes of action and the discovery of new resistance traits have become critical elements for increased agricultural productivity and for effective management of weed resistance (2,3). For example, auxinic herbicides, based on phenoxyalkanoic acid, benzoic acid, pyridine carboxylic acid, and quinoline carboxylic acid chemical scaffolds mimic the plant hormone auxin (indole-3-acetic acid (IAA)) 3 and are used extensively in agronomic and noncrop applications for broadleaf weed control ( Fig. 1) (4,5). These molecules elicit the same type of growth and developmental responses as IAA but, because of higher stability in the plant, result in longer-lasting and stronger effects such as plant overgrowth (4,5).
Of the auxinic herbicides, 2,4-dichlorophenoxyacetic acid (2,4-D) ( Fig. 1) was the first to be commercialized in 1945 and is the most widely used phenoxyalkanoic acid herbicide with ϳ46 million pounds applied in the United States per year, predominantly in the Midwest, Great Plains, and the Northwest United States (6,7). Related to 2,4-D, 4-(2,4-dichlorophenoxy)butyric acid (2,4-DB) ( Fig. 1) has also been used to control annual and perennial broadleaf weeds since 1958 (8,9). After foliar application, 2,4-DB is taken up by the leaves and roots and converted through peroxisomal ␤-oxidation to the active herbicide 2,4-D (8 -10). At the molecular level, 2,4-D binds to the auxin receptor F-box protein TIR1, which facilitates interaction between the receptor and corepressor Aux/IAA proteins (11,12). This leads to ubiquitination and degradation of the Aux/IAA proteins to modulate downstream interactions with auxin response factors that control transcription of auxin responsive genes (13,14). Although both IAA and 2,4-D target the auxin receptor, 2,4-D is metabolized more slowly than IAA, which enhances herbicidal effects through elevated expression of auxin responsive genes leading to plant death (6,15,16). For agricultural biotechnology applications, herbicide tolerance traits have relied on the identification of enzymes that either chemically inactivate the herbicide or prevent inhibition of a target by the herbicide (17)(18)(19)(20)(21)(22)(23)(24). For example, isolation of a microbial aryloxyalkanoate dioxygenase that cleaves 2,4-D provides tolerance to this auxinic herbicide and is the basis for 2,4-D-resistant crops currently entering the market (23). Access to tolerance traits with distinct modes of action is critical for reducing the emergence of herbicide-resistant weeds (2)(3)(4)(5)(6).
Given the chemical similarity between phenoxyalkanoic acids (i.e. 2,4-D and 2,4-DB) and auxins (IAA and IBA) (Fig. 1), we examined the potential of selected Arabidopsis GH3 proteins to modify either 2,4-D or 2,4-DB. The IBA-specific Arabidopsis GH3.15 protein (AtGH3.15) displayed high catalytic activity with 2,4-DB, which was comparable with that of IBA, and the X-ray crystal structure of the enzyme in complex with the herbicide shows how the molecule binds in the active site. When grown on 2,4-DB, A. thaliana T-DNA insertions in AtGH3.15 and 35S:FLAG-AtGH3.15 overexpression lines show hypersensitivity and tolerance, respectively, in root elongation assays. These findings suggest a potential use for AtGH3.15, and perhaps other GH3 proteins, as herbicide-modifying enzymes that employ a mode of action that differs from available auxinic herbicide-resistance traits.

Biochemical analysis of AtGH3.15
As the most active GH3 protein tested with 2,4-DB, AtGH3.15 was further characterized for its amino acid substrate profile and with other phenoxyalkanoic and phenylalkyl acids. As noted above, biochemical analysis of AtGH3.15 yielded steady-state kinetic parameters for 2,4-DB that were comparable to those obtained for IBA with glutamine (Table 1). QTRAP MS analysis confirmed formation of the 2,4-DBglutamine conjugate in vitro. Incubation of AtGH3.15 with 2,4-DB, ATP, and glutamine leads to formation of the conjugate (deprotonated molecular ion (M-H) Ϫ m/z ϭ 376.2) (Fig. S1). Assays in the absence of protein or any one substrate did not yield a peak corresponding to the conjugated product. To confirm that the amino acid preference of AtGH3.15 was the same with 2,4-DB as with IBA, the substrate profile was examined using 2,4-DB and each amino acid (Fig. S2). The amino acid profile was the same for AtGH3.15 with 2,4-DB as with IBA (33) with cysteine, histidine, methionine, glutamine, and tyrosine having the highest activity. Steady-state kinetics with cysteine, histidine, methionine, glutamine, and tyrosine were determined and confirm that, as with IBA, glutamine is the preferred amino acid for AtGH3.15 with 2,4-DB (Table 2).

Herbicide modification by an acyl acid amido synthetase
C4 position does not reduce catalytic efficiency; however, extension of this position reduces catalytic activity, as observed with 4-(4-methoxyphenoxy)butanoic acid. Comparison of the catalytic efficiencies of 4-phenoxybutyric acid, 4-phenylbutryic acid, 5-phenylvaleric acid, and 5-(4-fluorophenyl)valeric acid also indicate that compounds longer in length from carboxylate to the substituted phenyl group are superior substrates.

Three-dimensional structure of AtGH3.15 in complex with 2,4-DB
To provide insight on how 2,4-DB interacts with AtGH3.15, the protein was crystallized in the presence of the ligand. The 2.15 Å resolution structure of the AtGH3.15⅐2,4-DB complex was solved by molecular replacement (Table 4; Fig. 3A). The overall fold of the resulting structure was similar (1.4 Å 2 root mean square deviation for 550 C ␣ -atoms) to that of the previously reported AtGH3.15⅐AMP complex (33) with the conformationally mobile C-terminal domain adopting the open active site conformation. Examination of the electron density maps in the active site revealed two large patterns of density in the acyl acid-binding site, which were subsequently modeled and refined as two molecules of 2,4-DB (Fig. 3B). Comparison with the position of AMP in the AtGH3.15⅐AMP complex indicates that only one 2,4-DB molecule is positioned in an orientation that points the reactive carboxylate group toward the phosphate group that undergoes the adenylation reaction (Fig. 3C). This 2,4-DB molecule stacks with Phe-166, forms a hydrogen bond contact with Ser-122 (which was modeled in two alternate side-chain conformations), and is situated in a space bordered by Met-162, Val-163, Phe-325, and Phe-332 (Fig. 3D). The second 2,4-DB molecule positions its carboxylate group away from the nucleotide-binding site and is situated deeper in the acyl acid-binding site (Fig. 3C). This ligand forms a charge-charge interaction between its carboxylate and the side-chain of Arg-214 (Fig. 3D). Ser-299 contributes a hydrogen bond interaction to the carboxylate. The substituted phenyl ring is positioned to form van der Waals contacts with Ile-143, Leu-181, and Phe-219. It is not clear if the binding of two 2,4-DB molecules in the acyl acid site of AtGH3.15 is biochemically relevant or is an artifact of crystallization. Because of the large size of the site, it is possible that binding of one ligand deeper in the pocket positions the second for efficient catalysis.
To determine whether the in vitro activity of AtGH3.15 with 2,4-DB had in planta effects, previously generated and characterized knockout and overexpression lines of AtGH3.15 (33)

Herbicide modification by an acyl acid amido synthetase
In contrast, AtGH3.15 is highly active with both IBA and its auxinic herbicide counterpart 2,4-DB (Table 1; Fig. 2). Additional biochemical analysis with a range of substrates (Fig. S3) shows that AtGH3.15 does not accept benzoic acid (dicamba), pyridine carboxylic acid (clopyralid, picloram, and triclopyr), and short side-chain phenoxyalkanoic acid (dichlorprop, mecoprop, MCPA) auxin herbicides as substrates. AtGH3.15 did use 2,4-D and 2,4,5-T, which differs from 2,4-D by one additional chlorine, but with catalytic efficiencies 50-to 100fold lower than that observed with either IBA or 2,4-DB (Tables  1 and 3; Fig. 2). The kinetic analysis with MCPB, an analog of 2,4-DB, and other longer side-chain phenoxyalkanoic acid and phenylalkyl acid substrates indicates that substitutions of the phenyl group are not as important as side-chain length for activity; however, changes to phenyl group substituents that lengthen the moieties, such as the methoxy group of 4-(4-methoxyphenoxy)butanoic acid, reduce catalytic efficiency ( Table 3).
The X-ray crystal structure of AtGH3.15 in complex with 2,4-DB (Fig. 3) provides insight on how this molecule is recognized largely through apolar surface contacts, although some hydrogen bond interactions contribute. The orientation of one 2,4-DB molecule in the active site with its reactive carboxylate toward the location of the ATP/AMP-binding site suggests how a productive first-half reaction leading to the adenylated reaction intermediate occurs. As noted in the results, the binding of two 2,4-DB molecules in the acyl acid site of AtGH3.15 may be biochemically relevant or an artifact of crystallization, but is not unprecedented. For example, a set of stacked alrestatin molecules in aldose reductase was proposed to contribute to ligand specificity (34). It is possible that the large size of the AtGH3.15 acyl acid-binding site and binding of two substrates in different orientations contributes to efficient catalysis. This detail requires additional detailed biochemical analyses. Overall, the AtGH3.15⅐2,4-DB complex is the first of a GH3 protein with a herbicide bound and shows how binding in the site is largely dictated by surface contacts.
The in planta effect of AtGH3.15 knockout or overexpression indicates that changes in expression alter sensitivity to 2,4-DB (Fig. 4). Previous work characterized these plant lines (33). In the root elongation assays with Arabidopsis seedlings (Fig. 4), knockout lines of AtGH3.15 showed hypersensitivity to treatment with 2,4-DB, whereas overexpression lines of AtGH3.15 displayed clear tolerance to 2,4-DB. As with other plants, metabolism of 2,4-DB to 2,4-D by ␤-oxidation in the peroxisome, a process similar to conversion of IBA to IAA, leads to auxinic herbicide effects in Arabidopsis (35)(36)(37). Interestingly, this experiment with AtGH3.15 and 2,4-DB, along with other reported studies of various GH3 proteins and their responses to different phytohormones such as IAA, IBA, and jasmonates (25,26,30,32,33), highlights differences between in vitro steady-state kinetics and in planta responses. The K m values reported for various GH3 proteins with their cognate plant hormone substrates are typically in the 300 -800 M range; however, overexpression and knockout plant lines of the different GH3-encoding genes exhibit growth phenotypes with phytohormone treatments in the range of 1 to 10 M that correspond to GH3 protein expression changes (25,26,30,32,33). These differences highlight the need for further investigations into the metabolism of these molecules, which may alter local concentrations within different tissues and cell types of the plant and the fluxes that control plant growth and development.
Overall, the biochemical and in planta analysis of AtGH3.15 suggests a possible model for how altered expression affects plant growth (Fig. 5). Loss of AtGH3.15 in the T-DNA insertion lines would remove background conjugation to 2,4-DB, allowing more of the herbicide to be metabolized in the peroxisome (35)(36)(37). This results in the shortened root phenotype compared with WT Arabidopsis seedlings. In contrast, overexpression of AtGH3.15 would increase 2,4-DB conjugate formation, which results in the observed tolerance to herbicide treatment and the longer root phenotype. This parallels the effect of treating Arabidopsis AtGH3.15 knockout and overexpression lines with the auxin IBA (33). Overall, the biochemical, structure, and in planta experiments suggest the use of AtGH3.15 as a possible resistance trait for 2,4-DB.

Herbicide modification by an acyl acid amido synthetase
Although monocots and leguminous plants are inherently tolerant to 2,4-DB application, they are not completely resistant and dicots remain susceptible (4 -9). Overexpression of AtGH3.15 in planta could potentially enhance the tolerance of plants to 2,4-DB application. There are several possible advantages to exploring AtGH3.15 as a possible 2,4-DB resistance trait. Compared with overexpression of IAA-specific GH3 proteins, which results in severe growth phenotypes, such as dwarfing (29), overexpression of AtGH3.15 in Arabidopsis yielded no detrimental growth changes (33). Moreover, the distinct amino acid substrate profile of AtGH3.15 versus the IAAspecific proteins ( Table 2; Fig. S2), which primarily use aspartate and glutamate, may help maintain inactive forms of 2,4-DB and contribute to tolerance (25)(26)(27)(28)(29)(30)(31)(32)(33). Amino acid conjugated forms of IAA have varied roles with the IAA-aspartate and IAAglutamate conjugates leading to hormone degradation and IAA-alanine and IAA-leucine conjugates providing storage forms of the auxin (31,38,39). The best-studied IAA and 2,4-D conjugates are those of aspartate and glutamate, which suggests that these molecules can be hydrolyzed back into free acid forms (39). With 2,4-D conjugates this contributes to maintaining the effect of the herbicide (39). Currently, there is a lack of information on the metabolic fates of IBA and auxinic herbicides conjugated to other amino acids. Potential glutamine, cysteine, histidine, methionine, and tyrosine conjugates of IBA and 2,4-DB formed by AtGH3.15 need to be more fully explored with regard to biological fate and herbicide action.
In addition to the activity of AtGH3.15 with 2,4-DB, the structure of this enzyme in complex with the herbicide serves as a starting point to engineer variants that modify 2,4-D with amino acids that are neither aspartate nor glutamate as a means of exploiting potential differences in herbicide metabolism. As 2,4-D is a widely used herbicide, the ability to engineer activity of AtGH3.15 with 2,4-D would result in a greater agricultural impact than with 2,4-DB. Before the discovery and commercialization of auxin herbicides, like 2,4-D, perennial weeds were particularly difficult to control (40); however, to reduce the development of weeds with herbicide resistance multiple different modes of action for tolerance traits are needed. For example, extensive reliance on glyphosate in early agricultural biotech crops led to selection of weed populations with inherent tolerance to the herbicide and spurred the development of new herbicide-resistance traits.
To date, the molecular basis for enzyme-based herbicide protection strategies rely on a limited number of mechanisms: the use of modified enolpyruvylshikimate-3-phosphate synthase to prevent inhibition by glyphosate and glufosinate (21), acetylation of herbicide (18,19), degradation of 2,4-D by aryloxyalkanoate dioxygenases (23), conversion of dicamba by monooxygenases (22), modification of acetolactate synthase to prevent inhibition by sulfonylurea herbicides, degradation of Oxynil herbicide by a nitrilase (24), and use of p-hydroxyphenylpyruvate dioxygenases for mesotrione and isoxaflutole tolerance (20). Amino acid conjugation of herbicides may provide an additional resistance mechanism.
The benefits of herbicide-tolerant crops and the selection of resistant weeds, highlights the importance of discovery and development of new modes of action for herbicide tolerance. Stacked traits, the ability to tolerate different herbicidal applications, are also important for the future of herbicide tolerance in genetically modified crops to help combat the selection of herbicide-resistant weeds in the future (41). The development of 2,4-DB-tolerant crops via expression of AtGH3.15, or the use of an engineered variant that efficiently conjugates 2,4-D, would help to broaden the tool kit of herbicidal tolerance modes of action.

Mass spectrometry
Reactions were performed in the presence and absence of ϳ20 g AtGH3.15 with 50 mM Tris (pH 8), 3 mM MgCl 2 , 2 mM ATP, 2 mM 2,4-DB, and 2 mM glutamine in a 200 l volume. The reactions were allowed to react for 10 min at room temperature and then placed at Ϫ20°C. They were directly infused into the mass spectrometer. The MS1 (Q1) scan was acquired with the QTRAP 6500 (SCIEX) in low mass (LM) electrospray ionization in negative ion mode at a capillary voltage of Ϫ4500 and a mass range of 50 -600 m/z.

Protein crystallography
Crystals of AtGH3.15 in complex with 2,4-DB were grown by vapor diffusion in hanging drops of a 1:1 mixture of protein (13 mg ml Ϫ1 ) and crystallization buffer (1.2 M potassium phosphate (dibasic)/0.8 M sodium phosphate (monobasic) and 0.1 M sodium acetate/acetic acid, pH 4.5) with 2.5 mM 2,4-DB at 4°C. Crystals were frozen in liquid nitrogen with mother liquor sup-

Herbicide modification by an acyl acid amido synthetase
plemented with 15% (v/v) glycerol as a cryoprotectant. Diffraction data were collected at the SBC 19-ID beamline of the Argonne National Laboratory Advanced Photon Source with indexing and scaling performed using HKL-3000 (42). Molecular replacement was performed using PHENIX (43) with the three-dimensional structure of AtGH3.15 (PDB ID: 6AVH) (33) as a search model. Model building and refinement were performed with COOT (44) and PHENIX, respectively. Data collection and refinement statistics are summarized in Table 4. Coordinates and structure factors were deposited in the Protein Data Bank (PDB ID: 6E1Q).

Arabidopsis knockout and overexpression lines and root elongation assays
Confirmation and characterization of the two homozygous T-DNA insertion lines (SALK_108265C and SALK_071953) in the At5g13370 gene that codes for AtGH3.15 were described previously (33). Generation and characterization of the three independent A. thaliana Col-0 lines overexpressing N-terminally FLAG-tagged AtGH3.15 under control of the cauliflower mosaic virus 35S promoter were also previously reported (33). Root elongation assays to examine the effect of AtGH3.15 expression changes on resistance to 2,4-D and 2,4-DB used seeds that were surface sterilized with 70% (v/v) ethanol for 5 min, 90% (v/v) ethanol for 1 min, and resuspended in 0.1% (w/v) sterile agar. Sterilized seeds were stratified at 4°C for 2-4 days and plated on Murashige and Skoog plates with 0.6% (w/v) agar and supplemented with 0.5% (v/v) sucrose. Treatments were performed at 1 M 2,4-DB and 20, 40, and 80 nM 2,4-D with mock-treated plates receiving equivalent amounts of 70% (v/v) ethanol (2,4-DB and 2,4-D were dissolved in 70% ethanol). Plates were sealed with 3 M Micropore tape and incubated at 22°C under continuous white light for 10 days. Seedlings were excised from media and measured using a ruler. Percent root length versus mock-treated was calculated as (root length of treated seedlings)/(average root length of mock-treated seedlings)*100.