Population structure and genetic diversity of Bromus tectorum within the small grain production region of the Pacific Northwest

Abstract Bromus tectorum L. is an invasive winter annual grass naturalized across the United States. Numerous studies have investigated B. tectorum population structure and genetics in the context of B. tectorum as an ecological invader of natural areas and rangeland. Despite the wealth of information regarding B. tectorum, previous studies have not focused on, or made comparisons to, B. tectorum as it persists in individual agroecosystems. The objectives of this study were to assess the genetic diversity and structure, the occurrence of generalist and specialist genotypes, and the influence of climate on distribution of B. tectorum sourced exclusively from within small grain production regions of the Pacific Northwest. Genetic diversity of B. tectorum sourced from agronomic fields was found to be similar to what has been observed from other land use histories. Six distinct genetic clusters of B. tectorum were identified, with no evidence to indicate that any of the genetic clusters were better adapted to a particular geographical area or climate within the region. Given the apparent random spatial distribution of B. tectorum genetic clusters at the spatial scale of this analysis, unique genotypes may be well mixed within region, similar to what was reported for other inbreeding weedy grass species.

Previous studies have solely investigated downy brome population genetics in the context of downy brome as an ecological invader.
Consequently, previous research has focused on the prevalence of common versus rare genotypes across the landscape, genetic differences between native populations of Eurasia and invasive populations of North America, and evidence of local adaptation to distinct ecosystems (Leger et al., 2009;Merrill, Meyer, & Coleman, 2012;, 2016Novak, Mack, & Soltis, 1991Scott et al., 2010). Novak et al. (1991) reported that a limited number of genotypes were found distributed widely across North America. In comparing native and introduced populations, total genetic diversity across the entire native range was higher than the introduced range. However, within a population, genetic diversity was greater in the introduced range , 2016. Genetic differences between native and introduced ranges can be explained by the founder effect reducing genetic diversity in the introduced range coupled with mixed populations of selfing individuals from diverse origins , 2016.
While widely spread genotypes across the introduced range can be attributed to generalists, evidence for local adaptation to an environment by specialist genotypes has been reported for downy brome. When local adaption was observed, variation in phenological traits including flowering time, vernalization requirements, and timing of mature seed set was identified as driving adaptation (Ball, Frost, & Gitelman, 2004;Meyer, Nelson, & Carlson, 2004;Rice & Mack, 1991a,b). Ramakrishnan et al. (2006) found that ecological distance better predicted genetic distance of populations than physical distance, indicating that similar habitats select for similar genotypes from widely dispersed genotypes. Downy brome has been observed invading new habitats as both broadly adapted generalist genotypes and pre-adapted specialist genotypes (Scott et al., 2010). When characterizing genotypes in the Intermountain West, historically invaded land has been largely occupied by generalist genotypes, while recently invaded land was dominated by distinct specialist genotypes (Merrill et al., 2012;Scott et al., 2010).
Previous studies have identified low genetic diversity within the species downy brome, with mean expected heterozygosity ranging from 0.002 to 0.336 within populations (Bartlett, Novak, & Mack, 2002;Meyer et al., 2013). While heterozygous individuals have been reported in the literature, outcrossing is exceedingly rare (Leger et al., 2009;Meyer et al., 2013;Valliant, Mack, & Novak, 2007). A common garden experiment was designed to encourage and quantify outcrossing at greater frequencies than would be expected in nature, and outcrossing was observed at 0.75% (Meyer et al., 2013).
The PNW small grain production can be divided into four unique transition land can support a crop in 2 out of 3 years of a rotation, and crop-fallow land can only support a crop in 1 of 2 years. When irrigation is available winter wheat is still grown, but rotational crops become more diverse and winter wheat is not a principle rotational crop (Huggins et al., 2012). As the amount of moisture, both through precipitation and irrigation, increases so does the intensity of the cropping systems. Within the PNW small grain production region downy brome generalist genotypes would be expected throughout the entirety of the region, while specialists would be expected at greater abundance and frequency between different agroecological zones.
A genotype-by-sequencing approach was used in this study to estimate population structure and determine whether the genetic state of downy brome in PNW agroecosystems is similar to previous studies where individuals were sourced from nonagronomic locations.
Study objectives were as follows: (1) to assess the genetic variability of downy brome sourced exclusively from within small grain production fields, (2) determine the frequency and occurrence of generalists versus specialist genotypes, and (3) determine the influence of climate on the distribution of genotypes.

| Sampling of plant materials
Downy brome is distributed ubiquitously within agronomic fields of the PNW. No large landscape features, such as mountain ranges, are present that could block gene flow. Climate of the region exists across a longitudinal gradient with annual precipitation in the region ranging from <300 mm to >600 mm, with precipitation increasing from the west to east. Mean annual temperature also varies on an west to east gradient, with the western portion at 11°C and decreasing to 5°C to the east based on a 30-year average (1971; 2000) (Huggins et al., 2012). To accommodate for studies of both population genetics and structure, and for future field studies investigating climate and phenology, a systematic random sampling design was used to efficiently maximize the geographical area represented (Strofer et al., 2007).
A 10-km grid was laid over the small grain production region, and a point was randomly assigned for sampling within each grid. One hundred and ninety total sampling points were generated. If the sampling point was not located in a small grain field, the sampling point was moved to the nearest small grain field. If there was no field within 3 km of the original sampling point, the location was discarded. Following re-assignment of the original sampling locations, 130 sampling locations were retained (Figure 1).
Due to limited resources, the number of sampling locations was emphasized at the expense of collecting fewer individuals at each location (Ward & Jasieniuk, 2009). Collecting a single individual from approximately evenly spaced locations is an appropriate sampling method under the following condition: (1) the species is evenly distributed across the entire study area, (2) there are no known barriers to gene flow, (3) multilocus genetic data are used, and (4) newer Bayesian genetic clustering techniques are employed to determine genetic structure (Guillot et al., 2005(Guillot et al., , 2009Manel et al., 2007;Strofer et al., 2007). The aforementioned criteria were satisfied given the biology of downy brome, the study location, and the methodology employed.
In June of 2010 and 2011, a single downy brome plant was collected as either mature panicles or a live plant from each of 130 re-assigned sampling locations. Each plant was collected at least 10 m from the field border. Live plants were transplanted into a greenhouse and allowed to grow until mature panicles could be collected.
Caryopses from collected panicles were later germinated to provide tissue for DNA extraction. On 21 March 2014, as plants were at the two-to three-leaf stage, a single ~4-cm leaf was collected from 95 (
A modified GBS protocol developed by Mascher et al. (2013) for use with semiconductor sequencing platform was followed. Amplicons were sequenced on an Ion Proton" sequencer using an Ion P1" Chip (Life Technologies, Carlsbad, CA). Sequencing data were obtained in FASTQ file format, and the file size ranged from 5 to 112 MBs with an average size of 45.6 MBs. Average sequence length was 100 bp, and all sequences were trimmed to 100 bp using FASTX to provide a uniform sequence length for SNP calling.

| SNP calling
SNP calling was conducted using Stacks (1.22, Cresko Laboratory, Eugene, OR) (Catchen et al., 2013). The Stacks program aligns identical or nearly identical sequence reads into "Stacks" across individuals, and a catalog file is written. Each locus from each individual is matched against the catalog to determine the allelic state at each locus in each F I G U R E 1 The small grain production region of the inland Pacific Northwest T A B L E 1 Accession ID number, GPS coordinates of collection locations, year of collection, and cluster membership as determined by DAPC of each accession individual, while filtering and discarding poor-quality reads. As there is no reference genome available for Bromus tectorum, the Perl script denovo_map.pl was used to call SNPs using default settings (Catchen et al., 2013). Called SNPs were filtered using the populations command in Stacks. Default parameters were used, with the exception of requiring a minimum stack depth of 5, and all loci to be found in 75% of individuals to ensure the validity of obtained markers.

| Analysis of population structure and genetic diversity
The output from stacks was analyzed in R (R Development Core Team 2014) using the package adegenet (Jombart, 2008;Jombart & Ahmed, 2011  can be considered "more informative SNPs" and will be referred to as such throughout the manuscript. To complement cluster assignments based upon DAPC, the fixation index (F ST ) between each genetic cluster was calculated (Nei, 1973) along with genetic distance using Nei's standard (Nei, 1972(Nei, , 1978 using the R package "adegenet." A dendrogram was then constructed from the resulting genetic distance matrix using the R package "ape" (Paradis, Claude, & Strimmer, 2004;Popescu, Huber, & Paradis, 2012).
These further analyses were conducted as they retain the full complement of filtered SNPs, in contrast to the DAPC analysis which only retains a subset of the available genetic markers.
Population genetic metrics were also calculated for downy brome accessions based upon the agroecological class that the samples were taken from. Given the geographical separation between sampling locations, it is unlikely that recent gene flow occurred between any of the accession. As such, any accessions grouped together for the purpose of calculating population genetic metrics cannot be considered true populations. However, calculating heterozygosity, genetic diversity, and genetic partitioning based upon cluster assignment, and the land use of the sampling locations, may aid in the detection of specialist or generalist genotypes across the landscape and compliment the DAPC analysis.

| Reduced representation sequencing
Raw reads per accession ranged from 51,740 to 1,030,188 (Table 2).
After trimming and filtering retained reads ranged from 741 to 13,985, per accession, from which 16,382 SNPs were initially called. SNPs that were then found in at least 75% of individuals were retained for further analysis, resulting in 384 SNPs being selected. The number of retained reads and SNPs was not uniformly retained among accessions (

| Discriminant analysis of principal components
Thirty-five principal components were retained, corresponding to roughly 85% of cumulative variance, and used to identify seven clusters as optimal based upon BIC value. Following determination of the optimal number of clusters, multiple DAPC simulations identified six principal components as ideal in assigning group membership without overfitting the model. Three linear discriminants were retained to calculate the probability of group membership, and individuals were assigned accordingly. All clusters with the exception of cluster six contained multiple individuals. Cluster 6 contains the T A B L E 1 (Continued) T A B L E 2 Accession ID number, raw reads before filtering, retained reads, total SNPs called from retained reads, SNPs remaining after filtering for polymorphisms present in at least 75% of individuals, and cluster assignment

Retained reads
Total SNPs

| Heterozygosity, genetic diversity, and genetic partitioning
Heterozygosity was calculated for each individual loci and averaged across all accession, and across each cluster to calculate withincluster genetic diversity (H S ), total diversity (H T ), and the ratio of genetic diversity partitioned among clusters (G ST and G' ST ) using the R Package "mmod" (Hedrick, 2005;Nei, 1973;Winter, 2012

| DISCUSSION
Consistent with other studies, downy brome collected from small grain production fields in the PNW does not appear to have greater genetic diversity than populations in nonagronomic settings (Ashley & Longland, 2009;Bartlett et al., 2002;Meyer et al., 2013;Novak et al., 1991;Ramakrishnan et al., 2002;Valliant et al., 2007). Previous studies utilizing 25 allozyme markers reported observed heterozygosity ranging from 0.000 to 0.002 and expected heterozygosity ranging from 0.0 to 0.032 (Bartlett et al., 2002;Novak et al., 1991;Valliant et al., 2007). Later studies utilizing seven simple sequence repeat (SSR) markers reported greater genetic diversity compared to previous work with allozymes with observed heterozygosity ranging from 0.000 to 0.011 and expected heterozygosity ranging from 0.018 to 0.547 (Ashley & Longland, 2009;Kao, Brown, & Hufbauer, 2008;Ramakrishnan et al., 2002). Compared to the allozyme and microsatellite data, Meyer et al. (2013) reported greater observed heterozygosity, 0.001-0.009, and expected heterozygosity, 0.149-0.336, using 93 SNP markers. Within the PNW small grain production region, the average observed heterozygosity, 0.05, was greater than previous research using allozymes and SSR makers but similar to Meyer et al. (2013). Average expected heterozygosity within the PNW, 0.085, was between what was reported from allozyme and SSR marker data sets (Table 3). Higher observed heterozygosity would be expected from Meyer et al. (2013) as the sampled populations had been chosen because high rates of outcrossing were expected, based upon previous sampling which indicated relatively high heterozygosity and genetic diversity within the populations.
G ST values from introduced B. tectorum populations have been previously reported as ranging from 0.241 to 0.582 (Bartlett et al., 2002;Novak & Mack, 2016;Novak et al., 1991;Valliant et al., 2007). founder events coexisting within a single location (Merrill et al., 2012;Scott et al., 2010). As this study used genotype to define clusters within a geographical region rather than defining populations by the geographical proximity of the sampling locations, greater population differentiation would be expected.
As the western PNW is considerably dryer and warmer than the eastern PNW, it was hypothesized that evidence of specialist genotypes would be found when comparing the eastern and western portions of the region. The small genetic partitioning values returned when comparing accessions by the agroecological class from which they were sourced indicates that land use class, which is predominately driven by climate, has limited influence on genetic partitioning. If strong genetic partitioning was found based upon the land class from which accessions were sourced, it would be evidence of specialist genotypes occupying specific niches based on climate or agronomic practices. Results do not suggest segregation of genotypes between the eastern and western portions of the region or by agroecological class. The lack of a strong or easy-to-interpret genetic cline may be an indication that climate is not a major driver of downy brome genotype distribution within the PNW. Downy brome might also be adaptable to a larger range of climates than represented within this study.
Alternatively, the influence of climate might be more subtle than was detectable within this study.
The DAPC-defined clusters describing downy brome genetic distribution were successful in identifying the ripgut brome individual. While some clusters contain greater numbers of individuals, it appears all clusters are distributed throughout the small grain production region and none of the genetic clusters can be described as specialist genotypes in relation to climatic variables or spatial distribution. Cluster distribution appears random, and a farm's location within the region is a poor indicator of what genotype(s) are likely to be found.
Compared to all other genetic clusters, cluster 3 appears to have a higher degree of genetic diversity when described by genetic distance ( Figure 3) and expected heterozygosity ( Table 4). Both of these measures were calculated utilizing all of the retained genetic markers after filtering SNPs for quality. When comparing the relation of cluster 3 to all other clusters described by the DAPC analysis, cluster 3 is also quite isolated from all other genetic clusters across the first linear discriminate (Figure 1a,b) but not across the second or third linear discriminant (Figure 1a-c). However, cluster 3 is not uniquely distributed across the study region, compared with other clusters (Figure 4). Therefore, although increased genetic diversity was reported, the diversity does not appear to be adaptively significant at a landscape scale.
Cluster 6, the ripgut brome outlier, is distinct from all other clusters across all linear discriminants used, and when comparing pairwise While genetic markers linked to neutral gene regions, and SNPs in particular, are well suited to neutral evolutionary process such as genetic drift and gene flow (Helyar et al. 2011), such genetic markers are poor at detecting active evolutionary processes (Narum et al., 2013). Previous studies have demonstrated neutral markers can fail to detect local adaptation of population to habitats (Narum et al., 2010;Storz et al., 2009). As previous literature has demonstrated flowering time as adaptively significant and influenced by local climate, the genes responsible for regulating flowering pathways are a promising target to investigate potential adaptation of downy brome to climate (Ball et al., 2004;Meyer et al., 2004;Rice & Mack, 1991a,b). Future work will look at the functions associated with discovered SNPs in conjunction with known genes associated with regulating phenology.
Research into the population genetics and structure of related species to down brome and species with similar life histories provides further context into the adaptation of selfing grass species to the selection imposed by agronomic settings. Green et al. (2001) compared diversity of the inbreeding annual or biennial weed Bromus sterilis L.
(barren brome) between farms located in the United Kingdom. Barren brome exists as an assemblage of unique but inbred biotypes within agronomic fields. Similar to what was found from sampling B. tectorum within PNW small grain fields, considerable spatial mixing of genotypes was found across all sampled farm fields (Scott et al., 2010). When low genetic diversity was found within a field, Green et al. (2001) attributed diversity to selection of locally adapted inbred biotypes.
Population genetics and structure have also been investigated within Poaceae genus Setaria which contains several inbreeding summer annual agronomic weed species with worldwide distribution.
Surveying genetic diversity and structure of Setaria viridis (L.) Beauv (Wang et al., 1995a,b). Wang et al. (1995a) reported a separation in genotypes between northern and southern groups within North America. However, at smaller geographical scales, including at the farm and state level, geographical patterns were difficult to detect with some areas exhibiting high degrees of population differentiation while others were genetically identical. Wang et al. (1995a) concluded that diversity within a region is likely a result of the number of independent introductions, and the intensity and duration of natural selection. Wang et al. (1995b) expanded the analysis of Setaria species to S. glauca (L.) P. Beauv. S. geniculata P. Beauv. and S. faberi Herrm. Within the introduced range of the United States, S. geniculata and S. glauca both exhibited lower genetic diversity than what was found within their native range and regional patterns of genetic partitioning, while S. faberi was nearly genetically identical worldwide based upon the isozyme markers used. In summary Wang et al. (1995a,b) described the observed diversity of Setaria species in the context of a review of genetic diversity of 499 plant species conducted by Godt and Hamrick (1998). While significantly different from "average" plant species, the low genetic diversity and high population differentiation of both Setaria and Bromus species are typical of self-pollinating, invasive grass species (Green et al., 2001;Novak et al., 1991).
Given the apparently low genetic diversity and the similar genetic structure of Bromus and Setaria species within invaded and agricultural land, high levels of genetic diversity may not be essential for colonizing species. However, the use within this study of neutral markers may have masked genetic diversity conferring local adaption to novel environments (Narum et al., 2010(Narum et al., , 2013Storz et al., 2009). Future work will look to identify nonneutral genetic markers, which may better describe the influence of climate and human management on distribution and evolution.
Within the small grain production region of the PNW, Bromus tectorum clusters are highly differentiated and randomly dispersed, suggesting that generalists rather than specialist genotypes predominated across the region. The current structure of diversity within the PNW is likely the result of several independent introductions, constrained by natural selection (Novak et al., 1991;Wang et al., 1995a,b).
Given that a limited number of genetic clusters were found within the PNW, management strategies could be developed to target differences in phenotype between clusters. Previous studies have identified differences in Bromus tectorum herbicide susceptibility, germination characteristics, and date of seed production (Ball et al., 2004;Hulbert, 1955;Klemmedson & Smith, 1964;Lawrence, Burke, & Yenish, 2014).
If traits exhibiting variation can be linked to genotype, management strategies can be developed to target the specific populations in a given field. This targeted weed management approach has been called for in the literature (Baucom & Holt, 2009) but has yet to be realized. However, clusters are likely intermixed at smaller spatial scales than surveyed in this study, which may limit the implementation of genotype-specific management strategies as well mixed and distinct genotypes could adapt quickly to management strategies.

| CONCLUSIONS
Analysis of population genetics and genetic structure from downy brome collected within an agronomic region indicates that the heterozygous state of downy brome is similar, if not marginally greater, to what has been reported in previous literature. Downy brome exists within the PNW small grain production region as a series of generalist genotype clusters with limited evidence of spatial adaptation, similar to what was reported Novak et al. (1991) in a broad survey of downy brome across North America. Given the apparent random spatial distribution of downy brome clusters at the spatial scale of this analysis, unique genotypes may be well mixed within small grain fields, similar to what was reported for Bromus sterilis (Green et al., 2001).
To expand further upon the current reported findings, future efforts should include more samples of individuals from the same field to increase the spatial resolution of genetic inferences. Additionally, collection of individuals from nearby rangeland and natural areas may allow for the control of climate and the comparison of land use among accessions. Finally, phenotyping of collected individuals in common garden studies across several years or locations would provide traits to be compared across individuals and elucidate the results of DAPC clustering by correlating the separation of genotypes with phenotyping.