New invasion or expansion: evaluating the genetic relationships of Bactrocera dorsalis (Diptera: Tephritidae) among detections in Florida

2.1 Specimens and DNA extractions

A total of 127 B. dorsalis captured in traps in Florida between 25 August 2014 and 16 August 2021 representing 10 different detection events (Table 1, Figure 1 [map]) were included in the study. The 115 B. dorsalis captured in Redland, Miami-Dade county between 27 August 2015 and 10 October 2015, were classified as the Redland outbreak population. Two additional flies trapped in Miami-Dade county (Palmetto Bay and Miami) in 2015 were treated as separate detection events from the Redland population for initial analyses. The study also included additional flies from Florida captured in 2014 (N = 1), 2016 (N = 1), 2017 (N = 3), 2018 (N = 4), and 2021 (N = 1). The flies from Florida were morphologically identified as B. dorsalis by Gary Steck. Legs from each fly were removed and used for DNA extraction using the DNeasy Blood and Tissue Kit (Qiagen, Valencia, California, USA) as described in Barr et al. (2012). To support morphological identification, flies were analyzed using internal transcribed spacer 1 (ITS1) region sequences as reported in Barr et al. (2021) (data not shown). ITS1 is used for distinguishing B. dorsalis from Bactrocera carambolae Drew & Hancock (Boykin et al. 2014; Drew and Hancock 1994).

Figure 1:

Collection locations of oriental fruit flies in Florida from 2014 to 2021. Location, year and number of flies detected is provided for each detection.

To compare the Florida 2015 outbreak population to other fly populations, 352 B. dorsalis collected from four Hawaiian islands and 24 fruit flies originating from southeast Asia were selected as reference populations (Table 2). The Hawaiian populations were selected to provide an estimate of SSR diversity using an introduced population and source exclusion. The flies from southeast Asia were included to provide an independent measure of SSR diversity of individual flies from more diverse potential source populations. Each of these flies had DNA extracted from the whole body. However, only legs were provided for the flies captured in Florida. The legless bodies were kept as vouchers by the Florida Department of Agriculture and Consumer Services.

2.2 COI analysis

DNA samples of flies from Florida were analyzed for the COI gene using a protocol reported in Barr et al. (2014a). Reactions were performed in 25 µl volumes containing 1 µl of template (or water for the negative control), 2.5 µl of 10X Ex Taq buffer (Takara Bio Inc., Kyoto, Japan), 2 µl of dNTP (2.5 mM each, Takara Bio Inc.), 0.125 µl of Ex TaqHS DNA polymerase (5 U/µl, Takara Bio Inc.), and 1 µl of each primer (at 10 µM working concentration). Polymerase chain reactions (PCR) were performed using cycling conditions of 3 min at 94 °C followed by 39 cycles of 20 s at 94 °C, 20 s at 53 °C, and 30 s at 72 °C, and a final extension of 5 min at 72 °C. The amplified fragment is referred to as the C3p790 fragment in Barr et al. (2014a) and includes a 790 bp region of the 3′ portion of the gene. The primers for the PCR were HCO-2198RC [5′-TGATTTTTTGGTCACCCTGAAGTTTA-3′] and PAT-K508 [5′-TCCAATGCACTAATCTGCCATATTA-3′].

PCR products were visualized using 1.2 % agarose gels of TAE buffer (BioRad, Hercules, California, USA) and purified with ExoSAP-IT (ThermoFisher, Applied Biosystems, Waltham, Massachusetts, USA) prior to Sanger DNA sequencing. The amplicons were sequenced using the two PCR primers and ABI BigDye^® Terminator v.3.1 chemistry at commercial centers Functional Biosciences (Madison, Wisconsin, USA) or GeneWiz (South Plainfield, New Jersey, USA). The forward and reverse sequences were assembled into contigs using the program Sequencher v5 (Genecodes, Ann Arbor, Michigan, USA) and aligned using MEGA7 (Kumar et al. 2016). Templeton, Crandall, and Sing (TCS) haplotype networks of the COI data were generated using PopART v1.7 (Clement et al. 2002; Leigh and Bryant 2015; Templeton et al. 1992). Tajima’s D statistic was estimated using Arlequin v3.5. Sequence data were submitted to NCBI GenBank (OP243539–OP243560).

2.3 SSR analysis

A total of nine microsatellites, 618A, BO-D48 (Li et al. 2007), BD1, BD19, BD39, BD42, BD76, BD85B (Aketarawong et al. 2006), and MS5 (Dai et al. 2004) were used, end-labeled with either 6-carboxyfluorescein (6-FAM) (ThermoFisher, Applied Biosystems, Waltham, Massachusetts, USA) or VIC^® (ThermoFisher, Applied Biosystems, Waltham, Massachusetts, USA) dye set. Polymerase chain reactions were performed in 15 µl reactions containing 1 µl DNA template, 1.5 µl of 10X buffer, 1.2 µl 25 mM dNTP mix, 0.3 µl of labeled 5′ primer (10 nmol; Applied Biosystems), 0.3 µl of unlabeled 3′ primer (10 nmol Eurofins Genomics LLC, Louisville, Kentucky, USA), and 0.08 µl taqDNA Polymerase (TaKaRa Ex Taq Hot Start Polymerase, Takara Bio USA, San Jose, California, USA). Amplification was performed on Applied Biosystem’s Gene Amp 9700 thermocyclers. Cycling conditions were 94 °C for 5 min followed by 39 cycles of 1 min at 94 °C, 1 min at 55 °C, 1 min at 72 °C, and a final extension at 72 °C for 30 min. An aliquot (10 µl) of PCR product was visualized on a 2.0 % TAE agarose gel pre-stained with ethidium bromide (0.4 μg/ml final concentration). Documentation of gels was performed using the GelDocIt^TS2 Imager (Analytik Jena GmbH, UVP, Beverly, Massachusetts, USA) and VisionWorksLS Imaging Software v 7.1 (Analytik Jena GmbH, UVP, Beverly, Massachusetts, USA). An aliquot (2 µl) of the remaining PCR product was diluted 1:10 in water for fragment analysis. Fragment analysis was performed on an ABI 3730XL DNA Analyzer with the ABI Data Collection Program (v 2.0) provided by Functional Bioscience (Madison, Wisconsin, USA).

PeakScanner v2.0 (ThermoFisher, Applied Biosystems, Waltham, Massachusetts, USA) was used to determine fragment size, and binning was performed using Microsatellite Toolkit v3.1 (Park 2001) and Tandem (Ruby v1.08) (Matschiner and Salzburger 2009). Population differentiation was determined by F _st (Weir and Cockerham 1984) for all reference populations using Arlequin (Excoffier and Lischer 2010), and a Bayesian Markov Chain Monte Carlo (MCMC) based approach was implemented using STRUCTURE v2.3.4 (Pritchard et al. 2000) using 25 independent runs (K = 2–5) with 100,000 generation burn-in over 1,000,000 generations. Individual B. dorsalis flies detected before and after the 2015 Redland event and an outgroup of flies representing the southeast Asian population were then compared to the Hawaiian and the 2015 Redland outbreak collection using Geneclass2 (Piry et al. 2004). Assignment tests also were performed on a fly collected from the 2015 Redland outbreak (E2015-4798) as a reference point and the two flies collected in 2015 from Miami-Dade county in Palmetto Bay and Miami to determine if the Redland outbreak site could be excluded as a source.

To observe if changes occurred in allele and homozygosity values over the outbreak, the Redland population was divided into three temporal lifecycles: f1 (17 August–7 September 2015), f2 (8–22 September 2015), and f3 (23 September–13 October 2015). This division of cycles was based on generational life cycles of B. dorsalis (Borchert et al. 2011). The allele frequency of each division was estimated using Arlequin and changes in heterozygosity were estimated.

3.1 COI analysis

COI sequences were successfully generated from 124 of the 127 extraction samples. After editing, the sequence length of data records was 765 bp and comparison in an alignment detected no evidence of insertions or deletions or disruptions to reading frame of translated amino acids. These 124 records included nine distinct COI sequence haplotypes (Table 3). A TCS network of the nine haplotypes is presented in Figure 2. The greatest distance between any two sequences in the network is 13 mutational steps and an uncorrected p-distance of 0.017.

Figure 2:

TCS haplotype network of the cytochrome oxidase I data generated from Bactrocera dorsalis collected in Florida in 2014–2021. The haplotypes are presented as circles and each circle size represents the number of flies with that haplotype. The black circles represent hypothetical haplotypes that connect the sampled haplotypes. Mutation (substitution) steps are represented by the lines in between haplotypes. The figure was generated using TCS method in PopArt (http://popart.otago.ac.nz).

The 112 samples that represent the Redland outbreak population in 2015 included three haplotypes. The most common haplotype was the C3pX326 (N = 98) followed by C3pX357 (N = 11), and C3pX359 (N = 3). The two additional flies detected in 2015 from Palmetto Bay and Miami shared the C3pX326 haplotype. Consequently, these 114 flies from Miami-Dade county in 2015 were treated as the 2015 outbreak population. The estimated mean diversity of the 2015 outbreak population was d = 0.001 (S.E. 0.001) using p-distance and 1,000 bootstraps. The data set had seven segregating sites, Theta of 0.00170, and nucleotide diversity of 0.00127. Estimates of Tajima’s D statistic were not significant (p > 0.10).

The single fly collected in Broward county in 2014 had a haplotype sequence (C3pX358, Table 3) that was not previously reported but was only one mutation different from a haplotype (C3p02) frequently collected in Hawaiian populations of B. dorsalis (Barr et al. 2014b). The 2014 haplotype has not been reported in previous studies including amongst any flies captured in Florida. The flies trapped in Florida from 2016 to 2021 do not share haplotype sequences with the 2015 outbreak flies. The four flies detected in 2016 and 2017 had distinct haplotypes from each other. The St. Petersburg fly trapped in 2016 had a new haplotype (C3pX364) separated by at least four mutations from the other flies trapped in Florida. This haplotype has not been reported in previous studies. The three flies captured in 2017 each had unique haplotypes. The flies from Lake county and Pinellas county had new haplotypes. The flies from Broward county (2017) and Seminole county (2021) had the haplotype (C3p01), which is common to Hawaii (Barr et al. 2014b), found frequently in California (Barr et al. 2014b), and reported from China and Tahiti (San Jose et al. 2018). The four flies trapped in 2018 all shared the C3pX373 haplotype, which has not been previously reported.

Based on the Barr et al. (2014b) study, Hawaii can be excluded as a possible source for 122 (out of 124) of the flies in the study including all the flies collected in the years 2015, 2016, and 2018, but Asia and Africa are not excluded as possible sources of these flies. Only three flies had DNA consistent with the Hawaiian populations: the flies caught in 2014 and 2017 from Broward county and the 2021 fly from Seminole county. Failure to exclude Hawaii suggests a link with that source, but alternative regions also are possible, such as Tahiti and Asia where similar types have been reported. Prior studies have reported high COI genetic variation for B. dorsalis and a lack of structure needed to reliably exclude or assign the origin of flies from Asia (Barr et al. 2014a; Garzón-Orduña et al. 2019; San Jose et al. 2018). Consistent with these prior results, the Florida flies include seven COI haplotypes that have not been previously reported.

3.2 SSR population structure of reference populations

PCR of the SSR markers for the reference populations was highly successful with 99.4 % of all 3,364 PCR reactions producing results for analysis (Table 4). A total of 65 alleles across nine loci was detected. A total of 487 B. dorsalis originating from the state of Hawaii (N = 352), southeast Asia (N = 24), and captured during the Redland 2015 infestation (N = 111) were successfully genotyped and was used as reference populations for comparing diversity and performing tests of structure. The Hawaiian reference collection includes four sub-populations of the Big Island (Hawaiʻi), Maui, Oʻahu, and Kauaʻi (Tables 3 and 4). The Redland 2015 reference population did not include the two flies found in Palmetto Bay and Miami in 2015 (Table 1). The southeast Asia individuals were used to examine heterozygosity values but not to test structure based on a priori collection data because that reference population represents a composite of flies from multiple collections and likely not a single natural population. All individuals were included in Structure models to sort according to genetic similarity.

B. dorsalis individuals originating from Asia show the highest diversity of alleles (x = 6.56), followed by the fruit flies collected during the Redland outbreak in 2015 (x = 4.57). However, heterozygosity was low (0.12) in the flies captured in Redland in comparison to Hawaii ( x ‾  = 0.453) and SE Asia (0.505). F _st Pairwise distance analysis was then performed on the reference populations (Table 5). The closest genetic relationships (smallest genetic differentiation) in the study were exhibited between populations collected from each of the sampled Hawaiian Islands. The highest genetic distance measured among the population of the Hawaiian Islands was seen between the furthest islands (Hawaiʻi and Kauaʻi, F _st = 0.0314). However, all Hawaiian fly populations were more closely related to flies collected from southeast Asia than to the flies collected in Redland, Florida.

3.3 SSR Bayesian structure analysis of individuals from each detection event

In addition to the Redland reference flies, 11 flies captured in Florida from 2014 to 2018 were included in SSR analyses. Two of the four flies collected from Miami Dade county in June 2018 failed to produce adequate PCR fragments for analysis. There was also difficulty in producing SSR fragments of the BD1 locus with >40 % of the Florida flies failing. This could be due to low DNA yield from the extraction process. BD1 was able to amplify a PCR fragment from all the Hawaiian reference DNA samples and Asian outgroup, all of which had DNA extracted from whole bodies. PCR fragments were obtained for nine specimens using the remaining eight markers. This set of nine flies includes the Palmetto Bay and Miami flies trapped in 2015 within the same county as the Redland flies. Structure analysis of SSR profiles from a data set including the 487 reference individuals and the nine additional Florida collections is reported in Figure 3.

Figure 3:

Bayesian clustering pattern for Bactrocera dorsalis reference population and flies captured between 2014 and 2021 in Florida. (Top) Assignment probabilities assuming an admixture ancestry model with independent allelic frequencies. Used a LocPrior distribution and organized data based on geographic distance and estimate (K = 2–5) in STRUCTURE v2.3.4 (Pritchard et al. 2002) for the Hawaiian Islands, southeast Asia and the Redland outbreak and each individual fly for population assignment. (Bottom) A blow up of the captured oriental fruit flies analyzed independent of reference population and compared to reference populations. Based on the observations seen in the assignment test, none of the other captured flies observed in the other capture events could possibly be a direct descendent of the flies detected during the Redland detection event of 2015.

Bayesian analysis using STRUCTURE was performed including all individuals. Sample sites were divided into six groups (Hawaiʻi n = 90; Maui n = 88; Oʻahu n = 91; Kauaʻi n = 94; southeast Asia n = 24; and Redland n = 113). Population (K) was estimated to be between 2 and 5. In all estimates of K, Hawaiʻi strongly separated out as a single unique population with variations observed between the four islands. In contrast, at the lowest population estimates (K = 2) Redland and samples from southeast Asia had a structure consistent with a single population.

The genetic similarity observed between Redland and the small composite collection of specimens used to represent southeast Asia tends to indicate a shared heritage at lower K values. However, this might not have any biologically significance as the Hawaiian population is strongly structured relative to the other flies in the dataset. At higher K values (K > 3), Redland begins to separate away from the representative southeast Asian flies and begins exhibiting a structure consistent with a single population. By using these higher K values, we can then compare individual capture events to Hawaiian Islands and Redland. The representative Redland 2015 fly (E2015-4996) and the two flies collected outside the Redland quarantine zone in 2015 (Palmetto Bay (E2015-4798) and Miami (E2015-5172)) were tested and found to 100 % match to flies collected in Redland at K = 1–4 (Figure 3). Flies from the other seven detection events did not exhibit enough shared alleles to allow grouping with flies found during the Redland detection event. One fly (E2017-2956) had high similarity to flies found on the Hawaiian Islands when K was set at 2 and 3, but less similarity under higher values of K. Finally, the fly collected in 2021 (E2021-4133-1) was highly similar to the population observed on the Hawaiian Islands across all values of K indicating a shared relationship.

3.4 SSR assignment tests for captured flies

SSR first generation assignment tests (Paetkau et al. 2004) were performed for 10 fruit flies detected in Florida before, during, and after the 2015 Redland outbreak (Table 6). One fly from Broward county, Plantation (E2014-6069) was captured before the infestation (and could not be progeny of the outbreak), two flies were captured during the outbreak (but outside Redland area), and seven detections were made after the Redland infestation was declared eradicated. In 2018, four flies were captured within close proximity of the Redland infestation of 2015, but only one fly (E2018-3066-1) yielded high quality SSR results for all alleles, and was the only fly used in assignment testing.

As expected, a Redland source could not be rejected for the first 2015 Redland fly captured for first generation assignment tests using Geneclass2. A Redland 2015 source also was not rejected for the two flies collected in 2015 from Palmetto Bay and Miami. The other tested flies all had the Redland reference population rejected as the probable source (familial genetic relationship) with statistical significance (p < 0.05).

The Broward county fly captured in 2014 (E2014-6069) predates the Redland outbreak and would be excluded as possible progeny of that outbreak. The assignment tests demonstrated that the four Hawaiian populations also are excluded as a source of this fly. Although the analysis rejects southeast Asia as a source for this fly, that Asian reference data set is not representative of the true variability of southeast Asia and cannot be used to exclude it as a source. Therefore, a southeast Asian source is possible for this fly. Although provided in Table 6, the rejection or failure to reject the southeast Asia population as the source is not based on sufficient sampling. Therefore, this value is not reported or interpreted for other flies in this section.

The Broward county fly from 2017 (E2017-2956) did not have Hawaii excluded as a source. This is consistent with the mitochondrial data that was a match to a Hawaiian source. The exclusion tests performed here do not include adequate material to exclude other sources such as Asia, Africa, and the Pacific islands for this fly.

Finally, four detections of B. dorsalis occurred north of Miami-Dade county and were genetically different from the 2015 outbreak. For the two flies from Pinellas county (St. Petersburg 2016, Clearwater 2017) and the fly from Lake county (Mount Dora 2017), Hawaii was rejected as a source. For the most recent detection, the 2021 detection in Seminole county (E2021-4133-1), Hawaii was not rejected as a source. This fly had a COI sequence that is commonly observed in Hawaii.