Prediction of glyphosate resistance level based on EPSPS gene copy number in Kochia scoparia

Glyphosate-resistant (GR) Kochia scoparia has evolved in dryland chemical fallow systems throughout North America and the mechanism involves 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS) gene duplication. Sugarbeet fields in four states were surveyed for K. scoparia in 2013 and tested for glyphosate-resistance level and EPSPS gene copy number. Glyphosate resistance was confirmed in K. scoparia populations collected from sugarbeet fields in Colorado, Wyoming, and Nebraska. The GR samples all had increased EPSPS gene copy number, with median population values up to 11. An empirical model was developed to estimate the level of glyphosate-resistance in K. scoparia based on EPSPS gene copy number. The results suggested that glyphosate susceptibility can be accurately diagnosed using EPSPS gene copy number, and further increases in EPSPS gene copy number could increase resistance levels up to 8-fold relative to susceptible K. scoparia. These trends suggest that continued glyphosate selection pressure is selecting for higher EPSPS copy number and higher resistance levels in K. scoparia. By including multiple K. scoparia samples lacking EPSPS gene duplication, our empirical model provides a more realistic estimate of fold-resistance due to EPSPS gene copy number compared to methods that do not account for normal variation of herbicide response in susceptible biotypes.

Adoption of glyphosate-resistant (GR) sugarbeet systems can result in improved weed 2 control and reduced sugarbeet injury compared to conventional sugarbeet systems [1]. 3 Glyphosate can provide weed control similar to or greater than conventional weed control 4 programs consisting of three applications of desmedipham, phenmedipham, triflusulfuron, and 5 clopyralid [2]. Prior to commercial introduction, net economic return was predicted to be 6 significantly greater for GR sugarbeet systems compared to conventional sugarbeet due to 7 reduced crop injury and better weed control [1]. GR sugarbeets were commercially introduced in 8 2007. By 2009, more than 85% of US sugarbeet hectares were seeded with GR cultivars, with 9 remaining areas seeded with conventional cultivars that had resistance to specific pests or 10 diseases that were not commercially available with the GR trait [3]. Sugarbeet growers have 11 significantly reduced tillage and increased net economic return since adoption of GR sugarbeet 12 [4]. 13 Kochia scoparia is a competitive weed that can cause substantial yield loss, and is 14 particularly a problem weed in sugarbeet [5,6]. Kochia scoparia is a C4 summer annual 15 broadleaf weed that can germinate and emerge early in the growing season and is tolerant to 16 heat, drought, and saline conditions [5]. Kochia scoparia has protogynous flowers in which the 17 stigmas usually emerge one week before pollen is shed and are receptive to foreign pollen which 18 can promote outcrossing between plants in close proximity [7]. It also produces copious amounts 19 of pollen for extended periods of time, which is generally an indication that the species is 20 naturally highly outcrossing [5]. Kochia scoparia stem breakage at the soil surface during 21 senescence allows for a tumbling seed dispersal mechanism that can contribute to high rates of 22 spread in the western US [8]. In Wyoming, K. scoparia densities as low as 0.2 plants m -1 of crop 23 row reduced sugarbeet root yield by 18% [9]. The outcrossing nature of K. scoparia combined 1 with prolific seed production results in genetically diverse populations that facilitate the 2 evolution of herbicide-resistance mechanisms [ Widespread adoption of GR sugarbeet systems in the US has resulted in significant glyphosate 7 selection pressure, and increasingly sugarbeet growers are reporting reduced K. scoparia control 8 with glyphosate. Therefore, the objectives of this study were to a) confirm whether glyphosate 9 resistance was present in K. scoparia collected from sugarbeet fields; b) determine whether GR     Greenhouse bioassay 9 Each K. scoparia accession was screened for susceptibility to glyphosate. Approximately 10 15 to 20 seeds were planted in 10 cm ×10 cm plastic pots filled with a 50:50 by weight mixture 11 of field soil and commercial potting mix. After planting, pots were placed in a greenhouse where 12 air temperature was maintained at 27 C and pots were watered several times per day with an 13 automated sprinkler system. Kochia scoparia emerged approximately three days after planting, 14 and the pots were thinned to three plants per pot shortly after emergence. 15 When K. scoparia averaged 10 cm in height, each accession was treated with five rates of binomial data (alive vs dead) was used to estimate the glyphosate dose causing 50% mortality 5 (LD50) for each kochia accession. The log-logistic model is of the form: Where Y is the probability of survival; x is the glyphosate dose in g ae ha -1 ; LD50 is the dose 8 required to cause 50% mortality; and b is the slope of the curve at the LD50. 9 EPSPS gene copy number assay 10 Based on the results of the greenhouse bioassay, a sub-set of 40 K. scoparia accessions 11 exhibiting a range of whole-plant resistance levels were assayed for EPSPS copy number. To   (S2 and S3 Figs). The LD50 from the greenhouse bioassay was regressed against the 20 median EPSPS copy number for each accession. Because the dose response study was conducted 21 at the accession level (that is, multiple individuals from each accession were used as 22 experimental units), the median EPSPS value was used to quantify the relationship between gene 23 copy number and whole-plant resistance. The mean copy number could be skewed significantly 1 by a few (or even one) high copy number individuals within a population, and therefore, 2 overestimate the number of high copy number plants for an accession. 3 We did not know a priori which type of regression model was appropriate to quantify the 4 relationship between EPSPS gene copy number and whole-plant glyphosate resistance. Based on 5 a preliminary visual evaluation, three different models were fit to the data; a three-parameter 6 rectangular hyperbolic model, a two-parameter rectangular hyperbolic model, and a simple linear Where Y is the LD50 from the greenhouse bioassay; X is the EPSPS copy number; Rmax is an 10 upper asymptote, or the maximum theoretical level of resistance at very high values of X; L is the 11 estimated LD50 when X=0; and K is the value of X that results in Y halfway between L and Rmax.

12
The two-parameter nonlinear model was the same, but with L=0; setting L=0 reduced the 14 After all three models were fit to the data, Akaike information criterion (AIC) was used to 15 determine which model provided the best fit to the data.

17
Greenhouse bioassay 18 Of the 65 K. scoparia accessions included in the bioassay, 12 had large standard errors 19 associated with the LD50 estimates. The LD50 results with and without these accessions are 20 provided in S1 Fig. Estimated LD50 values ranged from less than 500 to nearly 4000 g ae ha -1 21 (Fig 1A). A majority of K. scoparia accessions showed a susceptible response (median LD50 for 22 all accessions tested was 906 g ae ha -1 ), even though this survey was biased toward K. scoparia 1 plants that were likely to have survived glyphosate application(s). Although the median LD50 of 2 906 g ae ha -1 is greater than the standard field use rate of 840 g ae ha -1 , this is not an indication of 3 a high level of resistance. Previous studies have shown that up to 53-fold more glyphosate is 4 required to control the same Chenopodium album biotype in the greenhouse compared to the 5 field [22]. Conversely, greater glyphosate efficacy has been observed in an outdoor environment response to a variety of environmental factors when screening GR K. scoparia [24]. Because the 12 LD50 can be influenced by a variety of factors, the LD50 values in our study (or any greenhouse 13 study) are not an absolute indicator of resistance, but rather a relative measure used to compare 14 accessions within the study for glyphosate response. 15 EPSPS gene copy number 16 Mean EPSPS copy numbers in the 30 accessions assessed for effect of EPSPS copy number on 17 whole-plant response ranged from 0.7 to 10.2 (Fig 1B), and K. scoparia accessions with 18 increased EPSPS copy number were identified in Wyoming, Nebraska, and Colorado (Fig 2). 19 Since the accessions were still presumed to be segregating for the GR trait, the median EPSPS 20 copy number may provide a more accurate estimate of gene copy level as it relates to resistance 21 level for the accession. Median EPSPS copy numbers ranged from 0.7 to 11.3 depending on K. 22 scoparia accession. The median EPSPS copy number was less than the mean for nearly all 23 accessions where a notable difference was present, indicating that a few individual plants had 1 much higher EPSPS copy numbers compared to the majority of plants within that accession.

2
Effect of EPSPS copy number on whole-plant response 3 A notable increase in LD50 was apparent for accessions with median EPSPS value of >2.5 4 (Figs 1A and 1B). Based on AIC, the 2 parameter Michaelis-Menten model fit was chosen as the 5 best model (of those tested) to describe the relationship between resistance level and EPSPS gene 6 copy number. This suggests there is a 'plateau' with respect to resistance level; that is, additional 7 gene copies provide limited increase in resistance level once a certain threshold has been 8 reached. To ensure that the 'trimmed' data set did not have a major impact on the empirical 9 relationship, the same models were fit to the full data set of 40 accessions, which included LD50  Rmax, or the theoretical maximum LD50 when the number of EPSPS gene copies is very 12 large, was 7486 g ae ha -1 (Fig 3). The Rmax parameter, as an estimate of LD50, is not very useful 13 in absolute terms. The LD50 value depends heavily on the environmental conditions during the 14 bioassay [18, 25], and can vary significantly from one experimental run to the next even when 15 using the same genotypes and experimental design. However, the LD50 ratio between resistant 16 and susceptible biotypes tends to remain relatively more stable than other responses such as GR50 17 calculated from dry weight [26]. A standardized estimate of resistance level can therefore be 18 calculated by dividing the Rmax parameter by the model estimate of LD50 for a plant with a single 19 EPSPS gene. Practically speaking, this gives an estimate of the "fold" resistance expected due to 20 increased EPSPS gene copies. The estimated LD50 for a K. scoparia individual with a single 21 copy of the EPSPS gene in our study was 897 g ae ha -1 . Our analysis suggests that increasing 22 EPSPS gene copies could potentially increase glyphosate resistance level by a maximum of 1 about 8.3-fold (7486 / 897).

2
The estimated maximum level of glyphosate resistance seems to be in line with previous 3 work on glyphosate resistance, which report 4-to 11-fold resistance in glyphosate-resistant K.  to what is observed in K. scoparia from dryland fallow systems. Gene duplication in the tested 20 GR samples has not risen to levels higher than 12 additional copies. Determining the EPSPS 21 copy number is a valuable assay for diagnosing glyphosate resistance in K. scoparia. If a sample 22 has increased EPSPS copy number, our results suggest that the sample is GR (Figs 1 and 3). If a 1 sample does not have increased EPSPS copy number, it is glyphosate-susceptible. , which may be difficult to observe under field conditions. Some accessions 7 had LD50 estimates with high standard errors and these were excluded from the analysis. We 8 interpret this as being due to experimental variation in the greenhouse bioassay, and partially due 9 to heterogeneity within the accessions for EPSPS copy number. This variance in LD50 estimate is 10 not due to somatic instability or loss of EPSPS gene duplication in K. scoparia, as the inheritance 11 of EPSPS gene duplication has been previously shown to be stable [20]. Our analysis here is 12 somewhat limited, since our highest median EPSPS copy number for any accession included in from this mechanism should be tested empirically with K. scoparia populations with higher 17 EPSPS copy numbers.

18
These trends suggest that continued glyphosate selection pressure is selecting for higher 19 EPSPS copy number, higher resistance levels, and multiple herbicide resistance in K. scoparia. 20 Recently a K. scoparia population from Kansas has been confirmed to be resistant to four 21 herbicide modes of action (PSII, ALS, glyphosate, and synthetic auxins) [10]. Known glyphosate 22 resistance mechanisms exceed those reported for any other herbicide and include target-site 23 14 mutations, target-site gene duplications, active vacuole sequestration, limited cellular uptake, and 1 rapid necrosis response [28]. Proper stewardship of glyphosate is critical, including use of other 2 herbicide modes of action, cultural and mechanical control practices, and preventing seed set on 3 surviving K. scoparia. 4 Our approach is different from many previous calculations of R:S ratio since we included 5 many K. scoparia accessions that exhibited a susceptible response. Typically, R:S ratios are 6 calculated using one or two 'known susceptible' biotype(s can vary widely among accessions not expressing the resistance mechanism. Similar levels of 19 variability would be expected among plants with multiple EPSPS copies. If a single K. scoparia 20 accession were used as our 'known susceptible' biotype, the R:S ratio of our most resistant 21 accession (LD50 = 3895) could range from 3 to 19. Using our empirical model, we provide an 22 estimate of the level of glyphosate resistance that is attributable to the resistance mechanism and 1 is less affected by variability contributed by other, unrelated genetic factors.