A Case of Mistaken Identity: Genetic and Anatomical Evidence Reveals the Cryptic Invasion of Xenopus tropicalis in Central Florida

Nonnative species present a major hazard to global biodiversity (Mack et al., 2000) and are one of the largest drivers of contemporary extinction (Bellard et al., 2016). Furthermore, nonnative species threaten human and ecosystem health (Juliano and Lounibos, 2005; Pejchar and Mooney, 2009) and cost the United States alone approximately $120 billion annually (Pimentel et al., 2005). Given these large costs, prevention of establishment and containment of extant invasions are key to preserving biodiversity (Molnar et al., 2008). Containment can require intensive and long-term management, aimed at controlling the spread of the invader (Meyer et al., 2011). Effective management of nonnative species often requires an approach tailored to the target species' biology (Lennox et al., 2015), making it imperative to correctly identify the species of interest. The failure to correctly identify a nonnative species is a phenomenon known as cryptic invasion (Geller et al., 2010).

Researchers often prioritize morphological features to assign taxonomic status, and as a result, cryptic invasions abound (Morais and Reichard, 2018). Although external morphology can be reliably used to distinguish certain species, closely related species often require molecular approaches to disambiguate (Bruschi et al., 2013). Otherwise, cryptic invaders may go unnoticed (Verloove, 2010; Hill, 2017) and known invaders may continue to proliferate (Thum et al., 2012). Taxonomic confusion can also reduce the reliability of models aimed at quantifying the target invader's potential for range expansion (Mori et al., 2018). This is particularly true for ecological niche models (ENMs; Hernandez et al., 2006), which rely on environmental data collected from within the target species' native range (Peterson, 2003).

In 2016, Hill et al. (2017) discovered a high-density breeding population of an African Clawed Frog of the genus Xenopus in west central Florida. Initially identified as Xenopus laevis, subsequent inspection of adult individuals revealed external morphology more consistent with members of the subgenus Silurana (Evans et al., 2015). In addition, experimental evidence suggests that individuals within the west central Florida population may be more adapted to higher temperature regimes and reduced thermal variation. Specifically, individuals from within the west central Florida population are characterized by an increased heat tolerance, reduced cold tolerance, and more narrow thermal breadth, relative to X. laevis (Goodman et al., 2019; Araspin et al., 2020). Here, we resolve the taxonomic confusion concerning the Florida population by using analyses of DNA sequence data and osteological data from microcomputed tomographic (CT) scans. These analyses reveal this population to be the world's only known nonnative population of X. tropicalis, thereby providing a unique opportunity for making comparisons with the invasion biology of the well-studied and widely invasive X. laevis. In addition, we provide detailed information on the current distribution of X. tropicalis in Florida and discuss the implications for future management.

We conducted all laboratory work at the Florida Museum of Natural History. We extracted genomic DNA from tissues (liver, muscle, or toe clips) by using Qiagen DNeasy kits, following their protocol for animals. Using polymerase chain reaction (PCR), we amplified an ∼520-base pair fragment of mitochondrial DNA that encodes part of the mitochondrial ribosomal 16S gene (95°C 30 s, 54°C 30 s, 72°C 1 min) by using 35 cycles and the oligonucleotide primers 16sar-L 16sbr-H (Palumbi et al., 2002). We used ExoSAP-IT (Affymetrixs) to purify all amplified PCR products and then shipped this product for Sanger sequencing at Genewiz. We deposited all sequences in GenBank (Appendix 1).

We created contigs in Geneious v.9.0.5 (Biomatters; http://www.geneious.com), aligned consensus sequences using MAFFT v.7.309 (Katoh and Standley, 2013), and checked manually for gross errors such as sequences in the wrong strand orientation. We then combined our sequences with the most complete taxonomic sampling of Xenopus to date (Evans et al., 2015). We trimmed the resulting block of aligned sequences so that all individuals had complete data, resulting in a final alignment of 119 samples, 498 base pairs in length with 113 informative sites. Using ModelFinder (Kalyaanamoorthy et al., 2017), we found that the best model of nucleotide substitution was GTR+F+I+G4. We created a maximum likelihood tree by using IQtree (Nguyen et al., 2015) with ultrafast bootstrapping (1,000 replicates; Hoang et al., 2018). We calculated the average pairwise distance between our Florida samples (N = 16) and nonidentical sequences of X. tropicalis from GenBank (N = 12), by using MegaX v.10.1.7 (Kumar et al., 2018). We applied a bootstrap (500 replicates) to estimate variance and assigned uniform rates of substitution within and between the Florida and native samples. There were no ambiguous positions in our final dataset.

To confirm the subgeneric identity, we used high-resolution CT. We performed all scans on ethanol-preserved specimens, at the University of Florida's Nanoscale Research Facility, by using a Phoenix v|tome|x M (GE Measurement & Control Solutions, Boston, MA, USA). We customized voltage and current for each specimen to balance resolution and intensity contrast. We scanned the entirety of four specimens and the anterior half of one additional specimen at an increased resolution (35.06 and 15.44 μm, respectively). We used GE's reconstruction software datos|x to convert x-ray images into tomogram images. We then stacked these images and imported them into VG STUDIOMAX v.3.3 (Volume Graphics, Heidelberg, Germany). Using the segmentation tools in VG, we isolated the skull and first presacral vertebrae as members of the subgenus Silurana, which includes X. tropicalis, are distinguished from members of the subgenus Xenopus by unfused nasal bones, an absent vomer, and fusion of the first two presacral vertebrae (Cannatella and Trueb, 1988; Evans et al., 2015). We uploaded all the resulting image stacks into MorphoSource (Appendix 2).

We conducted all surveys with the use of modified minnow traps, baited with chicken liver (Measey and Tinsley, 1998), excepting one pond where we relied on visual surveying because of landowner constraints. We placed traps in each respective water body around sunset and removed them the following morning, with an approximate trapping time of 12 h. The number of traps deployed at each site varied depending upon the size of the pond and the number of traps available. We deployed a minimum of 1 trap and a maximum of 30 traps per site during each sampling occasion. The large area of many ponds within the invaded area precluded us from standardizing effort across all sites. We recorded trap locations with the use of a Garmin Etrex GPS (3-m accuracy). We selected sites through the use of satellite imagery (Google Earth^TM). We prioritized water bodies in proximity to previous records collected by the Florida Fish and Wildlife Conservation Commission (FWC) (Hill et al., 2017) and based upon verified reports from residents in the area. Between May 2018 and October 2019, we sampled a total of 43 water bodies a minimum of two times. During each survey, we noted the life stage of any individuals of Xenopus observed. We designated water bodies as breeding sites if we observed any larvae of Xenopus during any survey. We delineated the extent of known occurrence by generating a minimum convex polygon (MCP) around all presence coordinates in ArcGIS 10.7.1 (ESRI, Redlands, CA). MCPs represent the smallest polygon around all points of interest such that no internal angle is greater than 180° (Mohr, 1947).

Our maximum likelihood phylogeny provides strong support that the Florida population is X. tropicalis (Fig. 1). The average pairwise distance between our Florida population and native populations from West Africa is 0.00212, translating to a similarity of 99.79%. The average genetic diversity for all X. tropicalis samples (N = 28) is 0.4%. The low intraspecific variation among specimens from the native range precludes us from identifying a source population for the Florida population.

The five specimens scanned exhibited osteological features associated with the subgenus Silurana, including unfused nasal bones, a lack of vomer bones, and fusion of the first two presacral vertebrae (Fig. 2). These features preclude the possibility of this population being X. laevis and corroborate the analysis of DNA sequences, indicating that these represent X. tropicalis, which is part of Silurana (Evans et al., 2015).

Of the 43 water bodies sampled, we detected the presence of X. tropicalis in 22 of these water bodies, including the single water body where we relied on visual surveys. We observed clear signs of breeding in 11 of these water bodies based on the presence of tadpoles of Xenopus. Including all verified reports and data collected previously by the FWC, the total number of documented occurrences increased to 26 sites, and the number of known breeding sites increased to 12 (Fig. 3; Appendix 3). The MCP around all known occurrence points represents an area of 1,630 ha.

Our results indicate that in contrast to previous reports (Krysko et al., 2016; Hill et al., 2017), the extant invasion of Xenopus in Riverview, Florida, is X. tropicalis and not the globally distributed nonnative species X. laevis. This represents the first population of X. tropicalis outside of its native range in West Africa. Although the vector and age of the introduction are unknown, the initial record of a Xenopus sp. in Hillsborough County dates back to the 1970s (Tinsley and McCoid, 1996). Researchers collected this individual near an animal import facility, close to, or within, Riverview (Godley, pers. comm.), and deposited it into scientific collections at the University of South Florida (McDiarmid, pers. comm.). Researchers initially identified this specimen as X. laevis. However, this was because of the species' ubiquity in global invasions rather than any formal identification (McDiarmid, pers. comm.), and an extended renovation of the University of South Florida's Science Center led to the destruction of the specimen (Mushinsky, pers. comm.). Given the proximity to currently occupied sites, this observation may represent the starting point of the extant invasion of X. tropicalis. This would imply an invasion detection lag of at least 34 yr. However, the imprecision of observation details and lack of any vouchered specimen make it impossible to determine the specific identity of this specimen. Indeed, both X. laevis and X. tropicalis are present in pet trade and are used in biomedical research (Beck and Slack, 2001; Measey, 2017). However, because X. tropicalis is diploid (opposed to tetraploid X. laevis; Kobel et al., 1996), it has become increasingly popular in genomics research (O'Rourke, 2007) and might possess a similar capacity for global release as X. laevis.

It remains unclear what impact the invasion of X. tropicalis in Florida is having on native species, although there is a growing body of literature on the invasive ecology of congener X. laevis. With at least 10 extant populations spanning four continents (Measey et al., 2012), X. laevis has been linked to reduced reproduction in native anurans (Lillo et al., 2011), altered macroinvertebrate composition (Courant et al., 2018a), and reduced amphibian species richness (Courant et al., 2018b). Xenopus tropicalis may present similar risks to Florida where native anurans are already at risk because of habitat degradation (Delis et al., 1996) and the presence of other invaders (Smith, 2005; Guzy et al., 2006). Both X. tropicalis and X. laevis are generalists, predominantly consuming aquatic macroinvertebrates (Measey, 1998; Imasuen and Aisien, 2016; Courant et al., 2017). However, both have distinct maturation rates, with X. tropicalis having a much shorter generation time (Hirsch et al., 2002). Generation time is an important life-history characteristic and is linked to an invasive species' ability to colonize new areas (Sakai et al., 2001). In addition, Allen et al. (2017) found that within amphibians and reptiles, species with fast life-history traits were more likely to successfully establish novel populations and to spread. It is unclear whether this disparity in life history between X. laevis and X. tropicalis will cause a concomitant disparity in the rate of invasion spread. Given the uncertainty of the source location and age of the X. tropicalis invasion in Florida, it is impossible to directly compare the spread rates of invasive X. tropicalis and X. laevis populations. However, even if the current X. tropicalis population dates to the initial vouchered specimen in 2013 (FLMNH 172054), this would still imply an annual rate of spread of <1 km, which is below the estimated rates of the French and Chilean X. laevis populations (Lobos and Jaksic, 2005; Fouquet and Measey, 2006).

In addition to life-history disparities, X. tropicalis and X. laevis occupy distinct thermal niches (Saito et al., 2016) and have nonoverlapping distributions in their respective native ranges (Furman et al., 2015). These differences are likely to affect the predicted habitat suitability in Florida for this Xenopus population. Indeed, ENMs have predicted most of Florida to be unsuitable for X. laevis (Measey, 2012; Rödder et al., 2017; but see Ihlow et al., 2016). Across its native and much of its invaded range, X. laevis is generally associated with Mediterranean or oceanic climates (Fouquet and Measey, 2006; Ihlow et al., 2016). Conversely, X. tropicalis is associated with tropical regions along the rain forest belt in sub-Saharan Africa (Rödel, 2000). This difference suggests that peninsular Florida may be at higher risk for further invasion of X. tropicalis than it would be for X. laevis. There have been at least three other confirmed releases of X. laevis in Florida (Krysko et al., 2016), including one release of ∼200 individuals in the greater Miami area (King and Krakauer, 1966). Despite these releases, none have resulted in the presence of an established breeding population. Although it is impossible to rule out low number of individuals released as the cause of these failures, X. laevis is known to have an ability to recover from low population bottlenecks (Measey and Tinsley, 1998; Measey, 2001). This fact, coupled with low predictions of climate suitability within the area, suggests that the novel environment was at minimum a contributing factor to these failures.

Along with the potential for mismatch in predicting the rate of and ability to spread, cryptic invasions may also impact the efficacy of management efforts. One control strategy that has received localized success at extirpating X. laevis is the use of chemical toxicants. Wildlife managers in Lacey, Washington, appear to have successfully eradicated a small X. laevis population by artificially increasing the salinity of an occupied water body (Boone, 2017). Near the University of California, Davis, managers eradicated a population through the use of the insecticide Thiodan (Zacuto, 1975). At Golden Gate Park in San Francisco, managers were able to extirpate an isolated population by using hydrated lime, which significantly elevated the pH of both the water and sediment to a level lethal to X. laevis (Larson, pers. comm.). Numerous surveys have been conducted in subsequent years, and no additional X. laevis individuals have been detected (Larson, pers. comm.), suggesting that eradication efforts were successful. In 2016, the FWC used the same strategy in an attempt to eradicate the X. tropicalis invasion in Florida, treating occupied ponds with hydrated lime (Sommers, 2016). Although the treatment did appear to have an acute impact, with the population suffering high mortality, managers observed frogs at the same sites within weeks, and X. tropicalis still occupied all previously treated sites at the time of this study. It remains unclear whether the differences in treatment success arise from interspecific differences in toxicant susceptibility—as X. tropicalis and X. laevis last shared a common ancestor >30 million yr ago (Feng et al., 2017)—or from differences in application. Future research should seek to determine which toxicant is best suited for managing X. tropicalis. In addition, care should be taken before application, as any toxicant used is likely to have spillover effects on native species (Baker et al., 2013).

Although the range of X. tropicalis in Riverview appears to be localized, our estimated range of occurrence (1,630 ha) is based on relatively few samples and is therefore likely to be an underestimation, representing the minimum invaded range of X. tropicalis. Given that the genome of X. tropicalis is fully sequenced (Hellsten et al., 2010), future research should focus on developing methods of eDNA detection. The use of eDNA has proven reliable for the detection of congener X. laevis (Secondi et al., 2016) and might facilitate a more exhaustive survey in areas surrounding the known invaded range.

In conclusion, these findings underscore the importance of using contemporary methods to obtain taxonomic validation of novel invaders and present a novel opportunity to compare the invasion trajectories of two closely related species, X. laevis and X. tropicalis.

This research was conducted with financial support from the Florida Fish and Wildlife Conservation Commission (FWC contract no. 13416-A3035). We thank D. Quinn, T. Dallas, and C. Cunningham of FWC for providing preliminary data and C. Watson and staff at the University of Florida, Institute of Food and Agricultural Sciences Tropical Aquaculture Laboratory for providing material support. Thanks to N. Claunch and O. Nathan for offering editorial suggestions. We also thank two reviewers for their helpful comments on this article. This study's methods comply with national and international guidelines and received Institutional Animal Care and Use Committee (IACUC) approval (University of Florida IACUC no. 201810274). This study adheres to the Association for the Study of Animal Behaviour/Animal Behavior Society Guidelines for the Use of Animals in Research. DNA sequence data are available at GenBank (accession nos. MT345772–MT345787). CT scan data are available through www.morphosource.org (see Appendix 2).