Into the weeds: Matching importation history to genetic consequences and pathways in two widely used biological control agents

Abstract The intentional introduction of exotic species through classical biological control programs provides unique opportunities to examine the consequences of population movement and ecological processes for the genetic diversity and population structure of introduced species. The weevils Neochetina bruchi and N. eichhorniae (Coleoptera: Curculionidae) have been introduced globally to control the invasive floating aquatic weed, Eichhornia crassipes, with variable outcomes. Here, we use the importation history and data from polymorphic microsatellite markers to examine the effects of introduction processes on population genetic diversity and structure. We report the first confirmation of hybridization between these species, which could have important consequences for the biological control program. For both species, there were more rare alleles in weevils from the native range than in weevils from the introduced range. N. eichhorniae also had higher allelic richness in the native range than in the introduced range. Neither the number of individuals initially introduced nor the number of introduction steps appeared to consistently affect genetic diversity. We found evidence of genetic drift, inbreeding, and admixture in several populations as well as significant population structure. Analyses estimated two populations and 11 sub‐clusters for N. bruchi and four populations and 23 sub‐clusters for N. eichhorniae, indicating divergence of populations during and after introduction. Genetic differentiation and allocation of introduced populations to source populations generally supported the documented importation history and clarified pathways in cases where multiple introductions occurred. In populations with multiple introductions, genetic admixture may have buffered against the negative effects of serial bottlenecks on genetic diversity. The genetic data combined with the introduction history from this biological control study system provide insight on the accuracy of predicting introduction pathways from genetic data and the consequences of these pathways for the genetic variation and structure of introduced species.


| INTRODUC TI ON
In the modern era of global trade, species are being inadvertently and deliberately introduced widely beyond their historic ranges (Crowl, Crist, Parmenter, Belovsky, & Lugo, 2008;Lockwood, Hoopes, & Marchetti, 2013;Mack et al., 2000). A crucial focus of evolutionary ecology of introduced species is to understand their pattern of spread and to identify their native origins and pathways of introduction to better prevent and manage biological invasions . Inferring the origins and spread of these exotic species is challenging and rarely are the true pathways or origins known. Thus, a fruitful approach may be to use documented introductions, such as those performed in classical biological control, as model systems to provide greater insights into population genetic analyses, as well as insight into the consequences of population movement and ecological processes for the genetic structure and variation of a species (Fauvergue, Vercken, Malausa, & Hufbauer, 2012;Marsico et al., 2010).
Classical biological control uses natural enemies (predators, parasitoids, and pathogens) to control invasive populations of weeds, and arthropod pests and disease vectors in the introduced range (Van Driesche, Hoddle, & Center, 2008). These natural enemies, as biological control agents, are often imported across disjunct geographic ranges for the long-term control of the target invasive species. In the modern era, these importations are well-regulated (Van Driesche et al., 2008) and well documented (but see Coulson, 1992;Marsico et al., 2010). Thus, they provide model systems to study the repercussions of invasion pathways and multiple introductions-including their effects on inter-and intraspecific hybridization, bottlenecks, inbreeding, genetic variation, and correlations of genetic diversity with population performance of the biological control agents (Fauvergue et al., 2012;Marsico et al., 2010;Roderick & Navajas, 2003).
To enhance the establishment and success of biological control agents, often multiple separate introductions are made, and large numbers of individuals are released (Van Driesche et al., 2008).
Multiple introductions here refer to introducing individuals from more than one population, or of more than one species, or both into the same geographic areas. Multiple introductions can increase the genetic diversity in an introduced population due to genetic admixture of different source populations (Bock et al., 2015;Dlugosch, Anderson, Braasch, Cang, & Gillette, 2015;Dlugosch & Parker, 2008;Rius & Darling, 2014;Szucs, Eigenbrode, Schwarzlander, & Schaffner, 2012). Alternatively, multiple introductions of more than one population could interfere with local adaptation, particularly in the native range (Rius & Darling, 2014;Verhoeven, Macel, Wolfe, & Biere, 2011). Additionally, hybridization can occur when more than one closely related species or strain is introduced, which can potentially lead to hybrid breakdown or hybrid vigor (Andersen & Mills, 2016;Arcella, Perry, Feder, & Lodge, 2014;Bean et al., 2013;Bitume, Bean, Stahlke, & Hufbauer, 2017;Mathenge et al., 2010;Szűcs et al., 2018Szűcs et al., , 2018. Hybrid vigor can result from positive epistatic interactions among loci, heterosis due to masking of deleterious alleles, or heterozygote advantage, whereas hybrid breakdown can occur from negative epistatic effects among loci and/or the underdominance of loci (Arcella et al., 2014;Edmands, 1999). Thus, the presence of multiple introductions and hybrids can greatly impact the growth and spread of introduced populations, and the efficacy of biological control programs.
The introduction of large numbers of individuals is critical to improve establishment success, as it buffers against demographic stochasticity and helps minimize loss of genetic variation (Fauvergue et al., 2012;Fraimout et al., 2017;Simberloff, 2009). Nonetheless, introduced populations often endure demographic bottlenecks (Dlugosch et al., 2015;Dlugosch & Parker, 2008;Estoup et al., 2016), which can decrease allelic richness and heterozygosity, with the latter depending on the rate of population growth following the initial bottleneck (Bock et al., 2015;Fauvergue et al., 2012;Nei, Maruyama, & Chakraborty, 1975). Certain alleles might increase or decrease in frequency by chance during bottlenecks, leading introduced populations to diverge from native populations (Dlugosch et al., 2015;Dlugosch & Parker, 2008). Genetic drift and inbreeding can also lead to increased homozygosity (Crow, 2010), which can be associated with reduced fitness (Bock et al., 2015) (but see Verhoeven et al., 2011). However, population bottlenecks do not always reduce genetic variation (Estoup et al., 2016;Goodnight, 1988;Kolbe et al., 2004;Taylor, Downie, & Paterson, 2011) or lead to genetic differentiation from the native population (Franks, Pratt, & Tsutsui, 2010), particularly if populations grow rapidly following introduction (Nei et al., 1975). Evaluating the effects of bottlenecks in population size on genetic diversity can enhance our understanding of the consequences of introductions and spread of species.
Although great efforts are taken to introduce many individuals from the native range to enhance establishment success, regulatory processes can make this difficult. Thus, the number of individuals (propagule size) imported to a region ranges widely from 10 to more than 1,000. While regulations vary by country (De Clercq, Mason, & Babendreier, 2011), each collection from the native range for release typically passes through quarantine to prevent unintentional introductions of other species (Hufbauer, Bogdanowicz, & Harrison, 2004). In many countries, such as the United States, further screening to characterize host range is often required for each new collection from the native range, which can mean many additional generations in quarantine even for agents that have already been approved. During this time, inbreeding and adaptation to the quarantine and mass-rearing environment can also occur (Freitas, Morales-Correa, Barbosa, & Fernandes, 2018;Hopper, Roush, & Powell, 1993;Hufbauer et al., 2004). Following quarantine screening, population size is typically increased as much as possible ("mass rearing") in order to release hundreds to thousands of individuals (e.g., see importation history section in this study). However, the proportion of individuals that survive in the field and contribute to the next generation may be low, resulting in another demographic bottleneck (Hufbauer et al., 2004).
Regulatory and logistical obstacles limit sampling from the native range; thus, biological control agents for release in new regions are often collected from a population already in use for biological control rather than revisiting the native range. This introduction process is analogous to the movement of invasive species, whereby an introduced population becomes the source of several secondary introductions, and is therefore acknowledged as a "bridgehead population" (Bertelsmeier & Keller, 2018;Dittrich-Schröder et al., 2018;Fraimout et al., 2017;Lombaert et al., 2010). Similarly, biological control agents frequently undergo serial importation steps, and thus serial bottlenecks in population size. By using the known introduction pathways from biological control programs, we can evaluate our ability to reproduce the introduction pathways by analyzing data from molecular markers.
Here, we examine the importation history, genetic diversity, and population structure of two closely related species introduced for biological control to gain insight into the consequences of population movement and ecological processes for the genetic structure and variation of these two species. Here, we ask: (1) Is there evidence of hybridization between these species, and (2) how do introduction processes affect the genetic variation and structure of these species? More specifically, (2a) are there indications of decreased heterozygosity and allelic diversity in the introduced populations relative to the native range, (2b) do increases in the number of individuals initially released or genetic admixture from multiple introductions result in increased genetic diversity, (2c) do populations with more introduction steps between them and the source population in the native range exhibit greater loss in genetic variation compared to populations with fewer introduction steps, and (2d) despite originating from the same initial populations, have introduced populations differentiated from the native range and from each other?
To address these questions, we use the documented importation history and polymorphic microsatellite loci of two weevils, Neochetina bruchi and N. eichhorniae Hustache (Coleoptera: Curculionidae) from their native and introduced ranges. These two weevils are the most widely used biological control agents of water hyacinth, Eichhornia crassipes (Hill, Coetzee, Julien, & Center, 2011), a floating aquatic plant native to South America. Water hyacinth is recognized as one of the world's worst invasive weeds (Hopper et al., 2017;Spencer & Ksander, 2005). Classical biological control of water hyacinth has been implemented across the globe, with some introductions resulting in significant reduction in water hyacinth cover and/or biomass, including parts of Australia, China, East Africa, the U.S. Gulf Coast, India, Mexico (Aguilar, Camarena, Center, & Bojorquez, 2003; F I G U R E 1 Partial importation history (a, b) compared to the introduction processes predicted by FLOCK analyses (c, d) of Neochetina bruchi and Neochetina eichhorniae, two weevils native to South America. Arrows depict the direction of the biological control releases and the date initially released, but do not point to the exact release site in that locality. Red markers are based on the GPS coordinates of the localities used in this study. Black lines and yellow-filled regions represent the routes of importation history that were tested with microsatellite markers. Green-filled regions and lines (routes) (arrows) were not tested with the genetic markers from this study, but represent relevant importation history to some of the tested regions. Abbreviations are detailed in Table 1 -50 FL 1972-50 FL 1980-50 FL 1983-50 FL 1991-50 FL 1993-50 FL 1975-50 FL 1996-50 FL 1989-50 FL 1996-50 FL 1985-50 FL 1983 (a) (b) 1974-75 1989 N. bruchi Akers, Bergmann, & Pitcairn, 2017), and the lower elevation regions of South Africa (Julien, Hill, Center, & Jianqing, 2000). Releases of N. bruchi and N. eichhorniae from the native range (South America) began in the early 1970s, with initial and subsequent releases in 30 and 32 countries, respectively. These weevils have contributed substantially to the control of water hyacinth in at least 13 countries (Julien et al., 2000).
Through their use as biological control agents, these two weevil species have often undergone multiple and serial introductions ( Figure 1). For example, in the United States, weevils of N. eichhorniae released into northern California underwent four sequential importation steps from the original Argentinian population in the native South American region. Native Argentinian weevils were released into USA: Florida in the 1970s, and the weevils in USA: Florida were used to found a population in USA: Louisiana, which were then used to found populations in USA: Texas. This USA: Texas population was the source for the northern California population released in the early 1980s (Stewart, Cofrancesco, & Bezark, 1988). Similarly, in South Africa, there were multiple introductions of each N. bruchi and N. eichhorniae with each new release being sourced from a different location to which they had been introduced for biological control (Cilliers, 1999), rather than directly from the native range.
These multiple introductions from the non-native range represent serial bottlenecks in population size that could potentially reduce genetic diversity and limit adaptive potential. Alternatively, these multiple introductions from different source populations could increase genetic diversity through genetic admixture of the different source populations (see Bock et al., 2015;Dlugosch et al., 2015;Dlugosch & Parker, 2008). The latter may occur particularly if each source population had sufficient time to diverge or adapt to the region of introduction, resulting in increased genetic differentiation from its source population.
Based on the importation history and documented releases of these two biological control agents, we proposed several hypotheses addressing our five study questions in turn. (1) We hypothesized that hybrids of these two species would occur, as they have frequently been co-introduced to the same geographic regions (Julien et al., 2000) and individuals with morphological characteristics of both species have been found (Hopper et al., 2017).

| Relevant importation and release history of N. bruchi and N. eichhorniae
Importation and release history were obtained from peer-reviewed literature, government reports (Cilliers, 1999;Confrancesco, 1981;Hendrich, Balke, & Yang, 2004;Julien et al., 2000;Manning, 1979;Stewart et al., 1988;Vanthielen et al., 1994;Winston et al., 2014), In 1980, following the quarantine and mass-rearing periods in USA: Florida, 50 N. bruchi adults were released from USA: Florida in Wallisville Reservoir, Texas, USA (Confrancesco, 1981). N. eichhorniae were found in this same reservoir as a consequence of westward migration from a biological control site in Louisiana (Confrancesco, 1981 Joaquin River Delta in California (Akers et al., 2017;Stewart et al., 1988). All other importation data pertinent to this study are summarized in Figure 1 and further detailed in the Supporting Information Appendix S1.

| Specimen collections and DNA extraction
We and (i) the Sungei Buloh Wetlands, Singapore (Table 1, Figure 1).
We were unable to collect weevils from Argentina for this study due to the current limitations on biological exports in that country.
Additional regions in the native (Argentina) and non-native (Benin, Zimbabwe, and Thailand) range were not surveyed, but have documented importation pathways to several of the above populations and are thus included in Figure 1.
Weevils were preserved in 95% ethanol immediately after collection in the field. Prior to DNA extraction, we made photographic vouchers and catalogued lateral, ventral, and dorsal photographs for all weevils and uploaded onto a public Google Drive folder (Hopper, 2018). We extracted DNA using a modified Chelex extraction method from (Hopper et al., 2017). Purified DNA extractions were stored at −20°C until amplification with PCR. A total of 438 weevils were processed for DNA extraction (Table 1).

| Microsatellite marker development, genotyping, and analysis
Potential microsatellite loci for N. bruchi and N. eichhorniae were identified using a Perl script, PAL_FINDER_v0.02.03 (Castoe et al., 2012), and Primer3 (Rozen & Skaletsky, 2000) to analyze 150-bp paired-end Illumina sequences from extracted DNA enriched for microsatellite loci at the Savannah River Ecology Laboratory (University of Georgia, USA). From this, primers were designed for 48 loci, using only those with tri-and tetranucleotides and those with at least six repeats. For each species, the final loci for analysis were tested on DNA extractions from 24 adult weevils ranging across several collection sites. A set of 10 and 11 microsatellite loci for N. bruchi and N. eichhorniae, respectively, met the criteria of selection, that is, pure repeat, polymorphism, and amplification by PCR. Following amplification by PCR, eight and 10 loci, respectively (Table S1), were kept for the statistical analysis due to the high occurrence of null alleles in two loci for N. bruchi and one locus in N. eichhorniae.
After the initial screening, primers were combined in three multiplex reactions per individual for each species. For each 96-well plate, we included a negative control (using water instead of DNA template) and an internal control of aliquoted DNA from an individual weevil that was used on every plate for the respective species.
PCR multiplex reactions were run separately for the two species to avoid cross-contamination. Pig-tails (Table S1: GT, GTT, or GTTT) were added to the 5′ end of each reverse primer, and one of four different universal tails (Blacket, Robin, Good, Lee, & Miller, 2012) was added to the 5′ end of each forward primer (Table S1). The system of universal tailed primers was used to introduce a fluorescent dye during the PCR according to Blacket et al., (2012) (Table S1). PCR was performed at the following conditions: 95°C for 15 min; 35 cycles of 94°C for 30 s, the optimum annealing temperature (Ta) of each primer (Table S1) for 1.5 min, 72°C for 1 min, and a final extension of 30 min at 60°C. Pro v. 5.6.2 (Drummond et al., 2012). We re-ran multiplex reactions We used the program GenAlex (Peakall & Smouse, 2006 and the R packages, "poppr" v. 2.5.0 (Kamvar, Brooks, & Grünwald, 2015;Kamvar, Tabima, & Grünwald, 2014) and "adegenet" (Jombart, 2008) to convert genotyping results into formats suitable for analysis in R (R Core Team, 2017). We calculated the null allele frequency (Brookfield, 1996) from the final datasets in the R package "popgenreport" (Adamack, Gruber, & Dray, 2014). As some statistical tests assume linkage equilibrium (LE) and Hardy-Weinberg equilibrium (HWE), we assessed deviations from LE with "poppr" and deviations from HWE across all sites for each locus (exact test) with the package "pegas" (Paradis, 2010). We constructed genotype accumulation curves with the R packages "poppr" and "vegan" (Oksanen et al., 2017) to test whether sufficient sampling had been performed for each species and collection site.

| Hybrid identification
To evaluate whether co-introduction of these two related weevils species resulted in hybridization, we first identified individuals for each species that had ambiguous markings on the elytra that contrasted the typical morphological characteristics for that species ( Figure S3). Then, we tested both sets of species-specific microsatellite markers on 12 weevils with ambiguous morphological characteristics, as well as on weevils that had the typical species-specific morphological characteristics for comparison. Hybridization is inferred from at least two of the species-specific markers from each species amplifying in the same individual (see Weigel, Peterson, & Spruell, 2003).

| Effects of introduction processes on genetic variation and population structure (2a-c) Genetic variation
As bottlenecks in population size can reduce genetic heterozygosity through processes of genetic drift and inbreeding, we estimated the average observed (H o ) and expected (H e ) heterozygosity, deviations from HWE (exact test), and the average "inbreeding coefficient" (F IS ) for each collection site across all loci with the R package "diveRsity" (Keenan et al., 2013). Here, we use F IS to estimate increases in homozygosity due to genetic drift caused by a larger population being separated into sub-populations, rather than due to consanguineous mating (Crow, 2010). Thus, we used "g 2 " to test for inbreeding within populations of each weevil species by using 1,000 permutations in the R package "InbreedR" (Stoffel et al., 2016). In populations with inbreeding, g 2 is significantly greater than zero, indicating correlated heterozygosity among pairs of loci (David, Pujol, Viard, Castella, & Goudet, 2007;Stoffel et al., 2016). We compared total and average allelic richness (accounting for sample size) and the number of private alleles among collection sites (R package "popgenreport"; Adamack et al., 2014).
To compare genetic diversity among the introduced and native populations, we tested for the effects of population (collection site) on genetic diversity by fitting linear mixed models (LMM) with the lmer function in the lme4 package (Bates, Maechler, Bolker, & Walker, 2015). Implementing an LMM accounts for the variability of the microsatellite loci by modeling locus as a random effect, and collection site as a fixed effect with allelic richness or expected heterozygosity as the response variables in separate models. Separate models were additionally used for each of the Neochetina spp. Stepwise model simplification (Crawley, 2013) was performed using likelihood ratio tests. Differences across collection sites were compared, based on 95% CI, using Tukey's posthoc test in the "multcomp" package (Hothorn, Bretz, & Westfall, 2008).
To determine whether the number of individuals released (propagule size) affects genetic diversity, we examined the number of weevils imported to each Florida, Texas, and California in the USA (described in the importation history) and the ratio of allelic richness retained from the native range in those states. To examine the influence of the number of introduction steps on the genetic diversity in populations of these two biological control agents, we combined the documented importation history (described previously and presented in Figure 1) and the genetic diversity data ( Table 2).
We counted the number of introduction steps based on the number of times the weevils were imported and exported since the initial export from the native range. We used linear models ( Jost's D analyses with the R package "popgenreport" (Adamack et al., 2014). Although F ST is one of the most utilized metrics in population genetic studies, it can be biased downward for loci with multiple alleles (Meirmans & Hedrick, 2011). Thus, we additionally present Jost's D (Jost, 2008), which measures the fraction of allelic variation among populations, but can be biased upwards (Meirmans & Hedrick, 2011).
We used three additional population structure analyses to infer population structure by determining the number of genetic clusters (populations) and to assign individuals to their appropriate genetic cluster. We validate the results from these analyses by using information gained from the documented importation history. We used: (a) a discriminant analysis of principal components (DAPC) (Jombart, Devillard, & Balloux, 2010) with the package "adegenet" in R, (b) an iterative reassignment of individuals with the FLOCK software (Duchesne & Turgeon, 2012), and (c) a Bayesian approximation with the STRUCTURE software STRUCTURE (Pritchard, Stephens, & Donnelly, 2000). The FLOCK software and DAPC do not assume HWE or LE, in contrast to the program, STRUCTURE (Pritchard et al., 2000). As some of the microsatellite loci and populations in this study significantly deviated from HWE and LE (Supporting Information Appendix S1), we present the methods for analysis in the STRUCTURE program and the results from STRUCTURE in the Supporting Information Appendix S5.
The FLOCK software (Duchesne & Turgeon, 2012) first randomly divides all of the genotypes into K genetic groups (ignoring the sample memberships) and then reassigns the genotypes at each iteration to the group with the highest probability of belonging, using the multilocus method of maximum likelihood described by Paetkau, Calvert, Stirling, and Strobeck (1995). FLOCK was run both to provide an estimate of the number of populations and to determine which of a potential set of genetic sources is most likely the true source of each introduced population of both species. We used the plateau record to determine the estimate of K as described in Duchesne and Turgeon (2012). Default parameter values were used (20 reallocations per run, 50 runs) for each k.
To identify the most likely sources of an introduced population, we followed a systematic search procedure. In summary, we ran FLOCK with the novel sample and all plausible source samples while k was set at 2. Based on the resulting allocation tables from this run, all of the possible sources that were not mainly allocated to the same cluster as the novel sample were discarded. This same procedure was applied iteratively until only one potential source sample remains. When selecting the initial set of candidate sources for the allocation tables, we discarded the samples that could not be realistically considered potential sources. Those decisions were based mainly a priori on strong historical evidence. The searching procedure is described more formally and in greater detail in Supporting Information Appendix S2. When the searching procedure did not produce an unambiguous output, it was complemented by visualization with a DAPC run with the same samples (Supporting Information Appendices S2 and S3). We compared the DAPC results and FLOCK runs to the importation history ( Figure 1).

| Hybridization
We confirmed hybridization between N. bruchi and N. eichhorniae by analyzing the species-specific markers on 12 individuals that had noticeable hybrid-like markings on their elytra ( Figure S3). teristics, we also noticed potential hybrids from the SA: George population, but these individuals did not amplify well for either set of markers likely due to poor DNA extractions. We could not analyze the prevalence of hybrids due to the sampling bias from collectors that selected individuals for each species based on distinct markings that separate the species.

| (2a-c) Consequences of introduction processes on genetic variation
Allelic richness and expected heterozygosity did not differ significantly among populations of N. bruchi (Table 3, Table 3).
Overall, we did not find a correlation between the number of individuals released (propagule size, Lockwood et al., 2013)  In addition to the absence of an effect of propagule size, we did not find a significant correlation between the number of introduction steps from the native range and allelic richness for N. bruchi (LM,
The results from the FLOCK runs are visualized in Figure 1d and

| D ISCUSS I ON
Here, we examined the genetic diversity in and among populations of two widespread biological control agents of water hyacinth, the weevils: N. bruchi and N. eichhorniae.   (Arcella et al., 2014;Bean et al., 2013) as well as affect the host-specificity of a biological control agent (Bitume et al., 2017;Mathenge et al., 2010).
If fitness of hybrids is low, it may be useful to determine the conditions under which hybrids form and try to minimize hybridization in regions where biological control programs are critical for the control of water hyacinth.
In addition to the occurrence of hybridization, we found evidence of genetic drift and inbreeding in several populations. From the importation history, there is documented evidence that these weevils went through demographic bottlenecks during the importation and release phases of the biological control programs. Subsequent drift or inbreeding following demographic bottlenecks can lead to increased homozygosity (Crow, 2010 populations that have undergone a demographic bottleneck (Bock et al., 2015;Nei et al., 1975). Alternatively, these results may have been artifacts of marker scoring (see David et al., 2007), the sampling or the markers used in this study (Selkoe & Toonen, 2006). Additional potential evidence of genetic drift or inbreeding was found in the California population of N. bruchi with F IS = 0.15 (Table 3) (Table 3). Although inbreeding can have detrimental consequences, it has also been known to promote local adaptation (Verhoeven et al., 2011).
Integrating the estimates of genetic diversity with the importation history for these biological control agents also permitted us to examine the consequences of propagule size and introduction processes on the genetic diversity of introduced populations. We did not find any evidence that initial propagule size or the number of introduction steps affected current day genetic diversity in populations of N. bruchi or N. eichhorniae. However, initial propagule sizes in this study system may have been higher than a specific threshold required for an effect to have taken place. Our initial hypothesis that populations with more introduction steps away from the native range would harbor lower genetic diversity than those populations with fewer steps was not supported. For example, we found inter- respectively. The high allelic richness and many private alleles found in the population in Uruguay supports the general trends that introduced populations typically undergo a loss in genetic diversity (Dlugosch et al., 2015;Dlugosch & Parker, 2008), but see (Estoup et al., 2016;Goodnight, 1988;Kolbe et al., 2004;Taylor et al., 2011).
We were also able to investigate the potential effects of genetic admixture on genetic diversity as a result from multiple introductions that occurred in this study system. The FLOCK allocation tables generally reflected the movement of weevils documented in the importation records ( Figure 1) and additionally clarified genetic sources where the importation history was unclear. In places such as  Table 1 DA eigenvalues  Figures S2 and S3). This indicates that significant divergence occurred among and between several of the introduced populations and the native population since the initial introductions in the 1970s. This supports previous studies on invasive species and biological control agents that demonstrate the divergence of populations from the native range (Zepeda-Paulo et al., 2015) but see Franks et al. (2010). Divergence of introduced populations from the native populations likely depends on the time since the initial introduction. For example, we sampled populations almost 50 years after the initial introductions, whereas Franks et al. (2010) sampled in the introduced range just 2 years after the initial releases.
One caveat that we acknowledge is that the genetic divergence between the introduced and native range may have been due to the fact we sampled from Uruguay rather than Argentina, where the actual initial source populations were exported from. However, based on the DAPC and FLOCK analyses, the populations from Uruguay for both species appear to be genetic sources for several of our populations. Thus, we feel confident that the genetic composition from weevils in Argentina compared to those in Uruguay is not very different.
In addition to the results demonstrating that genetic drift and inbreeding occurred in several populations, we speculate that divergence has also occurred due to local adaptation to some of the regions of introduction. Rapid local adaptation has been observed in invasive species (Sotka et al., 2018) as well as in biological control agents (Phillips et al., 2008). Many of the introduced regions that we tested in this study have colder climates than that occurring in South We support the recommendation that population genetic analyses be performed prior to the selection and release of biological control agents (see Rauth, Hinz, Gerber, & Hufbauer, 2011). The genetic diversity and genetic composition may have implications for the population growth of the biological control agents and their success in controlling the target weed or pest. Although these weevils have shown tremendous success in reducing water hyacinth in a number of countries (Julien et al., 2000), less than optimal levels of control has been found in regions with cooler temperatures, including some of the high altitude areas in South Africa (Hill & Olckers, 2001;May & Coetzee, 2013) and in the Sacramento-San Joaquin River Delta, in northern California (Hopper et al., 2017). The lower efficacy of biological control in these regions could be due to climatic mismatch and/or the inability to thrive and adapt to the local area based on the genetic diversity and composition as influenced by importation methods and the selected source populations.

ACK N OWLED G EM ENTS
We thank Charles Sawyer, Vincent Spadone, and Somanette Rivas for their assistance with voucher preparation and DNA extrac- Agriculture. USDA is an equal opportunity employer and provider.

CO N FLI C T O F I NTE R E S T
None declared.