Management Implications
The identification of three distinct Nymphoides cristata (crested floating heart) lineages here—and a fourth interspecific hybrid lineage (Nymphoides cristata × Nymphoides aquatica) reported elsewhere—has implications for the management of this invasive species. Studies of control techniques, spread, and impacts of N. cristata should explicitly consider the genetic diversity identified in this study. For example, herbicides are currently the most common option for controlling N. cristata, including submersed applications of diquat, endothall, and florpyrauxifen-benzyl; foliar applications of endothall, imazamox, and imazapyr; and foliar combinations of flumioxazin and glyphosate. However, herbicide testing has not explicitly considered whether N. cristata lineages differ in their responses. As differences in herbicide response have been found among distinct clonal genotypes of other aquatic plants (e.g., 2,4-D and fluridone in Myriophyllum spp.; fluridone in Hydrilla verticillata), we posit that herbicide trials for N. cristata should include representatives of the four distinct lineages identified thus far. Similarly, N. cristata has been identified as a candidate for biological control. However, there are no biocontrol agents currently in operation. The identification of distinct lineages in the United States suggests at least three independent introductions. A genetic survey of N. cristata in its native range could help inform the search for biological control agents, especially if introductions can be traced to distinct geographic origins. Further, any biological control agents identified should be tested on the distinct genetic lineages identified in the United States, as biocontrol efficacy can vary at the subspecific level. Finally, it is possible that the distinct lineages could have distinct environmental preferences or tolerances that could be important for predicting their spread across the landscape.
Introduction
Genetic variation can influence plant invasions and management (i.e., Barrett Reference Barrett, Jain and Botsford1992; Bossdorf et al. Reference Bossdorf, Auge, Lafuma, Rogers, Siemann and Prati2005; Sakai et al. Reference Sakai, Allendorf, Holt, Lodge, Molofsky, With, Baughman, Cabin, Cohen, Ellstrand, McCauley, O’Neil, Parker, Thompson and Weller2001; Ward et al. Reference Ward, Gaskin and Wilson2008). Additionally, genetic variation can facilitate adaptation to new environments encountered after introductions and during range expansion (e.g., Lee Reference Lee2002; Prentis et al. Reference Prentis, Wilson, Dormontt, Richardson and Lowe2008). For example, distinct genetic populations or lineages of invasive plants may vary in their response to herbicides (e.g., Chorak and Thum Reference Chorak and Thum2020; Kay Reference Kay1992; Kurniadie et al. Reference Kurniadie, Widianto, Widayat, Umiyati, Nasahi and Kato-Noguchi2021; Netherland and Willey Reference Netherland and Willey2017; Williams et al. Reference Williams, Harms, Knight, Grewell, Futrell and Pratt2020). Similarly, genetic lineages can differ in the effect of biological control agents (e.g., Blossey et al. Reference Blossey, Häfliger, Tewksbury, Dávalos and Casagrande2018; Bruckart et al. Reference Bruckart, Cavin, Vajna, Schwarczinger and Ryan2004; Sobhian et al. Reference Sobhian, Ryan, Khamraev, Pitcairn and Bell2003; Williams et al. Reference Williams, Friedman, Gaskin and Norton2014).
The amount and distribution of genetic variation in invasive plants will be primarily influenced by the number of introductions; the genetic structure and diversity of source populations; and the extent of admixture, sexual reproduction, and asexual reproduction in the introduced range. At one extreme, introduction of a single genet followed by exclusively asexual reproduction will result in essentially no genetic diversity in the introduced range (e.g., Hollingsworth and Bailey Reference Hollingsworth and Bailey2000; Le Roux et al. Reference Le Roux, Wieczorek, Wright and Tran2007; Loomis and Fishman Reference Loomis and Fishman2009; Zhang et al. Reference Zhang, Zhang and Barrett2010). At the other extreme, multiple introductions from genetically distinct source populations, followed by admixture and sexual reproduction, can generate greater and novel genetic diversity in introduced populations (Facon et al. Reference Facon, Pointier, Jarne, Sarda and David2008; Kolbe et al. Reference Kolbe, Glor, Rodríquez Schettino, Lara, Larson and Losos2004; Lavergne and Molofsky Reference Lavergne and Molofsky2007). Population genetic and genomic descriptions of invasive species can help determine the number of distinct lineages present in the introduced range and provide insight into the extent or potential for admixture of distinct lineages, which in turn can inform studies on herbicidal and biological control development.
Crested floating heart [Nymphoides cristata (Roxb.) Kuntze] is a floating-leaved aquatic plant native to Asia that has spread across the southeastern United States. The first introduction of N. cristata appears to have occurred in south Florida circa 1996, when plants escaped from cultivation for the aquatic garden trade (Burks Reference Burks2002). Since then, N. cristata has been observed in Louisiana (2012), Mississippi (2016), North Carolina (2017), South Carolina (2006), and Texas (2014) and is considered an invasive/noxious species in several of the inhabited states (Thayer and Pfingsten Reference Thayer and Pfingsten2024). The scattered distribution of N. cristata populations throughout the southeastern United States may reflect multiple introductions (Burks Reference Burks2002), but no genetic analysis has been conducted to date to test whether introduced N. cristata consists of one or more distinct lineages.
Nymphoides cristata is capable of both asexual and sexual reproduction, although the prevailing mode of reproduction remains largely unknown. In its native range, N. cristata reproduces sexually through a gynodioecious breeding system derived from heterostyly, wherein female plants with reduced, sterile stamens, rely on the bisexual-morph plants for pollen (Nair Reference Nair1973). In addition, bisexual plants from the native range have shown self-compatibility in an experimental setting (Nair Reference Nair1973). However, asexual reproduction via vegetative propagation is likely responsible for the majority of biomass (Nair Reference Nair1973; Sculthorpe Reference Sculthorpe1967), a strategy common in the Nymphoides genus (Gettys et al. Reference Gettys, Della Torre, Thayer and Markovich2017; Sivarajan and Joseph Reference Sivarajan and Joseph1993). Similarly, N. cristata is thought to reproduce primarily through vegetative means in the introduced range (fragmentation, daughter plants, tubers, and rhizomes) (Burks Reference Burks2002; Willey and Langeland Reference Willey and Langeland2011). Spread and range expansion of the species is likely facilitated via fragmentation caused by contact with boat motors, wave action, and mechanical harvesting (Burks Reference Burks2002; Willey et al. Reference Willey, Netherland, Haller and Langeland2014). Nymphoides cristata is also capable of producing seeds (Burks Reference Burks2002; Gettys et al. Reference Gettys, Della Torre, Thayer and Markovich2017), and has hybridized with a native species, big floating heart [Nymphoides aquatica (J.F. Gmel.) Kuntze] (Harms et al. Reference Harms, Thum, Gettys, Markovich, French, Simantel and Richardson2021), but the extent of sexual reproduction in U.S. populations remains unknown.
In this study, we conducted a population genomic survey of introduced N. cristata in the southern United States to determine how much genetic diversity is in the introduced range. For example, if the N. cristata invasion results from asexual spread of a single genetic clone, then we would expect to find no population genomic variation in our survey (barring somatic mutation and genotyping error). In contrast, identification of multiple distinct genetic lineages would suggest multiple introductions from different sources and the potential for intermediate genotypes if distinct lineages have hybridized.
Materials and Methods
Sample Collection
Samples were obtained from a previous study on N. cristata (see Harms et al. Reference Harms, Thum, Gettys, Markovich, French, Simantel and Richardson2021), using preserved DNA samples and dried plant tissues. Briefly, Harms et al. (Reference Harms, Thum, Gettys, Markovich, French, Simantel and Richardson2021) collected 1 to 13 plants per water body, depending on the size of the infestation, with small populations minimally sampled (i.e., only a few plants). Within water bodies, plants were sampled 3 to 5 m apart to avoid repeatedly collecting the same plant. The final dataset consisted of 62 samples (i.e., leaves from N. cristata plants) from 12 different water bodies throughout the southeast United States, including lakes, ponds, roadside ditches, and canals (Table 1). We acknowledge that our sampling strategy limits our inference regarding diversity within water bodies. However, the main focus of this study was to evaluate evidence for one versus multiple genetic lineages in the United States, and given the likelihood of local clonal reproduction, we prioritized the number of water bodies examined over the number of individuals per water body.
Table 1. Water bodies where Nymphoides cristata were collected for use in this study, including genetic lineage assignments (Lineage ID) and number of individuals (N) collected from each water body.
DNA Sequencing
We used next-generation, genotyping by sequencing (GBS) to generate a single-nucleotide polymorphism (SNP) dataset. Before sequencing, DNA was extracted from plant tissues using the Qiagen DNeasyPlant Kit (Qiagen, Valencia, CA) following the standard plant protocol. A Qubit fluorometer (ThermoFisher Scientific, Waltham, MA) was used to confirm genomic DNA content was high enough for sequencing (≥60 ng total gDNA), then all extracts were sent to the University of Minnesota Genomic Center for library assembly and sequencing. The sequencing library was prepared for double digest restriction-site associated DNA sequencing (ddRAD; Peterson et al. Reference Peterson, Weber, Kay, Fisher and Hoekstra2012) using the restriction enzyme pair PstI and MspI and size-selected for 101-bp fragments using the PippinHT system. The library was sequenced on an Illumina NovaSeq 6000 (Illumina, San Diego, CA) system targeting approximately 2.5M single-end reads per sample.
Bioinformatics and Filtering
Following sequencing, we processed the raw sequence data using a bioinformatics pipeline to produce an SNP dataset for downstream diversity analyses. First, the reads were demultiplexed by barcode, and adapters were trimmed using gbstrim, a custom script designed to preprocess GBS data generated by UMGC (Garbe Reference Garbe2022). The demultiplexed reads were then passed to Stacks v. 2.55 (Catchen et al. Reference Catchen, Hohenlohe, Bassham, Amores and Cresko2013), where the process_radtags module removed low-quality reads (i.e., “Phred” score < 25). Next, the core de novo pipeline was executed in Stacks with the minimum stack depth (m), mismatch distance between loci within an individual (M), and number of mismatches between loci in the catalog (n) parameters set at 5, 4, and 4, respectively. These were selected following a parameter optimization procedure similar to that outlined in Paris et al. (Reference Paris, Stevens and Catchen2017). After catalog creation, SNP identification and genotyping were also performed in Stacks, retaining only one biallelic SNP per RAD locus. The resulting variant-calling files were exported for further filtering and analysis. All bioinformatic work in this study was performed on the Montana State University Tempest computing cluster.
To ensure only high-quality variants were included in downstream analysis, the genetic data were filtered with the R package vcfR v. 1.13.0 (Knaus and Grunwald Reference Knaus and Grunwald2017). To reduce the amount of missing data, we excluded loci that were absent in >25% of the individuals and removed any individuals with >75% missing genotype calls. We also removed loci with unusual read depth (under 10th percentile or over 90th percentile) and set the minimum minor allele frequency to 0.05. After filtering, the final genetic dataset contained 62 N. cristata samples genotyped at 2,242 SNPs.
Data Analysis
To avoid violating any model assumptions associated with the clonality of N. cristata, we used a model-free approach to explore the amount and distribution of genetic diversity. A Euclidean genetic distance matrix was generated in adegenet v. 2.1.7 (Jombart and Ahmed Reference Jombart and Ahmed2011) for use in principal components analysis (PCA). The PCA was completed in the ade4 package (Dray and Dufour Reference Dray and Dufour2007) to visualize genetic variation among samples and determine whether distinct genetic groups of N. cristata exist. We also generated a genetic dissimilarity matrix using poppr v. 2.9.3 (Kamvar et al. Reference Kamvar, Tabima and Grunwald2014) to summarize the actual number of SNP differences between individuals. The distribution of N. cristata across the invaded range was mapped with ggplot2 (Wickham Reference Wickham2016). All data analyses were conducted in R v. 4.2.1 (R Core Team 2021).
Results and Discussion
Our survey of N. cristata diversity identified three distinct genetic groups (Figure 1). We refer to those groups as genetic lineages, collections of closely related individuals distinguished by the genomic variants (i.e., SNPs) inherited from a common ancestor. Genetic variation was found both within and between lineages, although the amount of between-lineage variation was far greater than within-lineage variation. Within lineages, individuals averaged approximately 35 SNP differences, while individuals compared across lineages differed by an average of 1,028 to 1,514 differences (Table 2).
Figure 1. Principal component analysis (PCA) of 62 Nymphoides cristata samples. PC1 and PC2 account for 96.1% of cumulative variation. Colored ellipses (nonstatistical) were added a posteriori to denote putative genetic lineages identified by ordination.
Table 2. Summary of individual-based genetic distance within and between lineages.a
a Reported values along the diagonal are the absolute number of single-nucleotide polymorphism (SNP) differences between individuals within a lineage. Between-lineage values represent the average number of SNP differences observed across lineages.
The CFH-2 lineage was the most common and widespread; it was found in Florida, Louisiana, and Texas and occurred in 10 of the 12 water bodies sampled overall (Figure 2). In contrast, CFH-3 was only found in one water body (Lake Marion, SC), and CFH-1 was only found in two water bodies, both in southeast Florida (Figure 2).
Figure 2. Distribution of Nymphoides cristata lineages across the introduced range. Points represent water bodies where samples were collected, and colors denote the genetic lineage(s) present, assigned according to the principal component analysis (PCA) analysis in Figure 1. Note there are two overlapping points representing the roadside canal in south Florida where CFH-1 and CFH-2 co-occurred.
The differences in relative abundance among lineages may be due to introduction dynamics and/or ecological differences among lineages. Nymphoides cristata was likely brought to the United States for trade as an aquatic garden ornamental and remains available for purchase from vendors in the industry. Although it is now illegal to possess, import, or distribute N. cristata in several southern states (FL, SC, NC, TX), the ornamental industry is a likely candidate for initial introduction(s) and subsequent range expansion. The relative abundance of the CFH-2 lineage may indicate that it was a preferential/popular lineage in the industry or that it was the first to escape cultivation. Indeed, N. cristata is clonally propagated for sale in the industry, although it was not possible to include commercially available ornamental samples in this study due to time and funding constraints. In addition, CFH-2 may be more vigorous and widely adapted in the introduced range than others, facilitating its spread across the southern states, while CFH-1 and CFH-3 remained isolated. Further investigation would be necessary to determine whether CFH-2 presents a greater management challenge. We recognize that the number of water bodies sampled and number of individuals sampled within a water body were limited and there may be more diversity in the introduced range than we detected.
Burks (Reference Burks2002) suggested multiple introductions, possibly escaped from ornamental water gardens, based on the scattered distribution of N. cristata across southern Florida. Indeed, we identified two distinct lineages in Florida: one that was restricted to southeastern Florida, and a second lineage that is widespread across Florida and the U.S. Gulf Coast (Figure 2). These may represent two independent introductions in Florida followed by range expansion of CFH-2. It is also possible that CFH-2 has been repeatedly introduced across the Gulf Coast states, with its widespread distribution representing numerous independent introductions. Further, we identified a third unique lineage found only in South Carolina (CFH-3), which could represent another introduction.
We cannot rule out with certainty the alternative hypotheses of a single introduction from a genetically variable source or accumulation of new mutations following introduction. However, we would have expected more within-water body lineage diversity if N. cristata was introduced from a genetically variable source (although we recognize that within-water body sample sizes were low). Similarly, we find clonal evolution postintroduction unlikely because of the relatively large number of allelic differences (1,028 to 1,514) separating the three distinct lineages combined with the relatively recent introduction (∼1996). Plant invasions resulting from multiple introductions are common (Dlugosch and Parker Reference Dlugosch and Parker2008), particularly for ornamental species like N. cristata, for which the plant trade increases the likelihood of repeated introductions (Dehnen-Schmutz et al. Reference Dehnen-Schmutz, Touza, Perrings and Williamson2007). Sampling efforts in the native range (i.e., southeastern Asia), along with additional sampling in the United States, could help clarify the number and location(s) of sources of N. cristata introduction. In addition, N. cristata has been identified as a good candidate for biological control, although natural enemies/potential agents have yet to be described (Harms and Nachtrieb Reference Harms and Nachtrieb2019). Identification of source populations in the native range might also yield natural enemies of this plant that could be tested as potential biocontrol agents (Bossdorf et al. Reference Bossdorf, Auge, Lafuma, Rogers, Siemann and Prati2005; Gaskin et al. Reference Gaskin, Bon, Cock, Cristofaro, Biase, Clerck-Floate, Ellison, Hinz, Hufbauer, Julien and Sforza2011).
The introduction of three distinct lineages provides opportunities for genetic admixture and the generation of novel genetic variants in the introduced range (e.g., Facon et al. Reference Facon, Pointier, Jarne, Sarda and David2008; Kolbe et al. Reference Kolbe, Glor, Rodríquez Schettino, Lara, Larson and Losos2004; Lavergne and Molofsky Reference Lavergne and Molofsky2007). Although interspecific hybrids between N. cristata and native N. aquatica have been identified in the Santee Cooper Reservoir system in South Carolina (Harms et al. Reference Harms, Thum, Gettys, Markovich, French, Simantel and Richardson2021), we did not find any evidence for sexual reproduction among the three lineages, as evidenced by the lack of genetically intermediate individuals (Figure 1). In addition, the average pairwise genetic distances between individuals were approximately 97% greater when comparing between lineages versus within lineages. It is possible that these lineages have sexually reproduced with one another, but that we did not sample any of their offspring. Alternatively, it is possible that the distinct lineages are capable of sexual reproduction but have not yet had sufficient opportunity. The three lineages were largely allopatric, but there was some overlap between two of them in south Florida. Finally, it is possible that the different lineages have some reproductive barriers (e.g., pre- or post-mating, pre- or post-zygotic) that limit sexual reproduction between them. Additional study of reproductive potential among the introduced lineages is warranted.
While we cannot rule out some degree of sexual reproduction within lineages, we hypothesize that the low genetic variation observed within each of the lineages primarily reflects sequencing/genotyping errors and that the N. cristata lineages identified here have primarily reproduced asexually throughout the southeastern United States. Although individuals within each lineage were not genetically identical per se, clonal genotypes are not expected to have identical genotypes across thousands of SNPs generated by ddRAD due to sequencing/genotyping error and somatic mutations (da Cunha et al. Reference da Cunha, Xue, Wright and Barrett2021; Reynes et al. Reference Reynes, Thibaut, Mauger, Blanfune, Holon, Cruaud, Couloux, Valero and Aurelle2021). The interpretation of the low within-lineage variation as clonal reproduction is consistent with previous field observations and reports of reproductive biology of N. cristata that hypothesize prolific vegetative propagation and spread in the invaded range (Burks Reference Burks2002; Harms and Nachtrieb Reference Harms and Nachtrieb2019; Willey and Langeland Reference Willey and Langeland2011). Inbreeding could also account for the within-lineage variation we observed. Nymphoides cristata has been proven to be self-compatible through artificial pollination of bisexual plants in an experimental setting (Nair Reference Nair1973). However, heterostyly provides a morphological-based incompatibility system in N. cristata. Although incomplete, this system is thought to promote reproduction between the female and bisexual plant morphs (i.e., dioecism) in the native range (Nair Reference Nair1973). Further sampling and detailed description of flower morphology in the introduced range could help decipher the reproductive capabilities of the introduced populations. Finally, we acknowledge that the low sample sizes within water bodies preclude an understanding of the relative extent of sexual versus asexual reproduction in any single population of N. cristata.
Acknowledgments
The authors would like to thank Greg Chorak, Del Hannay, and Ashley Wolfe for providing thoughtful comments to improve the article. We thank Cassidy Kempf, Michael Coulon, Tim Bister, Jacob Green, Casey Moorer, John Riser, and Carl Bussells for assistance with sample collection.
Funding statement
This work was supported by the U.S. Army Engineer Research and Development Center Aquatic Plant Control Research Program Cooperative Agreement W912HZ-18-2-0010, and the Montana Agricultural Experiment Station (Project MONB00249).
Competing interests
The authors declare no conflicts of interest.
Open access
