Effects of over-expressing a native gene encoding 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS) on glyphosate resistance in Arabidopsis thaliana

Widespread overuse of the herbicide glyphosate, the active ingredient in RoundUp®, has led to the evolution of glyphosate-resistant weed biotypes, some of which persist by overproducing the herbicide’s target enzyme, 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS). EPSPS is a key enzyme in the shikimic acid pathway for biosynthesis of aromatic amino acids, lignin, and defensive compounds, but little is known about how overproducing EPSPS affects downstream metabolites, growth, or lifetime fitness in the absence of glyphosate. We are using Arabidopsis as a model system for investigating phenotypic effects of overproducing EPSPS, thereby avoiding confounding effects of genetic background or other mechanisms of herbicide resistance in agricultural weeds. Here, we report results from the first stage of this project. We designed a binary vector expressing a native EPSPS gene from Arabidopsis under control of the CaMV35S promoter (labelled OX, for over-expression). For both OX and the empty vector (labelled EV), we obtained nine independent T3 lines. Subsets of these lines were used to characterize glyphosate resistance in greenhouse experiments. Seven of the nine OX lines exhibited enhanced glyphosate resistance when compared to EV and wild-type control lines, and one of these was discarded due to severe deformities. The remaining six OX lines exhibited enhanced EPSPS gene expression and glyphosate resistance compared to controls. Glyphosate resistance was correlated with the degree of EPSPS over-expression for both vegetative and flowering plants, indicating that glyphosate resistance can be used as a surrogate for EPSPS expression levels in this system. These findings set the stage for examination of the effects of EPSPS over-expression on fitness-related traits in the absence of glyphosate. We invite other investigators to contact us if they wish to study gene expression, downstream metabolic effects, and other questions with these particular lines.


Introduction
The overarching goal of our current research is to test for phenotypic effects of overproducing EPSPS in multiple lines of transgenic Arabidopsis thaliana. Brotherton et al. [32] noted that Arabidopsis is highly sensitive to glyphosate, and the availability of resistant lines would be useful for studying mechanisms of herbicide resistance. As recommended by the Weed Science Society of America, we use the term "resistance" to designate the emergence of heritable resistance relative to wild-type plants, rather than the term "tolerance", which refers to a natural, inherent ability to withstand herbicide treatment [33]. This model system allows us to isolate effects of overproducing EPSPS from other possible confounding factors that can occur in glyphosate-resistant weed biotypes, such as different genetic backgrounds and multiple mechanisms and types of herbicide resistance (e.g., [34,35]). Our initial approach was similar to that of Klee et al. [36], who cloned an EPSPS gene from Arabidopsis thaliana, fused it to the CaMV35S promoter, and reintroduced the gene back into Arabidopsis. These authors showed that over-expressing Arabidopsis lines had superior callus growth and seedling survival compared to control lines when exposed to glyphosate, but they did not report data for mature plants, nor did they examine other phenotypic traits. In the present study, we created multiple transgenic lines that are homozygous for an over-expression construct (OX lines), along with control lines that are homozygous for an empty-vector construct (EV lines). Our specific questions in this first stage of the project were: 1. To what extent does glyphosate resistance vary among independent lines that share the OX construct vs. the EV construct?
2. Is greater resistance to glyphosate correlated with enhanced EPSPS gene expression, as expected?
3. Are differences in glyphosate resistance among lines consistent for both vegetative and reproductive plants?
4. Are differences in glyphosate resistance among lines repeatable in different experiments?
Briefly, we identified OX lines with enhanced glyphosate resistance and expected increases in EPSPS gene expression when compared to EV and wild-type control lines. Our results indicate that OX lines are more resistant to glyphosate in both vegetative and flowering plants across multiple experiments and that resistance correlates with the level of EPSPS over-expression. In a forthcoming paper, we will report variation in fitness-related traits among these OX, EV, and wild-type lines.
We invite other investigators to contact us if they wish to study additional aspects of gene expression, downstream metabolic effects, and other questions with these particular lines.

Development of over-expressed and empty vector lines
We developed Arabidopsis lines with independent insertions of over-expressed (OX) and empty vector (EV) constructs. Arabidopsis has two native EPSPS loci (AT1G48860 and AT2G45300 [14]) that are highly expressed throughout development (http://bar.utoronto.ca/ efp/cgi-bin/efpWeb.cgi). We chose AT1G48860 due to its purported higher glyphosate resistance than AT2G45300 in Arabidopsis [37] and the G-clone (G09183, a gateway entry clone) was ordered from the Arabidopsis Biological Resource Center (ABRC) at Ohio State University. The entry clone was introduced into the binary expression vector pB2WG7 carrying the Bar gene conferring resistance to Basta (glufosinate) by LR recombination to generate the EPSPS construct, under the control of CaMV35S promoter (Fig 1). The CcdB (control of cell death B) gene from Escherichia coli present in the empty vector (EV) construct has no known function in plants. The OX construct was confirmed by sequencing, and both constructs were introduced to Agrobacterium tumefaciens strain GV3101. Then, GV3101 bearing either CaMV35S::EPSPS or the empty vector (Fig 1) was used to transform Arabidopsis thaliana (Columbia ecotype, Col-0) by the floral dip method [38].
Transgenic progeny were selected by spraying soil-grown seedlings on 3 consecutive days with 0.04% (w/v) Basta supplemented with 0.005% SilwetL-77 solution. Independent T1 lines with presumed single insertion loci were identified by 3:1 segregation of Basta resistance in the T2 generation. These T2 lines were allowed to self and homozygous T3 lines (those that segregated 4:0) were used in subsequent experiments. Using this procedure, 9 independent OX lines and 9 independent EV lines were generated. We also created OX and EV lines in a sid2 (eds16, ics1) mutant of Arabidopsis that has reduced levels of defense-associated salicylic acid [39]. These sid2 OX lines with enhanced glyphosate resistance and comparable empty vector lines also are available for study on request, but are not included in the current study.

Glyphosate resistance
Dose-response experiment. To compare glyphosate resistance among the OX, EV, and wild-type, we conducted four independent experiments (Table 1). In February 2015, we began a dose-response experiment using 9 empty vector lines (EV), 9 over-expressed lines (OX), and the wild-type line. The experimental design included nine dosage treatments with three plants per line in each treatment, for a total of 513 plants. Seeds were germinated in the Biological Sciences Greenhouse at Ohio State University, Columbus, Ohio, in plastic flats with moistened Sun Gro Metro-Mix 360 (www.sungro.com). Each flat was divided into evenly spaced rows and lines were randomized within this grid, with one plant per line in each flat. Approximately 3-4 seeds were planted on the soil surface in each position and hand-watered using a fine mist. Upon germination, seedlings were thinned to one plant per position. The flats were rotated around the greenhouse weekly to minimize environmental variation and were hand-watered twice daily as needed. The greenhouse was maintained at 18-21/23-26 C (night/day) and supplemental lights (400-watt metal halide) were used for 14 hours per day. To maintain soil fertility, the flats were sprayed weekly with nutrient solution (180 ppm 20:10:20; www.jrpeters. com), starting two weeks after seedling emergence.
At 35 days after planting, when the plants were~5-7 cm in diameter, each tray was sprayed with one of 9 treatments: 0x, 0.001x, 0.004x, 0.0156x, 0.0625x, 0.25x, 1x, 4x, and 16x, where 1x is the recommended application rate for many agricultural weeds, 840 g ae/ha, and equates to 0.6725% glyphosate (v/v; AquaMaster 1 , 648 g L -1 , Monsanto Co.; St. Louis, Missouri). Each At 14 days after treatment, the plants were visually assessed for damage on a scale from 0 (no damage) to 10 (dead). By 21 days, the plants showed signs of drought stress so they were not used for scoring resistance.
To quantify relative differences in glyphosate resistance among lines, we calculated the effective dose 50 (ED50; as in [40,41]). This dose corresponds to the midpoint between the lower and upper limits of the observed plant health. Using the drc package in R [40], the ED50 of each line was estimated using the four-parameter log-logistic function given in Eq 1: where e is the ED50, the lower limit is c, the upper limit is d, and the parameter b denotes the relative slope around e. Our model constrained the lower limit to 0 (zero) and the upper limit to 10, which effectively reduced to the three-parameter log-logistic function. To avoid a statistical lack of fit to the model, data from the three lowest doses (0x, 0.001x, 0.004x) were designated as a zero dose. The selectivity index was calculated to allow statistical comparisons for ED50 estimates among lines at p<0.05 (Table 2). Discriminating dose experiments. Three additional experiments (Experiments A, B, and C) were conducted to confirm results from the dose-response experiment and test for consistent relative resistances when glyphosate was applied to rosettes vs. flowering plants (Table 1). These experiments included 6 OX lines, 4-7 EV lines, and the wild-type line, all of which were subjected to discriminating dosages of either 1.0x or 0.5x glyphosate ( Table 1). Three of the original OX lines were not used because two had low resistance to glyphosate (labeled OX8, OX9; Table 2), and a third (OX7) produced plants that were severely deformed, with abnormal rosettes and very short internodes at flowering. We also eliminated two randomly chosen EV lines (labelled EV8, EV9; Table 2) because these were deemed unnecessary. Seeds were germinated in moistened Sun Gro Metro-Mix 360 soil, with one seedling per pot (4.5 or 6 cm square pots), and one pot per line arranged in random positions within trays. The experiments included 15-24 plants per line (i.e., 15-24 trays; Table 1). All trays were irrigated from the bottom every 2 or 3 days and were rotated weekly to account for variation in environmental conditions. Experiments A and B were started in a growth room with 8 hrs of light for one month and then transferred to a greenhouse with supplemental light as above and ambient day lengths (13-15 hrs). For Experiment B, the pots were fertilized twice using 25 mL of nutrient solution (180 ppm 20:10:20; www.jrpeters.com) and then were sub-irrigated with this nutrient solution approximately every 3 weeks. Supplemental light was used to provide 16 hours of daylight. Fertilizer was not used in Experiments A or C. Plants in Experiment C were started over four consecutive days to stagger the timing of leaf samples used for EPSPS gene expression (see below).
Plants in Experiments A and B were sprayed at 44 days after planting, prior to flowering, while those in Experiment C were sprayed at 35 days when they had started flowering (Table 1). Glyphosate treatments of 0.5x or 1.0x were applied as described above. Resistance to glyphosate was recorded 21 days after spraying in all experiments. Visual damage scores were "0" for dead plants, "1" for almost dead, "2" for likely to die, "3" for likely to survive, "4" for partly damaged, and "5" for those that were mostly green and had green meristems (center of rosette). This scoring system was deemed to be more efficient and easier to carry out than that used in the dose-response experiment, which involved scores of 1-10. For Experiment A, we also weighed the fresh, above-ground, living biomass of each surviving plant. A one-way ANOVA with Tukey's multiple range tests was performed to compare above-ground biomass

Gene expression of EPSPS
Quantitative real-time PCR was used to estimate gene expression of EPSPS relative to the native Actin7 gene (e.g., [42]). We compared EPSPS expression levels among 6 OX lines, 3 EV lines, and wild-type plants from Experiment C, before glyphosate was applied (Table 1). Because gene expression is likely to vary over time, we staggered the planting and RNA sampling of these lines over four days. Thus, on days 1-4, we planted six plants per line each day, for a total of 24 plants per line. Three plants per line were sampled from rosettes, 28 days after planting, and the other three were sampled at the flowering stage, 35 days after planting, to obtain composite samples (see below). This provided independent estimates of gene expression from different plants at two distinct developmental stages. Then, all 24 plants per line were sprayed with glyphosate at 35 days after planting, as described above.
On each day of sampling, we collected one healthy, young leaf (~1 cm 2 leaf area,~75-100 mg) per plant and combined the three leaves per line into a composite sample. Thus, for each line and each stage of development, we obtained four composite samples (replicates), which were frozen immediately in liquid nitrogen and used for RNA extraction. RNA was extracted using RNeasy Plant Mini Kit (Qiagen), quantified using a Nano drop spectrophotometer (ND-1000, Thermo Scientific), and treated by DNase I (Invitrogen). The cDNA was created by a Reverse-transcriptase (RT) reaction using Promega Reverse-Transcription System. For the real-time PCR, all primers were designed using PRIMER 3 (version 0.4.0) (http://bioinfo.ut.ee/ primer3-0.4.0/) and synthesized by Eurofins MWG Operon LLC. Primer sequences for EPSPS were 5'-TCGTGCTGTAGTTGAAGGATG-3' (forward) and 5'-GCGGTAAGTGGACGCATT-3' (reverse); Primer sequence for Actin7 were 5'-CAGTGTCTGGATCGGAGGAT-3' (forward) and 5'-TGAACAATCGATGGACCTGA-3' (reverse). Three reaction mixes of 25ul for real-time PCR was prepared for each composite sample with iQ™ SYBR 1 Green Supermix from BioRad proportioned according to its user guide. A two-step program on the real-time PCR cycler (Bio-Rad CFX96 Touch™) with annealing temperature at 60 C was used with each composite sample. Efficiencies for the EPSPS (96.1%) and Actin7 (104%) primer pairs were determined by slopes of dilution curves with R 2 >0.99. The lack of primer dimers and specificity of each primer was demonstrated through gel analysis, melt curves, and sequencing of products.
The Relative Normalized Quantification of EPSPS in each composite sample was calculated based on the modified 2 -ΔΔCT method [43,44]. Wild-type samples were used as the control and EPSPS and Actin 7 were the target and reference (ref) genes, respectively. To calculate the relative quantification, the modified equation used is: ciency, and Cq is the quantification cycle. To summarize, the sampling procedure entailed 10 lines x 2 developmental stages x 4 days on which we collected one independent composite sample per line, for a total of 80 composite samples.
The final Relative Normalized Quantification of each OX, EV, or wild-type line at each growth stage (vegetative vs. flowering) was calculated by averaging the values of composite samples from different days and weighting the values such that the wild-type line had a value of 1.0. In some lines, no usable data were obtained for one of the four days of sampling, resulting in final sample sizes of 3-4, except for one flowering line, OX1, which had two samples. For each growth stage, the average expression levels of the 6 OX, 3 EV, and the wild-type lines were compared using a two-way ANOVA to account for differences among sampling dates (blocks), with Tukey's multiple range tests (IMB SPSS Statistics ver. 19.0 for Windows). Although our final sample sizes were lower than planned, differences in gene expression associated with the OX construct were clear and consistent, as shown below.

Results and discussion
In this study, our main goal was to obtain and characterize independent Arabidopsis lines that overproduce EPSPS and exhibit enhanced glyphosate resistance. When transgenes are inserted into a plant's genome, their effectiveness is expected to vary among independent insertion events, such that only a subset of the lines will be useful for further research. Here, all of the EV lines were similar in terms of lacking resistance to glyphosate, as expected, with no visible developmental problems associated with the transgene construct. In contrast, we did observe variation in glyphosate resistance and morphology among the nine OX lines, as discussed below. This variation most likely results from position effects of transgene insertion or insertion of a transgene multimer at the insertion locus, although mutations generated by Agrobacterium during the production of T1 plants also may have contributed.
We found that transgenic insertion of the OX construct conferred greater resistance to glyphosate in 7 of the 9 independent T3 lines, as compared to the EV and wild-type lines ( Table 2), representing a relatively high success rate. One of the seven resistant lines, OX7, had a high ED50 (6.02 kg ae/ha), but was eliminated from further study because many plants had deformed rosettes and shortened internodes when flowering. Among the six remaining OX lines, we observed two general levels of glyphosate resistance (Table 2; Figs 2 and 3). Three lines, OX1, OX2, and OX3, had ED50 values ranging from 1.4 kg ae/ha to 1.7 kg ae/ha, while the other three had ED values from 4.8 kg ae/ha to 5.9 kg ae/ha; these differences were significant at p<0.05 ( Table 2). Levels of glyphosate resistance in the single-dose experiments were consistent with the dose-response experiment, but differences among the OX lines were less pronounced (Fig 3). In Experiment A, with a 1x glyphosate treatment, surviving biomass of line OX6 was significantly greater than the biomass of OX1, OX2, and OX3 ( Fig  3A). Likewise, in Experiment B (0.5x treatment), lines OX4, OX5, and OX6 performed significantly better than OX2 and OX3. In Experiment C (0.5x treatment), OX4 and OX5 were significantly more resistant to glyphosate than OX1 and OX2. In summary, levels of glyphosate resistance among the OX, EV, and wild-type lines were consistent across the four experiments (Table 2; Fig 3).
We observed enhanced gene expression in the leaves of vegetative rosettes and flowering plants in the OX lines (Fig 4), as expected for leaf samples with a construct driven by the CaMV35S promoter. Vegetative plants in the OX lines had~26-49-fold greater EPSPS gene expression than wild-type controls, while somewhat lower levels of gene expression were found in flowering plants from the OX lines (Fig 4). Among the OX lines, gene expression levels were lower in OX1, OX2, and OX3 compared to OX4, OX5, and OX6, which is consistent with lower glyphosate resistance observed in OX1, OX2, and OX3 (Table 2; Fig 3). Thus, we conclude that greater glyphosate resistance in the OX lines can be attributed to the CaMV35S:: EPSPS transgene, and variation in levels of over-expression among OX lines were correlated with levels of glyphosate resistance. Some studies have shown that 35S is not expressed strongly in reproductive tissues of other species [45], but we did not test for this effect here.
The long-term goal of our research is to examine other phenotypic effects of overproducing EPSPS in Arabidopsis in the absence of glyphosate, including studies of plant growth, competitive ability, and lifetime fitness. These ongoing and future studies are expected to improve our understanding of weed biotypes that have evolved glyphosate resistance by overproducing EPSPS. In addition, we note that The Scotts Company (Marysville, OH, USA) is developing a glyphosate-resistant turf grass, Poa pratensis, with over-expression of an EPSPS gene from Arabidopsis thaliana driven by an ubiquitin promoter from rice, Oryza sativa [46,47]. Therefore, insights from transgenic Arabidopsis lines that overproduce EPSPS may be relevant to both cultivated and weedy taxa that also overproduce this unique enzyme in the shikimic acid pathway.