DISCUSSION
Genomic-based epidemiology using markers specific to a characterized resistant haplotype of EPSPS duplication from the Central Great Plains identified three unique EPSPS haplotypes in GR kochia populations from across western North America. Haplotypes 1 and 2 both showed increased copy numbers of the MGE previously identified, but haplotype 2 does not have insertion of the MGE next to EPSPS at the same position as identified in haplotype 1. EPSPS duplication in haplotype 2 may have occurred through a similar genetic mechanism as haplotype 1 involving insertion of an active mobile genetic element next to EPSPS followed by tandem duplication. Resequencing and assembly of the EPSPS locus in haplotype 2 will be needed to determine the precise duplication mechanism that occurred and to determine whether the MGE, which had increased copy number in haplotype 2, is associated with the EPSPS duplication. We consider haplotype 3 to represent a third independent origin because it had increased EPSPS copy number and no increase in MGE, indicating that EPSPS duplication may also have occurred through yet another mechanism such as a double-strand break initiated by a different MGE. These haplotypes, based on either the presence or absence of type I and type II elements and ratio of MGE, provide evidence that glyphosate resistance evolved multiple times in geographically distinct locations, with three independent origins supported by the data.
Haplotype 3 was found in populations from OR, ID, and WY. From the neighbor-joining tree, population WY1R (Northern Plains) was closest to OR5R (Pacific Northwest) with support of 41% (Figure 2), but this pair was distant from the other Pacific Northwest populations and not clustered with Pacific Northwest populations in the STRUCTURE plot (Figure 3). The shared haplotype 3 between geographically isolated northern Wyoming and the Pacific Northwest could indicate two separate origins of haplotype 3 (increased-EPSPS without increased MGE), or it could indicate lack of resolution in the population genetics data to resolve gene flow from independent origins of resistance. Further characterization of the duplicated resistant locus is needed to determine whether populations from the Northern Plains (WY1R) and the Pacific Northwest have a shared or separate origin of duplicatedEPSPS genes. Other Northern Plains populations (WY and southern MT) contained individuals with either haplotype 2 or haplotype 3, indicating gene flow via migration between populations and/or dynamic MGE changes in copy number over time.
We predicted that population genetics analysis would show clear geographical structure if the three glyphosate resistance evolutionary events occurred and were followed by rapid dispersal and introgression within a region. Aside from a large grouping of Central Great Plains populations and a second grouping of Pacific Northwest populations, we were not able to identify clear geographical structure for the three regions corresponding to the three EPSPS haplotypes. The STRUCTURE analysis supported K=3, although populations were not consistently assigned to three groups corresponding geographically to the regions containing the three EPSPS haplotypes and most populations contained some presence of all three groups. Some strong signals of relatedness were detected between populations from geographically isolated locations, such as OR4R and KS10R (Figures 2, 3). While we consider it to be unlikely that the same duplication mechanism is evolving independently multiple times within a region on different genetic backgrounds, the SSR data may have insufficient resolution to identify population genetic relationships and extensive long-distance gene flow via seeds may make regional differences harder to detect. This aligns with previous population genetics studies in kochia that have found high genetic diversity within individuals and little population structure (Friesen et al., 2009; Kumar et al., 2019; Mengistu & Messersmith, 2002).
Haplotype distribution was consistent with division by mountainous geographic barriers, with haplotype 1 distributed across the Central Great Plains associated with the earliest report of glyphosate resistance in kochia, 2007, in Kansas (Waite et al., 2013); haplotype 2 located within the Northern Plains including northern Montana and Alberta; and haplotype 3 located in sugar beet fields in the Pacific Northwest and north-central Wyoming (Figure 1). Since glyphosate resistance evolved recently, the geographic patterns of haplotype distribution (via seed dispersal) also occurred more recently and appear to be introgressing at a more regional scale into diverse genetic backgrounds (via pollen flow).
The presence of more than one EPSPS haplotype within populations provides strong evidence for gene flow among populations. This was observed in some northern Colorado populations, with some haplotype 2 individuals present in populations that were mostly haplotype 1 (e.g., CO3R from Cope, CO showing high EPSPS , no Type I or II, and high MGE, like samples from northern Montana and Alberta), and some haplotype 3 individuals present in populations mostly containing haplotype 1 (Eaton, CO) (Figure 1, Table S1). These two populations showing admixture in CO indicates gene flow has occurred in this area.
A curious result in the population relatedness data was that several populations from Texas (TX3R, TXR4, TX5R), a population from Oregon (OR1R), and the furthest north population from Alberta, Canada (AB1R) were grouped with relatively low (>20%) bootstrap support in the phylogeny tree (Figure 2). Further research will be needed to define the relationship of these populations and to determine whether long-distance gene flow is occurring in kochia, either through natural tumbling dispersal or through human-mediated seed migration. Some other weed species also demonstrate little population structure. Palmer amaranth (Amaranthus palmeri ), an obligate outcrossing weed species subjected to widespread glyphosate selection pressure, also maintains high levels of genetic diversity and little population structure (Küpper et al., 2018; Molin, Patterson, & Saski, 2020a). Kochia has protogynous flowers and pollen that is both wind and insect dispersed (Blackwell & Powell, 1981), facilitating high outcrossing levels. Kochia is also one of the primary tumbleweed species of the western US and Canada. This uncommon mode of seed dispersal may serve to further homogenize genetic diversity across long geographic distances.
Single origins of herbicide resistance followed by substantial geographical distribution by pollen and seed mediated gene flow is known to have a major contribution to resistance frequency in multiple weed species (Beckie et al., 2019). Multiple independent origins of glyphosate resistance with little population structure was found in glyphosate-resistant populations of common morning glory (Ipomoea purpurea ) (Kuester, Chang, & Baucom, 2015). A genomics-based approach in the same species found evidence for parallel genetic responses in genomic regions encoding herbicide detoxification in, while other genomic regions showed divergent patterns of selection (Van Etten, Lee, Chang, & Baucom, 2020). The high sequence similarity of the extrachromosomal DNA containing the EPSPS gene in Palmer amaranth (Koo et al., 2018; Molin, Yaguchi, Blenner, & Saski, 2020b) across widespread populations supports the hypothesis of a single origin followed by dispersal (Molin et al., 2020a; Molin, Wright, Lawton-Rauh, & Saski, 2017; Molin et al., 2018). Glyphosate-resistant populations of flaxleaf fleabane from across multiple Australian states were highly related, supporting a high frequency of seed movement (Minati, Preston, & Malone, 2020). Multiple independent origins of glyphosate resistance were detected in horseweed in California, with localized movement of resistant individuals accounting for spread on regional levels correlating with groundwater regulations that encouraged more glyphosate use and less use of other herbicides (Okada et al., 2013).
In summary, we used genomic-based epidemiology to track the mutations underlying one specific origin of glyphosate resistance in kochia and showed that at least three independent origins of glyphosate resistance have evolved in kochia, followed by substantial regional gene flow to spread the resistance alleles to new genetic backgrounds. Due to the tumbling dispersal of kochia, intercepting seed movement across the landscape has high potential to mitigate the negative impact of herbicide resistance spreading from an initial origin. With the kochia reference genome now available (Patterson et al., 2019), the population genomics approach used by Kreiner et al. (2019) can be used in kochia to study population divergence and origins of resistance (Martin et al., 2019). Sequencing and assembly of the duplicated region from haplotypes 2 and 3 would provide insights and new markers to further investigate the evolutionary dynamics of the EPSPS tandem duplication in kochia across western North America.