Smooth pigweed (Amaranthus hybridus) and unresolved Amaranthus spp. from Brazil resistant to glyphosate exhibit the EPSPS TAP-IVS substitution

Abstract

The presence of glyphosate-resistant smooth pigweed (Amaranthus hybridus L.) biotypes has increased in southern Brazil in recent years, presenting the triple amino acid substitution TAP-IVS in 5-enol-pyruvylshikimate-3-phosphate synthase (EPSPS) previously found in Argentina. Some of these biotypes have morphological characteristics of A. hybridus and redroot amaranth (Amaranthus retroflexus L.). The present study aimed to identify, through molecular markers, the herbicide-resistant species of Amaranthus from Brazil that have the TAP-IVS substitution and to analyze the occurrence of pollen-mediated gene flow (PMGF) as the source of the TAP-IVS substitution in these biotypes. Six biotypes were evaluated using internal transcribed spacer (ITS) sequences, of which two (AMACHY-S and CAMAQ-R) were molecularly classified as A. hybridus and four (AMACRET-S, AMACVI-S, ARRGR-R, and SAOJER-R) were unclassified. Interestingly, all the glyphosate-resistant biotypes (ARRGR-R, SAOJER-R, and CAMAQ-R) had the TAP-IVS substitution and an increase in EPSPS relative copy number; however, only CAMAQ-R was confirmed as A. hybridus. Although the biotypes ARRGR-R and SAOJER-R are closely related to A. hybridus and green pigweed (Amaranthus powellii S. Watson), their species identity could not be resolved. The biotype SAOJER-R also was resistant to acetolactate synthase (ALS)-inhibiting herbicides due to a tryptophan to leucine substitution at position 574 in ALS. The evaluation of 119,746 seedlings in an intraspecific hybridization study of A. hybridus indicated an outcrossing frequency of 0.09%. In contrast, an absence of interspecific hybridization (A. hybridus × unclassified biotype, AMACVI-S) was found after screening 111,429 offspring. Unclassified biotypes might be derived from one or more ancient hybridization events and subsequently evolved the triple mutation independently. Alternatively, such biotypes could have evolved from recent hybridization events, which occur at a frequency below the level of detection in our study.

Introduction

The genus Amaranthus (Amaranthaceae) encompasses 75 species distributed in tropical and temperate regions (Costea et al. 2001a). Around 65 species are monoecious (Bayón 2015), and 10 are dioecious (Sauer 1955). All species are diploid with 32 or 34 chromosomes, except spleen amaranth (Amaranthus dubius Mart. ex Thell.), which is tetraploid (Grant 1959). Within Amaranthus, 20 species are weeds widely spread worldwide (Costea et al. 2001b). Weedy species of this genus are often dominant weeds and challenging to control due to their high growth rate, fecundity, stress tolerance, genetic variability, and ability to evolve herbicide resistance (Faccini and Vitta 2005).

Smooth pigweed (Amaranthus hybridus L.) is native to South America and is currently recognized as a major weed in agricultural production systems (Larran et al. 2018). It can reduce soybean grain yield by 82%, and the presence of one A. hybridus plant per square meter can result in a 6.4% yield reduction (Soltani et al. 2017; Zandoná et al. 2022). The importance of this species is increasing, including in Brazil, as resistance cases are reported (García et al. 2020; Perotti et al. 2019; Resende et al. 2020; Whaley et al. 2006).

Currently, A. hybridus populations have been identified with resistance to inhibitors of photosystem II, acetolactate synthase (ALS), protoporphyrinogen oxidase, 5-enol-pyruvylshikimate-3-phosphate synthase (EPSPS), and to the synthetic auxin herbicides (Heap 2023). Amaranthus hybridus biotypes from Brazil were identified with resistance to the EPSPS inhibitor, glyphosate, and to ALS-inhibiting herbicides (Mendes et al. 2022; Resende et al. 2020). However, the mechanism of resistance to ALS-inhibiting herbicides in biotypes from Brazil has not been characterized. Interestingly, glyphosate-resistant biotypes from southern Brazil have the same triple amino acid substitution Thr-102-IIe, Val-103-Ala, and Pro-106-Ser (TAP-IVS) (Mathioni et al. 2022) that was reported for the first time in Argentina (Perotti et al. 2019). However, limited information is available about the dispersion of the EPSPS triple mutation in Brazil populations.

The rapid distribution of A. hybridus resistant biotypes with the EPSPS TAP-IVS substitution in southern Brazil has been associated with high seed production, distribution through agricultural machinery, and seed crop contaminants (Penckowski et al. 2020). It is worth noting that farmers and field technicians have observed significant morphological variation in glyphosate-resistant Amaranthus biotypes, including those suspected to be redroot pigweed (Amaranthus retroflexus L.), making it difficult to accurately identify and classify certain species.

Glyphosate resistance caused by the triple TAP-IVS substitution is thought to be a rare event, which may indicate that its large spread is also caused by pollen-mediated gene flow (PMGF). Morphological admixtures that have been observed among Amaranthus species in South America support interspecies PMGF. Previous studies have confirmed interspecies gene flow among Amaranthus species, including between monoecious and dioecious species (e.g., Gaines et al. 2012; Tranel et al. 2002; Trucco et al. 2005).

Consequently, the goal of this research was to address the role of PMGF in the dissemination of the EPSPS TAP-IVS triple substitution in Amaranthus populations. Our specific objectives were (1) to classify the species of Amaranthus from Brazil with resistance to herbicides; (2) identify the occurrence of resistance to ALS-inhibiting herbicides and glyphosate and identify the target site–resistance mechanisms; and (3) determine A. hybridus outcrossing rates in both intra- and interspecific crosses.

Material and Methods

Plant Materials

Mature seeds of resistant Amaranthus spp. populations (ARRGR-R, SAOJER-R, and CAMAQ-R) were harvested from three regions in the state of Rio Grande do Sul, Brazil, near Arroio Grande (29.32°S, 52.5°W), São Jerônimo (29.57° S, 51.43° W), and Camaquã (30.51°S, 51.48°W). The populations were collected in glyphosate-resistant (GR) soybean fields during the 2019 to 2020 crop season. These areas had been treated with glyphosate more than 5 yr, approximately. In addition, seeds of four putative susceptible biotypes (AMACHY-S, AMACVI-S, EEA-S, and AMARET-S) were collected nearby. For all populations, seeds were obtained from 20 plants and pooled.

The progeny from each population was screened in a greenhouse at a temperature of 25 ± 3 C and a photoperiod of 14/10 h (day/night). The seeds were germinated in plastic trays with substrate based on expanded vermiculite, peat, perlite, and rice husk. Seedlings were transplanted individually into 200-ml pots and grown until the 4-leaf stage, when the herbicides were sprayed. Up to 10 plants of each biotype were used for the screening. The glyphosate doses were 0 (control) and 1,080 g ha⁻¹ (Roundup Original^® DI, Monsanto of Brazil Lt da, São José dos Campos, SP, Brazil). Herbicides were applied using a greenhouse spray chamber (Generation III, DeVries Manufacturing, Hollandale, MN, USA). The spray volume used was 200 L ha⁻¹, applied with a TJ8002E nozzle at a pressure of 280 kPa and a speed of 1.2 m s⁻¹ (TeeJet Technologies, Wheaton, IL, USA). Plants that survived 1,080 g ha⁻¹ of glyphosate were selected 21 d after herbicide application and self-pollinated. Progeny were subjected to a second generation of glyphosate selection and self-pollination to obtain more homogenous populations.

Morphological and Molecular Identification

The morphological traits of the biotypes were analyzed following two keys for the Amaranthaceae. The key described by Vasconcellos (1986) evaluated the male flower (three to five stamens), fruit dehiscence (dehiscence or indehiscent), presence of intra-axillary spines, and size of bracts relative to ovary. The synopsis of the subgenus Amaranthus in Venezuela described by Pinto and Velásquez (2010) was used to evaluate stems (glabrous or pubescent) and bract form (spinescent or lanceolate) and size (bracts larger or smaller than sepals).

Two different approaches were performed for the molecular identification of Amaranthus species. The first is based on the fragment sizes of intron 1 of EPSPS (Wright et al. 2016). For amplification of intron 1, three primer pairs—AW473_F × AW483_R, AW471_F × AW482_R, and AW477_F × AW493_R—were selected for A. hybridus, A. retroflexus, and slender amaranth (Amaranthus viridis L.), respectively (Supplementary Table 1). Supplementary Table 1 describes the PCR conditions and respective fragment sizes (bp). PCR products were analyzed for the presence or absence of a fragment and their size in a 1.5% agarose gel stained with ethidium bromide.

The second molecular approach was the amplification of a 648-bp fragment of an internal transcribed spacer (ITS), which constitutes a conserved region located between two subunits of ribosomal RNA (rRNA) that differentiate nine species of Amaranthus (Murphy and Tranel 2018). The amplification of the ITS5-ITS4 region was obtained with specific primers from the 18S, 5.8S, and 28S conserved regions of rRNA using primers ITS5_F and ITS4_R (Wetzel et al. 1999). The primers, PCR conditions, and size of PCR products (bp) are listed in Supplementary Table 1. The region amplified was purified (ExoSAP-IT^TM, Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA), and the purified product was used for Sanger sequencing. A multiple alignment was performed with ITS sequences from biotypes described in the “Plant Materials” section plus all 72 sequences of the ITS region described in Murphy and Tranel’s (2018) study that were obtained from the National Center for Biotechnology Information (NCBI, https://www.ncbi.nlm.nih.gov). Subsequently, a neighbor-joining tree was constructed from the alignment. The multiple alignment was generated using MAFFT, and a phylogenetic tree was created using Interactive Tree of Life (iTOLv6), following the protocol described by Murphy and Tranel (2018).

Dose–Response Assay

Herbicide susceptibility and resistance were quantified for each population using the progeny from the second generation of glyphosate selection described earlier. The resistance levels to glyphosate, chlorimuron, and imazethapyr were evaluated through a dose–response curve experiment with a completely randomized design. For glyphosate, factor A consisted of the susceptible AMACHY-S and resistant ARRGR-R, SAOJER-R, and CAMAQ-R biotypes. Factor B was the doses of glyphosate (Roundup Original^® MAIS, 480 g ae L⁻¹, Monsanto of Brazil) with 0, 8.4, 16.8, 33.7, 67.5, 135, 270, and 540 g ha⁻¹ for the susceptible biotype and 0, 540, 1,080, 2,160, 4,320, 8,640, 17,280, and 34,560 g ha⁻¹ for the resistant biotypes. For imazethapyr, factor A consisted of the susceptible AMACHY-S and SAOJER-R resistant biotype, and factor B was imazethapyr (Imazetapir Nortox, 212 g ae L⁻¹, Nortox S.A., Arapongas, PR, Brazil) with doses of 0, 3.3, 6.6, 13.2, 26.5, 53, 106, and 212 g ha⁻¹ for the susceptible biotype and 0, 106, 848, 1,696, 3,392, 6,784, 13,568, and 27,136 g ha⁻¹ for the resistant biotype. In the chlorimuron dose–response assay, factor A was the susceptible AMACHY-S and SAOJER-R resistant biotype, and factor B was the doses of chlorimuron (Clorim, 250 g ai kg⁻¹, UPL of Brazil, Ituverava, SP, Brazil) with 0, 0.3, 0.6, 1.2, 2.5, 5, 10, and 20 g ha⁻¹ for the susceptible biotype and 0, 20, 80, 160, 320, 640, 1,280, and 2,560 g ha⁻¹ for the resistant biotype. Surfactant at 0.5% v/v (Dash^®, BASF S.A. Guaratinguetá, SP, Brazil) and oil mineral at 1.0% v/v (Assist^®EC, BASF S.A.) were applied with imazethapyr and chlorimuron, respectively. Resistant and susceptible populations were treated as previously described. Each treatment had four replicates with one plant per pot, and the experiment was carried out twice.

The evaluations of herbicide control and fresh and dry shoot weights were carried out 21 d after the application of treatments (DAT). Herbicide control was visually evaluated using a percentage scale, in which 0% corresponds to the absence of plant injury and 100% to complete necrosis and plant death.

Mutations in the EPSPS and ALS Coding Sequences

Young leaf tissues from resistant and susceptible biotypes were sampled and frozen in liquid nitrogen. DNA was extracted using a modified cetyltrimethylammonium bromide (CTAB) protocol (Doyle and Doyle 1987). The DNA was diluted to a final concentration of 50 ng μL⁻¹. Forward gly.a.3F and reverse gly.a.4R primers were used to amplify a fragment of approximately 195 bp of EPSPS (Lorentz et al. 2014), encompassing the EPSPS triple amino acid substitution. The PCR conditions are described in Supplementary Table 1.

Four primer pairs were designed based on the A. hybridus ALS sequence (GenBank MH036304.1) using Primer3plus (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi) (Supplementary Table 1). These primers amplify fragments containing all the known plant ALS mutations associated with herbicide resistance. The PCR conditions are described in Supplementary Table 1. PCR products were analyzed by agarose gel electrophoresis, and purified with ExoSAP-IT^TM. Sanger sequencing was used to detect target-site mutations for EPSPS and ALS. The obtained sequences were analyzed using A. hybridus (GenBank MG595171.1) as a reference for EPSPS and Arabidopsis thaliana (L.) Heynh. (GenBank X51514.1) as a reference for ALS.

Relative Genomic EPSPS Gene Copy Number

Quantitative real-time PCR (qPCR) was used to estimate the relative genomic EPSPS copy number. The ACTIN gene was used as a reference (Lorentz et al. 2014), and the primer sets of forward gly.a.3F and reverse gly.a.4R were used to amplify a fragment of approximately 195 bp of the EPSPS gene (Lorentz et al. 2014). The primers and the qPCR conditions are described in Supplementary Table 1. Amplifications were carried in a QuantStudio 3 Real-Time PCR system (Applied Biosystems, Thermo Fisher Scientific). Melting curves were performed at the end of the cycling to control primer amplification specificity. Amplification efficiency was independently calculated in each reaction and was considered in data processing, according to Perotti et al. (2019). Twelve plants from each resistant (ARRGR-R, SAOJER-R, and CAMAQ-R) and susceptible biotype (AMACHY-S and AMACVI-S) were processed in triplicate. Data were analyzed by the 2^−(ΔΔCt) method (Livak and Schmittgen 2001), with the susceptible biotype (AMACHY-S) as the standard control, to obtain relative copy number.

Pollen-mediated Gene Flow (PMGF) Study: Wind-mediated Gene Flow

The experiments were conducted during the 2020 to 2021 and 2021 to 2022 cropping seasons at the Federal University of Rio Grande do Sul, Porto Alegre, RS, Brazil. The experimental design followed the PMGF wind-mediated gene flow described by Murray et al. (2002) and Walsh et al. (2012), which consisted of 12 receptor plants (susceptible plants, -S) surrounded by one donor plant (resistant plant, -R), without the presence of a crop species and physical barrier. Two experiments were performed using the CAMAQ-R population as a donor plant. The receptor plant was from the AMACHY-S population for Experiment 1 and from the AMACVI-S population for Experiment 2. The distance between the resistant plant (donor) and the susceptible plants was 0.3 to 0.6 m to avoid physical contact between plants. Each set of 13 plants formed a repetition, and in total, eight repetitions with 4.5-m distance between them were assembled. Experiments 1 and 2 were separated by 145 m (Supplementary Figure 1), isolated by native vegetation of 3 m height without the presence of Amaranthus species. Each experiment was repeated (2020 to 2021 and 2021 to 2022).

To ensure overlapping flowering periods, seeds of the donor-resistant biotype (CAMAQ-R) and the receptor-susceptible biotypes (AMACHY-S and AMACVI-S) were germinated at six periods with intervals of 7 d. Resistant biotypes were screened for glyphosate at the 3- to 4-leaf stage at a dose of 1,080 g ha⁻¹ (Roundup Original^® MAIS). In both years, plants were kept in a greenhouse until the beginning of the development of the floral structure. After that, plants were transplanted into 11-L pots and allocated as described above. Plants were irrigated daily. Seed collection (progeny) for the first year began in December 2020 and ended in early February 2021. For the second year, the collections were initiated at the end of March 2022 and finalized in May 2022. Seed harvesting was performed weekly, in bulk, within each repetition.

Screening for Resistant Hybrids

The glyphosate-resistance trait was used as a phenotypic and genotypic marker to screen plants from the receptor plants. The screening was carried out with a seed-soaking bioassay adapted from Zelaya and Owen (2005). Seeds were incubated in 24-well cell culture cluster plates (CoStar 3524, Corning Incorporated, Porto Alegre, RS, Brazil) with a germination paper of 16 mm in diameter. Twenty seeds were soaked with 500 µl of distilled deionized water (untreated control) or with the herbicide treatments. Each plate was wrapped with plastic film to avoid evaporation. Plates were placed in a germinator chamber at 30 C and 14-h diurnal conditions and 20 C and 10-h nocturnal conditions.

The experiment was arranged in a factorial completely randomized design. The progenies from the crosses for the intraspecific hybridization of A. hybridus (CAMAQ-R × AMACHY-S) and for interspecific hybridization of A. hybridus × unclassified biotype (CAMAQ-R × AMACVI-S) comprised factor A, along with the negative (seeds of susceptible biotypes) and positive controls (seeds of resistant biotype) in herbicide solution. Factor B comprised the control solution (dose 0) and glyphosate solution (dose 3.2 mM). The dose of 3.2 mM was selected by a previous study performed with a dose–response assay according to Zelaya and Owen (2005). Plant survival was evaluated at 14 d of incubation based on the positive and negative control observations. Seed germination, seedling hypocotyl, and radicle length were recorded at 2 wk after treatment. Germination was defined by radicle protrusion from the seed. Gene flow frequency calculations were performed using the number of glyphosate-surviving seedlings in relation to the total number of viable seeds, based on the no-herbicide control. The seedlings surviving the herbicide solution were selected and stored in 1.5-ml Eppendorf microtubes. These materials were stored in a freezer (−20 C) until genomic DNA extraction. DNA extraction and amplification of EPSPS were performed according to previous studies, and the triple substitution was used to confirm hybridization.

Statistical Analysis

The ANOVAs of all dose–response curves were analyzed for significance (P < 0.05). The dose–response curves were fit by a four-parameter log-logistic nonlinear regression model:

([1]) $$y = c + \{ d - c/1 + \exp [b(\log (x) - \log (e))]\} $$

where y is the variable response; x is the herbicide dose; parameter b is the slope; parameter d is the upper limit; parameter c is the lower limit; and parameter e is the effective dose for 50% control (ED₅₀). The data were analyzed using the drc package from the statistical software R v. 4.2.1 (Knezevic et al. 2007). This package provides the curve fit, equation parameters, and resistance factor (RF) significances. Log-logistic significant parameters (b, d, c, and e) (P-value < 0.05) mean good model fit. ANOVA (P < 0.05) was also carried out for relative genomic EPSPS gene copy number; if statistical significance was found, means were compared using Tukey’s test (P < 0.05). The frequency of PMGF ranged from 0 to 1 and followed a binomial distribution (dead or alive). For PMGF, a power analysis for binomial probabilities was performed to determine the statistical precision of hybridization with the sample size used in this experiment. The theoretical frequency used was 0.001 at α = 0.05 for hybridization and was compared with the observed hybridization frequencies as described in Jhala et al. (2011).

Table 1. Morphological identification and two molecular approaches to identify the Amaranthus biotypes from the present study.

^a Morphological identification keys used: Amaranthaceae of Rio Grande do Sul (Vasconcellos 1986) and a synopsis of the Amaranthus subgenus in Venezuela (Pinto and Velásquez 2010).

^b First molecular approach: amplification of intron 1 of EPSPS (Wright et al. 2016).

^c ITS, internal transcribed spacer. Second molecular approach: amplification of the ITS5-ITS4 region (Murphy and Tranel 2018).

Results and Discussion

Morphological and Molecular Identification

The available morphological keys allowed for the identification of the EEA-S biotype as A. viridis (Supplementary Table 3). The presence of a glabrous stem, flowers with four tepals, male flower with three stamens, and non-spinescent bracts (Pinto and Velásquez 2010; Vasconcellos 1986) was decisive in the differentiation and morphological identification of this biotype as A. viridis. The other biotypes have pubescent stems, flowers with five tepals (except AMACVI-S and AMARET-S, which have four tepals), and bracts of varying shapes and sizes. For instance, the bracts of AMACHY-S, ARRGR-R, SAOJER-R, and CAMAQ-R range from spiky to a well-defined apex, while the bracts of susceptible populations AMACVI-S and AMARET-S have less-defined features (Supplementary Table 3).

The size of bracts is the key characteristic to differentiate between A. hybridus and A. retroflexus (Vasconcellos 1986). Differentiating between A. hybridus and A. retroflexus based on morphological characteristics, such as bract size, is challenging. The morphological evaluation through the keys described by Vasconcellos (1986) and Pinto and Velásquez (2010) did not allow the classification of six out of seven biotypes evaluated. The morphological similarities between the two species make distinguishing them difficult, especially in the vegetative phase (Horak et al. 1994). Hybridization between Amaranthus species can also lead to the formation of hybrids with intermediate characteristics that cannot be reliably distinguished using traditional morphological strategies (Murray 1940). As a result, A. hybridus and A. retroflexus may have been, in some cases, misidentified for a long time in Brazil.

In the genotypic study, amplification of intron 1 of the EPSPS allowed for differentiation of only the EEA-S biotype as A. viridis. The ARRGR-R, SAOJER-R, CAMAQ-R, AMACHY-S, AMACVI-S, and AMARET-S biotypes each yielded PCR products for both A. hybridus– and A. retroflexus–specific primer sets. However, AMARET-S had weak bands for both sets of primers (Supplementary Figure 2). This suggests that methods described by Wright et al. (2016) were validated against a limited range of plants/samples and geographic origins and may not represent the Amaranthus species found in Brazil.

The amplification of the ITS region allowed the classification of three out of seven biotypes. Phylogenetic analysis classified AMACHY-S and CAMAQ-R biotypes as A. hybridus and EEA-S as A. viridis (Figure 1). ITS sequences have been widely used to resolve phylogenetic issues and to select candidate plant barcodes (Baldwin et al. 1995). The biotypes ARRGR-R, SAOJER-R, and AMACVI-S appeared to be most closely related to A. hybridus and A. powellii. Although the ITS sequences of these three biotypes were clearly within the hybridus/powellii/retroflexus clade, their species identity was not resolved. The biotype AMARET-S is closely related to prostrate pigweed (Amaranthus albus L.) and mat amaranth (Amaranthus blitoides S. Watson) (Figure 1). The biotype AMACHY-S was used as susceptible biotype in further studies because its species identity was better resolved. Through morphological analysis and two different molecular study strategies, it was possible to verify that only EEA-S showed identical classification results between evaluations (Table 1). These results show that identifying Amaranthus species using phenotypic traits alone is challenging, and molecular strategies may be necessary. Based on our results, the ITS proved to be the most promising strategy for identifying Amaranthus species. Despite its limitations, many studies have shown that ITS is advantageous for identifying species of this genus (Murphy et al. 2017; Xu et al. 2017). Phylogenetic trees based on ITS suggested that Amaranthus species from China could be divided into five sections and two subsections. Amaranthus retroflexus and A. powellii formed a clade together, and foxtail amaranth (Amaranthus caudatus L.) and A. hybridus were closely related (Xu et al. 2017). Palmer amaranth (Amaranthus palmeri S. Watson) was successfully detected in mixed samples using ITS with a quantitative PCR method (Murphy et al. 2017). The increasing cases of Amaranthus species with resistance to glyphosate and ALS-inhibiting herbicides in Brazil have aroused interest in identifying them. Molecular techniques are reliable; however, it is necessary to obtain positive controls related to known samples of each species to verify distinguishable single-nucleotide polymorphisms to validate the identification of biotypes from southern Brazil.

Figure 1. Radial phylogenetic tree of reference Amaranthus with internal transcribed spacer (ITS) sequences extracted from Murphy and Tranel (2018) and ITS sequences of AMACHY-S, AMACVI-S, AMARET-S, EEA-S, ARRGR-R, SAOJER-R, and CAMAQ-R biotypes from this study. Asterisks (*): green, biotypes susceptible to glyphosate; red, resistant biotypes with the TAP-IVS mutation in EPSPS gene; from ALS gene: biotypes susceptible to (black) and biotypes resistant to (blue) acetolactate synthase (ALS)-inhibiting herbicides.

Confirmation of Resistance and Resistance Level

Herbicide ED₅₀ values for all biotypes are presented in Table 2. The glyphosate ED₅₀ values of control at 21 DAT for AMACHY-S, ARRGR-R, SAOJER-R, and CAMAQ-R biotypes were 19, 9,157, 9,857, and 5,880 g ha⁻¹, respectively. These results indicate that the ARRGR-R, SAOJER-R, and CAMAQ-R biotypes are 477-, 513-, and 306-fold less sensitive to glyphosate, respectively, than the AMACHY-S biotype, presenting a high RF level (RF > 300) (Table 2). The RF of biotypes with resistance to glyphosate identified in Argentina in 2018 was 314 (Perotti et al. 2019). Furthermore, RF of the fresh and dry shoot weight measurements also showed a high RF level for ARRGR-R, SAOJER-R, and CAMAQ-R biotypes, where the RF was 852, 780, and 485 g ha⁻¹ for fresh shoot weight and 361, 547, and 405 g ha⁻¹ dry shoot weight, respectively (Table 2). Biotypes of A. hybridus previously identified with resistance to glyphosate in Brazil showed a dose of glyphosate to reduce the shoot weight ranging from 3,019 to 3,316 g ha⁻¹ (Resende et al. 2020). Thus, the RFs from biotypes studied in the present study were similar to those found in Argentina and other reports of glyphosate-resistant populations in Brazil. This experiment confirmed the resistance of ARRGR-R, SAOJER-R, and CAMAQ-R to glyphosate.

Table 2. Parameters of the logistic equation and resistance factors (RFs) of weed control, fresh shoot weight, and dry shoot weight at 21 d after the application of treatments for glyphosate, imazethapyr, and chlorimuron.

^a Amaranthus putative resistant: ARRGR-R, SAOJER-R, and CAMAQ-R; susceptible: AMACHY-S.

^b b, slope; c, lower limit; d, upper limit; e, herbicide dose (g ha⁻¹) to control 50% of the variable (ED₅₀); lower/upper confidence interval (CI) of the variable e (ED₅₀); RF, resistance factor (resistant/susceptible).

* Statistically significant (P < 0.05).

The ED₅₀ control values at 21 DAT for AMACHY-S and SAOJER-R biotypes for imazethapyr were 0.8, and 3,759 g ha⁻¹, respectively (Table 2). For chlorimuron, the values were 0.4 and 337 g ha⁻¹, respectively (Table 2). These results indicate that the SAOJER-R biotype is 4,659- and 418-fold less sensitive to imazethapyr and chlorimuron than the AMACHY-S biotype, respectively, with high RF levels (Table 2). Another study showed that A. hybridus with resistance to imazethapyr was 200- to 600-fold (Whaley et al. 2006). In the present study, fresh and dry shoot weight measurements also showed a high RF level for the SAOJER-R biotype; RF was 4,147 g ha⁻¹ for fresh shoot weight and 2,652 g ha⁻¹ dry shoot weight for imazethapyr and 317 g ha⁻¹ for fresh shoot weight and 209 g ha⁻¹ dry shoot weight for chlorimuron (Table 2). This experiment confirmed the cross-resistance of SAOJER-R to ALS-inhibiting herbicides.

Multiple resistance to ALS- and EPSPS-inhibiting herbicides in A. hybridus was first verified in 2014 in Argentina (Heap 2023). In 2018, a population of A. hybridus was found in Brazil with multiple resistance to glyphosate and chlorimuron (Heap 2023). However, in those cases, the mechanism(s) of resistance has not been characterized (Mathioni et al. 2022; Resende et al. 2020). Recently, a study identified and characterized multiple resistance to ALS and EPSPS inhibitors for the first time in populations of A. hybridus in Argentina (García et al. 2020). García et al.’s (2020) study showed resistance to imazamox with ED₅₀ > 700 g ha⁻¹ and RF of 39. The present study also showed high-level resistance rate to the imazethapyr herbicide with RF of 4,658.

Target-Site Mutation and Increase in Copy Number

The triple amino acid substitution TAP-IVS in the EPSPS gene was confirmed in the biotypes with glyphosate resistance (Figure 2A). Sequences of susceptible and triple-mutated biotypes were aligned with the resistant A. hybridus from Argentina (García et al. 2020; Perotti et al. 2019) (Supplementary Figure 3). Our results also showed ATA¹⁰², GTC¹⁰³, and TCA¹⁰⁶ codons for the resistant biotypes, the same as those identified in biotypes from Argentina; on the other hand, ACA¹⁰², GCG¹⁰³, and CCA¹⁰⁶ were verified for the susceptible biotype (Figure 2A). The chromatogram of the CAMAQ-R biotype presents double peaks in the first nucleotide of the TCA¹⁰⁶ codon (Figure 2A) that were also verified in the different self-fertilized generations G1 to G2, showing that the locus was not heterozygotes (Supplementary Figure 4). Similarly, the double peaks were found in a previous study with A. hybridus in Brazil (Mathioni et al. 2022). The analysis of relative genomic EPSPS gene copy number indicates that there is more than one EPSPS locus in ARRGR-R, SAOJER-R, and CAMAQ-R biotypes (Figure 2B). These results are similar to those found in a study conducted by Perotti et al. (2019). Thus, CAMAQ-R probably has at least one locus from a susceptible population, which explains the double peaks in the chromatogram (Figure 2A and 2B). The presence of multiple EPSPS copies could also explain the ambiguity in species identification using the EPSPS marker.

Figure 2. (A) Chromatogram with susceptible biotype without mutation (AMACHY-S), and biotypes with the triple amino acid substitution Thr-102-IIe, Val-103-Ala, and Pro-106-Ser (TAP-IVS) mutation (ARRGR-R, SAOJER-R, and CAMAQ-R) in the EPSPS gene. (B) EPSPS relative copy number in susceptible (AMACHY-S and AMACVI-S) and resistant biotypes (ARRGR-R, SAOJER-R, and CAMAQ-R). The results were normalized using the AMACHY-S sample as the reference. EPSPS relative copy number was relative to the ACTN gene. An asterisk (*) indicates statistical difference according to Tukey’s test (P < 0.05) compared with AMACHY-S. Bars indicate standard error of the mean.

The mutation at the Pro-106 position has already been identified to confer resistance in many weeds (De Carvalho et al. 2012; Gaines et al. 2012). It should be noted that mutations in this locus result in low levels of resistance, generally ranging from 2- to 10-fold, when present in populations without other mechanisms of resistance (Baerson et al. 2002; Takano et al. 2019; Wakelin and Preston 2006). Furthermore, these substitutions do not confer an adaptive cost in plants (Sammons and Gaines 2014). Single substitutions found at positions Gly-101 and Thr-102 confer a high level of resistance to glyphosate due to the reduced size of the glyphosate binding site. However, this reduction also affects the binding of the enzyme with its substrate, phosphoenolpyruvate (Funke et al. 2009). The double mutation Thr-102-Ile and Pro-106-Ser, named TIPS, identified in goosegrass [Eleusine indica (L.) Gaertn.] (Yu et al. 2015) and hairy beggarticks (Bidens pilosa L.) (Alcántara-de la Cruz et al. 2016), also provides a high level of resistance to glyphosate. In Paraguay, the replacement of Thr-102-Ile and Pro-106-Thr (TIPT) was identified in greater beggarticks (Bidens subalternans DC) and confers high-level resistance to glyphosate (Takano et al. 2020). The triple mutation (TAP-IVS) was identified in biotypes of A. hybridus in Argentina and Brazil (García et al. 2020; Mathioni et al. 2022; Perotti et al. 2019). Due to the difficulty of occurrence of triple mutation in the same gene, because it is a rare event, it is suggested that the resistant biotypes existing in southern Brazil are due to the dissemination of seeds of resistant biotypes from Argentina. However, the origin of the EPSPS TAP-IVS substitution is unknown, and it is still possible to consider that its origin is in Brazil, even though it was reported earlier in Argentina.

Based on alignment with the A. thaliana reference sequence (GenBank X51514.1), the positions Ala-122, Pro-197, Ala-205, Ala-376, Arg-377, Trp-574, Ser-653, and Gly-654 associated with ALS mutations were analyzed (Figure 3). The sequencing of ALS revealed the presence of a predicted tryptophan to leucine substitution at position 574 (Trp-574-Leu) (Figure 3) in the individuals resistant to ALS herbicides (SAOJER-R biotype). This point mutation is well known to confer resistance to ALS-inhibiting herbicides in roughfruit amaranth (Amaranthus rudis Sauer), great ragweed (Ambrosia trifida L.), kochia [Bassia scoparia (L.) A.J. Scott], and rough cocklebur (Xanthium strumarium L.) (Tranel and Wright 2002). In addition, this mutation has already been described in A. hybridus biotypes from the United States (Whaley et al. 2006), and it is the first time described in Brazil. Thus, SAOJER-R has the EPSPS TAP-IVS substitution and ALS Trp-574-Leu target-site mechanisms conferring resistance to glyphosate and ALS inhibitors, respectively.

Figure 3. Sequence alignment of the ALS gene using Arabidopsis thaliana (GenBank X51514.1) as the reference and acetolactate synthase (ALS)-susceptible AMACHY-S and ALS-resistant SAOJER-R Amaranthus biotypes. The sequences presented here show the eight positions with known ALS mutations associated with resistance to ALS-inhibiting herbicides. Red nucleotide bases indicate the codon with mutation site position Trp-574-Leu.

Intra- and Interspecific Hybridization

Initially, we established experimental conditions to evaluate for response to glyphosate utilizing a seedling assay that permitted large-scale screening of individuals. The seedling assay revealed the dose of 3.2 mM glyphosate to distinguish susceptible from resistant biotypes (Supplementary Table 2; Supplementary Figure 5), similar to a study with waterhemp [Amaranthus tuberculatus (Moq.) Sauer] (Zelaya and Owen 2005). Moreover, the results were obtained after 14 days of incubation, which is faster than conventional whole-plant assays.

In wind-mediated gene flow trials, hybridization was detected under field research conditions between resistant and susceptible Amaranthus biotypes (Table 3). The intraspecific hybridization was evaluated in 82,206 and 37,540 individuals for the first and second seasons (2020 to 2021 and 2021 to 2022), respectively. The number of survivors were 73 and 34 for the experiments carried out in 2020 to 2021 and 2021 to 2022, respectively (Table 3). Thus, the observed rates for intraspecific hybridization between CAMAQ-R (A. hybridus) × AMACHY-S (A. hybridus) in 2020 to 2021 and 2021 to 2022 were 0.089% and 0.091%, respectively. Hybrids from the cross of A. hybridus biotypes were confirmed with EPSPS sequencing (Supplementary Figure 6), and the triple mutation was found in all 12 samples sequenced from each season (data not shown).

Table 3. Frequency of pollen-mediated gene flow (PMGF) between resistant and susceptible Amaranthus hybridus biotypes, and between a resistant A. hybridus biotype and an unclassified species of Amaranthus under field research conditions at the Federal University of Rio Grande do Sul in seasons 2020–2021 and 2021–2022.

^a Minimum sample size for theoretical frequency (0.001) of gene flow considering analysis power of 95% is 55,870, according to Jhala et al. (2011).

^b Seedlings that survived 3.2 mM of glyphosate.

^c PMGF frequency was evaluated as number of surviving seedlings/number of emerged seedlings.

^d Sample power for 95% confidence interval (1 − β), α = 0.05.

^e Estimated with 95% confidence, based on simple binomial distribution.

Even though Amaranthus hybridus is considered a predominantly self-pollinating species due to the proximity of the male and female flowers, it can be pollinated by wind, and the male and female flowers are separate (Murray 1940). Our study showed that gene flow might be a relevant contribution to the dispersion of TAP-IVS substitution from A. hybridus in South America. Pollen spread contributes to a higher frequency of herbicide-resistant plants compared with the expected initial frequency of herbicide-resistant individuals (10⁻⁸ to 10⁻⁹) in herbicide-unselected populations, and it plays an important role in resistance evolution by increasing the incidence of herbicide-resistant weeds (Beckie et al. 2019). This is the first documentation of dispersal by PMGF of alleles from the triple mutation in A. hybridus. However, the source of the EPSPS TAP-IVS substitution remains unknown.

Interspecific hybridization was not observed in the study with CAMAQ-R (A. hybridus) × AMACVI-S (unclassified genotype) in a total of 84,769 and 26,660 individuals analyzed in seasons 2020 to 2021 and 2021 to 2022, respectively (Table 3). However, the possibility of such hybridization cannot be entirely excluded. It should be considered that it might occur at a lower frequency, which needs to be evaluated in a larger sample size (Table 3). For example, the interspecific hybridization among species within the Amaranthus and Ambrosia genera enables the transference of herbicide-resistance alleles, even at relatively low levels, which requires evaluating a large number of individuals (Jhala et al. 2021).

Interspecific hybridization was reported in A. palmeri, transferring glyphosate resistance to A. hybridus at a rate of 0.01% (Gaines et al. 2012). Introgression between A. tuberculatus and A. hybridus was verified only in one direction, from A. hybridus to A. tuberculatus (Trucco et al. 2009). High rates of hybridization of 5.9% were found between A. tuberculatus and A. hybridus (Trucco et al. 2005). Hybridization between other species of Amaranthus was also reported (Gaines et al. 2012; Oliveira et al. 2018). Therefore, the observed hybridization in Amaranthus species appears to play a significant role in explaining instances of resistance spread, as observed in the findings in this study.

Gene flow of traits is a process that depends on many factors, such as particular genes and alleles, genetic inheritance of the trait, species ploidy, fitness costs, and species ecological traits (Jasieniuk et al. 1996). The interspecific transference of herbicide resistance within Amaranthus is mostly seen in dioecious species, which are different from A. hybridus, A. powellii and A. retroflexus, that are monoecious species. The transference of mesotrione resistance due to cytochrome P450 metabolism in A. tuberculatus to A. palmeri, both obligate outcrossing species, occurred under field conditions at a rate of 0.1%. When intraspecific transfer was evaluated, gene flow varied in distance and direction. The 90% reduction in PMGF was 13.1 and 26.1 m in each assessment year. PMGF under field conditions might be accelerating the increase of herbicide resistance in important weeds for the United States, such as A. palmeri and A. tuberculatus (Oliveira et al. 2018). However, it should be noted that similar studies are important to understand the dynamics of dispersion of resistance to glyphosate and ALS inhibitors in A. hybridus.

We showed that some biotypes that have the EPSPS TAP-IVS substitution and have morphological characteristics of A. hybridus potentially are hybrids of A. hybridus and A. powellii (Figure 1), explaining the morphological variation observed in the field by farmers. This can mean that these biotypes may be derived from one or more past hybridization events. In this case, the triple mutation could have been in Brazil for a long time, or the triple mutation may not be such a rare event for EPSPS. Additionally, a hybrid biotype used in the study showed resistance to ALS inhibitors caused by a target-site mechanism, with the presence of the Trp-574-Leu mutation described for the first time in Brazil for Amaranthus. It is possible to infer that the predominantly autogamous A. hybridus has a cross-fertilization rate that contributes to the spread of resistance to glyphosate and other herbicides. PMGF is a pathway for the spread of the EPSPS TAP-IVS substitution and dispersal of resistance to glyphosate. Furthermore, PMGF may favor the accumulation of resistance mechanisms through the pyramiding of genes and the acquisition of novel adaptive traits.