Contrasting plant ecological benefits endowed by naturally occurring EPSPS resistance mutations under glyphosate selection

Abstract Concurrent natural evolution of glyphosate resistance single‐ and double‐point EPSPS mutations in weed species provides an opportunity for the estimation of resistance fitness benefits and prediction of equilibrium resistance frequencies in environments under glyphosate selection. Assessment of glyphosate resistance benefit was conducted for the most commonly identified single Pro‐106‐Ser and less‐frequent double TIPS mutations in the EPSPS gene evolved in the global damaging weed Eleusine indica. Under glyphosate selection at the field dose, plants with the single Pro‐106‐Ser mutation at homozygous state (P106S‐rr) showed reduced survival and compromised vegetative growth and fecundity compared with TIPS plants. Whereas both homozygous (TIPS‐RR) and compound heterozygous (TIPS‐Rr) plants with the double TIPS resistance mutation displayed similar survival rates when exposed to glyphosate, a significantly higher fecundity in the currency of seed number was observed in TIPS‐Rr than TIPS‐RR plants. The highest plant fitness benefit was associated with the heterozygous TIPS‐Rr mutation, whereas plants with the homozygous Pro‐106‐Ser and TIPS mutations exhibited, respectively, 31% and 39% of the fitness benefit revealed by the TIPS‐Rr plants. Populations are predicted to reach stable allelic and genotypic frequencies after 20 years of glyphosate selection at which the WT allele is lost and the stable genotypic polymorphism is comprised by 2% of heterozygous TIPS‐Rr, 52% of homozygous TIPS‐RR and 46% of homozygous P106S‐rr. The high inbreeding nature of E. indica is responsible for the expected frequency decrease in the fittest TIPS‐Rr in favour of the homozygous TIPS‐RR and P106S‐rr. Mutated alleles associated with the glyphosate resistance EPSPS single EPSPS Pro‐106‐Ser and double TIPS mutations confer contrasting fitness benefits to E. indica under glyphosate treatment and therefore are expected to exhibit contrasting evolution rates in cropping systems under recurrent glyphosate selection.


| INTRODUC TI ON
From an evolutionary perspective, recurrent use of herbicides on large plant populations is a strong selection pressure for the enrichment of herbicide resistance mutations (Beckie & Tardif, 2012;Gressel & Levy, 2006;Jasieniuk et al., 1996;Maxwell & Mortimer, 1994;Powles & Yu, 2010). The ability of herbicide resistance mutations to spread and increase in frequency is a function of the survival and reproductive success of plants harbouring those mutations under herbicide selection (Cousens & Mortimer, 1995;Neve et al., 2014). As a result, the spread of particular resistance mutations in environments under herbicide selection will depend on the relative fitness (W) of resistant (R) and susceptible (S) genotypes (i.e. the so-called resistance benefit) (Simms & Rausher, 1987). A herbicide resistance benefit is a measure of the efficiency of a resistance trait in protecting plants from the lethal effect of a particular herbicide dose in a particular environment. Empirical estimations of resistance benefits associated with herbicide resistance mutations are lacking in the literature where efforts have mostly focussed on assessing survival but not reproductive traits (but see Beckie & Morrison, 1993a, 1993bSou Sheng et al., 2016;Yanniccari & Gigón, 2020).
Insights into the adaptive resistance evolution to the very widely used herbicide glyphosate are key to predict weed infestations in extensive cropping areas (Baucom, 2019;Gaines et al., 2019). This understanding has benefited from knowledge of molecular and biochemical resistance mechanisms conferred by resistance-endowing target site point mutations and their associated fitness costs (Funke et al., 2009;Gaines et al., 2019;Han et al., 2017;Li et al., 2018;Sammons & Gaines, 2014;Sammons et al., 2018;Vila-Aiub et al., 2019;Yu et al., 2015). However, predicting the evolutionary trajectories of specific glyphosate resistance mutations requires quantifying resistance fitness benefits of particular EPSPS resistance mutations.
Glyphosate resistance evolution is complicated by the occurrence of multiple EPSPS resistance mutations, and this makes important the quantification of the relative fitness benefit associated with each of these glyphosate resistance mutations. Distinctive to glyphosate resistance is the possibility of occurrence of multiple EPSPS resistance mutations in a single allele within a single plant (Gaines et al., 2019). Two naturally occurring EPSPS mutations at Thr-102 and Pro-106 have been reported. Whereas the first EPSPS variant was identified in E. indica from Malaysia including the Thr-102-Ile +Pro-106-Ser (i.e. TIPS) (Yu et al., 2015), the second variant was recently found in Bidens subalternans from Paraguay involving the Thr-102-Ile +Pro-106-Thr (TIPT) (Takano et al., 2020). Interestingly, a rare triple glyphosate resistance EPSPS mutation involving the Thr-102-Ile +Ala-103-Val +Pro-106-Ser codons (TIAVPS) has evolved in Amaranthus hybridus from Argentina (Perotti et al., 2019).
Evolution of E. indica with single (Pro-106-Ser) and double (TIPS) glyphosate resistance target site EPSPS gene mutations has been recently documented (Yu et al., 2015). The objective of this study is the estimation of resistance fitness benefits associated with these mutational events. Different to previous studies on resistance fitness cost in the absence of glyphosate (Han et al., 2017;Vila-Aiub et al., 2019), comparison of E. indica plants sharing a common genetic background and growing in a nonlimiting resource environment under glyphosate selection has enabled, in this current study, an accurate estimation of the glyphosate resistance benefit of EPSPS target site Pro-106-Ser vs TIPS mutations and the associated effects on alleles frequencies over time. These results help understand the differential selective advantage and spread rate of these single and double glyphosate resistance EPSPS mutations in environments under glyphosate selection.
Considering the identified three alleles in the collected field population, genetic recombination could potentially result in six genotypes but only four were identified within the single E. indica population (Table 1) (Yu et al., 2015): glyphosate-susceptible plants (homozygous WT) and glyphosate-resistant plants with either the homozygous Pro-106-Ser mutation (P106S-rr) or TIPS (TIPS-RR) mutation. A fourth genotype was comprised by plants segregating only at the Thr-102 position but not at the Pro-106 position rendering plants as compound heterozygous for the TIPS mutation (TIPS-Rr) (Table 1) (Yu et al., 2015). Neither the genotype heterozygous for the Pro-106-Ser mutation (WT/r) nor heterozygous for the TIPS mutation (WT/R) were identified in this population (Yu et al., 2015).
The sub-populations used in this study arise from the genotyping of plants (n = 7-12) which were bulk selfed in isolation in glasshouse conditions to produce seeds which resulted in three purified subpopulations containing plants with homozygous genotypes of WT, P106S-rr or TIPS-RR. Progeny plants (n = 10-12) from each of these purified sub-populations were DNA genotyped using dCAPS markers to confirm their genotype and homozygosity prior to use in the experiments detailed below (Yu et al., 2015). The TIPS-Rr individuals were identified in seedlings derived from a bulked progeny of Rr × Rr crossing and immediately used in the following experiments described below.

| Resource allocation to vegetative growth
An experiment was conducted to quantify the growth of aboveground tissues (stems and leaves) and roots during early growth of individual seedlings in pots (Ø = 20 cm, height = 20 cm) treated with 1080 g/ha glyphosate. Ten to fifteen plants per genotype (WT, P106S-rr, TIPS-RR) were exposed to glyphosate at the 4-5 leaf stage, and then at weekly intervals, the aboveground foliage and the root biomass were separately harvested from individual plants, washed with tap water, oven-dried at 60°C for 7 days and weighed. Inflorescences from each individual were threshed to separate seeds from chaff and rachis material, and total seed mass and number were TA B L E 1 EPSPS mutations, alleles and genotypes identified in the glyphosateresistant Eleusine indica population used in this study quantified. Experiments were repeated for further examination of seed production in plants (n = 10-20) of all genotypes (P106S-rr, TIPS-RR, TIPS-Rr) exposed to the single recommended glyphosate field dose of 1080 g/ha. The number of seeds (S n ) produced per plant was estimated as: where TS w denotes the total seed weight produced per plant and S w represents the averaged weight of three aliquots of 100 seeds per individual.

| Statistical analysis
Variations in plant survival and associated seed number production over an increasing gradient of glyphosate were analysed by dose-response model with the package drc (Ritz et al., 2015) in R (R-CoreTeam, 2017). The following three-parameter log-logistic function (LL.3 in drc) (Knezevic et al., 2007) was fitted to the data: where y denotes plant response (i.e. survival or seed number production) at glyphosate rate x, d is the upper limit, b is the slope at e which accounts for the glyphosate dose causing a decrease of 50% survival (LD 50 ) or seed number production (Ysn 50 ) between the upper limit d and lower limit. The model fitted with the function drm (dose-response model) (in drc, "dose-response curve" package in R) included glyphosate dose and genotype as independent variables. The glyphosate resistance index (RI) is calculated as the ratio of LD 50 (or Ysn 50 ) estimates between the resistant (R) (TIPS-RR or P106S-rr) and susceptible (S) (WT) genotypes (IR = LD 50 (R) / LD 50 (S)).
Experiments conducted to assess vegetative and reproductive response to the single recommended glyphosate dose (1080 g/ha) were subjected to two-way analysis of variance (ANOVA) to determine main genotype (WT, P106S-rr, TIPS-RR, TIPS-Rr) and glyphosate (control, 1080 g/ha) effects on traits, using InfoStat statistical software (Di Rienzo et al., 2020). Means were separated using Tukey's HSD (honestly significant difference) test (α = .05). ANOVA output is provided ( Figure S1).  fitted three-parameter log-logistic models in which the glyphosate concentration (×) was fixed at 1080 g/ha (eq. 2) (Figure 1). For the TIPS-Rr, mean survival was assessed from three pooled experiments in which plants were also exposed to glyphosate at 1080 g/ha under similar environmental conditions ( Figure 2). The number of seeds produced by P106S-rr, TIPS-RR and TIPS-Rr was estimated from dedicated experiments conducted with glyphosate at the field dose ( Figure 6), except for WT whose fecundity was estimated from the fitted regression model ( Figure 5).

| Predicted changes in the frequency of EPSPS alleles under glyphosate selection
Changes in the frequency of wt, r and R alleles were modelled assuming an environment under recurrent glyphosate selection with the field dose (1080 g/ha) over E. indica generations. Eleusine indica (2n= 2x =18) is an annual species and the prediction model assumed a single glyphosate treatment per generation or calendar year. Initial allelic frequencies were elaborated to resemble a glyphosate-susceptible population at the onset of the glyphosate selective process and therefore assumed as 0.999, 1 × 10 −6 and 1 × 10 −12 for the wt, r (single mutation) and R (double mutation) alleles, respectively. The assumed initial frequency of the nuclearencoded EPSPS r and R alleles are associated with theoretical and empirical estimations of herbicide resistance mutation frequencies (Casale et al., 2019;Diggle & Neve, 2001;Gressel & Segel, 1990;Jasieniuk et al., 1996).
The combination of the EPSPS wt, r and R alleles found in the field-collected E. indica population define six genotypes ( Figure   S2). As the heterozygous WT/r and WT/R genotypes were not detected in the field population, the matrix of relative fitness was completed assuming that the heterozygous genotypes for the glyphosate resistance r and R alleles exhibit incomplete dominance (0.5) compared with the estimated relative fitness of the homozygous WT and TIPS-RR genotypes (Huffman et al., 2016) ( Figure   S2).
The expected frequency of wt, r and R alleles in the next generation after glyphosate selection are REF (Allendorf et al., 2009): where f denotes de frequency of alleles wt, r and R, w is the relative fitness of the three homozygous WT, P106S-rr and TIPS-RR and three heterozygous WT/r, WT/R and TIPS-Rr genotypes and w is the average fitness of the E. indica population. Calculations were conducted assuming no overlapping generations in a large population, and absence of migration or back mutation.
The average fitness (w) of the E. indica population is calculated as the sum of the products of the frequency of each genotype multiplied by its relative fitness frequency (Allendorf et al., 2009): When the estimated frequencies of EPSPS wt, r and R allelic frequencies showed no further generational changes and reached equilibrium, the frequency of EPSPS genotypes were estimated using a model that accounts for a departure from Hardy-Weinberg due to nonrandom mating condition to reflect the E. indica inbreeding system (Futuyma, 2013). As the wt allele was predicted to be lost in the population after glyphosate selection (see Results), the genotypic frequencies were estimated using an EPSPS locus with the r and R alleles: where P106S-rr, TIPS-RR and TIPS-Rr denote the EPSPS genotypes, r and R are the predicted frequency of EPSPS glyphosate resistance alleles after glyphosate selection and F is the inbreeding coefficient.
A high inbreeding coefficient (F) was assumed (F = 0.97) to reflect the species reproductive biology.

| Plant survival
The logistic model provided a significant data fit (p < 0.0001) associated with variations in survival to increasing glyphosate doses for all genotypes. As expected, glyphosate dose treatments revealed differential survival of P106S-rr versus TIPS-RR genotypes (Figure 1).  Figure 2).

| Resource allocation to vegetative growth
Vegetative growth of glyphosate-surviving plants (1080 g/ha) showed significant differences between the genotypes. Surviving P106S-rr plants exhibited a major reduction in shoot/leaves ( Figure 3c) and root (Figure 3d) growth following glyphosate treatment compared with surviving TIPS-RR (Figure 3c,d). The impaired vegetative growth associated with the P106S-rr genotype was correlated with a lower relative growth rate (RGR) of shoots and roots over 3 weeks after glyphosate treatment which accounted for about 99% aerial and root biomass reduction compared with glyphosate untreated P106S-rr plants (Figure 3a,b). At plant maturity (i.e. 15 weeks after seed germination), P106S-rr plants showed some growth recovery from glyphosate, displaying a 32% reduction in total vegetative biomass compared with untreated plants (Figure 4a).

TIPS-RR plants showed a marginal decrease in RGR displaying
about 15% reduction in both the aboveground and root biomass compared with glyphosate untreated TIPS-RR plants 3 weeks after glyphosate treatment (Figure 3c,d vs 3a,b). At maturity, TIPS-RR showed a reduction in vegetative biomass that ranged from 20% to 13% compared with glyphosate untreated plants (Figure 4a,b).

Unlike homozygous TIPS-RR, heterozygous TIPS-Rr plants treated
with the glyphosate field dose (1080 g/ha) displayed no apparent reduction in vegetative growth as evidenced by the similar total biomass produced compared with untreated plants at maturity (Figure 4b).
It is noteworthy highlighting the notable growth reduction associated with TIPS-RR but not with TIPS-Rr when compared to WT in the absence of glyphosate treatment (Figures 3a,b and 4a,b), denoting an adaptive fitness cost associated with the homozygous TIPS mutation. Days since germination Root biomass (g plant -1 ) P106S = 0.01

| Resource allocation to reproduction
seed production showed by TIPS-RR plants was notably lower than WT and P106S-rr in the absence of glyphosate treatment and remained unaffected until glyphosate doses increased above 8500 g/ ha ( Figure 5). Thus, the Ysn 50 for TIPS-RR was 56,514 g/ha. The estimated resistance index based on seed production as compared to WT for the P106S-rr and TIPS-RR was P106S = 7.3 ± 1.5 and 113 ± 53, respectively.
A detrimental effect of glyphosate on reducing the reproductive potential of P106S-rr genotype was observed. P106S-rr plants showed a 55% reduction in the ability to produce seeds when exposed to the glyphosate field dose (x = 100,000 seeds) compared with untreated plants (x = 220,000 seeds) (Figure 6a). In contrast, TIPS-RR plants showed no significant reduction in the number of seeds produced when exposed to the field rate of glyphosate (x = 130,000 seeds) compared with untreated plants (x = 160,000 seeds). Conversely, seed number in TIPS-RR was notably lower than WT and P106S-rr in glyphosate free environment (Figure 6a).
Although showing a 27% reduction in the number of seeds

| Predicted changes in the frequency of EPSPS alleles under glyphosate selection
The frequency of the wt allele started to decline after three generations under glyphosate use coinciding with a steep increase in

| D ISCUSS I ON
The results of this study indicate that the glyphosate resistance sin-

| A higher glyphosate resistance benefit is associated with the TIPS than P106S mutation
Recent published studies have revealed that, at the enzyme level, the Escherichia coli expressed E. indica homozygous Pro-106-Ser and Thr-102-Ile/Pro-106-Ser TIPS mutations differ in their susceptibility to glyphosate inhibition (Yu et al., 2015). Whereas the glyphosate inhibitory concentration (IC 50 ) of Pro-106-Ser was 87 μM, the IC 50 value was c. 53,000 μM for the TIPS variant (Yu et al., 2015).
Similarly, various studies determining both the IC 50 and inhibition constant (K i ) demonstrate that the resistance EPSPS maize Pro-106-Ser variant is more glyphosate susceptible than the TIPS variant (Alibhai et al., 2010;T Funke et al., 2009;Robert Douglas Sammons & Gaines, 2014). Molecular modelling by Funke et al., (2009)

| Evolutionary and ecological significance of results
Understanding the evolutionary ecology of target site mutations endowing glyphosate resistance in agricultural weeds is fundamental to predict the trajectory of resistance evolution. The likelihood of spread and fixation of novel herbicide resistance mutations in agroecosystems depends on their impact on plant survival and fecundity (i.e. fitness) under both the presence (resistance benefit) or absence (resistance cost) of herbicide selection (see reviews by Bergelson & Purrington, 1996;Jasieniuk et al., 1996;Vila-Aiub et al., 2009, 2011.
It is clear that the value of herbicide resistance benefit plays an evolutionary role in favouring the rapid genetic fixation of the fittest resistance alleles. Hence, differences in resistance benefits among resistance traits define the rate of enrichment and equilibrium frequency in areas under herbicide selection.
In an environment under persistent glyphosate selection, our estimations predict that the EPSPS resistance r (Pro-106-Ser) and R (TIPS) glyphosate resistance-endowing alleles will increase their frequency, as expected, at the expense of the glyphosate susceptible wt allele which is nearly lost after 10 years of selection. However, despite the associated lower resistance benefit, the rate of enrichment of the r allele will be higher than the TIPS R allele for the first 10 years of glyphosate selection, driven by the likely higher initial frequency (1 × 10 −6 ) before selection compared with the rarer frequency of the TIPS R allele (1 × 10 −12 ). The TIPS R allele is predicted to reach the highest frequency of 0.53 after 20 years of glyphosate selection in stable equilibrium with the resistance r allele.
The studied E. indica population collected from the field has been shown to comprise 49% of plants with the heterozygous TIPS-Rr genotype, 34% P106S-rr, 16% WT and 1.6% homozygous TIPS-RR (Yu et al., 2015). However, our modelling exercise predicts that, in a highly inbreeding species like E. indica and after 20 years of glyphosate selection, only 2% of individuals would display the heterozygous TIPS-Rr genotype, despite the highest associated fitness benefit. In contrast, homozygous plants with the TIPS-RR and P106S-rr genotypes would comprise 52% and 46% of the population, respectively.
The discrepancies between observed and predicted EPSPS genotypic frequencies may represent unstable transient frequencies observed in the field that may lead to the loss of WT, and a small fraction of TIPS-Rr at the expense of TIPS-RR and P106S-rr over generations. Alternatively, the higher frequency of plants segregating at the Thr-102 position (TIPS-Rr) observed in the field may be the result of sequential mutational events occurring at higher rates than expected where the Thr-102-Ile mutation integrates in plants already harbouring the homozygous P106S-rr mutation (Sammons & Gaines, 2014;Yu et al., 2015).
If other alternative weed control tools are not implemented (Vila-Aiub, 2019), given the differences in survival, growth and reproductive fitness between TIPS-Rr, TIPS-RR and P106S-rr, and considering equally similar starting low frequencies in the field, any attempt to control E. indica populations harbouring these genotypes with increasing glyphosate doses would select and enrich faster for the TIPS-Rr followed by TIPS-RR and then P106S-rr genotypes.
Although the high inbreeding observed in E. indica would dilute the presence of the fittest heterozygous TIPS-Rr in favour of the homozygous TIPS-RR and P106S-rr.

ACK N OWLED G EM ENTS
This research was financially supported by the Australian Grains Research and Development Corporation (GRDC) and the Argentinean National Agency for Scientific and Technological Promotion (ANPCyT-FONCyT-PICT 2015/1544).

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.