In-Field Assessment of EPSPS Amplification on Fitness Cost in Mixed Glyphosate-Resistant and Glyphosate-Sensitive Populations of Palmer Amaranth (Amaranthus palmeri)

Abstract Comparing fitness of herbicide-resistant and herbicide-susceptible weed biotypes is important for managing herbicide resistance. Previous research suggests there is little to no fitness penalty from amplification of the 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS) gene (a mechanism of glyphosate resistance) in Palmer amaranth (Amaranthus palmeri S. Watson) in controlled studies in the greenhouse or growth chamber. A field study was conducted in North Carolina at three locations naturally infested with A. palmeri to determine vegetative, reproductive, and germination fitness of plants with and without EPSPS amplification grown season-long with cotton (Gossypium hirsutum L.). Seed number was not correlated with EPSPS copy number. However, when plants were binned into two groups, those having an EPSPS copy number ≥2 (relative to reference genes) and those having an EPSPS copy number <2, plant fresh weight and seed number were 1.4 and 1.6 times greater, respectively, for plants with fewer than 2 EPSPS copies. Amaranthus palmeri height and seed germination, and yield of cotton, did not differ when comparing the two binned groups. These data suggest that A. palmeri plants with EPSPS amplification are relatively less fit in the absence of glyphosate, but this reduced fitness does not translate into differences in interference with cotton.


Introduction
Glyphosate resistance in Palmer amaranth (Amaranthus palmeri S. Watson) was first confirmed in Georgia in 2005 (Culpepper et al. 2006) and continues to threaten numerous crops in the United States (Heap 2022;Webster 2013). Characteristics of A. palmeri that contribute to its ability to dominate fields include high photosynthetic capacity via the C 4 photosynthetic pathway (Ehleringer 1983), rapid growth (Horak and Loughin 2000;Sellers et al. 2003), drought tolerance mechanisms (Place et al. 2008;Wright et al. 1999), shade-adaptive capabilities (Jha and Norsworthy 2009;Jha et al. 2008), prolonged germination (Steckel et al. 2004;Ward et al. 2013), immense fecundity (Schwartz et al. 2016;Webster and Grey 2015), and wide genetic variation (Chandi et al. 2013). These traits give A. palmeri a competitive advantage over most crops (Bensch et al. 2003;Burke et al. 2007;MacRae et al. 2013;Massinga et al. 2001; Monks and Oliver 1988). For example, 8 A. palmeri plants m −1 row are capable of reducing cotton (Gossypium hirsutum L.) yield by as much as 92% (MacRae et al. 2013;Morgan et al. 2001;Rowland et al. 1999). When coupled with widespread herbicide resistance, these characteristics cause significant yield loss and increased management costs Klingaman and Oliver 1994;Morgan et al. 2001).
In some cases, herbicide resistance comes at a penalty, often referred to as a fitness cost. For example, Yanniccari et al. (2016) reported GR biotypes of perennial ryegrass (Lolium perenne L.) were shorter, had less leaf area and shoot biomass, and produced fewer seeds than GS biotypes. Triazine-resistant smooth pigweed (Amaranthus hybridus L.) is less efficient at photosynthesis and is shorter than susceptible wild-type species (Ahrens and Stoller 1983;Jordan 1996). Wu et al. (2018) reported a significant fitness cost in waterhemp [Amaranthus tuberculatus (Moq.) Sauer] associated with ALS resistance but not with resistance to atrazine, PPO-inhibiting herbicides, HPPD-inhibiting herbicides, or glyphosate. However, the researchers did note that A. tuberculatus resistant to glyphosate via amplification of the EPSPS gene decreased in frequency, while glyphosate resistance due to Pro-106-Ser EPSPS codon substitution increased over the course of the six generation (3-yr) experiment. Numerous greenhouse experiments have been conducted to determine whether glyphosate resistance imparts a fitness penalty in A. palmeri. Giacomini et al. (2014) reported no significant relationship between both final plant biomass and seed production with EPSPS copy number, although a negative relationship between EPSPS copy number and seed production was noted for one of the four pseudo-F 2 families tested. Similarly, researchers from Argentina and Australia found no correlation between EPSPS copy number and plant height or biomass for this weed (Vila-Aiub et al. 2014).
Determination of fitness costs associated with herbicide resistance is challenging due to the need to control for genetic background and because of the potential environmental dependence for manifestation of the fitness costs. (Keshtkar et al. 2019;Vila-Aiub et al. 2011). Furthermore, control of genetic background is necessary to directly relate pleiotropic effects on fitness cost with genes that confer resistance (Bergelson and Purrington 1996;Vila-Aiub et al. 2011). Otherwise, the differences in fitness between a resistant and susceptible biotype might simply be explained by differences in other traits. It is critical to keep in mind that the question about fitness cost due to resistance trait is not just theoretical, and ultimately, it is intended to determine the likelihood of the trait persisting or increasing in agroecosystems. The reality is that under field conditions, and especially in a dioecious species, genome architecture can change constantly, so the genetic context in which the resistance trait acts will likely differ from population to population and even from plant to plant . Therefore, studying multiple populations can help our understanding of the impact of the herbicide-resistance trait on the species and the range of fitness costs, if they exist, that can influence the spread of resistant biotypes (Bravo et al. 2017(Bravo et al. , 2018Keshtkar et al. 2019). Although studies have indicated no fitness penalty associated with EPSPS amplification in A. palmeri under controlled conditions, fitness costs under field conditions have not been documented. The objective of this experiment was to study fitness costs of EPSPS amplification in A. palmeri under field conditions.

Materials and Methods
The experiment was conducted in North Carolina in 2014 near Clayton at the Central Crops Research Station (35. 67°N, 78.51°W ) and in a grower's field near Mount Olive (35.20°N, 77.96°W ). The experiment was also conducted near Rocky Mount, NC, in 2015 at the Upper Coastal Plain Experiment Station (35.89°N, 77.67°W). Soil at Clayton was a Dothan loamy sand (fine-loamy, kaolinitic, thermic Plinthic Kandiudults) with 0.22% humic matter and pH 5.5. Soil at Mount Olive was a Lakeland sand (thermic, coated Typic Quartzipsamments) with 0.41% humic matter and pH 5.1. A Goldsboro fine sandy loam (fine-loamy, siliceous, subactive, thermic Aquic Paleudults) with 0.56% humic matter and pH 6.2 was present at Rocky Mount.

Plant Material
Each field site was dedicated to weed science research and naturally infested with a segregating field population of A. palmeri at densities greater than 100 plant m −2 and a long history (>15 yr) of recurrent glyphosate use. Expression of glyphosate resistance varied across populations and was approximately 40% to 60% at Clayton and Rocky Mount and 67% to 80% at Mount Olive. Level of glyphosate resistance was estimated from the population's response to glyphosate in previous experiments at each field site. As mentioned previously, control of genetic background is necessary to associate fitness cost with genes that confer resistance (Bergelson and Purrington 1996;Vila-Aiub et al. 2011); genetic background in this experiment was controlled by utilizing these three field sites naturally infested with A. palmeri segregating for EPSPS amplification.
One hundred to 175 A. palmeri plants per population were randomly selected at 3 wk after cotton planting and were covered with plastic cups to allow a broadcast application of glufosinate (Liberty® 280 herbicide, Bayer CropScience) at 656 g ai ha −1 plus acetochlor (Warrant® herbicide, Monsanto, St Louis, MO, USA) at 1,260 g ai ha −1 to control all other weeds. Cups were removed within 30 min of application. Other than the selected A. palmeri plants, weeds were removed by hand weeding throughout the season. Experimental fields were approximately 0.2 to 0.4 ha in size. The final plant density of A. palmeri was 1 plant 19 m −2 , which constituted a single plot; plots were spaced approximately 2.6 m apart. At 6 wk after planting, a single recently emerged leaf, approximately 1 cm −2 in size, was removed from each A. palmeri plant and stored on dry ice before being shipped to the University of Illinois to determine EPSPS copy number relative to a 1-copy reference gene (CPS, which encodes the large subunit of 664 Cahoon et al.: Amaranthus palmeri fitness carbamoylphosphate synthetase) number using quantitative PCR as described by Chatham et al. (2015). Amaranthus palmeri plants with an EPSPS copy number <2 were considered GS, whereas plants with EPSPS copy number ≥2 were considered GR. Amaranthus palmeri height was recorded at 15 wk after planting, with aboveground fresh weight determined at physiological maturity in mid-September of each year before seed rain. Cotton yield within 1 m of the A. palmeri plant and between 1 and 2 m from the A. palmeri plant was determined by hand picking cotton when bolls were fully open in October or November. Amaranthus palmeri reproductive inflorescences from female plants were removed, dried, and threshed to determine seed production.

Germination Assay
Seed production per plant was determined by counting seeds in 0.06-to 0.081-g subsamples to calculate 100-seed weight of the cleaned seed. One hundred seeds from each female plant were seeded in separate 25 by 53 cm greenhouse trays containing potting soil mix (Sun Gro® Fafard 4P Mix, Agawam, MA, USA). The greenhouse was maintained at 35 ± 5 C with metal-halide lighting (400 μmol m −2 s −1 ; Hubbell Lighting, Greenville, SC, USA) supplementing natural light for 14 h daily; greenhouse trays were lightly irrigated twice daily. To determine germination percentage, the number of emerged A. palmeri seedlings was recorded at 14 d after seeding.

Statistical Analysis
Data for EPSPS copy number and A. palmeri height, plant biomass, seed number, and germination percentage were subjected to ANOVA using PROC GLIMMIX in SAS (SAS Software v. 9.4, SAS Institute, Cary, NC, USA) to test for main effect of population, glyphosate-resistance binary designation (yes or no), and the interaction of population and resistance designation. Seed cotton yield within 1 m of the weed and from 1 to 2 m away from the weed was also subjected to ANOVA using PROC GLIMMIX in SAS along with cotton yield from a weed-free section. Significance between data for GR and GS biotypes was determined using a standard t-test at P ≤ 0.05. Differences in cotton yield in presence of GR and GS biotypes and the weed-free control were separated using Fisher's protected LSD test at P ≤ 0.05. Regression coefficients were constructed using the PROC REG in SAS for the following relationships: EPSPS copy number versus plant height, EPSPS copy number versus biomass, EPSPS copy number versus seed number, and EPSPS copy number versus seed germination at P ≤ 0.05. Additional regression coefficients included seed production versus plant height, seed number versus biomass, and plant height versus biomass.

Results and Discussion
Average EPSPS copy number for A. palmeri plants with ≥2 EPSPS copies was 41.3, 35.9, and 36.2 at Mount Olive, Clayton, and Rocky Mount, respectively (Table 1). EPSPS copy number for A. palmeri plants with <2 EPSPS copies from these respective populations was 0.8, 0.5, and 0.7, respectively. Fewer female A. palmeri plants having <2 EPSPS copies relative to females having ≥2 EPSPS copies were present at the end of the season at Mount Olive, while roughly the same number were present at Clayton and Rocky Mount (Table 1). During the period of time when these experiments were conducted, fields often had varying frequencies of GR A. palmeri. EPSPS copy number was not correlated with A. palmeri height, plant biomass, seed number, or seed germination (Table 2). These results are consistent with previous research (Giacomini et al. 2014;Vila-Aiub et al. 2014) showing no relationship between EPSPS copy number and biology of this weed, including seed production. While A. palmeri height was not correlated with plant biomass or seed number, plant biomass and seed number were correlated (Table 2).
We considered the lack of correlation between EPSPS copy number and fitness-associated traits may have occurred because EPSPS amplification affects those traits more in a qualitative rather than a quantitative fashion. Stated differently, we hypothesized that plants with, for example, 10 or 100 EPSPS copies would incur fitness costs relative to wild-type plants, but the relative fitness among plants expressing increased copy number would be similar. Therefore, we treated EPSPS copy number as a binary variable, with those having an EPSPS copy number ≥2 (relative to reference genes) and those having an EPSPS copy number <2 making up the two groups. Analyzed this way, the main effect of EPSPS copy number was significant for A. palmeri plant biomass and seed number (Table 3). The interaction of EPSPS copy number and population was not significant for these measurements. Amaranthus palmeri height and seed germination were not affected by EPSPS copy number or the interaction of EPSPS copy number and population.
When pooled over the three populations, EPSPS copy number was 37.8 for plants having ≥2 EPSPS copies and 0.6 for plants having <2 EPSPS copies (Table 4). Amaranthus palmeri height for plants having ≥2 EPSPS copies (141 cm) and plants having <2 EPSPS copies (144 cm) were not different. Amaranthus palmeri plant biomass and seed number were 1.4 and 1.6 times greater for plants having <2 EPSPS copies (n = 17) compared with plants having ≥2 copies (n = 29). Seed size was similar for the two A. palmeri biotypes, with 100-seed weights of 32.9 and 36.4 mg for plants having <2 EPSPS copies and plants having ≥2 copies, respectively (data not shown). Seed germination did not differ for A. palmeri plants having <2 EPSPS copies or those having ≥2 copies, and no difference in cotton yield was observed after season-long interference with the two biotypes regardless of distance from the A. palmeri plant (Table 5). Cotton yield in the presence of a single A. palmeri plant was 62% of weed-free cotton within 1 m of the weed and 87% at a distance of 1 to 2 m from the weed. These data are the only published results in the peer-reviewed literature with a mixed population of A. palmeri segregating for EPSPS amplification under field conditions in cotton demonstrating a fitness penalty for EPSPS copy number associated with seed production. Evaluation of fitness, especially in dioecious species, is difficult due to the changing genome architecture, which can differ from population to population . Furthermore, we considered the possibilities of a quantitative or qualitative relationship between fitness cost and EPSPS amplification. Because the frequencies of glyphosate-resistance mechanisms changed over time, Wu et al. (2018) hypothesized A. tuberculatus plants resistant to glyphosate via Pro-106-Ser EPSPS codon substitution were more fit than biotypes resistant to the herbicide by amplification of the EPSPS gene, which suggests a fitness cost may exist. While differences in plant biomass and seed production were noted between plants having EPSPS copy number <2 and those having ≥2 copies, interference with cotton and possibly other crops will not be affected by resistance. While there is a significant difference in seed production between plants having EPSPS copy number <2 and those having ≥2 copies, the amount of seed produced by plants having ≥2 copies remains substantial.
Another important consideration is that in a species such as A. palmeri with high genetic variability and obligate outcrossing, the genetic architectures in which the resistance allele(s) operate are also variable within populations. Furthermore, other traits that affect fitness are also under selection, and they could additively (or synergistically/antagonistically) increase or mitigate the fitness change caused by the resistance allele(s) Leon and van der Laat 2021). In fact, Bravo et al. (2017Bravo et al. ( , 2018 documented in commercial farms the existence of A. palmeri populations in which GR biotypes were taller and produced more biomass than GS biotypes. They also found that the differences in growth were at least partially due to increased nutrient-use efficiency in the former compared with the latter. Later, Leon and van der Laat (2021) provided evidence that those differences were the result of selection pressure in the different production systems from which the populations originated. Therefore, interpreting the presence or absence of fitness costs in A. palmeri and other dioecious or outcrossing species with high genetic diversity must be done cautiously, because the final fitness that weeds exhibit in the field is highly determined by genes other than those controlling herbicide resistance.
It is not clear why we did not observe a quantitative relationship between fitness cost and EPSPS amplification but did observe a significant fitness cost when EPSPS amplification was analyzed as a qualitative trait. Intuitively, one would expect more copies would result in more fitness cost due to greater resource drain on the plant or greater perturbation of the shikimate pathway and, therefore, a quantitative relationship. However, one possibility is that increased copies of EPSPS do not confer a fitness cost. Instead, perhaps one   or more genes that are present within the EPSPS amplicon and that might play a role in fostering DNA amplification (Molin et al. 2020) act in a qualitative fashion to impart a fitness cost. A practical goal for determining reproductive fitness costs of herbicide resistance is predicting the utility of herbicide rotation as a herbicide-resistance mitigation strategy. In this study, we did not compare fitness between A. palmeri males having <2 EPSPS copies and males having ≥2 copies. Also, because increased EPSPS copy number is not inherited as a single-gene trait (Gaines et al. 2010), it is difficult to extrapolate how our observed fitness cost in A. palmeri seed production would decrease the frequency of GR individuals over time if the population was not exposed to glyphosate. However, if one simply assumes that all GR and GS females produce only GR and GS progeny, respectively, and one further assumes no impact by relative male fitness and no seed dormancy, then even after 5 yr, a population starting with 75% GR individuals would still have about 20% GR individuals. With seed dormancy, the decrease in GR frequency would occur even more slowly. However, a return to glyphosate use would dramatically increase glyphosate resistance, but perhaps it could be an effective option for managing a population of GR A. palmeri if used in rotation only once every 5 to 10 yr.