Management Implications
We show that the majority of genetic variation in Myriophyllum spicatum (Eurasian watermilfoil) and Myriophyllum spicatum × Myriophyllum sibiricum (hybrid watermilfoil) is structured among lakes, whereas genetic variation within lakes tends to be low. That genetic variation tends to occur among lakes means that managers should consider differences in genetic composition as a source of variation in the growth and herbicide response among lakes. Previous laboratory and field studies demonstrate that genotypes can differ in their growth and herbicide response properties, and it therefore stands to reason that the efficacy of specific control tactics may differ among lakes with different genotypes.
Whereas the majority of genetic variation in this study was structured among lakes, some lakes contained genetic diversity. Because it is possible that different genotypes occurring in the same lake could exhibit different herbicide responses, managers should consider the potential impact of local diversity on management outcomes. For example, genetic monitoring in diverse lakes might identify increases in the relative frequencies of specific genotypes following management actions. Such a signature could identify putatively problematic genotypes (e.g., herbicide resistance) and trigger focused laboratory or field study on specific genotypes to facilitate decision making. At the very least, herbicide studies of diverse lakes need to test genotypes separately, instead of as composite samples.
Although lakes tend to harbor different genotypes, it is clear that genotypes can spread from lake to lake. The identification of shared genotypes among lakes indicates that genotyping of Myriophyllum populations can provide a practical decision-making tool, especially when combined with quantitative laboratory and/or field data on growth and herbicide response properties. For example, one genotype that was found in multiple lakes has been determined in previous studies to exhibit fluridone resistance. Therefore, fluridone should not be a primary control tactic in these lakes. Genotypes that are shared across lakes may be prioritized for herbicide characterization.
Introduction
Recent studies demonstrate that genetic variation can be relevant to aquatic plant management outcomes. For example, fluridone-sensitive versus fluridone-resistant genotypes of hydrilla [Hydrilla verticillata (L. f.) Royle] can be identified by DNA substitutions in the phytoene desaturase gene (Michel et al. Reference Michel, Arias, Scheffler, Duke, Netherland and Dayan2004), and genetic screening of H. verticillata populations can therefore be used to predict fluridone efficacy (Benoit and Les Reference Benoit and Les2013). Similarly, genetic variation for growth and herbicide response properties has been identified in Myriophyllum spp. (e.g., Berger et al. Reference Berger, Netherland and Macdonald2012; LaRue et al. Reference LaRue, Zuellig, Netherland, Heilman and Thum2013; Taylor et al. Reference Taylor, McNair, Guastello, Pashnick and Thum2017; Thum et al. Reference Thum, Wcisel, Zuellig, Heilman, Hausler, Tyning, Huberty and Netherland2012) and fanwort (Cabomba caroliniana A. Gray; Bultemeier et al. Reference Bultemeier, Netherland, Ferrell and Haller2009). However, population genetic studies of within- and among-population genetic variability are still lacking for managed submerged aquatic plant species.
Understanding patterns of within- versus among-lake genetic variation can provide important information for managers. For example, the extent of within-population genetic variation may influence the potential for managed populations to locally adapt to environmental conditions and control tactics. Similarly, among-population variation may influence whether specific control tactics work equally effectively in different locations. For example, similar management outcomes may be expected in lakes that are dominated by the same genotype(s), whereas lakes with different genetic composition may have different management outcomes with the same control tactic if the composite genotypes differ in their growth and herbicide response properties. Finally, among-population patterns of genetic variation can lead to insight on the extent to which genes or genotypes move among water bodies (e.g., through human-aided or natural dispersal). Understanding this can help managers anticipate how particularly invasive or resistant genotypes might spread across the landscape.
A key factor that should influence the extent of within- and among-lake genetic diversity in aquatic plants is the relative extent of asexual propagation versus sexual reproduction. Asexual propagation is often the dominant mode of reproduction (Hutchinson Reference Hutchinson1975; Sculthorpe Reference Sculthorpe1967), and plays an important role in the establishment, growth, and maintenance of aquatic plant populations (Philbrick and Les Reference Philbrick and Les1996). Many aquatic plant species are characterized by low within- and among-population variation and broad distributional ranges of clones, which could be caused by bottlenecks associated with colonization, dominance of a generalist and high-fitness genotype, or loss of genetic variation via selection of a specific genotype across a homogeneous environment (Barrett et al. Reference Barrett, Eckert and Husband1993).
Despite extensive clonal propagation, most aquatic plant species retain the ability to reproduce sexually (Barrett et al. Reference Barrett, Eckert and Husband1993; Philbrick and Les Reference Philbrick and Les1996). Although within-population variation is often low, many aquatic plant species exhibit high among-population genetic differentiation (Santamaría Reference Santamaría2002), implying that sexual reproduction plays an important role in generating variation at the landscape scale. This pattern may be related to dispersal and recruitment of sexual propagules (seeds), stochastic founder events by vegetative propagules from genetically variable sources, or adaptive responses to local selection pressures (competition among clones) (Santamaría Reference Santamaría2002).
Eurasian watermilfoil (Myriophyllum spicatum L.) is a heavily managed invasive aquatic species in the United States. Myriophyllum spicatum is a prolific vegetative reproducer, but can sexually produce germinable seed (Aiken et al. Reference Aiken, Newroth and Wile1979; Hartleb et al. Reference Hartleb, Madsen and Boylen1993; Patten Reference Patten1955; Thum and McNair Reference Thum and McNair2018; Xiao et al. Reference Xiao, Wang, Xia and Liu2010). Myriophyllum spicatum appears to have been introduced to the United States from China (Moody et al. Reference Moody, Palomino, Weyl, Coetzee, Newman, Liu, Xu, Harms and Thum2016). At least two distinct lineages of M. spicatum occur in the United States (Zuellig and Thum Reference Zuellig and Thum2012) and may represent independent introductions. Myriophyllum spicatum also hybridizes with its native sister species, northern watermilfoil (Myriophyllum sibiricum Kom.), and hybrids (Myriophyllum spicatum × Myriophyllum sibiricum) are also frequently managed across the United States (Zuellig and Thum Reference Zuellig and Thum2012). It is clear that at least some of the genetic variation among M. spicatum × M. sibiricum genotypes is relevant to management-related traits, such as vegetative growth and herbicide response. For example, one M. spicatum × M. sibiricum genotype exhibits clear resistance to fluridone, whereas other M. spicatum and M. spicatum × M. sibiricum genotypes do not (Berger et al. Reference Berger, Netherland and Macdonald2012, Reference Berger, Netherland and MacDonald2015; Thum et al. Reference Thum, Wcisel, Zuellig, Heilman, Hausler, Tyning, Huberty and Netherland2012). Similarly, M. spicatum × M. sibiricum genotypes vary in their vegetative growth rates and response to 2,4-D (Taylor et al. Reference Taylor, McNair, Guastello, Pashnick and Thum2017).
Although it is clear that M. spicatum and M. spicatum × M. sibiricum exhibit genetic variation, it is unclear how genetic variation is structured within and among populations. Therefore, we aimed to answer two questions: (1) What is the genetic diversity of M. spicatum? (2) How is genetic diversity structured within and among lakes? We address these questions using microsatellite markers for two sets of lakes; one from Michigan and one from Minnesota.
Materials and Methods
We sampled 41 lakes in Michigan and 62 lakes in Minnesota for this study during the growing seasons (May to September) in 2017 and 2018. As part of a related project, some of our study lakes were sampled multiple times; in both years and/or early (around May) and late (around September) in a single year. In cases in which a lake was sampled in both years, we only included samples from 2017 in our analyses. However, in cases in which a lake was sampled twice in a single year, we combined those samples for our analyses.
In Minnesota, study lakes were chosen from the lakes reported in 2017 to contain M. spicatum (which includes M. spicatum × M. sibiricum; http://www.dnr.state.mn.us/invasives/ais/infested.html) across the state to represent a range of sizes, durations of infestation, and geographic distribution. The number of lakes to sample in a given county was determined by the approximate proportion of infested lakes in the state that occurred in that county. Minnesota lakes were chosen independently of their management history. In contrast, roughly half of the Michigan lakes sampled for this study have active management for M. spicatum. Lakes in Michigan were chosen based on the willingness of volunteers (aquatic plant managers and citizen volunteers) to collect samples. In both states, the study lakes represented broad geographic coverage.
Sampling covered the entire littoral zone in each lake. For most Minnesota lakes, sampling locations were 100 predetermined random points spread across the littoral zone of the lake (≤4.6-m deep). A few lakes were sampled based on a fixed grid. For Michigan lakes, samples were collected from either: (1) predetermined aquatic vegetation assessment sites evenly spread across the littoral zone, (2) predetermined locations using the Cooperative Lakes Monitoring Program’s Exotic Aquatic Plant Watch protocol (Bednarz et al. Reference Bednarz, Wandell, Steen, Dimond and Latimore2015), or (3) meandering shoreline surveys with plants collected at least 10 m apart to avoid unintentionally sampling the same ramet. Plants were sampled using an aquatic vegetation rake, with one Myriophyllum plant (top 15 cm with meristem) haphazardly chosen from each rake toss. All plant samples were preserved either by flash-freezing fresh meristematic tissue or by drying with silica gel.
These survey methods commonly resulted in collection of more than 50 plants per lake. When more than 20 plants were collected from a lake, we randomly subsampled ~20 plants for genetic data collection; when fewer than 20 plants were collected, we processed all of them (see Table 1 for sample sizes).
Lakes sampled for this study with the number of individuals of each genotype for Myriophyllum spicatum, Myriophyllum spicatum × Myriophyllum sibiricum, and Myriophyllum sibiricum from each lake reported as genotype:number of individuals.
a Sample excluded from the analysis of molecular variance because n < 5.
DNA extractions were performed using the Qiagen DNeasy Plant Mini Kit (Valencia, CA) following the standard plant protocol. We performed duplicate DNA extractions on ~10% of all samples to assist with scoring of microsatellites.
We genotyped eight microsatellite loci developed by Wu et al. (Reference Wu, Yu and Xu2013) (Myrsp 1, Myrsp 5, Myrsp 9, Myrsp 12, Myrsp 13, Myrsp 14, Myrsp 15, and Myrsp 16). Each microsatellite locus was amplified using the protocols detailed in Wu et al. (Reference Wu, Yu and Xu2013). Fluorescently labeled microsatellite PCR products were sent to University of Illinois–Urbana-Champaign’s Core Sequencing Facility for fragment analysis on an Applied Biosystems 3730xl sequencer.
Microsatellites were scored using GeneMapper v. 5.0 (Applied Biosystems, Foster City, CA). Myriophyllum spicatum, M. sibiricum, and M. spicatum × M. sibiricum are hexaploid, and exact genotypes cannot be determined, because the numbers of allele copies are ambiguous. Therefore, we treated microsatellites as dominant, binary data (i.e., presence or absence of each possible allele at each locus) using the R package polysat (Clark and Jasieniuk Reference Clark and Jasieniuk2011).
We delineated distinct multilocus genotypes (hereafter referred to simply as “genotypes”) using Lynch distances and a threshold of 0 in polysat (Clark and Jasieniuk Reference Clark and Jasieniuk2011). We identified genotypes as M. spicatum, M. sibiricum, or M. spicatum × M. sibiricum using visual groupings in a principal coordinates analysis performed in polysat, judiciously choosing samples from that analysis to identify using restriction enzyme banding patterns for the internal transcribed spacer (Grafé et al. Reference Grafé, Boutin and Pick2015).
For each taxon in a lake, we calculated genotypic diversity as Nei’s genetic diversity (Reference Nei1987) corrected for sample size.
We also performed hierarchical analysis of molecular variance (AMOVA) for M. spicatum, M. spicatum × M. sibiricum, and M. sibiricum individually using GenAlEx v. 6.5 (Peakall and Smouse Reference Peakall and Smouse2006). We excluded from the AMOVA any taxon that had fewer than five individuals within a lake (see Table 1). This means that in some cases, one or more taxa were excluded from the AMOVA, even if another taxon was included for the lake.
Results and Discussion
Overall levels of genetic diversity were highest for M. sibiricum, lowest for M. spicatum, and intermediate for M. spicatum × M. sibiricum. In the 41 lakes sampled in Michigan, we found 18 unique M. spicatum genotypes, 21 unique M. spicatum × M. sibiricum genotypes, and 17 unique M. sibiricum genotypes. In the 62 lakes assessed in Minnesota, we found 6 unique M. spicatum genotypes, 52 unique M. spicatum × M. sibiricum genotypes, and 85 unique M. sibiricum genotypes (Figure 1; percent variation explained is 68.7% for axis 1 and 9.6% for axis 2). Because there are differences among regions in sample size, and because some genotypes are more common than others, we calculated the effective number of genotypes per taxon in each region as:
$${1 \over {\mathop \sum \nolimits_{i = 1}^s p_i^2}}$$
Principal coordinates analysis of all unique genotypes identified in Michigan (filled shapes) and Minnesota (open shapes). Squares, Myriophyllum spicatum (EWM); triangles, Myriophyllum spicatum × Myriophyllum sibiricum (HWM); circles, Myriophyllum sibiricum (NWM). Percent variation explained is 68.7% for axis 1 and 9.6% for axis 2.
where p i is the frequency of genotype i. This can be interpreted as the number of equally frequent genotypes that would equal the same overall observed genotypic diversity and is analogous to the effective number of alleles (Kimura and Crow Reference Kimura and Crow1964). For M. spicatum, the effective number of genotypes was 4.7 for Michigan and 1.2 for Minnesota. For M. spicatum × M. sibiricum, the effective number of genotypes was 8.2 and 9.8 for Michigan and Minnesota, respectively. Finally, the effective number of M. sibiricum genotypes was 8.2 and 25 for Michigan and Minnesota, respectively.
Within-lake diversity was lowest for M. spicatum, highest for M. sibiricum, and M. spicatum × M. sibiricum exhibited intermediate within-lake diversity. For M. spicatum, only 4% of the total genetic variation was distributed within individual lakes (Table 2), and the majority of lakes had only a single genotype (Figure 2, top). In contrast, 36% of the total genetic variation for M. sibiricum was distributed within lakes (Table 2), and most lakes with M. sibiricum had more than one genotype (Figure 2, bottom). For M. spicatum × M. sibiricum, 16% of the total genetic variation was distributed within lakes (Table 2). However, although some M. spicatum × M. sibiricum lakes had multiple genotypes, the majority of lakes contained only a single M. spicatum × M. sibiricum genotype (Figure 2, middle). These patterns suggest that clonal reproduction is the dominant form of reproduction locally for Myriophyllum.
Hierarchical analysis of molecular variance for Myriophyllum spicatum, Myriophyllum spicatum × Myriophyllum sibiricum, and Myriophyllum sibiricum.
a Fixation indices shown are significant at P < 0.001, based on 999 permutations.
Genotype diversity frequencies for Myriophyllum spicatum (EWM), Myriophyllum spicatum × Myriophyllum sibiricum (HWM), and Myriophyllum sibiricum (NWM). Genotype diversities were calculated as Nei’s (Reference Nei1987) diversity corrected for sample size. Black bars are populations from Michigan; white bars are populations from Minnesota.
Although genetic variation is generally low within lakes, there are instances where lakes contain genetic diversity. This can occur at the taxonomic level, as mixtures of M. spicatum, M. spicatum × M. sibiricum, and M. sibiricum, or as mixtures of genotypes within a taxon. Because it is possible that different genotypes occurring in the same lake could exhibit different herbicide responses, managers should consider this local diversity and at the very least quantify the growth, spread, and control efficacy of different genotypes separately. Furthermore, managers must recognize the potential for temporal changes in growth, spread, and control efficacy to occur if genotypes with different traits increase or decrease in relative frequency. For example, Parks et al. (Reference Parks, McNair, Hausler, Tyning and Thum2016) found a significant change in the proportion of M. spicatum versus M. spicatum × M. sibiricum in a large lake before and after large-scale auxinic herbicide treatments. Incorporating genetic data into routine aquatic vegetation survey and monitoring efforts could therefore assist in determining whether specific genotypes within a genetically diverse lake exhibit disproportional changes in their frequencies or abundance. Such observational information could trigger a focused laboratory study on specific genotypes of concern, which in turn could facilitate decision making on control tactics in specific lakes.
Among-lake genetic diversity was higher than within-lake diversity for all three taxa. In other words, lakes tended to be genetically different from one another and harbor different genotypes. This means that, although local reproduction appears to be largely clonal, sexual reproduction is important at landscape scales.
Myriophyllum spicatum had the highest among-population variation (96%). Of this, 58% was attributed to genetic variation among populations within regions, whereas 38% was attributed to genetic differences among regions. Thus, while lakes tend to be genetically different, there is also regional population structure. This is reflected by the relatively higher genetic diversity of M. spicatum in Michigan compared with Minnesota. In Minnesota, we found a single, widespread M. spicatum genotype that occurred in 40 lakes (Table 1). Effective diversity (the number of equally frequent genotypes that would equal the same total diversity) of M. spicatum in Minnesota was only 1.2, because the widespread genotype made up 93% of M. spicatum occurrences. Although this same genotype was found in Michigan, it made up only 39% of all M. spicatum occurrences, and several other M. spicatum genotypes were at appreciable frequencies in Michigan (>10%; effective diversity was 4.7). Therefore, the among-region variation for M. spicatum is due to the presence of M. spicatum genotypes occurring in Michigan but not in Minnesota.
Myriophyllum sibiricum also had high among-population variation (64%). However, in contrast to M. spicatum, M. sibiricum did not have any significant among-region variation. This indicates that lakes harbor different genotypes of M. sibiricum, but there is no regional population structure. That is, genetic distances among genotypes in different regions are not greater than genetic distances among genotypes within regions. Effective genetic diversity was higher in Minnesota (25) compared with Michigan (8.2). It is unclear whether this reflects an actual difference between the two regions or is an artifact of the sampling strategies or sizes in the two regions.
Myriophyllum spicatum × M. sibiricum also exhibited high among-population variation (84%). Of this, the vast majority (75%) could be attributed to differences among lakes within regions, whereas a smaller fraction (9%) could be attributed to regional population structure. The two regions did not share any M. spicatum × M. sibiricum genotypes, which suggests that independent hybridization events explain the high among-population variation. Further, the intermediate among-region component of variation compared with M. spicatum and M. sibiricum likely reflects the regional population structure of M. spicatum and the lack of regional population structure of M. sibiricum.
General explanations for our observed pattern of low within- but high among-lake variation include low dispersal and recruitment of seeds, stochastic founder events of clones from genetically variable sources, or adaptive responses to local selection pressures (competition among clones) (Santamaría Reference Santamaría2002). Distinguishing among these alternative hypotheses has important implications for management. For lakes where a single genotype is dominant, one potential explanation is stochastic colonization of a single genotype (i.e., founder effect). Similarly, a lake may initially be genetically diverse, but reductions in population size by management efforts could result in genetic drift via a strong population bottleneck. In both of these cases, the dominant genotype is stochastic. In contrast, a genotype may be dominant in a lake because it outcompetes other genotypes (i.e., is locally adapted). Of particular importance to managers would be whether a genotype comes to dominate a population because it is resistant to historical management efforts (i.e., herbicide-resistance evolution). Therefore, an important avenue for further research is to understand whether there are temporal dynamics of genetic diversity, and whether a single genotype occurs by chance versus nonrandomly.
That genetic variation tends to occur primarily among lakes means that managers should consider differences in genetic composition as a source of variation in the growth and herbicide response among lakes. Because genotypes can differ in their growth and herbicide properties (e.g., Berger et al. Reference Berger, Netherland and Macdonald2012, Reference Berger, Netherland and MacDonald2015; Netherland and Willey Reference Netherland and Willey2017), it stands to reason that the efficacy of particular control tactics may differ in lakes with different genotypes. Myriophyllum spicatum × M. sibiricum populations in particular have received increasing attention in recent years over concerns that they may exhibit resistance. Although it is clear that some M. spicatum × M. sibiricum genotypes exhibit resistance to some herbicides (Berger et al. Reference Berger, Netherland and Macdonald2012; LaRue et al. Reference LaRue, Zuellig, Netherland, Heilman and Thum2013; Netherland and Willey Reference Netherland and Willey2017; Taylor et al. Reference Taylor, McNair, Guastello, Pashnick and Thum2017; Thum et al. Reference Thum, Wcisel, Zuellig, Heilman, Hausler, Tyning, Huberty and Netherland2012), it is still unclear how commonly M. spicatum × M. sibiricum genotypes exhibit resistance to one or more herbicides. Therefore, future research should prioritize quantitative studies on growth and herbicide response of different genotypes in both the laboratory and field.
Although among-lake genetic diversity was relatively high, several genotypes were shared among lakes. One M. spicatum genotype (8) was found all across Minnesota and Michigan (Figure 3; Table 1). Three other M. spicatum genotypes were found in two lakes each. Four M. spicatum × M. sibiricum genotypes were shared among lakes in Minnesota, and an additional three M. spicatum × M. sibiricum genotypes were shared among bays in Lake Minnetonka. Similarly, four M. spicatum × M. sibiricum genotypes were found in multiple lakes in Michigan (Figure 4; Table 1). Unlike M. spicatum, no M. spicatum × M. sibiricum genotype was found in both Michigan and Minnesota. Finally, two M. sibiricum genotypes were found in two lakes each in Michigan, and one M. sibiricum genotype was found in both Michigan and Minnesota (Table 1). In many cases, lakes that shared genotypes were geographically close to one another, but this was not always the case.
Myriophyllum spicatum (EWM) genotypes that were shared among lakes in our study. Genotype numbers correspond to those in Table 1.
Myriophyllum spicatum × Myriophyllum sibiricum (HWM) genotypes that were shared among lakes across the entire study region (A) and among bays in Lake Minnetonka, MN (B). Genotype numbers correspond to those in Table 1.
A key question for managers is whether the subset of genotypes that have spread to multiple lakes is random or nonrandom with respect to management history. One possibility for shared genotypes is stochastic dispersal; plants may be spread from lake to lake based on the movement patterns of dispersal vectors (boats, birds, etc.) as opposed to any particular traits that they possess. Alternatively, some genotypes may be more likely to spread and establish based on properties related to their growth or management response. For example, the widespread distribution of a genotype may represent a general-purpose and high-fitness genotype or a genotype that is adapted to a specific environmental factor that is shared among lakes, such as commonly employed control tactics.
The identification of shared genotypes among lakes indicates that genotyping of Myriophyllum populations can provide a practical decision-making tool, especially when combined with quantitative laboratory and/or field data on growth and herbicide response properties. For example, we found one M. spicatum × M. sibiricum genotype in six lakes across Michigan. Interestingly, this genotype is the same genotype as a known fluridone-resistant genotype isolated from Townline Lake, Michigan (Berger et al. Reference Berger, Netherland and Macdonald2012, Reference Berger, Netherland and MacDonald2015; Thum et al. Reference Thum, Wcisel, Zuellig, Heilman, Hausler, Tyning, Huberty and Netherland2012) that also appears to exhibit diquat resistance (Netherland and Willey Reference Netherland and Willey2017). Although experiments are necessary to determine whether plants with the same microsatellite genotype exhibit the same properties, it seems prudent for managers to assume that they do unless demonstrated otherwise. Thus, for genotypes that are known to exhibit properties that are of concern to managers (e.g., herbicide resistance), genetic survey and monitoring hold promise to efficiently determine whether lakes proposing to use a herbicide harbor a genotype known to be resistant to it. Similarly, genotypes that are found in multiple lakes should be prioritized for laboratory study of growth and herbicide response so that the information can be used to determine the best control tactic(s) and to coordinate that information across multiple sites.
While our study provides some hope for combining genetic delineation of clones (genotypes) with herbicide characterization, it is important to recognize current limitations associated with identifying and tracking Myriophyllum genotypes. We distinguished genotypes with a strict threshold of zero differences across eight polyploid microsatellite markers. It is possible that scoring errors and/or somatic mutations could lead to identifying individuals that share ancestry through clonal descent as distinct genotypes. It is also possible that individuals that share ancestry through clonal descent would share similar growth and herbicide response characteristics. Pashnick and Thum (Reference Pashnick and Thum2020) analyzed a subset of previously determined genotypes and found that scoring error rates for molecular markers generated by genotyping by sequencing were very low. Therefore, replacing microsatellites with single-nucleotide polymorphism (SNP)-based markers to identify, track, and characterize lineages is likely to increase accuracy in distinguishing genotypes and may thus increase the efficiency of identifying lineages to characterize. Further, identifying diploidized SNP-based markers would facilitate genetic analyses based on allele frequencies, whereas our current approach is limited to treating polyploid markers as dominant data and distinguishing unique multilocus genotypes.
We have shown here that the majority of genetic variation in M. spicatum and M. spicatum × M. sibiricum is structured among lakes, but that some lakes contain genetic diversity and some lakes share genotypes. Because genotypes can differ in their growth and herbicide response properties, managers should consider differences in genetic composition as a source of variation in the growth and herbicide response among lakes. Similarly, because genotypes co-occurring in the same lake could exhibit different herbicide responses, managers should consider the potential for temporal change in herbicide control efficacy. For example, population-level efficacy of a specific herbicide is expected to decrease over time if a resistant genotype increases in relative frequency over time. Finally, the identification of shared genotypes among lakes indicates that genotyping can provide a practical tool to identify specific genotypes of interest (e.g., known herbicide-resistant genotypes that have spread to multiple lakes) and to prioritize widespread and common genotypes for characterization.
Acknowledgments
We thank Jake Britton, Arda Bucher, Rick Buteyn, Leslie Clark, Jamiee Conroy, Mike Devareene, Terry Dugan, Deborah Emmer, Michael Finney, Barb Gajewski, Mike and Kathy Gallagher, Breanne Grabill, Ann Hruska, Dan Haukuess, Paul Hausler, Dan Hayes, Bill Henning, Craig Kivi, Mark Luttenton, Rick Meeks, Marilyn Merit, Lori Mroczek, Matt Novotny, Budan Price, Doug Pullman, John Ransom, Chris Riley, M. Rose, Jeff Sanborn, Biu Smith, Rick Sotolongo, Tom Tisue, Gage Vanorder, Charlie Walmsley, and Kim Yeip for collection of the Michigan samples. Anna French, Jeff Pashnick, Emma Rice, and Leah Simantel assisted with the molecular data collection. T. J. Ostendorf, Jacob Olson, Alex Franzen, Kyle Blazek, Matt Gilkay, and Matt Manthey assisted with field collection and data entry in Minnesota. Jim McNair, the Thum laboratory group, and two anonymous reviewers provided comments that improved the original article. This project was supported by the Minnesota Aquatic Invasive Species Center, with funding provided by the Minnesota Environmental and Natural Resources Trust Fund, as recommended by the Legislative-Citizen Commission on Minnesota Resources, and the Michigan Invasive Species Grant Program (www.michigan.gov/invasives). Additional support was provided by the Montana Agricultural Experiment Station (Project MONB00249) and by the Minnesota Agricultural Experiment Station USDA National Institute of Food and Agriculture, Hatch grant MIN-41-081. No conflicts of interest have been declared.
