Resistance to a nonselective 4-hydroxyphenylpyruvate dioxygenase-inhibiting herbicide via novel reduction-dehydration-glutathione conjugation in Amaranthus tuberculatus.

Metabolic resistance to 4-hydroxyphenylpyruvate dioxygenase (HPPD)-inhibiting herbicides is a threat in controlling waterhemp (Amaranthus tuberculatus) in the USA. We investigated resistance mechanisms to syncarpic acid-3 (SA3), a nonselective, noncommercial HPPD-inhibiting herbicide metabolically robust to Phase I oxidation, in multiple-herbicide-resistant (MHR) waterhemp populations (SIR and NEB) and HPPD inhibitor-sensitive populations (ACR and SEN). Dose-response experiments with SA3 provided ED50 -based resistant : sensitive ratios of at least 18-fold. Metabolism experiments quantifying parent SA3 remaining in excised leaves during a time course indicated MHR populations displayed faster rates of SA3 metabolism compared to HPPD inhibitor-sensitive populations. SA3 metabolites were identified via LC-MS-based untargeted metabolomics in whole plants. A Phase I metabolite, likely generated by cytochrome P450-mediated alkyl hydroxylation, was detected but was not associated with resistance. A Phase I metabolite consistent with ketone reduction followed by water elimination was detected, creating a putative α,β-unsaturated carbonyl resembling a Michael acceptor site. A Phase II glutathione-SA3 conjugate was associated with resistance. Our results revealed a novel reduction-dehydration-GSH conjugation detoxification mechanism. SA3 metabolism in MHR waterhemp is thus atypical compared to commercial HPPD-inhibiting herbicides. This previously uncharacterized detoxification mechanism presents a unique opportunity for future biorational design by blocking known sites of herbicide metabolism in weeds.

Introduction

Metabolic detoxification plays an important role in conferring crop tolerance and weed resistance to herbicides that target a wide range of sites of action in plants (Kreuz et al., 1996). Herbicide metabolism‐based resistance in weedy plant species threatens effective and sustainable crop production (Yu & Powles, 2014; Rigon et al., 2020). Therefore, identifying herbicide detoxification pathways is essential to assist in the biorational design of new herbicides that avoid or overcome weed resistance mechanisms. Herbicide detoxification occurs in four sequential phases: Phase I (oxidation), Phase II (conjugation), Phase III (transport) and Phase IV (compartmentation) (Kreuz et al., 1996; Coleman et al., 1997; Barrett, 2000; Werck‐Reichhart & Feyereisen, 2000; Siminszky, 2006; Riechers et al., 2010; Délye et al., 2013). In most Phase I reactions, cytochrome P450 monooxygenase (P450) enzymes produce hydroxylated or dealkylated forms of herbicides. Plant P450 proteins are encoded by large, diverse multigene families consisting of c. 350 genes in rice and 250 genes in Arabidopsis, and are involved in numerous processes such as secondary product biosynthesis, fatty acid metabolism, hormone homeostasis and xenobiotic detoxification (Nelson et al., 2004; Paquette et al., 2009; Mizutani & Ohta, 2010; Schuler, 2015; Dimaano & Iwakami, 2021). Phase II metabolic reactions are catalyzed by glutathione S‐transferases (GSTs; EC 2.5.1.18) and uridine diphosphate‐dependent glycosyltransferases, forming GSH conjugates and glucose conjugates, respectively. Phase III processes involve vacuolar transport of herbicide conjugates (sequestration) mediated by ABC transporters, whereas Phase IV involves further transformation of conjugates through partial degradation, secondary conjugations or incorporation into the cell wall (Riechers et al., 2010; Zhang & Yang, 2021). Identifying these above‐mentioned phases of herbicide detoxification and the specific biochemical reactions occurring in resistant weeds is critical for designing new resistance‐breaking herbicide chemistries (Kaundun, 2021).

Among commonly used herbicides in corn are those that inhibit 4‐hydroxyphenylpyruvate dioxygenase (HPPD, EC 1.13.11.27). HPPD catalyzes the formation of homogentisic acid, the precursor of plastoquinone (PQ) and vitamin E, from catabolism of tyrosine (Schulz et al., 1993; Secor, 1994; Moran, 2005; Santucci et al., 2017). HPPD inhibitors, also known as pigment inhibitors, bleach meristematic tissues in sensitive weeds via carotenoid depletion, which results from the lack of PQ as a cofactor for phytoene desaturase (van Almsick, 2009). HPPD‐inhibiting herbicides possess several desirable features, such as broad‐spectrum activity against dicot weeds, natural crop selectivity, low application rate, low mammalian toxicity, and pre‐ or postemergence treatment flexibility, and display synergistic effects when combined with photosystem II‐inhibiting herbicides (Mitchell et al., 2001; Matringe et al., 2005; Ndikuryayo et al., 2017; O’Brien et al., 2018).

Resistance to HPPD‐inhibiting herbicides is a multigenic trait (Huffman et al., 2015) and has been reported in several populations of weedy Amaranthus species, especially in waterhemp (Amaranthus tuberculatus Moq. Sauer), a troublesome and competitive weed in corn and soybean (Hausman et al., 2011; Nandula et al., 2019; Heap, 2021). Waterhemp is a dioecious, small‐seeded, annual dicot, capable of producing up to a million seeds per female plant, even when grown under suboptimal growth conditions (Steckel, 2007). Multiple‐herbicide‐resistant (MHR) waterhemp populations and tolerant corn detoxify commercial, corn‐selective HPPD inhibitors by oxidative metabolism (Fig. 1) (Mitchell et al., 2001; Ma et al., 2013; Kaundun et al., 2017; Küpper et al., 2018). Specifically, hydroxylation of the cyclohexanedione ring predominates for the two triketones, mesotrione (Fig. 1a) (Ma et al., 2013) and tembotrione (Fig. 1c) (Küpper et al., 2018). In the pyrazole, topramezone (Fig. 1d), detoxification in HPPD inhibitor‐resistant waterhemp and tolerant corn occurs by hydroxylation of the isoxazoline ring or N‐demethylation of the pyrazole ring, respectively (Grossmann & Ehrhardt, 2007; Lygin et al., 2018). Furthermore, these Phase I oxidative reactions are likely catalyzed by P450 enzymes (Meunier et al., 2004; Siminszky, 2006; Dimaano & Iwakami, 2021).

Fig. 1

Structures of 4‐hydroxyphenylpyruvate dioxygenase (HPPD) inhibitors. Commercial HPPD‐inhibiting herbicides: (a) mesotrione, (b) sulcotrione, (c) tembotrione and (d) topramezone; and noncommercial HPPD inhibitors: (e) syncarpic acid‐1 (SA1), (f) SA2 and (g) SA3. The blue arrows in mesotrione, sulcotrione and tembotrione point towards sites of oxidative metabolism reported in HPPD‐inhibitor‐resistant waterhemp populations and tolerant corn (a–c), whereas the blue and red arrows in topramezone point towards the site of oxidative metabolism in HPPD‐inhibitor‐resistant waterhemp and tolerant corn, respectively (d). Syncarpic acids are characterized by alternating methyl groups (R‐(CH3)2; shaded in green) and ketone substituents (shaded in orange) in their cyclohexanetrione ring (e–g). SA2 and SA3 possess one and two trifluoromethyl substituents (shaded in blue) in their aromatic rings, respectively (f, g).

We had previously theorized that noncommercial HPPD inhibitors with limited plant metabolism could be utilized as chemical probes to investigate potential resistance mechanisms other than, or in addition to, enhanced oxidative metabolism in MHR populations (Ma et al., 2013). Considering paradigms for metabolism (Mitchell et al., 2001; Durand et al., 2006; Grossmann & Ehrhardt, 2007) and known resistance mechanisms to commercial HPPD inhibitors (Ma et al., 2013; Kaundun et al., 2017; Küpper et al., 2018; Lygin et al., 2018), our main interests were to: investigate whether nonselective, noncommercial syncarpic acids (SAs) are metabolized by HPPD inhibitor‐resistant waterhemp, given that known structural sites for oxidative metabolism are theoretically blocked (Beaudegnies et al., 2009); and inform future biorational herbicide design to effectively manage resistant weeds (Kaundun, 2021). Before our current study, SA analogs with characteristic 2‐benzoyl or 2‐heteroaroyl cyclohexane‐1,3‐dione templates of triketones (Beaudegnies et al., 2009) were tested (such as SA1, SA2 and SA3; Fig. 1e–g) to find new crop‐selective herbicides with high unit efficacy (i.e. low field application rates). SA3 (IUPAC name: 6‐[2,4‐bis(trifluoromethyl)benzoyl]‐2,2,4,4‐tetramethylcyclohexane‐1,3,5‐trione) is a noncommercial, nonselective, synthetic compound derived from the natural product, leptospermone (IUPAC name: 2,2,4,4‐tetramethyl‐6‐(3‐methylbutanoyl)cyclohexane‐1,3,5‐trione). Leptospermone was originally discovered as an allelochemical produced by the bottlebrush plant (Callistemon citrinus) and is a progenitor of mesotrione (Beaudegnies et al., 2009). HPPD is a target site for β‐triketones extracted from Leptospermum scoparium (Dayan et al., 2007). SA3 has two trifluoromethyl groups (Fig. 1g; shaded in blue) that play important roles in pharmaceutical and agrochemical design, whereby incorporation into drug candidates provides enhanced chemical and metabolic stability, improved lipophilicity and bioavailability, and increased protein‐binding affinity (Wang et al., 2014). The cyclohexanetrione ring of SA3 is considered robust to metabolism since it contains alternating methyl (R‐(CH3)2) groups (Fig. 1g; shaded in green) and ketone substituents (Fig. 1g; shaded in orange), rendering the molecule electron‐deficient, and hence presumed as recalcitrant to metabolism (Beaudegnies et al., 2009).

Recent advances in separation techniques, including gas chromatography and liquid chromatography combined with mass spectrometry (LC‐MS), have facilitated detection of a wide range of primary and secondary metabolites in plants (Fiehn, 2002; Hall et al., 2002; Theodoridis et al., 2008). However, the application of such techniques in herbicide metabolism research has not been utilized for determining the fate of nonselective, noncommercial HPPD inhibitors in plants. Herbicidal activity of soil‐applied and foliar‐applied SA3 was previously assessed using several grass and broadleaf weed species (not including waterhemp) (Beaudegnies et al., 2009). SA3 demonstrated greater levels of weed control compared to several related 2‐benzoyl and 2‐heteroaroyl syncarpic acids tested, but also severely injured corn and cereal crops due to lack of metabolism (Beaudegnies et al., 2009). However, we hypothesized that detection of SA3 metabolites via untargeted metabolomics may reveal structural sites amenable to enzymatic modification and assist in postulating a metabolic pathway for SA3 detoxification in MHR waterhemp. Therefore, the goal of this research is to better understand resistance mechanisms and potential detoxification pathways in waterhemp for the presumed metabolically blocked SA3 molecule, which is not selective in corn or cereal crops (Beaudegnies et al., 2009). We used a combination of targeted analysis of SA3 metabolism during a time course and an untargeted metabolomics approach to test the hypothesis that SA3 resistance in waterhemp is metabolism‐based and to uncover the metabolic pathway that confers resistance.

Materials and Methods

Plant materials

The four waterhemp populations utilized in this research are the same as those described previously (Kaundun et al., 2017; Lygin et al., 2018; O’Brien et al., 2018). Two are sensitive to HPPD‐inhibiting herbicides: standard sensitive population (SEN) (Kaundun et al., 2017) and Adams County mesotrione‐sensitive but atrazine‐resistant (ACR) (Patzoldt et al., 2005; Hausman et al., 2011). The other two populations are resistant to postemergence applications of mesotrione, tembotrione and topramezone: Stanford, Illinois Resistant (SIR) (Hausman et al., 2011) and Nebraska (NEB) (Kaundun et al., 2017). In addition, SIR and NEB are resistant to the foliar‐applied herbicides atrazine and acetolactate synthase‐inhibitors, but are sensitive to glyphosate and glufosinate (Hausman et al., 2016; Heap, 2021). For comparison with waterhemp, hybrid corn (DKC 63‐14 RR) was used.

Waterhemp plants were grown at 28°C : 22°C, day : night temperature, and at a 16 h : 8 h photoperiod as described earlier (Lygin et al., 2018). In brief, stratified seeds were germinated in 12 × 12 cm trays with a commercial potting medium (Sun Gro Horticulture, Bellevue, WA, USA) in the growth chamber (Controlled Environments Limited, Winnipeg, MB, Canada), transplanted into 80 cm3 pots in the glasshouse upon emergence, and then transplanted into 950 cm3 pots containing a 3 : 1 : 1 : 1 mixture of potting mix : soil : peat : sand. Slow‐release fertilizer (Osmocote; The Scotts Company, Marysville, OH, USA) was added to this mixture. Corn seeds were planted 2.5 cm deep in the same soil mixture. Plants at a height of 10–12 cm were transferred to a growth chamber for 24 h before conducting herbicide metabolism studies.

Chemicals and reagents

Unless otherwise stated, all solvents were HPLC grade (Thermo Fisher Scientific Inc., Waltham, MA, USA), formic acid was MS grade (Fisher Scientific, Loughborough, UK), and analytical‐grade syncarpic acids 1–3 were supplied by Syngenta Ltd (Cambridge, UK).

Whole‐plant resistance study

Whole plants of the four waterhemp populations (SEN, ACR, SIR and NEB) were subjected to dose–response analysis with SA3 in the glasshouse at the Syngenta Research Station, Jealott’s Hill, Bracknell, UK, using similar methods and glasshouse conditions as previously described (Lygin et al., 2018). Pots (10 cm in diameter) containing a commercial potting medium of sandy loam soil included 30 waterhemp plants (7 cm) from each population. Plants were treated with nine rates of SA3, with five replicates per treatment, to determine effective doses (ED) that provided 50% or 80% control. A limited amount of formulated SA3 was available; consequently, it was not possible to apply SA3 rates > 3200 g ha−1 to obtain all ED50, ED80 and resistance index (RI) values for the two MHR populations. Spray solutions including the following spray adjuvants: 1% (v/v) Agridex (Helena Chemical, Collierville, TN, USA) and ammonium sulfate at 2.5% (w/v). Plants were assessed for visual percentage control compared to an untreated control for each corresponding population at 21 d after treatment. ED50 and ED80 estimates and RIs (Table 1) for the two MHR populations, NEB and SIR, were generated relative to the two HPPD‐inhibitor‐sensitive populations, SEN and ACR, by fitting a parallel line regression analysis of logit‐transformed percentage control on the log10(rate).

Table 1

Effective dose (ED) estimates of 50% and 80% waterhemp control with syncarpic acid‐3 (SA3) in four populations.

Population	ED50 (g ai ha−1)	RI		ED80 (g ai ha−1)
		ACR	SEN
SIR	> 3200 a (nd)	> 67.3 (nd)	> 40.0 (nd)	> 3200 (nd)
NEB	1441 (987–2194)	30.3 (17.2–57.3)	18.0 (10.5–32.9)	> 3200 (nd)
SEN	80.0 (55.5–113.1)	–	–	454 (322–652)
ACR	47.5 (32.2–68.1)	–	–	267 (192–382)

nd, not determined.

Resistance indices (RI) of multiple‐herbicide‐resistant populations (SIR and NEB) to SA3 are indicated relative to two 4‐hydroxyphenylpyruvate dioxygenase‐inhibitor‐sensitive populations, ACR and SEN, based on ED50 values only. Confidence limits at 95% are listed in parentheses next to ED50 and ED80 values. Data were calculated based on the dose–response curves depicted in Fig. 2.

Since a limited amount of formulated SA3 was available, it was not possible to apply a high enough rate to obtain all ED50, ED80 and RI values for SIR and NEB.

Fig. 2

Dose–response analysis of syncarpic acid‐3 (SA3) in waterhemp populations. Linear regression plot of the response to SA3 in two multiple‐herbicide‐resistant populations: from Nebraska, USA (NEB), and Stanford, Illinois, USA (SIR), and two 4‐hydroxyphenylpyruvate dioxygenase‐inhibitor‐sensitive populations: standard sensitive (SEN) and Adams County, Illinois, USA (ACR; atrazine‐resistant). Waterhemp plants (7 cm) from each population were treated with nine rates of SA3 at five replications per treatment. Spray solutions included the following spray adjuvants: 1% (v/v) Agridex (Helena Chemical) and ammonium sulfate at 2.5% (w/v). Plants were assessed for visual percentage of control compared to an untreated control for each corresponding population, 21 d after treatment. ED50 and ED80 estimates and resistance indices (shown in Table 1) were obtained as described in the Materials and Methods section by fitting a parallel line regression analysis of logit‐transformed percentage control on the log10(rate).

SA3 metabolism in excised waterhemp leaves

Excised leaf assay time course

The excised leaf assay was carried out with four waterhemp populations according to protocols described previously, with minor modifications (Ma et al., 2015; Lygin et al., 2018). Two leaves (third and fourth youngest) from each waterhemp plant were collected in a container with water, and leaf petioles were cut again (c. 3 mm) with a razor blade under water, then placed in 1.5 ml tubes (VWR Scientific, Batavia, IL, USA) containing 200 µl of 0.1 M Tris–HCl buffer (pH 6.5) to equilibrate for 1 h. Leaves were then transferred to new 1.5 ml tubes containing SA3 incubation solution (0.3 mM SA3 in 0.1 M Tris–HCl (pH 6.5)). The youngest corn leaf from 10 to 12 cm plants was processed as above with waterhemp leaves for comparison.

After 2 h of incubation in SA3 solution, leaves were washed with 0.1 M Tris–HCl buffer (pH 6.5) and placed in 1.5 ml tubes containing 500 µl of one quarter‐strength Murashige and Skoog’s basal salts liquid medium (MP Biomedicals, LLC). Leaves were harvested at five time points (2, 4, 8, 12, and 16 h after treatment (HAT) with SA3), briefly rinsed in deionized water and then dried with tissue paper. Tissue fresh weights of the leaves were recorded, followed by placing leaves in 2 ml tubes with screw caps and then immediately freezing in liquid nitrogen. Mean uptake of SA3 from the incubation solution was c. 10% after 2 h. Frozen leaves were stored at −80°C until further analysis. Each waterhemp population per time point consisted of a total of nine replicates (i.e. excised leaves from different plants) from two independent experiments.

Metabolite extraction for SA3 quantification

Leaf samples obtained from excised leaf assays were ground to a powder in liquid nitrogen using a mortar and pestle. The internal standard, syncarpic acid‐2 (50 µl, 100 µM) (Fig. 1f), was added to the samples. Extraction was carried out twice with 80% methanol (containing 0.1% formic acid) with agitation at 10°C for 12 h. Next, samples were centrifuged at 10 000  (10 min, 4°C) using a refrigerated centrifuge (5417R; Eppendorf, Hamburg, Germany). Supernatants from the first and second extractions were combined and concentrated under vacuum (SpeedVac, Farmingdale, NY, USA) at 40°C for 2 h. Phase separation was achieved by adding methylene chloride and water (2 : 1, v/v), and tubes were gently inverted, centrifuged at 3000  for 1 min and then incubated at 23°C for 4 h. Finally, samples were centrifuged for 10 000  for 10 min, and the lower methylene chloride layer was transferred to a fresh 1.5 ml tube and then concentrated under nitrogen gas for 20 min. Experimental blank extracts (without herbicide) were prepared identically. Immediately before analysis, the extract was filtered through a 0.45 µm nylon filter (Fisher Scientific), reconstituted in 250 μl of water : acetonitrile (1 : 1, v/v) containing 0.1% formic acid and stored in a polyspring glass insert (Thermo Scientific) that was sealed in a 2.0 ml HPLC glass vial (Fisher Scientific). Samples were randomized before analysis.

Quantification of SA3 using HPLC‐PDA

HPLC separations were performed with a Waters Alliance separations module (model 2695) coupled with a Waters 996 photodiode array detector (PDA; Waters Corp.). The HPLC device was operated at a flow rate of 1.0 ml min−1 with a C18 column (Alltima, 250 × 4.6 mm, 5 μm particle size; 30°C; Hichrom Ltd) equipped with a Javelin filter (4 mm; Thermo Scientific). The mobile phase solvents consisting of HPLC‐grade water (A) and HPLC‐grade acetonitrile (B) and were acidified with 0.1% (v/v) MS‐grade formic acid. The injection volume was 50 μl. Analytes were eluted with the following gradient program: 0–10% B (3 min), 10–60% B (8 min), 60–80% B (3 min), 80–95% B (3 min) and 95% B (3 min, isocratic) before returning to 0% B. The peak of SA3 was detected with the PDA at 280 nm. The recovery of SA3 was c. 80% based on the mean recoveries obtained from test samples in preliminary experiments (data not shown). The peak area of recovered SA3 was normalized based on the percentage recovery of internal standard after the extraction procedure and then calculated using a standard curve (Supporting Information Fig. S1). This normalized value was divided by the tissue fresh weight and then expressed as a percentage of the average SA3 concentrations of the 2 HAT samples (separately for each population and repeated experiment).

Untargeted metabolomics in excised leaves and whole waterhemp plants

Excised leaf assay at two time points

A separate excised leaf assay using ACR and SIR populations was carried out as described in ‘the Excised leaf assay time course section’. Experimental blank extracts (without herbicide) were prepared identically. Leaves were harvested from three biological replicates for each of the SA3‐treated and untreated waterhemp populations at 4 and 12 HAT.

Whole‐plant treatments

The four waterhemp populations analyzed using HPLC‐PDA were also tested for SA3 metabolism using whole plant samples with methods described previously (Lygin et al., 2018). The third‐ and fourth‐youngest leaves (from 10 to 12 cm waterhemp plants) were treated with 50 µl of 0.3 mM SA3 (in 0.1 M Tris–HCl (pH 6.5) containing 0.1% (v/v) nonionic surfactant) (Fig. S2). Each treated leaf received 25 µl SA3 solution (3.2 µg SA3 per leaf) applied as c. 0.3 µl droplets with a Hamilton glass syringe. At 12 and 24 HAT, treated leaves were harvested (including the petioles), washed in 10 ml of 20% methanol to remove unabsorbed SA3, fresh weights were recorded, leaves were then flash frozen in liquid nitrogen and stored at −80°C until extraction and further analysis. Five biological replicates of each population × time point after treatment were used.

Metabolite extraction for untargeted metabolomics

Metabolite extraction for untargeted metabolomics was performed on leaf samples obtained from the excised leaf and whole‐plant assays. Methods employed were modified from a protocol used previously to study unlabeled topramezone (Lygin et al., 2018). These modifications included the extraction of fresh leaves instead of freeze‐dried tissues and analysis of methanolic extracts without additional partitioning steps (Lygin et al., 2018). Experimental blank extracts (without herbicide) were prepared identically. Frozen tissue was extracted once with 1 ml of 80% methanol (containing 0.1% formic acid) overnight at 10°C, then once with 0.5 ml of 95% methanol for 2 h. Combined extracts were centrifuged, concentrated under nitrogen gas and then reconstituted in 1 ml acetonitrile : water (1 : 1, v/v) containing 0.1% formic acid. Immediately before analysis, samples were filtered through a 0.45 µm nylon filter (Fisher Scientific) and sealed in a 2 ml HPLC glass vial (Fisher Scientific). Samples were randomized before further analysis.

Untargeted metabolomics: LC‐MS and data processing

Untargeted metabolomics of waterhemp leaf extracts was carried out using a Dionex Ultimate 3000 series HPLC system (Thermo Scientific, Germering, Germany) coupled to a Q Exactive Hybrid Quadrupole‐Orbitrap Mass Spectrometer (Thermo Scientific, Bremen, Germany) as described previously (Elolimy et al., 2019). The chromatographic analysis was conducted in a randomized sequence order with pooled quality control (QC) samples injected at the beginning of the analysis to equilibrate the analytical platform and after every 10 test samples to evaluate the stability of the experimental procedure (Sangster et al., 2006; Dunn et al., 2011; Godzien et al., 2015; Wehrens et al., 2016). Raw data files obtained in full‐MS mode (samples, procedural blank and QC) and data obtained in full‐MS followed by data‐dependent MS2 were processed by Compound Discoverer software v.3.1.305 (Thermo Scientific) for initial data processing, including peak detection, peak alignment and peak integration. Detailed procedures of metabolite peak processing are described in Methods S1.

Since the negative electrospray ionization mode (ESI (−)) yielded greater peak abundance of putative SA3 metabolites compared to the positive electrospray ionization mode (ESI (+)), a separate data analysis was carried out on the ESI (−) data using MS‐Dial v.4.24 (with open source publicly available EI spectra library) following the parameters previously described (Tsugawa et al., 2015). Details of data preprocessing are presented in Methods S1. The putative SA3 metabolites, all unknowns (Table 2) (Sumner et al., 2007), were inspected for MS2 fragmentation patterns, and normalized peak areas were exported for comparisons among waterhemp populations. Before multivariate statistics, a compound detection rate of 50% relative to the QC samples was set (Broadhurst et al., 2018), leaving a total of 2580 and 2930 compounds for the excised leaf assay and the whole plant assay samples, respectively.

Table 2

Major first‐generation product ion (MS2) mass spectra of syncarpic acid‐3 (SA3) and its putative metabolites detected in waterhemp populations using LC‐MS in negative ionization mode.

Compound	RT (min)	Elemental composition	Theoretical m/z [M−H]	Experimental m/z [M−H]−	m/z error (ppm)	MS2 ions, m/z
						Precursor	Fragment a
SA3 b	23.50	C19H16F6O4	421.0880	421.0886	1.3679	421.0870	213.0140
							281.0030
							80.9638
							196.0360
							257.0029
M437 (hydroxy‐SA3) c	20.33	C19H16F6O5	437.0829	437.0827	−0.5811	437.0832	213.0137
							181.0505
							367.0433
							80.9637
							339.0472
							196.0719
							257.0023
M424 (reduced‐SA3) c	23.44	C19H18F6O4	423.1036	423.0940	−22.6994	423.0676	255.2335
							59.0123
							213.0139
							281.0029
							381.0561
							257.0029
M405 (dehydrated‐SA3) c	17.41	C19H16F6O3	405.0931	405.0559	−91.8984	405.0863	281.0033
							213.0140
							257.0022
							96.9578
							361.0678
M527 (SA3‐Cys) c	17.03	C22H23F6NO5S	526.1128	526.0761	−69.7884	526.0743	405.0548
							281.0033
							120.0106
							213.0137
							185.0376
							257.0047
M582 (SA3‐Cys‐Gly) c	17.92	C24H24F6N2O6S	581.1186	581.0890	−50.9801	581.0890	405.0542
							175.0231
							115.0021
							281.0029
							213.0136
							257.0027
M584 (SA3‐Cys‐Gly) c	17.12	C24H26F6N2O6S	583.1343	583.0684	−113.0639	583.0669	407.0335
							175.0230
							115.0021
							281.0022
							213.0138
							365.0240
							96.9585
M713 (SA3‐GSH) c	16.12	C29H33F6N3O9S	712.1769	712.1409	−50.5855	712.1048	405.0541
							306.0759
							281.0031
							254.0768
							213.0134
							181.0599
							257.0027

RT, retention time of the compound; Elemental composition, putative molecular formula; m/z, mass‐to‐charge ratio of the compound detected using LC‐MS; m/z error, difference between theoretical and detected m/z values.

MS2 ions is bold type were also detected in SA3.

Major MS2 ions detected.

Identified compound.

Unknown compounds: identification level according to Metabolomics Standards Initiative (Sumner et al., 2007).

Statistical analysis

SA3 quantification

The concentration of SA3 was calculated using a standard curve generated with authentic SA3 parent (Fig. S1; supplied by Syngenta, UK) and the amount of remaining parent SA3 was expressed as a percentage of its concentration at 2 HAT. ANOVA was carried on the pooled samples (n = 9) using proc glimmix in Sas 9.4. Data were analyzed via a generalized linear model fit to a beta distribution with logit link function. Fixed effects were population (SIR, NEB, ACR and SEN), HAT (2, 4, 8, 12 and 16) and their interaction. Random effects included experiment and replicate nested within experiment. Means were separated by Fisher’s protected least significant difference with α = 0.05. Graphs were created using GraphPad Prism 8 (GraphPad Software Inc., San Diego, CA, USA).

Multivariate statistics of metabolomics data

For multivariate analysis, the peak areas of metabolite features of waterhemp leaf extracts were normalized via log transformation. Peak areas were then Pareto scaled to moderate the influence of compounds produced at high concentrations and to recognize the effects of compounds produced at relatively lower concentrations in the overall characterization of waterhemp metabolites (van den Berg et al., 2006). To visualize trends in the global metabolomic data among the waterhemp populations, data were represented by unsupervised modeling through principal components analysis (PCA) using Simca v.15 (Umetrics, Umeå, Sweden). Next, to determine the discriminating compounds in the SA3‐treated samples used in this study, orthogonal projections to latent structures discriminant analysis (OPLS‐DA) was conducted, again in Simca v.15 (Umetrics). OPLS‐DA maximizes the covariance of the matrix of predictor data, X (metabolite features), and the response block, Y (treatment), thus providing a more powerful prediction of among‐group variation compared to PCA (Worley et al., 2013). Following this analysis, hierarchical clustering using Ward’s linkage method was conducted on the X variables. Pearson’s correlation analysis and generation of correlation plots were performed for putative SA3 metabolites using the corrplot package in RStudio (R Core Team, 2013). Visual comparisons of the relative abundance of putative SA3‐metabolites were carried out using GraphPad Prism 8 software (GraphPad Software Inc.).

Results

MHR waterhemp populations are highly resistant to foliar‐applied SA3

In preliminary glasshouse studies, SA1 and SA2 did not control sensitive waterhemp at rates of up to 600 g ha−1 (data not shown). SA3 (Fig. 1g) was then tested and demonstrated higher foliar activity than either SA1 or SA2 in sensitive waterhemp. Foliar SA3 treatments ranging from 12.5 to 3200 g ha−1 were applied (Fig. 2), then effective dose (ED50 and ED80) values and estimated resistance : sensitive ratios, or RIs, were determined (Table 1). The MHR waterhemp populations (SIR and NEB) responded differently to SA3 compared with the HPPD‐inhibitor‐sensitive populations (ACR and SEN), with SIR exhibiting the highest level of resistance (Fig. 2; Table 1). ACR was the most sensitive population based on ED50 and ED80 values (Table 1). Relative to ACR, the calculated RIs for NEB and SIR to SA3 were c.  30 and > 67, respectively, and RIs were 18 and > 40, for NEB and SIR, respectively, relative to SEN (Table 1). The magnitude of resistance to SA3 in the SIR population compared with NEB is consistent with previous research investigating foliar‐applied topramezone in these same populations (Lygin et al., 2018).

SA3 is metabolized faster by MHR waterhemp populations

The two MHR waterhemp populations (SIR and NEB) metabolized SA3 more rapidly than the two sensitive populations (ACR and SEN) (Fig. 3), and the least amount of parent SA3 was quantified in SIR at each time point (Fig. 3a). This amount of SA3 parent, however, did not differ significantly between SIR and NEB at earlier time points in the assay (Fig. 3b) and at the latest time point examined (Fig. 3c). Resistance to SA3 in MHR waterhemp is thus conferred by enhanced metabolism. As a result, experiments were designed to detect and identify SA3 metabolites with increased abundance in MHR waterhemp to further investigate elevated rates of metabolism and elucidate the pathway of SA3 detoxification.

Fig. 3

Metabolism of syncarpic acid‐3 (SA3) in waterhemp populations. (a) Quantification of SA3 remaining expressed as the percentage of the 2 h after treatment (% of 2 HAT) in excised leaf samples of two multiple‐herbicide‐resistant populations from Nebraska (NEB) and Stanford, Illinois, USA (SIR), and two 4‐hydroxyphenylpyruvate dioxygenase‐inhibitor‐sensitive populations: standard sensitive (SEN) and Adams County, Illinois, USA (ACR; atrazine‐resistant). (b) Comparison of SA3 remaining at 4 HAT. (c) Comparison of SA3 remaining at 16 HAT. The third‐ and fourth‐youngest leaves from waterhemp plants (10–12 cm) were treated with 200 µl of 0.3 mM SA3 in 0.1 M Tris (pH 6.5). Relative concentrations of SA3 extracted from the treated leaves are plotted on the y‐axis, using SA2 as an internal standard as described in the Materials and Methods section. Vertical bars represent the SE of the treatment mean. Different letters indicated above the error bars denote a significant difference at α = 0.05.

Mass spectral analysis of SA3

LC‐MS of an authentic SA3 standard provided first‐generation product ion (MS2) mass spectra in the negative ion mode (Fig. S3; Table S1). The parent compound eluted at c. 23 min with an organic solvent composition of c. 90%. The precursor ion for SA3 is m/z 421.0916 [M−H] and its major fragments ions are as follows: m/z 213, m/z 257, m/z 196 and m/z 281 (Fig. S3; Table S1). The fragment ion m/z 213 corresponds to the CF3‐substituted aromatic ring of SA3, m/z 281 is probably formed from fragmentation of the cyclohexanetrione ring, m/z 257 is a carboxylic acid derivative of the CF3‐substituted benzoyl ring of SA3 derived from fragmentation in the bridge carbon (C7’) of SA3, whereas m/z 196 is probably the cyclohexanetrione portion of SA3 (Fig. S3). We therefore expected that SA3 metabolites are likely to possess these key fragment ions.

Untargeted metabolomics of waterhemp leaf extracts detected SA3 metabolites

Since SA3 is a noncommercial herbicide without literature reports of its possible metabolic fate in plants or soil, detection of as many metabolites as possible using untargeted metabolomics was necessary to provide evidence of putative enzymatic detoxification of SA3 in MHR waterhemp. Untargeted metabolomics using LC‐MS in both ESI (+) and ESI (−) mode was carried out to detect SA3 metabolites following the discovery of elevated rates of metabolism of SA3 in SIR and NEB compared to sensitive populations (Fig. 3). Fig. 4 illustrates a PCA scores plot (Fig. 4a) and loadings plot (Fig. 4b) of > 2000 metabolite features of ACR and SIR excised leaves treated with SA3 at 4 and 12 HAT. Variation in the metabolite profiles of these two waterhemp populations is explained by PC1 and PC2 at 25.3% and 13.2%, respectively. PC1 is explained by differences in metabolites between the populations, whereas PC2 is explained by metabolite variation between time points. SIR exhibited an increased level of metabolic diversity compared to ACR (Fig. 4a,b), and SA3 (highlighted in red in the loadings plot; Fig. 4b) is characteristic for ACR. Matching the HPLC‐PDA results, untargeted metabolomics of excised leaves from ACR and SIR revealed decreasing amounts of SA3 from 4 to 12 HAT, with SIR exhibiting significantly less parent herbicide at the later time point (Fig. 4). OPLS‐DA, followed by hierarchical clustering of the metabolite features between the SA3‐treated and nontreated control samples, yielded four main groups of compounds (Fig. S4; Table S2). Group 3 metabolite features (in pink; Fig. S4) represent the most discriminating compounds in SA3‐treated samples. Of 100 metabolite features in this group, seven revealed signature MS2 fragment ions that can be traced back to MS2 data of SA3 (Table S1; Fig. S3). These metabolite features were detected in ESI (−) mode and are unknowns (Table 2).

Fig. 4

Principal components analysis (PCA) of compounds detected in syncarpic acid‐3 (SA3)‐treated and nontreated excised leaf extracts of waterhemp populations ACR and SIR. Treated samples were incubated with 0.3 mM SA3 for 2 h and harvested at 4 and 12 h after treatment (HAT). (a) Scores scatter plot of the SA3‐treated samples. Scores are colored according to population and labeled according to HAT. The quality control (QC) samples are indicated by yellow diamonds. (b) Loadings scatter plot of compounds in the treated leaves of whole waterhemp plants from two populations analyzed using untargeted metabolomics via ultraperformance liquid chromatography–mass spectrometry (Thermo Scientific). SA3 is highlighted in red in the loadings plot. PC1 and PC2, respectively, explain 25.3% and 13.2% of the total variation. X, metabolite features detected in SA3‐treated excised leaves of the two waterhemp populations; SIR, multiple‐herbicide‐resistant waterhemp population from Stanford, Illinois, USA; ACR, 4‐hydroxyphenylpyruvate dioxygenase‐inhibitor‐sensitive but atrazine‐resistant waterhemp population from Adams County, Illinois, USA.

Reprocessed LC‐MS data in the ESI (−) mode of these samples were inspected for SA3‐related MS2 features, and the same major fragment ions in these seven discriminating unknown compounds were identified. OPLS‐DA results of the metabolite features in SA3‐treated leaves obtained from the whole‐plant assay of four waterhemp populations are shown in Fig. 5. The model explained c. 25% of the metabolite variation among waterhemp populations, and clearly separated ACR from NEB and SIR (Fig. 5a). Upon close inspection of whole‐plant metabolite features discriminating for resistant populations SIR and NEB (Fig. 5a), putative SA3 metabolites previously detected in excised leaf samples were detected again (highlighted in red; Fig. 5b).

Fig. 5

Orthogonal partial least squares discriminant analysis of compounds detected in syncarpic acid‐3 (SA3)‐treated and nontreated whole plant leaf extracts of waterhemp populations ACR, SEN, NEB and SIR. (a) Scores scatter plot of SA3‐treated and nontreated control samples of two multiple‐herbicide‐resistant waterhemp populations (Stanford, Illinois, USA (SIR), and Nebraska, USA (NEB)) and two 4‐hydroxyphenylpyruvate dioxygenase‐inhibitor‐sensitive waterhemp populations (Adams County, Illinois, USA (ACR; atrazine‐resistant), and standard sensitive (SEN)). Leaf samples were treated with 25 µl of 0.3 mM SA3 and harvested 12 and 24 h after treatment (HAT). Scores are colored according to population and labeled according to HAT. (b) Loadings scatter plot of compounds in the treated leaves from whole waterhemp plants analyzed using untargeted metabolomics via liquid chromatography–mass spectrometry (Thermo Scientific). Putative SA3 metabolites are highlighted in red in the loadings plot. PC1 and PC2, respectively, explain 18.1% and 6.5% of the total variation. X, metabolite features detected in SA3‐treated leaves from whole plants of the four waterhemp populations; Y, the four waterhemp populations compared to determine SA3 resistance mechanisms in this study.

Phase I metabolism of SA3 does not occur in the CF3‐substituted ring

Mass spectral analysis of SA3 and its metabolites did not reveal evidence of metabolism of the CF3‐substituted aromatic ring (Fig. 6). Putative SA3 metabolites exhibiting a mass spectral fragment ion m/z 213 indicate an intact CF3‐substituted aromatic ring (Table 2; Fig. 6b–h). These putative SA3 metabolites displayed other unifying fragment ions m/z 281 and m/z 257 that are also major fragment ions of SA3 (Fig. S3; Table S1). However, none of these metabolites possessed the ions m/z 229, m/z 256 or m/z 273 (Fig. S5), supporting our hypothesis that metabolism of the CF3‐substituted aromatic ring of SA3 does not occur. Among these putative SA3 metabolites is a hydroxy‐SA3 metabolite (m/z 437 [M−H]) detected in all treated samples (Table 2). A proposed structure of this metabolite is characterized by hydroxylation of C9 or C10 of the cyclohexanetrione ring of SA3 (Fig. 6b), with major fragment ions m/z 213, m/z 181, m/z 367, m/z 339 and m/z 257 in the ESI (−) mode. However, multivariate statistical data indicated this metabolite is not associated with resistance, given that its relative abundance across the samples examined was consistently greater in the ACR population compared with MHR populations (Figs 4, 5).

Fig. 6

Representative first‐generation product ion (MS2) mass spectra of seven putative metabolites of syncarpic acid‐3 (SA3) in resistant waterhemp populations. (a) SA3, (b) M438 (hydroxy‐SA3), (c) M424 (reduced SA3), (d) M406 (dehydrated M424), (e) M713 (SA3‐GSH), (f) M582 (SA3‐Cys‐Gly), (g) M584 (SA3‐Cys‐Gly) and (h) M527 (SA3‐Cys). Data were acquired using ultraperformance liquid chromatography–mass spectrometry.

Without strong evidence for a hydroxy‐SA3 metabolite discriminating for resistance, the course of our research pivoted towards identifying other metabolites formed in MHR waterhemp. Further inspection and structural elucidation of the discriminating compounds for SA3‐treated samples revealed two putative Phase I metabolites, M424 and M406 (Table 2). The proposed structures for these metabolites are summarized in Fig. 6(c,d). These two metabolites contain the major fragment ions m/z 213, m/z 257, m/z 96 and m/z 281 in the ESI (−) mode. M424 is a putative reduced‐SA3 metabolite, whereas M406 is putatively formed from dehydration of M424 (Fig. 7). M406 contains an α,β‐unsaturated carbonyl, a reactive electrophilic site (Farmer & Davoine, 2007), which is potentially a substrate for GST‐mediated conjugation (Fig. 7).

Fig. 7

Proposed routes of syncarpic acid‐3 (SA3) metabolism in waterhemp. We postulate that Phase I metabolism occurs in two ways: alkyl hydroxylation of SA3, presumably catalyzed by cytochrome P450 monooxygenase, forming a hydroxy‐SA3 metabolite (M438); and a putative reductase‐catalyzed transformation of SA3 into M424 (SA3‐reduced), which undergoes dehydration to form M406 (SA3‐dehydrated). Phase II metabolism of SA3 occurs by conjugation of M406 with reduced glutathione, followed by vacuolar transport via tonoplast ABC transporters in Phase III, then catabolism within the vacuole of M713 to its amino acid constituents, M584 (SA3‐Cys‐Gly dipeptide) and M527 (SA3‐Cys) in Phase IV.

Phase II metabolism of SA3 results in formation of a glutathione conjugate

Putative Phase II metabolites of SA3 likely consist of higher molecular weights and shorter retention times relative to SA3 (Table 2; Fig. 6e–h). A putative SA3‐GSH conjugate, minus an oxygen (M713), and its amino acid catabolites (M582, M584 and M527) were detected using untargeted metabolomics, which were also identified as characterizing compounds for MHR waterhemp populations (Figs 4, 5; Table 2). M713 includes the fragment ion m/z 306, which corresponds to a deprotonated GSH molecule (Fig. 6e; Table 2). Further inspection of the MS2 fragment ions from the M713 metabolite elucidated the anion at m/z 272 that corresponds to the deprotonated γ‐glutamyl‐dehydroalanyl‐glycine moiety of GSH. The other product ions include m/z 128, 143, 160, 179, 210 and 254 (Fig. 6e), all of which are known fragment ions of GSH (Dieckhaus et al., 2005), suggesting that M713 is a putative SA3‐GSH conjugate (minus an oxygen). Two putative cysteine–glycine conjugates of SA3 (SA3‐Cys‐Gly) are shown as M582 (C24H24F6N2O6S; RT = 17.9 min; m/z 581 [M−H]) and M584 (C24H26F6N2O6S; RT = 17.1 min; m/z 583 [M−H]). M582 and M584 differ by the presence of a double bond and include the product ion m/z 175, which corresponds to deprotonated l‐Cys‐Gly (Table 2; Fig. 6f,g). The putative cysteine conjugate of SA3 (SA3‐Cys) is shown as M527 (C22H23F6NO5S; RT = 17.0 min; m/z 526 [M−H]) (Table 2; Fig. 6h). M527 includes the fragment ion m/z 120, which corresponds to deprotonated Cys (molar mass: 121.15 g mol−1; formula: C3H7NO2S).

Comparisons of the peak areas of the putative Phase II metabolites of SA3 are shown in Fig. 8. For the excised leaf extracts, SIR exhibited greater abundance of the four putative SA3‐GSH‐derived conjugates relative to ACR (Fig. 8a–d). The relative abundances of S‐conjugates of SA3 were positively correlated (Fig. S6). In whole‐plant extracts, although the SA3‐GSH conjugate M713 and its catabolites occurred in varying abundances among waterhemp populations at each time point (Fig. 8e–h), SIR and NEB exhibited greater levels of SA3‐GSH conjugate M713 (Fig. 8e) compared to sensitive populations ACR and SEN. A trend was also noted that levels of the four metabolites in SIR were initially higher than NEB but decreased from 12 to 24 HAT, while metabolite levels increased in NEB from 12 to 24 HAT (Fig. 8e–h). This trend indicates the initial rate of SA3‐GSH conjugate formation occurs more rapidly in SIR than in NEB, and consequently that SA3‐GSH conjugate catabolism occurs earlier during the time course in SIR compared to NEB. Putative SA3‐Cys‐Gly conjugates M582 (Fig. 8f) and M584 (Fig. 8g) were also more abundant in both resistant populations compared to ACR at 24 HAT.

Fig. 8

Comparisons of the normalized peak area of four putative Phase II metabolites of syncarpic acid‐3 (SA3) in multiple‐herbicide‐resistant and 4‐hydroxyphenylpyruvate dioxygenase (HPPD)‐inhibitor‐sensitive waterhemp populations. (a–d) Mean normalized peak areas derived from the excised leaf assay at 4 h after treatment (HAT) and 12 HAT; (e–h) derived from the whole plant assay at 12 and 24 HAT. SIR, multiple‐herbicide‐resistant population from Stanford, Illinois, USA; NEB, multiple‐herbicide‐resistant population from Nebraska, USA; ACR, HPPD‐inhibitor‐sensitive but atrazine‐resistant population from Adams County, Illinois, USA; SEN, standard sensitive population.

SA3 detoxification occurs via multiple pathways

The proposed mechanism by which SA3 is metabolized in MHR waterhemp includes several enzymes (Fig. 7). Phase I metabolism occurs in two ways: via alkyl hydroxylation of a methyl group attached to the cyclohexanetrione ring, producing the putative hydroxy‐SA3 metabolite, M438; and via reduction of the C1 ketone of SA3, catalyzed by a putative reductase. We theorize the reduced SA3 molecule undergoes dehydration by an unknown mechanism (possibly dehydratase‐catalyzed), forming the metabolite M406 containing an α,β‐unsaturated carbonyl group, which in theory is a reactive electrophile that can be attacked by nucleophiles such as GSH (Farmer & Davoine, 2007).

Although an atypical metabolite of HPPD inhibitors in plants (Ma et al., 2013; Kaundun et al., 2017; Küpper et al., 2018; Lygin et al., 2018), an SA3‐GSH conjugate (M713) was detected and was characteristic for MHR waterhemp, indicating that GSH conjugation is a Phase II detoxification mechanism for SA3 in resistant waterhemp. Despite several attempts, putative S‐conjugates of SA3 were not detected in in vitro GST assays using crude protein extracts from SIR and NEB (data not shown). However, the lack of a GSH conjugate was expected in vitro since an electrophilic site in parent SA3 is not present (i.e. no leaving group) to allow direct nucleophilic displacement by the thiolate anion of GSH, nor does a site exist for a 1,4‐Michael addition to occur. Following presumed transport of the SA3‐GSH conjugate into the vacuole by ABC transporters (Phase III), enzymatic transformation of M713 into its amino acid catabolites (Phase IV) by peptidases in the vacuole (Lamoureux & Rusness, 1981; Wolf et al., 1996; Martin et al., 2007) is supported by the detection of SA3‐Cys‐Gly dipeptide and SA3‐Cys conjugates that are also associated with SA3 resistance (Figs 4, 5, 7, 8).

Discussion

Metabolic resistance to HPPD‐inhibiting herbicides is a growing concern in crop protection research (Yu & Powles, 2014; Nandula et al., 2019). Of particular concern are metabolic resistance mechanisms that confer cross resistance in weeds to multiple, distinct herbicide classes (Rigon et al., 2020), and as our study has demonstrated, even to a noncommercial herbicide. Identifying the specific enzymes and which metabolic phases they catalyze is essential for herbicide design. The commercial triketones are a subclass of HPPD‐inhibiting herbicides characterized by a cyclohexane ring with keto groups on C1 and C3 and another keto group on C7, positioned β to these C1 and C3 keto groups (Fig. 1a–c) (Secor, 1994; Mitchell et al., 2001; Beaudegnies et al., 2009). These herbicides act as competitive tight‐binding inhibitors to HPPD (Mitchell et al., 2001; Moran, 2005). Successful weed control by natural triketones is largely attributed to their chemical structure; an alternating position of the saturated side chain, high degree of steric hindrance, asymmetry, molecular planarity, enantiomer and higher lipophilicity of the alkyl chain could influence inhibition of enzyme activity (Meazza et al., 2002). In our current study using a noncommercial triketone, Phase I metabolism of SA3 occurred by two distinct mechanisms. The first occurred via putative P450‐catalyzed alkyl hydroxylation of a methyl group attached to the cyclohexanetrione ring, forming a hydroxy‐SA3 (Fig. 6b). The second putative Phase I mechanism is reduction of SA3 followed by elimination of water, forming a reduced‐SA3 and dehydrated‐SA3 metabolite, both of which were associated with resistance in MHR waterhemp. Reduction of SA3 was not an expected Phase I metabolic transformation initially hypothesized for SA3 after identifying rapid metabolism by MHR waterhemp (Fig. 2), primarily because reductases are not usually associated with herbicide detoxification in plants (Riechers et al., 2010) and a hydroxy‐SA3 metabolite was already detected (Fig. 6b). However, our untargeted metabolomics data support the formation of a reduced SA3 and dehydrated SA3 metabolite, thereby introducing a Michael acceptor site required for Phase II GSH conjugation.

One of the enzyme groups that can potentially catalyze such reduction reactions in xenobiotics are the aldo‐keto reductase (AKR) enzymes (Jin & Penning, 2007; Penning, 2015). The AKR superfamily of enzymes are classified as oxidoreductase enzymes that are widely distributed in prokaryotes and eukaryotes (Barski et al., 2008; Simpson et al., 2009), and typically catalyze NAD(P)(H)‐dependent reduction of aldehydes and ketones to primary and secondary alcohols, respectively, under both normal or stress conditions. AKRs are encoded by stress‐regulated genes with broad substrate specificity (Jin & Penning, 2007) and play a central role in cellular responses to osmotic, electrophilic and oxidative stress. Although AKRs in plants have not been widely explored compared to other Phase I metabolic enzymes, several studies on plant AKRs have been reported. However, since they are known Phase I metabolism enzymes for xenobiotics (Jin & Penning, 2007; Barski et al., 2008; Simpson et al., 2009; Penning, 2015), AKRs are candidate enzymes for SA3 reduction to M424 (Fig. 7). To date, the only case of herbicide detoxification associated with AKR activity in resistant weeds involves glyphosate (Nisarga et al., 2017; Vemanna et al., 2017; Pan et al., 2019; McElroy & Hall, 2020; Ramu et al., 2020). Protein‐docking methods (Vemanna et al., 2017) have demonstrated that AKR enzymes can bind glyphosate, and glyphosate has a carbonyl that is amenable to reduction by AKRs (Sanli et al., 2003; Penning, 2015). AKR genes from Pseudomonas (PsAKR1) and rice (Oryza sativa; OsAKR1) improved glyphosate tolerance when overexpressed in bacteria and tobacco (Nicotiana tabacum) (Nisarga et al., 2017; Vemanna et al., 2017). However, these genes were experimentally derived, and the precise contribution of AKR‐mediated glyphosate detoxification in resistant Echinochloa remains unclear (McElroy & Hall, 2020). In other plant species, AKR17A1 from the cyanobacterium Anabaena sp. PCC7120 catalyzed metabolism of the herbicide butachlor to dicarboxylic acid and phenol (Agrawal et al., 2015). Importantly, since the SIR and NEB populations examined in the current study are glyphosate‐sensitive, it is possible that an AKR enzyme different from the AKRs characterized in resistant Echinochloa (Pan et al., 2019) catalyzed the reduction of SA3 to metabolite M424 (Fig. 7).

Biotransformation of xenobiotics via GSH conjugation is a key Phase II detoxification mechanism (Coleman et al., 1997; Edwards et al., 2000; Theodoulou et al., 2003; Riechers et al., 2005; Mano et al., 2019; Gao et al., 2020). The formation of S‐conjugates can occur spontaneously between GSH and an electrophile at alkaline pH but is most often catalyzed by GSTs (Cooper & Hanigan, 2018). GSTs catalyze the reaction of the thiolate anion of GSH with an electrophile producing the corresponding glutathione S‐conjugate (Cooper & Hanigan, 2018). In our current study, GSH conjugation was evident by detection of a SA3‐GSH conjugate and its catabolites (SA3‐Cys‐Gly and SA3‐Cys conjugates) (Table 2; Fig. 6e–h). There are no reports of triketone herbicides metabolized to form S‐conjugates in plants; however, GSH conjugation readily occurs for other herbicides such as the photosystem II inhibitor, 2‐chloro‐4‐ethylamino‐6‐isopropylamino‐s‐triazine (atrazine) (Shimabukuro, 1967; Lamoureux et al., 1970; Shimabukuro et al., 1970) (Fig. S7a), and the very‐long‐chain fatty‐acid elongase inhibitor, 2‐chloro‐N‐(2‐ethyl‐6‐methylphenyl)‐N‐[(2S)‐1‐methoxypropan‐2‐yl]acetamide (S‐metolachlor) (O'Connell et al., 1988) (Fig. S7b). GSH conjugates of atrazine and S‐metolachlor are formed via direct nucleophilic displacement of chlorine via the thiolate anion of GSH (Shimabukuro et al., 1970; O'Connell et al., 1988), without the requirement for Phase I activation. This reaction can either occur enzymatically by GSTs or nonenzymatically (Edwards et al., 2000). Alternatively, we theorize that SA3 forms a GSH conjugate via Michael addition of GSH to the cyclohexane ring (Fig. 7). Unlike atrazine and S‐metolachlor, however, SA3‐GSH conjugation is likely to be enzymatic and can only occur after Phase I reductive–dehydration reactions (Fig. 7), which is supported by the lack of SA3‐GSH conjugates formed in vitro with crude protein extracts from MHR waterhemp (data not shown). The proposed GSH conjugation mechanism for SA3 (Fig. 7) is based on the possible reactivity of the intermediate (e.g. dehydrated SA3 metabolite, M406), which structurally resembles the electrophilic Michael acceptors E9 and E10, as previously reported (Mayer & Ofial, 2019) (Fig. S8). Although Phase II metabolites via GSH conjugation (and their catabolites) were detected by LC‐MS in our research, the reaction probably did not occur spontaneously, nor without SA3 first transformed into an electrophilic metabolite. Overall, these findings also imply that formation of the SA3‐GSH catabolites (SA3‐Cys‐Gly dipeptide and SA3‐Cys conjugates) are enzymatically driven by vacuolar peptidases (Lamoureux & Rusness, 1981; Wolf et al., 1996; Martin et al., 2007) (Fig. 7).

Our research demonstrated metabolism of a noncommercial herbicide, SA3, in MHR waterhemp via a novel reduction–dehydration–GSH conjugation process that is not active in corn or sensitive waterhemp. The highly evolved, complex Phases I and II metabolic machinery of MHR waterhemp is peculiar, and the lack of previous field exposure to SA3 (Ma et al., 2013) implies resistance selection by commercial HPPD‐inhibiting herbicides, other herbicides or stress‐inducing factors (Dyer, 2018). Future proteomic work will aim to identify the putative reductase, dehydratase and GST enzymes that enable SA3 detoxification by analysis of subcellular Amaranthus protein fractions. One strategy that may prove fruitful in identifying detoxification enzymes for SA3, and other compounds that exhibit undefined Phase I reduction mechanisms (Küpper et al., 2018), is to utilize combined untargeted detection of proteins and metabolites diagnostic for resistance, thus removing biases towards an existing mode of herbicide detoxification (Aebersold & Mann, 2003; Sindelar & Patti, 2020).

The present research discovered a novel detoxification mechanism involving Phase I reduction, which was enabled via a data‐driven approach acquired through traditional herbicide resistance experiments combined with untargeted metabolomics. Our findings revealed metabolites formed from SA3 and posit detoxification mechanisms and enzymes catalyzing these reactions. Future research will include proteomic and transcriptomic studies to identify proteins and genes, respectively, conferring SA3 resistance in dioecious, weedy Amaranthus. Improved biorational design of resistance‐breaking compounds (Kaundun, 2021) while heeding their environmental fate is an overarching, long‐term goal of this research. Herbicide safener technology (Riechers et al., 2010) and genetic modification of crops to tolerate the phytotoxic effects of nonselective herbicides, such as SA3, through transgenic and/or gene‐editing technology are avenues for future research (Green, 2014; Duke, 2015). However, the benefits gained from these cutting‐edge technologies should outweigh possible risks, such as evolution of novel weed resistance mechanisms.

Author contributions

DER, JCTC and SSK conceptualized and designed the research; S‐JH performed the glasshouse dose–response experiments; JCTC performed the metabolism‐based experiments; SAS and AVL assisted with metabolism experiments and data analysis; JCTC analyzed and interpreted the data, with assistance from JAM; JCTC and DER wrote the manuscript. All authors read and discussed the manuscript.

Fig. S1 Standard curve of syncarpic acid‐3 (SA3) for quantifying remaining SA3 parent in excised leaves of four waterhemp populations.

Fig. S2 Flow diagram of metabolite profiling carried out on waterhemp populations treated with the nonselective, 4‐hydroxyphenylpyruvate dioxygenase‐inhibiting herbicide syncarpic acid‐3 (SA3).

Fig. S3 First‐generation product ion (MS2) mass spectra of syncarpic acid‐3 (SA3).

Fig. S4 Orthogonal partial least squares discriminant analysis (OPLS‐DA) of compounds detected in excised leaf extracts of waterhemp populations ACR and SIR treated with 0.3 mM syncarpic acid‐3 (SA3).

Fig. S5 Possible, but not detected, structures of fragment ions derived from metabolism of the aryl ring of syncarpic acid‐3 (SA3).

Fig. S6 Correlation plot of the putative metabolites of syncarpic acid‐3 (SA3) detected in leaf extracts of waterhemp populations SEN, ACR, NEB and SIR.

Fig. S7 Chemical structures of herbicides detoxified in plants via GST‐catalyzed glutathione conjugation.

Fig. S8 Chemical structures of Michael acceptors.

Methods S1 LC‐MS, data processing and statistical analysis of metabolomics data.

Table S1 Mass spectral fragment ions and their relative abundance of a noncommercial, nonselective, HPPD‐inhibiting herbicide, syncarpic acid‐3 (SA3).

Table S2 The 10 most discriminating compounds for syncarpic acid‐3 (SA3)‐treated excised leaf samples.

Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Click here for additional data file.