Standing genetic variation fuels rapid evolution of herbicide resistance in blackgrass

Abstract

Repeated herbicide applications in agricultural fields exert strong selection on weeds such as blackgrass (Alopecurus myosuroides), which is a major threat for temperate climate cereal crops. This inadvertent selection pressure provides an opportunity for investigating the underlying genetic mechanisms and evolutionary processes of rapid adaptation, which can occur both through mutations in the direct targets of herbicides and through changes in other, often metabolic, pathways, known as non-target-site resistance. How much target-site resistance (TSR) relies on de novo mutations vs. standing variation is important for developing strategies to manage herbicide resistance. We first generated a chromosome-level reference genome for A. myosuroides for population genomic studies of herbicide resistance and genome-wide diversity across Europe in this species. Next, through empirical data in the form of highly accurate long-read amplicons of alleles encoding acetyl-CoA carboxylase (ACCase) and acetolactate synthase (ALS) variants, we showed that most populations with resistance due to TSR mutations-23 out of 27 and six out of nine populations for ACCase and ALS, respectively-contained at least two TSR haplotypes, indicating that soft sweeps are the norm. Finally, through forward-in-time simulations, we inferred that TSR is likely to mainly result from standing genetic variation, with only a minor role for de novo mutations.

Keywords: Alopecurus myosuroides; blackgrass; herbicide resistance; rapid adaptation.

Fig. 1.

Reference genome of an A. myosuroides individual from a German herbicide-sensitive population. (A) Histogram of relative DNA content from flow cytometry of propidium iodide-stained nuclei of A. myosuroides and the reference standard Secale cereale cv. Daňkovské (diploid genome size = 16.9 pg). (B) Circos plot of the A. myosuroides genome, with colored lines connecting anchor pairs (genes in collinear regions) with synonymous substitution rates (K_S) > 0.5. Numbers represent megabases. (C) K_S distributions for paralogs within the A. myosuroides, Hordeum vulgare (23), and Oryza sativa (24) genomes and for ortholog pairs shared by the three species. Divergence time, expressed as Mya, was estimated based on 7.0 × 10⁻⁹ as the substitution rate in grasses (25). (D) Syntenic relationships between the chromosomes of A. myosuroides and H. vulgare (Top) and O. sativa (Bottom).

Fig. 2.

Population structure analysis of 47 European A. myosuroides populations with 109,924 genome-wide ddRAD-seq markers. (A) Maximum-likelihood tree. Branch ends are marked in country colors. TSR mutations for ALS and ACCase in each individual are indicated in the outer and inner rings, respectively. (B) Principal component analysis (PCA) showing the first two eigenvectors, with explained genetic variance in parentheses. Colors reflect country-specific origin of the populations. (C) Heatmap of fixation index (F_ST) values in contrast between the different populations, ranging from close to 0 (blue) to 0.06 (dark red). Populations are clustered by similarity of F_ST patterns, and colors of the branch tips indicate countries of origin of the populations. (D) Admixture proportions with ancestry groups of K = 9. Admixture proportions with ancestry groups of K = 7 and K = 8 can be found in SI Appendix, Fig. S4. Each bar corresponds to one individual, grouped by country [Austria (AT), Belgium (BE), Switzerland (CH), Germany (DE), France (FR), Luxembourg (LX), Netherlands (NL), Poland (PL), United Kingdom (UK)].

Fig. 3.

Haplotype analysis of the complete ACCase gene (13.2 kb). (A) Network and maximum likelihood (ML)-tree of 44 haplotypes from the sensitive reference population UK01413 (HerbiSeed standard), which has not been under herbicide selection. The color code in all networks and trees (A–C) indicates different target-site resistance (TSR) mutations, with haplotypes that have wild-type sequence at known TSR positions in green. Likely wild-type haplotypes of origin for TSR mutations are indicated (WT). (B) Network and ML-tree of 44 haplotypes from the British population UK06481, which shows a selection pattern characteristic of an emerging soft sweep for TSR mutations. (C) Network and ML-tree of 46 haplotypes from the French population FR03200, with a predominant soft sweep pattern for the TSR mutation Ile2041.2_Asn_A. (D) Schematic representation of alternative origins of soft sweep patterns: recombination vs. independent mutation events. (E) In the FR03200 population, two distinct wild-type haplotypes (WT 1 and WT 2) have independently sustained the same TSR mutation, giving rise to haplotypes TSR 1 and 2. In addition, wild-type haplotype WT 2 has given rise to a second TSR haplotype (TSR 3). Positions of TSR Ile1781.1_Leu_T and TSR Ile2041.2_Asn_A mutations are marked with red and blue circles, respectively.

Fig. 4.

Simulations of different scenarios for adaptation. Equations from Hermisson and Pennings (34) were used to derive (A) expectations for the probability of adaptation via a selective sweep from beneficial target-site resistance (TSR) mutations in general and (B) from standing genetic variation in particular. Probabilities of sweeps with different effective population sizes (N_e) are estimated as a function of the strength of selection.

Fig. 5.

Simulations of expected allele frequencies for TSR alleles arising from standing genetic variation or de novo mutation. The distribution of mutations was generated with a generic gene model that has the same number of exons and introns, and ratio between coding and noncoding sequences as the ACCase gene. Mutations in introns and noncoding regions were considered to be neutral, while exons had a ratio of 0.25/0.75 (neutral/deleterious) mutations according to Messer and Petrov (52), with selection coefficients (s) for deleterious mutations drawn from a gamma distribution with E[s] = −0.000154 and a shape parameter of 0.245 (51). Five hundred of one thousand simulation runs are shown for an effective population size (N_e) of (A and B) 42,000 individuals and (C and D) 84,000 individuals. Continuous lines represent mutations originating from standing genetic variation; de novo TSR mutations are shown with dashed lines. Colors indicate the total number of TSR mutations per population. (A and C) Standing genetic variation scenario, with TSR mutations preexisting in the populations before herbicide selection. Shown is the increase in TSR allele frequencies under herbicide selection of up to 30 generations, with one herbicide application per generation. The right panel shows a truncated Y axis at 0.005 TSR allele frequencies. Notice that some TSR de novo mutations have also arisen in runs that had preexisting TSR alleles. (B and D) De novo mutation scenario. Any TSR mutation that might have arisen before the start of selection has been lost again, so that no TSR mutations are present at generation 0 of selection.