Geographical variation in resistance to acetyl-coenzyme A carboxylase-inhibiting herbicides across the range of the arable weed Alopecurus myosuroides (black-grass)

Authors


Author for correspondence:
Christophe Délye
Tel: +33 380 693 185
Email: delye@dijon.inra.fr

Summary

  • The geographical structure of resistance to herbicides inhibiting acetyl-coenzyme A carboxylase (ACCase) was investigated in the weed Alopecurus myosuroides (black-grass) across its geographical range to gain insight into the process of plant adaptation in response to anthropogenic selective pressures occurring in agricultural ecosystems.
  • We analysed 297 populations distributed across six countries in A. myosuroides’ main area of occupancy. The frequencies of plants resistant to two broadly used ACCase inhibitors and of seven mutant, resistant ACCase alleles were assessed using bioassays and genotyping, respectively.
  • Most of the resistance was not endowed by mutant ACCase alleles. Resistance and ACCase allele distribution patterns were characterized by mosaicism. The prevalence of resistance and of ACCase alleles differed among countries.
  • Resistance clearly evolved by redundant evolution of a set of resistance alleles or genes, most of which remain unidentified. Resistance in A. myosuroides was shaped by variation in the herbicide selective pressure at both the individual field level and the national level.

Introduction

Plants are sessile organisms that cannot elude selective pressures and must consequently evolve adaptive traits to withstand them. Human activities, especially those related to agriculture and industry, have generated strong selective pressures that have promoted plant adaptation in a number of species. The consequences of geographical variation of anthropogenic selective pressures on plant adaptation have been investigated particularly in short-lived plants, for which adaptation occurs within a reasonable length of time. Such studies identified continuous variation in adaptation when anthropogenic selective pressures varied along a geographical gradient, as is the case for deposited atmospheric pollutants (Vergeer et al., 2008). On the other hand, adaptation to geographically discontinuous selective pressures, such as the occurrence of high concentrations of heavy metals at highly disjunct industrial sites, leads to multiple independent local adaptation events (Pauwels et al., 2005). Thus, geographical variation in the selective pressure intensity is a major causal factor of geographical variation in plant adaptive response, especially when gene flow connecting the populations is low (Linhart & Grant, 1996; Kawecki & Ebert, 2004).

Weed adaptation to agricultural ecosystems provides another intriguing system to investigate the variation of plant adaptive response over geography. Agricultural landscapes consist of a mosaic assembly of heterogeneous environments. In particular, arable fields are subjected to powerful anthropogenic disturbances such as soil cultivation or herbicide applications (Booth & Swanton, 2002). These disturbances vary abruptly in time and among fields. Yet, despite the drastic selective pressures exerted by agricultural practices, plant species referred to as weeds have successfully adapted to agricultural ecosystems. Weeds are essentially annual plant species that are abundant over broad geographical ranges (Baker, 1974). Because they compete with crop plants for environmental resources, weeds are destroyed by growers. The most effective way of destroying weeds is using herbicides, which can kill up to 98% of the sensitive weed plants (Foster et al., 1993) and are, accordingly, heavily relied upon by growers (Rüegg et al., 2007). Herbicides are thus a powerful driver of weed evolutionary adaptation.

Weeds can adapt to the herbicide selective pressure by evolving resistance mechanisms (for reviews, see Tranel & Wright, 2002; Délye, 2005). Herbicide resistance is an adaptive trait that confers a high fitness advantage in modern agricultural ecosystems, where it is crucial for the persistence of many weed populations. Evidence that herbicides can act as a driving force for the selection of resistant genotypes in the field has now been documented in over 190 weed species (Heap, 2009). The genetic basis of herbicide resistance has been the subject of numerous studies (Tranel & Wright, 2002; Délye, 2005); however, comparatively little is known about the geographical variation in herbicide resistance. Herbicide applications are designed by growers at the level of each individual field. As a consequence, herbicide selective pressure varies among agricultural fields, although the history of herbicide application may be documented. Growers also take into account herbicide use regulations and particularities of agriculture that vary among countries. Given that herbicide selective pressure varies from one field to another, local evolution of resistance is expected. Besides, as most weed species have geographical ranges spanning across different countries, regional or national variation in the herbicide selective pressure may also influence the evolution of resistance to herbicides. Evolution of resistance is therefore not a homogeneous process at the level of the species, and geographical heterogeneity is expected at different geographical scales. Yet, to our knowledge, no study has characterized geographical variation in herbicide resistance by sampling across the biogeographical range of a weed species, and only a handful have done so at a more local level (Warwick & Marriage, 1982; Menchari et al., 2006; Baucom & Mauricio, 2008).

Herein, we shall consider resistance of Alopecurus myosuroides (black-grass) to herbicides inhibiting acetyl-coenzyme A carboxylase (ACCase), one of the most intensively studied cases of resistance to herbicides. A. myosuroides is a diploid, mostly allogamous, wind-pollinated grass naturally occurring in Europe and the Mediterranean region (Fig. 1; Van Himme & Bulcke, 1975). It is the most important weed in central and northern Europe to have evolved herbicide resistance (Moss et al., 2007). ACCase inhibitors (reviewed in Délye, 2005) have been broadly used to control A. myosuroides across its geographical range in Europe since the end of the 1980s. As a consequence, by the early 2000s, numerous populations had evolved resistance to these herbicides (Délye et al., 2007; Moss et al., 2007). Two kinds of resistance mechanisms have evolved in A. myosuroides (reviewed in Délye, 2005). Target site-based resistance (TSR) in A. myosuroides is endowed by at least seven different mutant ACCase alleles, conferring a reduced sensitivity solely to ACCase inhibitors. Non target site-based resistance (NTSR) is governed by other genes causing a reduction in the number of ACCase inhibitor molecules reaching ACCase. NTSR can confer resistance to ACCase inhibitors as well as to herbicides with other modes of action. Its polygenic genetic basis is poorly understood.

Figure 1.

 Geographical distribution of Alopecurus myosuroides in Europe and western Asia (after Van Himme & Bulcke, 1975). The main area of occupancy of A. myosuroides is shaded, with a dotted border. Severe infestations of A. myosuroides outside this area are anecdotic. The continuous black line shows the limit of the species’ extent of occurrence. The countries where A. myosuroides populations were sampled in this study are France (Fra), the United Kingdom (UK), Belgium (Bel), the Netherlands (NL), Germany (Ger) and Turkey (Tur).

The biogeographical range of A. myosuroides encompasses different countries (Fig. 1) with differences in agriculture as well as in the use of ACCase inhibitors (Table 1). Previous studies investigating the geographical structure of resistance in A. myosuroides revealed independent appearances of mutant, herbicide-resistant ACCase alleles (Délye et al., 2004) and an absence of geographical structure of TSR with no isolation by distance at either a broad scale (France) or a local scale (a 120 × 150 km French region) (Menchari et al., 2006). Other studies conducted on the same populations showed NTSR was the predominant resistance mechanism in France (Délye et al., 2007), and suggested NTSR evolved in a manner similar to that known for TSR (Petit et al., 2010). All these studies suggested that geographical variation in A. myosuroides resistance to ACCase inhibitors was at a very local geographical scale, and was mostly shaped by local variation in herbicide application programmes. However, only populations from France were considered. National particularities of herbicide use arising from differences in herbicide regulation can be expected to influence local herbicide selective pressure. We thus hypothesized that there was a range-wide lack of geographical structure variation in herbicide resistance in A. myosuroides and that spatial heterogeneity in herbicide resistance was instead associated with variation in the herbicide selective pressure at both the local and national levels. Finally, we present the first biogeographical assessment of selection of TSR vs NTSR in response to locally varying herbicide selection pressure.

Table 1.   National variation in some agricultural characteristics in the Alopecurus myosuroides geographical range and global resistance features of A. myosuroides populations in each country
CountryAgricultural characteristicsACCase inhibitor useNo. populations investigatedAverage % of TSRaAverage resistance classb
IntensitySpecializationIntensityAssociation with other modes of actionFenoxapropClodinafop
  1. ACCase, acetyl-coenzyme A carboxylase; TSR, target site-based resistance; na, herbicide not assayed for populations from this country.

  2. aPercentage of ACCase alleles genotyped (see also Table 3).

  3. bResistance classes: class 1, 0% resistant plants; class 2, 1–20% resistant plants; class 3, 21–50% resistant plants; class 4, > 50% resistant plants. The average resistance class value is computed over all populations from a given country analysed with each herbicide (fenoxaprop or clodinafop). The number of populations analysed with each herbicide is given in brackets.

FranceHighCroppingHighRare6625.23.5 (66)2.9 (66)
UKHighCroppingHighFrequent8614.23.9 (84)3.7 (86)
BelgiumHighLivestockMediumFrequent411.83.2 (41)2.9 (33)
GermanyMediumCroppingMediumFrequent753.83.2 (75)2.6 (43)
The NetherlandsHighLivestockMediumFrequent142.12.6 (14)na
TurkeyMediumCroppingMediumFrequent156.61.5 (15)1.3 (10)

Materials and Methods

Alopecurus myosuroides populations

A total of 297 Alopecurus myosuroides Huds. populations collected across five European countries and Turkey (Table 1; Fig. 1) were used in this study. They were provided by H. Menne (Bayer CropScience, Frankfurt am Main, Germany; 234 populations) or M. Salas (DuPont Solutions SAS, Puteaux, France; 63 populations). The populations were collected in 2001 (15 populations), 2002 (83 populations), 2003 (122 populations), 2004 (68 populations) or 2005 (nine populations). They were collected in fields sown with winter wheat, a crop frequently infested by A. myosuroides. Each field was cultivated by a distinct farmer. The fields were selected, in the main, because growers reported poor control of A. myosuroides infestations by ACCase-inhibiting herbicides applied during the growing season. A single bulk seed sample issued from c. 50–75 plants was collected in each field as described by Menchari et al. (2006).

Herbicide sensitivity assessment

In Europe, selection for resistance to ACCase inhibitors in A. myosuroides has mostly been caused by two molecules, fenoxaprop and/or clodinafop. Both molecules can be degraded by enzymes of general plant metabolism (Tal et al., 1993; Frear & Swanson, 1996), and therefore select for both TSR and NTSR. The sensitivity of 100 A. myosuroides seedlings per population to each molecule was assessed using bioassays described elsewhere (Salas et al., 1999; Menne et al., 2008). Bioassays enable the detection of resistant plants regardless of the resistance mechanism (TSR or NTSR). The frequency of resistant plants for each population and herbicide combination was determined as previously described (Moss et al., 2007). Populations were subsequently assigned to one of the following four resistance classes according to the frequency of resistant plants observed: class 1, 0% resistant plants; class 2, 1–20% resistant plants; class 3, 21–50% resistant plants; class 4, > 50% resistant plants.

Genotyping mutations within ACCase

About 100 seeds from each population were allowed to germinate as previously described (Menchari et al., 2006). For each population, DNA was extracted, following a rapid procedure described elsewhere (Menchari et al., 2006), from 48 seedlings taken at random among those germinated. DNA extracts were diluted 50-fold and kept at −20°C before genotyping.

Seven mutant ACCase alleles endowing herbicide resistance are known in A. myosuroides (Supporting Information, Table S1). In the following, they will be referred to as C-L1781, T-L1781, C-C2027, T-C2027, N2041, G2078 and A2096. Genotyping was performed using the TaqMan® (Applied Biosystems, Foster City, CA, USA) technology (Holland et al., 1991). Each assay targeted the wild-type allele and one of the mutant alleles. PCRs were performed on Dual 384-Well GeneAmp® PCR System 9700 thermocyclers (Applied Biosystems) as previously described (Giancola et al., 2006) using FAM and VIC reporter dyes. Primer and probe sequences are given in Table S1. Results were analysed using the SDS 2.0 Software Workspace (Applied Biosystems).

Variation of resistance among countries

Variations in the distribution of populations into resistance classes in each country (TSR and NTSR) and in the prevalence of mutant ACCase alleles in populations in each country (TSR) were analysed using mosaic plots. These enable the visualization of contingency tables and aid in the detection of structure that may not be obvious from a purely numerical output, such as test statistics (Meyer et al., 2008). The observed values of a contingency table are pictured by area-proportional boxes arranged in a rectangular mosaic. Country was the first splitting variable used for the construction of mosaic plots. Independence or resistance classes or mutant allele frequencies among countries were tested using Pearson’s χ2-test. A colour-shading of the boxes based on the maximum of the absolute values of the Pearson residuals generated by 10 000 permutations (Meyer et al., 2008) allowed for the visualization of residuals from a given statistical test (Friendly, 1994; Meyer et al., 2008). Mosaic plots were constructed with the R software (http://www.r-project.org) using the vcd package (Meyer et al., 2006).

Local evolution of resistance within each country

Geographical distances (in km) were computed between all pairs of populations from the locations of the fields sampled. Spatial autocorrelation analyses of the distribution of populations into resistance classes to fenoxaprop or clodinafop were performed within each country. Isolation by distance among populations within each country was assessed using the frequencies of the eight ACCase alleles detected (one wild-type and seven mutant alleles). Genetic distances between all pairs of populations were computed as pairwise FST, estimated as Weir and Cockerham’s Fst (Weir & Cockerham, 1984). The relationship between genetic and geographical distances was assessed by the regression of Fst/(1 − Fst) on the natural logarithm (loge) of geographical distances, as proposed by Rousset (1997). Analyses were performed with the R software using the Ade4 package (Dray & Dufour, 2007).

Local and national variation in agricultural selective pressures

The record of cultural practices and herbicide applications was surveyed in most of the fields from which the 297 A. myosuroides populations studied were issued. The main cultural practices considered in the surveys were the crop rotation, the type of soil cultivation and the herbicide spraying programme. The correlations between the proportion of winter crops in the crop rotations, the diversity of crops in the crop rotation computed as Simpson’s dominance index (Simpson, 1949) or the frequencies of herbicide applications and the herbicide resistance class or the frequencies of the mutant ACCase alleles detected were estimated using the Spearman coefficient computed with the FREQ procedure in the SAS software (SAS Institute Inc., Cary, NC, USA). The frequency of herbicide application over the period surveyed was compared among countries using Student’s t-tests computed with Statistica 6.0. (StatSoft, Inc., Tulsa, OK, USA)

The proportion of winter crops in the crop rotations and the diversity of crops in rotations, computed as Simpson’s dominance index, were compared among countries using Student’s t-tests computed with Statistica 6.0. National agriculture and herbicide use data were obtained from the European Environment Agency or communicated by the following experts: Prof. Dr Peter Zwerger, Julius Kühn-Institut, Braunschweig, Germany; Mr Jean Marot, Comité Régional PHYTO, Louvain-la-Neuve, Belgium; Dr Stephen Moss, Rothamsted Research, Harpenden, UK; Dr Christian Gauvrit and Jacques Gasquez, INRA, Dijon, France.

Results

Prevalence of resistance (TSR + NTSR) and variation among countries

The sensitivity of 100 A. myosuroides seedlings to fenoxaprop was assessed in each of the 297 populations studied but two from the UK. The sensitivity of 100 A. myosuroides seedlings to clodinafop was assessed in 238 populations. It was not assessed in all 14 populations from the Netherlands and in 32, eight and five of the populations from Germany, Belgium and Turkey. Thus, 236 populations were analysed using both herbicides (Table 2). A total of 48, 57 and 30 populations were assigned to class 1 (0% resistant plants) for fenoxaprop, clodinafop and both herbicides, respectively. For both herbicides, resistance was therefore not generalized, although its prevalence was high: the majority of populations were assigned to class 4 (> 50% resistant plants), while very few populations were assigned to class 2 (1–20% resistant plants) (Table 2). However, the distribution of A. myosuroides populations within resistance classes was dissimilar between herbicides (Pearson’s χ2 test, χ2 = 8.3, P = 0.04), as a result of a lower frequency of populations in clodinafop resistance class 4 (Table 2). The distribution of populations across resistance classes was homogeneous among years of sample collection for both fenoxaprop (χ2 = 16.1, P = 0.19) and clodinafop (χ2 = 14.4, P = 0.27). For both herbicides, populations in each of resistance classes 1, 3 and 4 were present in each country.

Table 2.   Distribution of Alopecurus myosuroides populations across herbicide resistance classes
CountryNo. populations assayedFenoxaprop resistance classa% class 4 populations without mutant ACCase
1234
  1. ACCase, acetyl-coenzyme A carboxylase; na, herbicide not assayed for populations from this country.

  2. aResistance classes: class 1, 0% resistant plants; class 2, 1–20% resistant plants; class 3, 21–50% resistant plants; class 4, > 50% resistant plants.

  3. bPercentage of the populations assayed assigned to the resistance class.

  4. cPercentage of the populations in class 4 where no mutant ACCase alleles were detected.

France6615.2b0.06.178.717.3c
UK842.40.02.495.221.3
Belgium4117.10.024.458.566.7
Germany7514.78.021.356.061.9
The Netherlands1442.90.014.242.933.3
Turkey1580.00.06.713.350.0
All populations29516.32.011.969.834.5
  Clodinafop resistance classa 
1234
France6631.80.013.654.62.8
UK864.71.215.179.020.6
Belgium3327.30.024.248.575.0
Germany4332.614.014.039.447.1
The Netherlands0nanananana
Turkey1090.00.00.010.0100.0
All populations23823.92.915.158.126.1

The distribution of populations across resistance classes was heterogeneous among countries for both fenoxaprop (χ² = 110.7, < 0.001) and clodinafop (χ² = 73.8, < 0.001). Mosaic plot analysis revealed a highly significant excess of populations in resistance classes 1 and 2 in Turkey and Germany, respectively (Fig. 2a). For clodinafop (Fig. 2b), mosaic plot analysis revealed a highly significant excess of populations in resistance classes 1 and 2 in Turkey and Germany, respectively, as observed for fenoxaprop, plus a highly significant deficit in populations in resistance class 1 in the UK (Fig. 2b).

Figure 2.

 Mosaic plot analysis of the distribution of Alopecurus myosuroides populations into classes of resistance to fenoxaprop (a) or clodinafop (b) among samplings, and of the prevalence of mutant acetyl-coenzyme A carboxylase (ACCase) alleles among countries (c). Analyses were performed using country as the first splitting variable. The size of each box is proportional to the numbers of populations in each resistance class (a, b) or to the number of ACCase alleles detected (c). The boxes are colour-coded as follows: dark red, the observed proportion in a country is significantly much lower (< 0.01) than expected under the hypothesis of a homogeneous distribution among countries; light red, the observed proportion is lower than expected (0.01 < < 0.05); light blue, the observed proportion is higher than expected (0.01 < < 0.05); dark blue, the observed proportion is much higher than expected (< 0.01). Countries are Germany (Ger), Belgium (Bel), France (Fra), the Netherlands (NL), Turkey (Tur) and the United Kingdom (UK). The prevalence of resistance to clodinafop was not assessed in the populations from the Netherlands (b).

Populations in different resistance classes were often present in adjacent locations in every country (Fig. 3), which suggested some mosaicism in the pattern of resistance distribution. Autocorrelation analyses revealed no spatial structure in the distribution of populations into resistance classes to fenoxaprop or to clodinafop overall and within each country (not shown).

Figure 3.

 Geographical structure of resistance to the acetyl-coenzyme A carboxylase (ACCase) inhibitors fenoxaprop (a) and clodinafop (b). Each of the 297 Alopecurus myosuroides populations studied is located by a coloured dot. Dot colour indicates the resistance class: white, class 1 (0% resistant plants); yellow, class 2 (1–20% resistant plants); orange, class 3 (21–50% resistant plants); red, class 4 (> 50% resistant plants). Black dots, populations not analysed using herbicide sensitivity assays.

Prevalence of TSR and variation among countries

A total of 14 256 plants from the 297 populations were genotyped for the presence of the seven mutations within ACCase that endow TSR. Overall, wild-type alleles were detected in all populations. A total of 3234 (11.34%) mutant ACCase alleles were detected (Table 3). The two L1781 alleles, comprising in total 44.0% of the mutant alleles, were predominant (Table 3). Of the 297 populations analysed, 169 (56.9%) contained mutant ACCase allele(s). The number of populations containing each type of mutant allele did not differ significantly among alleles, except for alleles T-C2027 and G2078, which were detected in a lower number of populations (Table 3). Mutant alleles were observed at very variable frequencies among populations (Table 3), thereby suggesting a global heterogeneity in their distribution.

Table 3.   Occurrence of acetyl-coenzyme A carboxylase (ACCase) alleles among 297 European Alopecurus myosuroides populations
 Wild-type alleleMutant alleles
C-L1781T-L1781C-C2027T-C2027N2041G2078A2096
  1. nd, not detected.

  2. aIdentical superscript letters within a line indicate proportions that are not significantly distinct according to the two-sample test for equality of proportions with continuity correction at the 5% level after Bonferroni correction.

  3. bPercentage of the haplotypes present in a single population.

All populations
 Overall frequency (% of 28 512 haplotypes genotyped)88.641.98AB a3.02C0.77D1.32E2.11A0.45F1.71B
 % populations where detected (out of 297)100.019.9A21.5A12.5ABC6.4B20.9A9.8C21.2A
 Minimal frequency when detected (%)b9.01.01.01.01.01.01.01.0
 Maximal frequency (%)b100.077.084.046.068.078.039.091.0
France
 Overall frequency (% of 6336 haplotypes genotyped)74.800.74A6.85B0.32C5.23D4.92D1.29A5.85BD
 % populations where detected (out of 66)100.010.6A30.3A10.6A18.2A28.8A16.7A31.8A
 Minimal frequency when detected (%)b9.01.01.01.01.01.01.01.0
 Maximal frequency (%)b100.021.063.06.068.078.039.091.0
UK
 Overall frequency (% of 8256 haplotypes genotyped)85.795.02A3.59B1.65C0.53D2.44E0.34D0.64D
 % populations where detected (out of 86)100.041.9A39.5AB17.4BCD8.1CD29.1ABE16.3DE23.3ABCDE
 Minimal frequency when detected (%)b12.01.01.01.01.01.01.01.0
Maximal frequency (%)b100.074.051.029.019.032.04.014.0
Belgium
 Overall frequency (% of 3936 haplotypes genotyped)98.170.17A0.02A1.39Bnd0.15And0.10A
 % populations where detected (out of 41)100.09.8A2.4A19.5And9.8And9.8A
Minimal frequency when detected (%)b54.01.01.01.0nd1.0nd1.0
 Maximal frequency (%)b100.02.01.046.0nd3.0nd1.0
Germany
 Overall frequency (% of 7200 haplotypes genotyped)96.191.19A0.39BCD0.11CDnd1.05A0.28D0.79AB
 % populations where detected (out of 75)100.06.7A2.7A8.0And12.0A5.3A18.7A
 Minimal frequency when detected (%)b23.01.01.01.0nd1.01.01.0
 Maximal frequency (%)b10077.028.03.0nd32.017.044.0
The Netherlands
 Overall frequency (% of 1344 haplotypes genotyped)97.860.71AB1.07Andnd0.29ABnd0.07B
 % populations where detected (out of 14)100.042.9A28.6Andnd21.4And7.1A
 Minimal frequency when detected (%)b92.01.01.0ndnd1.0nd1.0
 Maximal frequency (%)b100.03.07.0ndnd2.0nd1.0
Turkey
 Overall frequency (% of 1440 haplotypes genotyped)93.400.13A6.00B0.07And0.20And0.20A
 % populations where detected (out of 15)100.06.7A20.0A6.7And13.3And20.0A
 Minimal frequency when detected (%)b14.02.02.01.0nd1.0nd1.0
 Maximal frequency (%)b100.02.084.01.0nd2.0nd1.0

Mutant ACCase allele distribution among countries was globally heterogeneous (χ2 = 4121.0, < 0.001). Mutant ACCase allele prevalence was highest in France (25.2% of the ACCase alleles detected) and lowest in Belgium (1.8% of the ACCase alleles detected; Table 3). Mosaic plot analysis confirmed that populations from Germany, Belgium and the Netherlands displayed a highly significant excess of wild-type alleles together with a significant deficit in most of the mutant alleles (Fig. 2c). Populations from the UK displayed a highly significant excess in C-L1781 and C-C2027 alleles. Populations from France displayed a highly significant excess in T-L1781, T-C2027, N2041, G2078 and A2096 alleles, plus a highly significant deficit in wild-type alleles. Populations from Turkey displayed a highly significant excess in T-L1781 allele.

As observed when considering the distribution of populations across resistance classes, the pattern of ACCase allele distribution in each country was characterized by mosaicism (Fig. 4). Isolation by distance was tested globally for all countries except for Turkey, which is geographically disjunct from the other countries, and within each country using the frequencies of the mutant ACCase alleles detected. Overall, a significant relationship was observed between pairwise geographical distances and Fst values (Fig. S1) (F = 32.73; P = 1.07 × 10−8) but this was not biologically interpretable (adjusted r2 = 8 × 10−4). Within countries, no significant relationship was observed between pairwise geographic distances and Fst values for Turkey (F = 0.8086, = 0.37), France (= 0.4682, = 0.49), the Netherlands (= 0.2966, = 0.59), Germany (= 0.1424, = 0.71) and Belgium (= 0.0945, = 0.76). A significant relationship was observed for the UK (= 48.5, = 3.86 × 10−12) but was not biologically interpretable (adjusted r2 = 0.013).

Figure 4.

 Geographical distribution of mutant acetyl-coenzyme A carboxylase (ACCase) alleles among the 297 Alopecurus myosuroides populations studied. Each population is located by a pie chart positioned as close as possible to its geographical origin. Pie charts show the respective frequencies of the mutant, resistant ACCase alleles detected in populations. White, wild-type allele; yellow, C-L1781; orange, T-L1781; red, C-C2027; purple, T-C2027; green, N2041; blue, G2078; grey, A2096. For clarity, populations where no mutant resistant ACCase allele was detected are shown as small, white circles.

Respective roles of TSR and NTSR in resistance

No mutant ACCase allele was detected in 95 of the 241 populations in fenoxaprop resistance classes 3 and 4, and in 52 of the 174 populations in clodinafop resistance classes 3 and 4 (Tables 1, 2). Populations in resistance class 3 or 4 to fenoxaprop or clodinafop that were devoid of mutant ACCase alleles were predominant in Belgium, Germany and Turkey (approximately two-thirds of the class 3 or 4 populations) but were comparatively rare in the UK (< 25% of the class 3 or 4 populations) and, in particular, in France (< 20% of the class 3 or 4 populations) (Table 2, Fig. S2).

History of the populations

Records of cultural practices and herbicide applications were obtained for 232 of the 297 fields sampled. Records ranged from 1 to 7 yr, with only 12 records ranging across >5 yr. The quality of the records was extremely heterogeneous, especially with respect to soil cultivation. We thus focused on crop rotations and herbicide spraying programmes. We only considered surveys ranging across 4 yr or more. With these criteria, only 94 surveys were analysed. They consisted of 59, 26, six and three surveys from France, Germany, Belgium and the UK, respectively. No survey from Turkey or the Netherlands was complete enough to be analysed. Because of the low number of interpretable surveys in the other countries, we only considered surveys from Germany and France.

Because only 10 of the 26 German populations considered were assayed with clodinafop, we estimated the correlations of the proportion of winter crops that favour A. myosuroides infestations in the crop rotations, the diversity of crops in the crop rotation computed as Simpson’s dominance index and of the frequencies of herbicide applications with the fenoxaprop resistance class and with the frequency of the mutant ACCase alleles detected. Analyses were conducted for all populations, and for populations in each country. No significant relationship was found between the proportion of winter crops in the crop rotations and resistance, nor between the dominance index and resistance.

No significant differences were observed between French and German fields in the proportion of winter crops in the crop rotations (= −1.78, = 0.079). Similarly, the diversity of crops in the crop rotation across the period recorded in the surveys computed as Simpson’s dominance index was not different between the two countries (= −0.17, = 0.866).

No significant relationship was found between resistance and the frequency of herbicide application considering all modes of action applied. Overall, the prevalence of mutant ACCase alleles and the fenoxaprop resistance class were found to increase with the mean number of ACCase inhibitor applications per year (= 0.344, < 0.001 and = 0.3, = 0.0024, respectively). This was mostly because of the 59 populations from France. For these populations, the prevalence of mutant ACCase alleles and the fenoxaprop resistance class were found to increase with the number of ACCase inhibitor applications (= 0.463, < 0.001 and = 0.339, = 0.0098, respectively). By contrast, for the 26 populations from Germany, no correlation was detected between the prevalence of mutant ACCase alleles or the fenoxaprop resistance class and the number of ACCase inhibitor applications (= 0.211, = 0.291 and = 0.055, = 0.783, respectively).

The number of applications per year of ACCase inhibitors over the period surveyed was not different between France (0.54) and Germany (0.48) (= −0.92, = 0.362). The total number of herbicide applications per year over the period surveyed was also not different between France (1.38) and Germany (1.25) (= −0.99, = 0.325).

Discussion

Herbicide resistance is an adaptive trait evolved by weeds in response to human action. It can vary within and among populations, regionally or at a broader geographical scale. Our aim was to determine the spatial level(s) at which variation in resistance could be detected across the A. myosuroides range, and, if possible, to link resistance variation with variation in agricultural selective pressure.

Previous studies compared the frequency of A. myosuroides plants resistant to fenoxaprop or clodinafop in a large number of populations collected either in fields selected at random or in fields where resistance was suspected. Similarly high frequencies of resistant plants were identified by the two sampling strategies (Délye et al., 2007; Moss et al., 2007). Thus, although the populations studied herein were mostly sampled in fields where resistance to ACCase inhibitors was suspected, they are representative of a random sample of populations from across the A. myosuroides range.

To date, the geographical structure of herbicide resistance has only been investigated at a local level, and only in a few studies (Warwick & Marriage, 1982; Menchari et al., 2006; Baucom & Mauricio, 2008). Several other studies addressed the prevalence of weed resistance to herbicides at broader scales, although not across the full range of a weed species (Bourgeois et al., 1997; Beckie et al., 1999; Owen et al., 2007; Owen & Powles, 2009). Yet, they did not consider herbicide resistance geographical structure. The comprehensive geographical scale across which herbicide resistance structure was characterized in A. myosuroides makes it possible for the first time to obtain a global overview of how resistance to herbicides evolves in an arable weed species.

Respective roles of TSR and NTSR in resistance

Mutant ACCase alleles only represented a minor fraction of the genes involved in resistance to ACCase inhibitors (Table 2, Fig. S2). NTSR mechanisms clearly endow most of this resistance. In a previous study considering 243 A. myosuroides populations from France, > 75% of the plants analysed were resistant to ACCase inhibitors because of NTSR mechanisms (Délye et al., 2007). Our study clearly shows that the role played by NTSR mechanisms is still more important in the rest of the A. myosuroides range.

Herbicide applications will strongly select for any mutation(s) enabling A. myosuroides plants to survive. Thus, at the initial stage of the selection for resistance, it is likely that many such mutations will be selected for in A. myosuroides. Because TSR to ACCase inhibitors is governed by a few alleles at the ACCase gene, while NTSR is considered to be governed by a large number of genes (Délye, 2005), most of the mutations selected for by herbicides are expected to cause NTSR. Besides, TSR to ACCase inhibitors can only be selected for by these herbicides, while NTSR can be selected for by herbicides with diverse modes of action (Délye, 2005). The tradeoff between the selective advantage and the fitness cost associated with each mutation considering the variation in time and space of the herbicide selective pressure will ultimately determine the respective prevalence of TSR and NTSR as well as the nature of the predominant mutations involved in each type of resistance. Whether the evolution of resistance resulted from the selection of new, spontaneous mutations or from A. myosuroides’ standing genetic variation remains to be investigated.

Local variation in resistance in A. myosuroides

At the spatial scale of our study, we observed an absence of structure in the distribution of the prevalence of resistance to fenoxaprop or clodinafop and in the spatial distribution of mutant ACCase alleles. In particular, we did not observe any pattern of isolation by distance, which would be expected if resistance had spread from a limited number of initial sources. Resistance to ACCase inhibitors instead varied according to a mosaic pattern throughout the A. myosuroides range, with sensitive populations neighbouring resistant ones. This suggested that resistance evolved independently within each field sampled.

An obvious possible cause for the mosaic structure of resistance would be a spatial heterogeneity of anthropogenic selective pressure, that is, cultural practices (Warwick, 1991). Cultural practices vary with the individual field, and this variation has been shown to have an influence on the evolution of resistance to ACCase inhibitors (Beckie et al., 2008). By contrast, an overall effect of cultural practices on resistance evolution was not detected. However, we did detect a positive correlation between the frequency of ACCase inhibitor applications and the prevalences of mutant ACCase alleles and of resistant plants. Previous works also showed that the frequency of ACCase inhibitor applications was the major factor explaining the prevalence of resistance to these herbicides in weed populations (Légère et al., 2000; Beckie et al., 2002; Délye et al., 2007; Menne et al., 2008). Our results thus suggest that herbicide resistance in A. myosuroides is primarily shaped by variation in the herbicide selective pressure at the scale of the individual field.

Redundancy in the evolution of resistance

Previous results obtained across a 120 × 150 km French area (Menchari et al., 2006) and across France (Délye et al., 2004) showed that TSR evolved by multiple, local, independent appearances of herbicide-resistant ACCase alleles. Our results confirmed this finding, and extended it to the whole A. myosuroides range. Adaptation via redundant evolution (Nosil et al., 2009) can be expected in geographically isolated plant populations undergoing similar selective pressures. It has, for instance, been proposed in the case of adaptation to geographically disjunct areas where anthropogenic pollutants exert a selection on plant populations (Pauwels et al., 2005). The dispersal of A. myosuroides seems to be mostly limited, with gene flow occurring within, at most, a few kilometers (Cavan et al., 1998; Menchari et al., 2006, 2007). Here, 98.9% of the distances separating two A. myosuroides populations were >20 km. At the timescale considered (c. 20 yr since the commercialization of ACCase inhibitors), these populations can thus be considered mostly independent, with each one evolving according to its own evolutionary trajectory. Evolution of TSR in A. myosuroides in different localities involved one or several of the same seven mutant ACCase alleles (Fig. 4, Table 3). Recent results suggested that NTSR evolved in a similar fashion (Petit et al., 2010). Thus, herbicide resistance in A. myosuroides probably evolved according to a replicated pattern, that is, via redundant evolution (Nosil et al., 2009). Redundant evolution had not previously been demonstrated to occur throughout the geographical range of a weed species having evolved resistance to herbicides.

Furthermore, as a variety of genes can confer resistance to ACCase inhibitors (Délye, 2005), resistance in A. myosuroides is also characterized by a high genetic redundancy (Goldstein & Holsinger, 1992), that is, the same resistant phenotypes can be produced by different genotypes. This implies cryptic genetic variation sensuPhillips (1996) in resistance of A. myosuroides to ACCase inhibitors, especially in NTSR. If A. myosuroides populations were to exchange migrants, then mixing of the different local resistant genotypes would generate new combinations of resistance genes, and likely novel patterns of resistance at the individual plant level.

The role of national particularities of agriculture

All known mutant ACCase alleles confer resistance to both fenoxaprop and clodinafop (Délye, 2005; Délye et al., 2008). However, the two areas where these alleles are most abundant (southwestern UK and especially central and eastern France; Fig. 4) did not exactly coincide with the area of highest resistance (Fig. 3). This suggests that the kind of resistance mechanisms selected for by herbicide pressure varies across broad regions in the geographical range of A. myosuroides. This is supported by our results showing clear differences in the prevalence of resistance to ACCase inhibitors and in the prevalence of mutant ACCase alleles among countries (Tables 2, 3; Fig. 2). Besides, the correlations detected between the frequency of ACCase inhibitor applications and the respective prevalences of mutant ACCase alleles and resistant plants in A. myosuroides populations were strong for populations from France but nonsignificant for populations from Germany, where the prevalence of mutant ACCase alleles was very low (Table 3). These differences were observed although a similar average frequency of ACCase inhibitor applications was recorded over a 4–7 yr period in each sampling. ACCase inhibitors were first marketed in the late 1970s in France and Germany, > 20 yr before the beginning of this study. As the early stages of selection are considered crucial to shape resistance (Neve et al., 2009), the period recorded in the surveys (most often 4–5 yr) is likely too short for the herbicide application records to fully explain the differences in resistance structure observed.

As discussed in the preceding section, cultural practices vary from one field to the next. They also vary with the intensity of agriculture. In this respect, the area of highest resistance prevalence in our samples collected in the early 2000s across A. myosuroides’ main area of occupancy in Europe superimposes nicely with areas of intensive agriculture in Europe at that time (Fig. S3). Yet, within areas of intensive agriculture, the respective roles of TSR and NTSR in resistance vary by country (Table 2, Fig. S2).

Acetyl-coenzyme A carboxylase inhibitors have been in use for c. 15–20 yr in all countries where samplings were performed. However, the way they were used, and the intensity, differed among countries (Table 1, Fig. S4). These differences are reflected in the structure of resistance to herbicides in A. myosuroides. In France and in the UK, where crop production is important (Fig. S5) and ACCase inhibitor use is intensive, a high prevalence of resistance is observed (Table 1). The higher prevalence of TSR observed in France is consistent with ACCase inhibitors having been used intensively alone against A. myosuroides while they were mostly used in association with a range of other molecules in the UK (Caseley et al., 1991; Table 1). NTSR mechanisms can confer resistance to herbicides with diverse modes of action (Délye, 2005), and such mechanisms have long been known to occur in A. myosuroides in the UK following intensive use of urea herbicides (Moss & Cussans, 1985). This is consistent with A. myosuroides showing a higher prevalence of resistance to ACCase inhibitors but a lower prevalence of mutant ACCase alleles in the UK than in France.

A situation similar to that in the UK occurred in Belgium, except that herbicide selective pressure was likely more moderate because farmers in Belgium specialize mainly in livestock production (Fig. S5), and hence there is a less stringent need for weed-free cereals. This is consistent with the lower, though substantial, prevalence of resistance to ACCase inhibitors essentially endowed by NTSR mechanisms in Belgium. In Germany, the Netherlands and Turkey, ACCase inhibitor use was lower than in the other countries (Fig. S4) and in association with other herbicides (Table 1), but urea herbicides were abundantly used and selected for NTSR (Niemann & Pestemer, 1984). This is consistent with the relatively low prevalence of resistance to ACCase inhibitors essentially endowed by NTSR mechanisms observed in these countries.

A previous study conducted at a local geographical scale suggested that the variation in resistance to herbicides could depend on both the individual field and a broader geographical scale (Baucom & Mauricio, 2008). Similarly, our results and the data about national particularities in herbicide use suggest that both variation in the herbicide selective pressure at the scale of the individual field, and national variation in herbicide use are important factors in shaping the evolution of herbicide resistance in A. myosuroides.

Concluding remarks

Unlike other, previously studied anthropogenic selective pressures, such as pollutants, herbicides exert a mosaic of local selective pressures, the nature and intensity of which vary among individual fields, with some homogeneous features at the level of countries. The geographical structure of resistance in A. myosuroides can thus be explained by variation in local cultural practices, essentially herbicide spraying programmes, which are influenced by the national particularities of agriculture. To our knowledge, no previous study has reported national particularities leaving their mark on the geographical structure of an adaptive trait in a plant. This result highlights the importance of considering the country of origin of weed populations where management practices have been designed, because using the response of a weed in one country as a model for weed evolution or weed management decisions in another country with a different agricultural history could be unreliable.

Another important finding in this study is the minor role played by mutant ACCase alleles in resistance to ACCase inhibitors, especially outside France. NTSR clearly endows nearly all of A. myosuroides resistance to these herbicides. Hardly any data are available concerning the genes governing this kind of resistance, although the differences we observed between the prevalences of resistance to fenoxaprop and to clodinafop confirm previous findings that NTSR does not confer systematic cross-resistance to both herbicides (Petit et al., 2010). To completely understand A. myosuroides adaptation to the selective pressure of herbicides, it is crucial to identify the genes endowing NTSR, bearing in mind that such genes are likely under redundant selection, and that regional or national differences in herbicide use probably influenced the kind of genes selected for at the regional or national level.

Acknowledgements

The authors are grateful to H. Menne (Bayer CropScience AG, Frankfurt am Main, Germany) and M. Salas (DuPont Solutions SAS, Puteaux, France) for the gift of the A. myosuroides populations and population data, to all colleagues mentioned in the Materials and Methods section for sharing their national expertise, and to two anonymous reviewers for judicious comments. This work was partly supported by a grant from INRA, Direction Scientifique Plante et Produits du Végétal, France.

Ancillary

Advertisement