Deep haplotype analyses of target-site resistance locus ACCase in blackgrass enabled by pool-based amplicon sequencing

Abstract

Rapid adaptation of weeds to herbicide applications in agriculture through resistance development is a widespread phenomenon. In particular, the grass Alopecurus myosuroides is an extremely problematic weed in cereal crops with the potential to manifest resistance in only a few generations. Target-site resistances (TSRs), with their strong phenotypic response, play an important role in this rapid adaptive response. Recently, using PacBio's long-read amplicon sequencing technology in hundreds of individuals, we were able to decipher the genomic context in which TSR mutations occur. However, sequencing individual amplicons are costly and time-consuming, thus impractical to implement for other resistance loci or applications. Alternatively, pool-based approaches overcome these limitations and provide reliable allele frequencies, although at the expense of not preserving haplotype information. In this proof-of-concept study, we sequenced with PacBio High Fidelity (HiFi) reads long-range amplicons (13.2 kb), encompassing the entire ACCase gene in pools of over 100 individuals, and resolved them into haplotypes using the clustering algorithm PacBio amplicon analysis (pbaa), a new application for pools in plants and other organisms. From these amplicon pools, we were able to recover most haplotypes from previously sequenced individuals of the same population. In addition, we analysed new pools from a Germany-wide collection of A. myosuroides populations and found that TSR mutations originating from soft sweeps of independent origin were common. Forward-in-time simulations indicate that TSR haplotypes will persist for decades even at relatively low frequencies and without selection, highlighting the importance of accurate measurement of TSR haplotype prevalence for weed management.

Keywords: ACCase; Alopecurus myosuroides; pbaa; Amplicon sequencing; HiFi long reads; herbicide resistance.

Figure 1

Workflow to generate and analyse long‐read amplicons in pools. (a) Leaf material is collected using a size template to ensure equal sample representation in each pool. Long‐range amplicon products are obtained by PCR with direct barcoding to individually tag each pool. Products are visualized by gel electrophoresis for quality control and validation of amplicon concentration measurements. (b) All population pools are combined in equal amounts in a single tube. A PacBio library is generated and sequenced in circular consensus mode on a Sequel II system. (c) Computational processing includes read‐consensus building, demultiplexing and filtering of raw reads. (d) pbaa clustering is used for variant detection and filtering (Kronenberg et al., 2021). The output is fasta files listing all haplotypes per population pool and including meta information on read coverage of each haplotype.

Figure 2

Unique ACCase haplotypes identified by pbaa in individuals compared to pools for the same population. (a) Maximum‐likelihood tree of haplotypes identified in the pool data set (200 samples), and haplotypes inferred in the individual data set (24 samples). Samples were collected in an agricultural field in Belgium (BE01585). For simplicity, in the individual data set, only unique haplotypes are shown. Tree labels indicate the data set of origin (Pool vs. Individual). Coloured tree tips show target‐site‐resistance (TSR) mutations. Curly brackets mark identical haplotype pairs found in both the individual and the pool data set from the same population. (b, c) Haplotype network representing the corresponding clade in the tree. pbaa can successfully recover haplotypes that differ only in one mutation (tick bar). (d) Haplotype counts per population in the individual data set. The number of haplotypes that could have been successfully identified in the pool data set is marked in green. Only a fraction of the low abundant ones could not be recovered (grey). (e) Correlation of haplotype frequencies in the pool data set versus the individual data set.

Figure 3

Comparison between conventional single‐nucleotide polymorphism (SNP) mapping and pbaa haplotype clustering. (a) TSR allele frequencies obtained by SNP mapping. Colours indicate different TSR mutations. (b) Haplotype frequencies were inferred using pbaa (Kronenberg et al., 2021). Colours refer to TSR and wild‐type haplotypes. (c) Correlation between allele frequencies and haplotype frequencies summarized per TSR amino acid position. Correlation coefficients and P values are shown separately in each TSR panel. BW, Baden‐Württemberg; NI, Lower Saxony; NW, North Rhine‐Westphalia; SH, Schleswig‐Holstein; SN, Saxony; ST, Saxony‐Anhalt; TH, Thuringia.

Figure 4

TSR haplotypes and the corresponding wild‐type haplotypes from which they arose in organic fields. (a, c, e) Wild‐type haplotypes, (b) a haplotype with the TSR mutation Ile1781.1Leu_T, (d) a haplotype with the TSR mutation Gly2096.2Ala_C, (f) a haplotype with the TSR mutation Trp2027.3Cys_T.

Figure 5

Simulations of the number of generations in which TSR alleles remain in A. myosuroides field populations in the absence of selection, assuming different selection coefficients, as estimated from fitness experiments (Du et al., ; Menchari et al., 2008). While homozygous individuals suffer the full consequences of deleterious TSR mutations, we simulated two different dominance coefficients for heterozygous allele states: an intermediate codominance of 0.5 (a–d) and a more recessive coefficient of 0.25 (d–h). The coloured numbers above the x‐axis indicate the average number of generations at which the mutations shown at the bottom are lost in the different scenarios. Means and 0.95 confidence intervals per parameter combination are shown.