The resistant Palmer amaranth population designated as MIS-D was used in this study. MIS-D field population was confirmed resistant to glyphosate, PPO- and ALS-inhibiting herbicides (Salas-Perez et al., 2017). The susceptible biotype (SS) was collected in Crawford County, Arkansas, and is susceptible to all herbicides tested in previous experiments (Salas et al., 2016). Plants from MIS-D field population that survived the recommended dose of fomesafen (264 g ai ha^-1) were used for sequencing the PPO2 gene. Four plants (two male and two female), were confirmed by gene sequencing to contain the mutation leading to alanine in 399 position (G399A), and were isolated to generate F1 populations. Two male and two female A. palmeri plants that were heterozygous and homozygous for the G399A mutation, respectively, were allowed to hybridize to produce R₃₉P and R₄₃P F1 populations, respectively. Plants were grown in the greenhouse maintained at 32/25 ± 3°C day/night temperature with supplemental light for 14 h. Seeds from SS, MIS-D (original field population), R₃₉P, and R₄₃P populations were used for bioassays in the greenhouse. Herbicide treatments were applied with a laboratory sprayer equipped with a 110° flat-fan spray nozzle (Teejet; Spraying Systems Co., Wheaton, IL, United States) delivering 187 L ha^-1 of herbicide solution at 310 kPa. The nozzle was set at 51 cm above the plant canopy. Plants were returned to the greenhouse immediately after herbicide treatment. The herbicides were applied when the plants were 5- to 7-cm tall.

Some plants from MIS-D in the bioassay experiment were assigned ID numbers and ∼25 mg of leaf tissues were collected before spraying fomesafen. After evaluation of plant response, the leaf tissues collected from survivors were labeled as R while those from plants, which did not survive the herbicide application were labeled as S. Total RNA was isolated using the Plant RNeasy Mini Kit (Qiagen), treated with DNAse (Thermo Fisher Scientific), and converted to cDNA using Access RT-PCR System (Promega). Full-length PPO2 were amplified using 1F (5′-GG GGTACCCGGGTAAACTGATCTTATGTTAATTC-3′) and 3R (5′- GGAATTCGAGCTCGCATGCTTACGCGGTCTTCTCAT CATC-3′) primer pair, purified, and sent for sequencing using internal primers. The PCR was conducted in a 40-μL volume that consisted of 2 μL (60 ng) of cDNA, 0.5 μM of each primer, and 20 μL of EmeraldAmp MAX PCR Master Mix (Takara). The PCR was run with the following profile: 98°C for 30 s; 40 cycles of 98°C for 1 min, 58°C for 1 min and 72°C for 2 min; followed by a final extension step for 10 min at 72°C. Overlapping internal primers used for sequencing are as follows: Apx2-1F (5′- GTAATTCAATCCATTAC CCAC CTT 3′), Apx2-1R (5′-TTCCATA CGTCGGGAAATGT-3′), Apx2-2F (5′-TGTTGGAACCATTTCTCTGG-3′), Apx2-2R (5′-GGGGATAAGAACTCCGA AGC-3′), Apx2-3F (5′-GATG CTGTGGTTGTCACTGC-3′), and Apx2-3R (5′-TTACGCGG TCTTCTCATCCAT-3′). The PCR product purified from agarose gel using Pure-Link Quick Gel Extraction Kit (Invitrogen) was sent to Eurofins Genomics for sequencing. The resulting overlapping fragments from all individual R and S plants were assembled into one sequence using Sequencher 5.4.6 (Gene Codes Corporation) to obtain the complete coding region of the PPO2 gene. The nucleotide sequences were translated into open reading frames using the online ExPASy translation tool.

Two structural models of A. palmeri PPO2 protein, S-PPO2 and R-PPO2, were built using the homology modeling workflow in YASARA (YASARA, version 17.4; YASARA Biosciences GmbH, Vienna, Austria) using protein data bank (PDB) entry 1SEZ of tobacco PPO2 as template. The S-PPO2 model used a modified version of the resistant PPO sequence in which position 399 was changed back to the wild-type glycine (G399 model). All other sequence positions were as found in the resistant sequence. The R-PPO2 model used the unmodified resistant PPO sequence (A399 model). Default settings and standard protein preparation steps were used. Fomesafen was initially modeled into the binding-site of the wild type S-PPO2 using the docking scripts and tools of YASARA. To gain insight into the effect of the G399A on fomesafen binding, the predicted binding model was then transferred into the R-PPO2 model, where G399 was changed to A399. The changes in protein-ligand interactions due to the G399A mutation were then examined and interpreted.

Based on the PPO2 sequence information obtained from the S and R plants, we developed a CAPS marker for detecting the mutation at 399 amino acid position. We found the mutation from wild type codon GGT to GCT at 399 position. RsaI restriction endonuclease recognizes the sequence GTˆAC, so a G to C mutation eliminates the restriction site. Genomic DNA was extracted from all R plants along with a few S plants and a 691-bp fragment was amplified using the primers, RsaIF (5′-CGTACCCCTTTCCGTTATGA-3′) and RsaIR (5′-CCGGGAAGATCTT TTTCCAT-3′). The amplicon was then subjected to RsaI digestion. Upon digestion, the wild-type amplicon would generate a 505-bp and 185-bp band, whereas the homozygous mutant amplicon would generate the intact 691-bp band (Supplementary Figure S3). A heterozygous plant for the G399A mutation would produce three bands; 691-, 505-, and 185 bp (Supplementary Figure S3). PCR conditions were similar to those described earlier with 30 cycles. Restriction digestions were carried out according to the manufacturer’s recommendations (New England Biolabs), and digestion patterns were viewed on 3% agarose gels with GelRed (electrophoresis at 90–100 V for 70–80 min). Thirty-five fomesafen-resistant Palmer amaranth accessions from a previous study (Salas-Perez et al., 2017) were genotyped to detect the presence of G399A mutation by CAPS assay.

Seeds from Palmer amaranth populations, MIS-D, R₃₉P, R₄₃P, and SS were sown in 15-cm pots filled with commercial potting soil (Sunshine Premix No. 1; Sun Gro Horticulture, Bellevue, WA, United States). Four days after planting, seedlings were thinned to five per pot. When seedlings were 5-7-cm tall, fomesafen (Flexstar, Syngenta, Greensboro, NC, United States) was applied at 16.5 (0.062X), 33 (0.125X), 66 (0.25X), 132 (0.5X), 197.5 (0.75X), 263 (1X), 527 (2X), 1053 (4X), and 2107 (8X) g ai ha^-1, where 1X represents the recommended dose. The SS population was sprayed with nine doses as well: 1, 2, 4, 8, 16.5, 33, 66, 132, and 263 g ai ha^-1, corresponding to a range of 1/256 to 1X of the recommended dose. Herbicide treatments included a non-ionic surfactant (Induce, Helena Chemical, Collierville, TN, United States) at 0.5% v/v. A non-treated control was included with each population. The overall effects of fomesafen (stunting, chlorosis, necrosis, and desiccation) were assessed visually on a scale of 0 (no control) to 100% (complete kill) relative to the non-treated control at 21 days after treatment (DAT). The shoot biomass was harvested, dried at 60°C for 72 h, and weighed. Data were analyzed by fitting a non-linear, three-parameter log-logistic regression model using the “drc” package in R 3.5.1 (Ritz et al., 2015). The model used is defined by: Y = d/[1+exp{b[log(x) – log(ED50)]}], where Y is the response (biomass or herbicide injury), d is the asymptote at upper limit, x is the fomesafen dose, and b is the slope around “ED50,” which is the value of x giving a 50% response of Y. The dose needed to cause 50% herbicide injury (ED₅₀) was calculated. The resistance factor was determined by comparing ED₅₀ values between the R and SS accessions.

The MIS-D population is cross-resistant to the 1X dose of various foliar-applied PPO herbicides (Salas-Perez et al., 2017). In this follow-up research, the cross-resistance profile of F1 population, R₃₉P, was investigated using twice the recommended dose of each PPO herbicide. Seeds from R₃₉P and SS populations were planted in cellular trays in the greenhouse. Herbicides representing each PPO herbicide family were applied to 5-cm tall seedlings. Twenty-five seedlings from each population were treated with acifluorfen (1120 g ai ha^-1), carfentrazone (560 g ai ha^-1), flumioxazin (141 g ai ha^-1), fomesafen (528 g ai ha^-1), lactofen (448 g ai ha^-1), and saflufenacil (49 g ai ha^-1). The herbicides were applied with the respective recommended adjuvants. Carfentrazone and flumioxazin treatments included 0.25% non-ionic surfactant (NIS) (v/v), whereas acifluorfen included 0.125% NIS. Lactofen was applied with 0.5% crop oil concentrate (v/v), whereas saflufenacil was sprayed with 1% methylated seed oil, and 1% ammonium sulfate (w/v). Herbicides were applied in a laboratory spray chamber fitted with 800067 flat fan nozzles calibrated to deliver 187 L ha^-1 at 269 kPa. At 21 DAT, injury, and mortality were evaluated visually.

Effects of G399A substitution were studied using A. tuberculatus backbone. The glycine at 399 position in A. palmeri PPO2 corresponds to G398 in A. tuberculatus PPO2, but the said mutation is referred as G399A in this experiment (although A. tuberculatus PPO2 backbone was used) to avoid confusion. Recombinant expressions of five other possible G399 substitutions (G399D, G399V, G399C, G399R, and G399S) based on a single nucleotide base change at position one or two of the G399 codon were also investigated. In addition, in vitro inhibition of the wild type and previously known PPO2 mutations conferring PPO resistance (ΔG210, R128L) were studied. The PPO2 gene from A. tuberculatus was synthesized de novo and subcloned into pRSetB plasmid (Invitrogen, CA, United States) using BamHI and HindIII restriction sites. The construct contained an N-terminal hexa-histidine tag to facilitate protein purification. The plasmid construct was transformed into E. coli strain BL21(DE3)pLysS (Novagen, EMD Millipore Corp., MA, United States) and was selected in LB agar plates supplemented with 100 μg mL^-1 ampicillin and 34 μg mL^-1 chloramphenicol. For pre-cultures, 3 mL of LB medium containing ampicillin and chloramphenicol was inoculated using a single colony picked directly from the agar plates, and then cultivated at 37°C with shaking at 200 rpm for 6 h (pre-culture). From this culture, 20 μL was inoculated into a 100-mL flask containing 20 mL LB medium and the same antibiotics. The culture was incubated overnight in a shaker (200 rpm, 37°C). An aliquot (100 μL) of the overnight culture was transferred into a 250-mL flask containing 100 mL ZYM-5052 autoinduction medium (preparation procedure is described in Supplemental Data) with ampicillin (100 μg mL^-1) and chloramphenicol (34 μg mL^-1). The main culture was incubated at 37°C for 5 h with shaking at 200 rpm, then for an additional 21 h at 25°C with shaking.

At the end of cultivation, the cells were harvested by centrifugation at 6000 x g for 30 min at 4°C. For every gram of pellet, 10 mL of PPO lysis buffer (50 mM NaH₂PO₄, 100 mM NaCl, 5 mM imidazole, 5% glycerol (v/v), pH 7.5, 20 mg mL^-1 lysozyme, 30 unit mL^-1 DNase I) containing protease inhibitors (complete EDTA free, Roche Diagnostics, Mannheim, Germany) was added. The suspension was sonicated (3 min, 90% amplitude, with 30 s rest intervals) in an ice-bath to avoid overheating the samples. The cell debris was removed by centrifugation at 38,000 × g for 30 min at 4°C and the supernatant was added with 2 mL of 5 M NaCl. To purify the enzyme, a 500-μL-bed volume of HisPur Ni-NTA resin (Thermo Fisher Scientific, Illinois, United States) was placed in a column and equilibrated with a buffer (20 mM NaH₂PO₄, 50 mM NaCl, 5 mM imidazole, 5 mM MgCl₂, 17% glycerol (v/v), 0.1 mM EDTA, pH 8.0). The supernatant, which contained the soluble recombinant protein, was passed through the resin. Non-specifically bound proteins were eluted using 5.6-mL wash buffer (20 mM NaH₂PO₄, 50 mM NaCl, 5 mM imidazole, 17% glycerol (v/v), adjusted to pH 7.5) and the target protein was eluted in 1 mL of elution buffer [20 mM NaH₂PO₄, 50 mM NaCl, 250 mM imidazole, 17% glycerol (v/v), pH 7.5]. The purified enzyme was quantified using Scandrop nano-volume spectrophotometer (Analytikjena, Life Science, Germany). The protein and the cell debris (2.5 μg) was electrophoresed using 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis to check protein quality and solubility.

Protoporphyrinogen IX oxidase was assayed using the fluorescence assay with excitation and emission wavelength of 405- and 630 nm, respectively. The assay was conducted in a 187-μL reaction mixture containing100 mM Tris–HCl, 1 mM EDTA, 5 mM DTT, 0.0085% Tween 80, and 15 μL of enzyme in resuspension buffer (0.05 M Tris–HCl, pH 7.3, 3.2 mM EDTA, 20% v/v glycerol). The concentration of enzyme differed for each mutant depending on the activity of the enzyme in the presence of saturating concentration of substrate and absence of PPO inhibitor. Maximum fluorometric readings were recorded for the amount of each enzyme. These values were used to calculate the enzyme activity (fluorescence units/minute) compared to the WT. Recombinant protein that showed enzyme activity in the absence of PPO inhibitor was subjected to in vitro PPO inhibitor dose-response assay. PPO inhibitors, purchased as Pestanal standards (Sigma-Aldrich, Steinheim, Germany) were dissolved in 80% DMSO. Inhibitors evaluated included diphenylether, pyrimidinedione, triazolinone, N-phenylphthalimide, phenylpyrazole, and thiadiazole herbicides. Ten PPO inhibitor concentrations (10 μL) ranging from 0.00005 M to 5.12 × 10^-12 M were added to the reaction mixture. The samples were incubated for 30 min at room temperature. Protoporphyrinogen IX (3.24 μM) was then added and the fluorescence was recorded for 30.25 min (33 cycles of 55 s cycle^-1) using a microplate reader (CLARIOstar, BMG LabTech, Germany). The substrate protoporphyrinogen IX was prepared from protoporphyrin using sodium amalgam reduction (preparation procedure described in Supplemental Data). The slope (per min) was recorded and the percent inhibition was calculated relative to the slope of the positive and negative control. The assay was conducted with two replicates. Inhibitor concentrations (M) required to reduce PPO2 activity by 50% (IC₅₀ values) were estimated using non-linear regression procedures (three-parameter or four-parameter log-logistic function). For each inhibitor, resistance factors were calculated by dividing the IC₅₀ values of mutant with the sensitive PPO2.

A field population, MIS-D, has high resistance to fomesafen (Salas-Perez et al., 2017). In the same study, we found that the majority of survivors did not carry the G210 deletion mutation (ΔG210) that confers resistance to PPO inhibitors (Patzoldt et al., 2006). To determine if another target site mutation is the basis for high-level resistance in this population, the PPO2 of five resistant (R) plants and three fomesafen-sensitive (S) plants was sequenced using cDNA. The full-length cDNA of 1608 bp codes for the dual targeting signal peptide, the membrane-binding domain, the FAD-binding domain, and the substrate-binding domain of mature mitochondrial PPO2 protein (based on published sequence from A. hypochondriacus, EU024569, and A. palmeri, MF583744). The sequences we obtained were 1,608 bp; therefore, we generated full-length cDNA sequences of PPO2 from R and S plants (GenBank accession numbers: MK408971, MK408972, MK408973, MK408974, MK408975, MK408976, MK408977, and MK408978). Comparison of the cDNA-inferred amino acid sequences of the PPO2 protein revealed three amino acid polymorphisms in R plants; however, only a single change in amino acid, glycine to alanine at position 399 (G399A) (numbered relative to the full-length sequence of A. palmeri, ATE88443), consistently differed between all R and S plants (Supplementary Figure S1). The glycine to alanine substitution resulted from a codon change of GGT to GCT at position 399. Except for G399A, other previously reported resistance-conferring mutations (ΔG210, R128G, or R128M) were not found in R plants.

Using the grain amaranth (A. hypochondriacus, EU024569) genomic sequence of PPO2 as reference, we found that the genetic code for G399 resides on exon 15 (Figure 1A). The corresponding change in amino acid (G399A) in R plants is located in the catalytic domain of PPO2 protein (Figure 1B) based on the resolved protein structure of Nicotiana tabacum (tobacco) PPO2 (Koch et al., 2004). The G399A mutation is not among the previously reported resistance-conferring mutations from PPO-resistant weeds (Patzoldt et al., 2006; Salas et al., 2016; Giacomini et al., 2017; Salas-Perez et al., 2017). An examination of various eukaryotic PPO proteins revealed that the glycine at position 399 is highly conserved among wild type PPO2 and PPO1 (Figure 1C and Supplementary Figure S2). We found only an alanine mutation at position 399 among the R plants sequenced. As PPO1 is also a binding target for PPO-inhibiting herbicides, we sequenced the PPO1 gene in the same R and S plants and did not find any polymorphisms among R plants from MIS-D population.

The high resolution X-ray crystal structure of N. tabacum PPO2 generated by Koch et al. (2004) provides three-dimensional protein structural information that is useful for understanding the role of critical amino acid residues in herbicide binding to PPO2 in plants. Here, as a first step toward rationalizing the impact of the G399A mutation on fomesafen binding, this protein structure (PDB entry 1SEZ) was used as a template to obtain structural models of the A. palmeri PPO2-fomesafen complex. In particular, the homology modeling workflow implemented in YASARA was used to derive models for both S-PPO2 and R-PPO2. The binding geometry of fomesafen in the binding-site of the S-PPO2 model was then predicted using in-house docking tools and transferred into the R-PPO2 model. The resulting visualization (Figure 2) suggests the following hypothesis for the molecular mechanism of G399A-induced fomesafen resistance in A. palmeri PPO2. The principal difference observed between the S-PPO2 and R-PPO2 models is the additional methyl (-CH₃) group of A399 in R-PPO2 (highlighted pink). Although it comprises just a single heavy atom, this methyl group is oriented directly into a buried part of the PPO binding site, making the binding pocket smaller. This creates close steric contacts (indicated by orange lines in Figure 2), significantly below the sum of van-der-Waals radii, with the chloro-substituent of the terminal 2-chloro-4-trifluoromethylphenyl ring and the 2-position of the central phenyl ring. These repulsive interactions could reduce the magnitude of the protein-ligand binding energy and therefore weaken the inhibition of the PPO2 enzyme by fomesafen.

A PCR-based assay was designed to determine the presence of G399A mutation in 35 fomesafen-resistant field populations that we characterized in a previous study (Salas-Perez et al., 2017). We surveyed the presence of the G399A mutation among 130 R plants representing these 35 populations using cleaved amplified polymorphic sequences (CAPS) assay (Supplementary Figure S3). Collectively, about 16% (21 of 130) of R plants carried the G399A mutation. Further, only 14% of field populations (5 of 35) had R plants with this mutation and only two resistant populations had R plants predominantly carrying the G399 mutation. Our previous study on these same populations revealed that 26 of 35 populations harbor the ΔG210 mutation, and that the ΔG210 was present in 49% of all R plants tested (Salas-Perez et al., 2017).

To determine the level of resistance to fomesafen in F1 population, we conducted the dose response assay with R₃₉P and R₄₃P F1 populations together with the original field population, MIS-D and SS. None of the resistant populations were killed 100% with 2X dose (528 g ai ha^-1) of fomesafen. The SS population was controlled 100% with 8 g ai ha^-1 (1/32X) (Figure 3). The amplitude of injury ratings among survivors of the 1X dose was high, ranging from total absence of symptoms to severe leaf necrosis and stunting. The estimated fomesafen dose that would cause 50% injury (ED₅₀) of MIS-D, R₄₃P, and R₃₉P ranged from 123 to 187 g ai ha^-1, whereas the ED₅₀ for SS was 12 g ai ha^-1. Thus, the levels of resistance to fomesafen in MIS-D, R₃₉P, and R₄₃P were 11-, 12-, and 16-fold, respectively, relative to the SS population (Figure 4 and Table 1). The G399A mutation confers dominant resistance to fomesafen as plants from R₄₃P and R₃₉P (derived from heterozygous male parent) manifests higher resistance, when compared to S plants (Figure 4). Also, the recommended dose of fomesafen (1X) did not control plants containing the G399 mutant allele either in homozygous or heterozygous state in screening survivors from MIS-D population (data not presented).

We performed a comprehensive In vitro enzyme activity assay of the wild type and recombinant PPO2 proteins with known resistance-conferring mutations ΔG210, R128L, and the novel mutation G339A using A. tuberculatus PPO2 gene. Constructs were made to generate recombinant protein with ΔG210, R128L, and G339A mutations; other possible exchanges at G399 position; and wild type PPO2. Four G399 protein variants (G399D, G399V, G399C, and G399R) were soluble, but lacked enzyme activity; no enzyme activity was detected even after adding large amounts of enzyme to the assay mix. G399S had negligible activity, necessitating the addition of a high amount of enzyme to the assay mix before any measurable activity was detected. Among these recombinant variants, only G399A was active. Compared to ΔG210 and R128L, G399A had a very low activity (Supplementary Table S1). Resistance factors (IC₅₀ values) were calculated after subjecting the recombinant protein to increasing concentrations of diphenylether, pyrimidinedione, triazolinone, N-phenyl-phthalimide, phenylpyrazole, and thiadiazole herbicides (Table 2). The G399A substitution had similar target resistance potency as ΔG210. The G399A and ΔG210 mutants exhibited higher IC₅₀ than the wild type supporting that G399A mutation confers resistance to PPO-inhibiting herbicides. The R128L mutation seemed to endow resistance to diphenylethers, N-phenyl-phthalimides, pyrimidionedione, and triazolinones, but not to the other PPO herbicides tested. Based on the IC₅₀ values, the G399A substitution afforded greater resistance level to most diphenylethers relative to ΔG210 or R128L.

The field population, MIS-D, was resistant to the 1X dose of carfentrazone, flumioxazin, fomesafen, lactofen, fluthiacet-methyl, and pyraflufen-ethyl. The 1X dose of acifluorfen and saflufenacil caused >90% mortality in MIS-D while the other PPO-inhibiting herbicides caused 31–80% mortality (Salas-Perez et al., 2017). The number of plants resistant to the 2X dose of acifluorfen, carfentrazone, flumioxazin, lactofen, and saflufenacil in R₃₉P was greater than the number of survivors at the 1X dose in the field population, MIS-D (Table 3), reflecting increased frequency (and level) of resistant plants in the F1 population. The frequency of resistant plants in R₃₉P at the 2X dose was ≥40% with carfentrazone, flumioxazin, and lactofen. Although the survival rate was <20% with acifluorfen, fomesafen, and saflufenacil, the 2X dose of these herbicides did not kill all plants. This indicates that increasing the herbicide dose to 2X does not achieve complete control of this population. Besides, farmers cannot use 2X dose in the field because it is off-label and would increase the residual activity of the herbicide, which could be detrimental to the following (sensitive) crop in a crop rotation system. Cross-resistance to various families of PPO-inhibiting herbicides suggests that G399A mutation endows broad resistance to PPO-inhibiting herbicides.