Investigating Liriomyza (Diptera: Agromyzidae) Populations From Northeastern Brazil: mtDNA Analyses of the Global Pests L. sativae and L. huidobrensis

Abstract Species of Liriomyza Mik (Diptera: Agromyzidae) occur worldwide and are economically important leafminers. However, populations of some pest species, although very similar morphologically, show highly divergent mtDNA sequences, suggesting that nominal species are in fact complexes of cryptic species. This study focuses on two globally invasive pests, L. huidobrensis (Blanchard) and L. sativae Blanchard, already known to be highly destructive in Brazilian crops, although only a few studies on morphological and genetic divergences of them have been made. A total of 63 sequences of the mitochondrial cytochrome oxidase I (COI) and cytochrome oxidase II (COII) from Brazilian populations of L. huidobrensis and L. sativae collected from six crops (gypsophila, chrysanthemum, melon, watermelon, tomato, and onion) in Northeastern Brazil were generated to investigate their genetic structure together with available sequences from the Americas, Europe, Asia, Africa, and Australia. Genetic structure was not found to be correlated to neither host plant nor geographical locality. Liriomyza huidobrensis showed an overall low intraspecific global genetic divergence in both genes. On the other hand, high intraspecific divergences for L. sativae and its phylogenetic position confirm a divergent clade currently found only in Brazil and suggest it may be a global complex of cryptic species. Considering the possibility of cryptic species (in the latter case), we provided detailed redescriptions of these Brazilian populations for future studies and local management of these global pests. Finally, our results also revealed a new synonym herein proposed, L. strigosa Spencer as a junior synonym of L. huidobrensis.

pregonite always lacking and surstyli of various shapes, fully reduced or completely fused with epandrium (Zlobin 1996). Other similar genera, including Haplopeodes Steyskal, Phytoliriomyza Hendel, and Metopomyza Enderlein, together with Liriomyza form a monophyletic group (Lonsdale 2011). Galiomyza Spencer was recently synonymized with Liriomyza due to the presence of an apically swollen and pigmented ejaculatory duct, considered a unique and consistent synapomorphy (Lonsdale 2017).
Liriomyza huidobrensis (Blanchard) and L. sativae Blanchard are two of the seven species considered polyphagous and are important examples of potential agricultural pests worldwide (Spencer 1973, Lonsdale 2011. Liriomyza huidobrensis was first described from Argentina, mining leaves on Cineraria sp. (Asteraceae) in Buenos Aires (Blanchard 1926), but has already been recorded from 365 species of 49 botanical families (Weintraub et al. 2017). This species is globally known as a pest species, however, it has been eradicated or currently controlled by natural enemies in some parts of the world. In South and Central America and Africa it is still a serious agricultural pest. Newly established populations can be devastating, and L. huidobrensis is currently the most destructive pest among agromyzids (Scheffer et al. 2014). Liriomyza huidobrensis and L. langei Frick are closely related and were once considered synonyms (Spencer 1973). Both species are morphologically very similar and it is not always possible to distinguish them only by external morphology; although analyses of the barcode region of mitochondrial cytochrome oxidase I has improved species identification (Weintraub et al. 2017). Liriomyza sativae was described from alfafa as a host plant in La Pampa, Argentina (Blanchard 1938), but occurs in many host plants (Spencer 1973). Its coloration is very varied, closely resembling other species, such as L. brassicae (Riley), L. eupatorii (Kaltenbach), L. helianthi Spencer, L. sabaziae Spencer, some morphs of L. langei (Lonsdale 2011), and its sister species L. trifolii (Burgess) (Scheffer and Lewis 2006).
Seventeen species of Liriomyza have been recorded from Brazil (Sousa et al. unpublished). Nevertheless, both L. huidobrensis and L. sativae have been recorded as the most severe Liriomyza pests on Cucurbitaceae, Fabaceae, Solanaceae, and Asteraceae (Esposito et al. 1992, Costa-Lima et al. 2010, Araujo et al. 2013, Ferreira et al. 2017, Parish et al. 2017. Other species have few host crop records, such as, L. trifolii on coffee (Silva et al. 2015) and L. brassicae on Brassicaceae (Parish et al. 2017). Despite their economic importance in Brazil, little is known about their biology, host plants, as well as, their genetic structure. For example, only recently, an analysis of partial cytochrome c oxidase subunit I sequences revealed a divergent lineage of L. sativae in Brazil (Parish et al. 2017).
Phylogenetic studies with DNA sequences have been conducted for Liriomyza identifications and population surveys (Scheffer 2000, Scheffer andLewis 2005). In some of the pest species studied (e.g., L. cicerina (Rondani), L. sativae, and L. trifolii), high mtDNA sequence divergence was found among populations, and authors suggest that they represent potential cryptic species (Scheffer andLewis 2005, 2006;Pérez-Alquicilla et al. 2018;Carapelli et al. 2018). Other molecular methods were also used to discriminate leafminer species, such as Polymerase Chain Reaction-Restriction Fragment-Length Polymorphism (PCR-RFLP) (Scheffer et al. 2001, Kox et al. 2005 and Mutiplex PCR method (Miura et al. 2004, Nakamura et al. 2013. The objective of our study was to investigate the genetic structure of L. huidobrensis and L. sativae from six cultivated crops in Northeastern Brazil, based on cytochrome oxidase subunit I and subunit II. The use of both markers was needed to compare our results to previously published works that used different genes or gene regions. Our results indicate that genetic structure was not found to be correlated to neither host plant nor geographical locality. Furthermore, comparisons of Brazilian populations of both species with populations from other regions of the world were conducted (which has never been conducted for L. huidobrensis) resulting in an overall low intraspecific global genetic divergence in L. huidobrensis and confirmation of a divergent clade of L. sativae currently found only in Brazil.

Specimen Sampling
Leaves with mines were removed from all crops and transported to the laboratory, where they were placed in plastic containers covered with organza mesh until adult emergence. Adults were then mounted, labeled, and identified. For the morphological identification, taxonomic keys were used (Boucher 2010;Lonsdale 2011Lonsdale , 2017. Male and female terminalia were clarified in KOH 10%, prepared in acetic acid, 95-70% ethanol, and placed on temporary slides for dissection and analysis. Larvae were mounted on permanent slides with butyl acetate. High resolution images of adults were made using Leica video camera DFC450C attached to a Leica M205 C stereoscope with the Programa Leica Application Suite software version 4.8.0. Digital images of the larval structures, pupae, and male terminalia were made using an optical microscope NIKONECLIPSE E200 MV R, with the software Zen 2 (version 2.0). Terminology followed Cumming and Wood (2017), with some exceptions: 'ori' and 'ors'
Polymerase chain reactions (PCR) were set up using 12.5 μl of the GoTaq G2 green Master Mix (Promega), 1.0 μl of each primer (10mM), 1.0-4.0 μl of DNA extract, and DEPC water to a total volume of 25 μl. PCRs were carried on a Veriti 96-well Thermal Cycler (Applied Biosystems), with the following program: 95°C for 3 min; 35 cycles of 95°C for 1 min, 53°C for 1 min, 72°C for 1 min; and 72°C for 5 min. The volume of 3.0 μl of the PCR product was submitted to an agarose gel electrophoresis for visualization of positive bands stained with GelRed (Biotium) under a UV transilluminator. Successfully amplified PCR products of the correct size were sent to Macrogen (South Korea) for purification and Sanger sequencing of both DNA strands.
Neighbor-joining and genetic distances were calculated using MEGA X v.10.0.5 (Kumar et al. 2018). Genetic distances between sequences were estimated using uncorrected distances (p distances) for fragment COI-3′ and COII or distances modeled by Kimura-2-parameter (K2P) for COI-5′ (given that most literature use this substitution model). Neighbor-joining trees with bootstrap support with 500 pseudoreplicates were calculated for the three separate alignments, following the same models used for the genetic distance calculations. Bayesian phylogenetic analyses were conducted on the separate matrices with MrBayes 3.2.6 (Ronquist et al. 2012) with 4 independent runs of 4 chains for 50 million generations sampling every 5,000. Convergence and parameter mixing were evaluated using Tracer 1.71 (Rambaut ). All markers were modeled under GTR+I+G based on AIC comparisons of likelihood scores of 24 different models with PartitionFinder 2.1.1 (Lanfear et al. 2016).
To evaluate genetic structuring among populations of focal species from different countries in the world, matrices of COI-5′ (26 and 181 sequences), COI-3′ (93 and 288 sequences), and COII (52 sequences of L. huidobrensis) including only L. huidobrensis or L. sativae, respectively, were constructed based on the above-mentioned ones. Haplotype networks of these three markers for each species were constructed using the TCS network method (Clement et al. 2002) with PopART v.4.8.4 (Leigh and Bryant 2015). Phylogenetic trees and haplotype networks were edited in Adobe illustrator CS6 v.16.0.0 (Adobe Systems Inc.).

Nomenclature
This paper has been registered in Zoobank (www.zoobank. org), the official register of the International Commission on Zoological Nomenclature. The LSID (Life Science Identifier) number of the publication is: urn:lsid:zoobank. org:pub:44890547-034A-4A05-942A-7DC726D756D7
Bayesian analyses showed similar results between species groups for three fragments of COI and COII genes (Figs. 2-4) (see complete trees in Supp Fig. T1-T3 [online only]). Posterior probabilities and NJ bootstrap values for Liriomyza species were 1.00 and 96-100%, in most cases, respectively. Exceptions to this high clade support were for L. sativae with 0.98 and 76% in the COI-3′ and 76% in the COI-5′ analyses. Present analyses confirmed the identification of Brazilian specimens as belonging to both species treated in this study, because they were each recovered in clades with conspecifics.
Haplotype networks of L. huidobrensis showed a single Brazilian haplotype for COI-5′ (10 sequences) shared with 2 sequences from Indonesia, with low levels of genetic differentiation from other sampled populations from Canada, and other countries in Africa and Asia (Fig. 5a). Similarly, a single Brazilian haplotype for COI-3′ (11 sequences) was found shared with sampled populations from Canada, Netherlands, South Africa, and countries in Asia (Fig. 5b) and also a unique Brazilian haplotype for COII (10 sequences), with few genetic differences from populations from Canada, Netherlands, Guatemala, and other countries in South America and Asia (Fig. 5c).
Haplotype networks for L. sativae revealed six different Brazilian haplotypes for COI-5′ (123 sequences), with low levels of genetic differentiation between them (Fig. 6a). However, these haplotypes show significant differences from the sampled populations of Australia, Mexico, and other countries from Asia. The COI-3′ revealed five different haplotypes from Brazil (23 sequences) and also suggested a fairly high genetic differentiation of Brazil populations from haplotypes found in Australia, Egypt, and countries throughout the Americas and Asia (Fig. 6b).

Diagnosis.
Wing length 1.75-2.6 mm (♂), 1.9-2.7 (♀); two ori directed inwardly or dorsally (anterior ori sometimes shorter; one in some females); two ors; head light yellow (some males, pale yellow), with back of head brown to base of outer vertical setae; antenna yellow (some specimens with pedicel and postpedicel brown); calypter grey, margin and fringe brown; halters yellow; legs with coxae and femora mostly yellow with light brown maculae, tibiae and tarsi dark brown; abdomen dark brown, sometimes yellowish medially on second tergite; male terminalia black (Fig. 7e); oviscape dark brown (Fig. 7k); epandrium Table 4. Range, mean, and standard deviation of COI-5′ interspecific sequence divergences (K2P) expressed as percentages for L. huidobrensis and L. sativae with one apical spine; surstylus with single subapical spine (Fig. 7g); basiphallus with a long gap before mesophallus; hypophallus welldeveloped; mesophallus sclerotized laterally; distiphallus cylindrical, with apex less sclerotized and basal margin angled, sclerotized and upcurved in lateral view (Fig. 7h-i); ejaculatory apodeme short and narrow, with sperm pump rounded (Fig. 7j). Host Plants. Linear mine in Gypsophila paniculata L. (Fig. 9a) and linear mine in Chrysanthemum morifolium Ramat. (Fig. 9b). Comments. Lonsdale (2011) observed paler specimens found mostly among males and a few females from South American countries, including Brazil, congruent with our specimens. He mentioned that these paler specimens have a wing length 1.9-2.2 mm in males, a dark brown abdomen, and sometimes a yellow medial line on the second tergite. Darker specimens, from Chile and populations introduced in western North America have more black instead of brown, with dorsal 1/6-1/4 of anepisternum yellow and wing length usually 2.3-2.6 mm in males. Paler specimens from Quebec, Canada, have a 1.9 mm wing length, and a laterally yellow abdomen with only large anteromedial spot on tergite 6 (Lonsdale 2017). Although the yellow lateral line on the abdomen is found in specimens herein studied, the anteromedial spot on tergite 6 was not observed. Spencer (1963) described Liriomyza strigosa from unreared specimens collected in Santa Catarina, Brazil, and mentioned that it could be easily recognized by the matt scutum and two rows of acrostichals. We analyzed the holotype of L. strigosa  deposited at the BMNH and observed that there are no differences between the male terminalia of L. huidobrensis  and L. strigosa (Fig. 10d). Both species have one strong spine at the epandrium and one on surstylus, a long membranous gap between basiphallus and distiphallus, as well as a membranous gap before the distiphallus. The opaque scutum and presence of two rows of acrostichals as in the original description are also found among populations of L. huidobrensis. Based on these observations, we propose L. strigosa as a junior synonym of L. huidobrensis.
Host Plants. Cucumis melo L. (Fig. 9c), Citrullus lanatus (Thunb.) Matsum. & Nakai (Fig. 9d), Solanum lycopersicum L. (Fig. 9e), and Allium cepa L. (Fig. 9f). Comments. Color variation was found among studied specimens with hind tibiae and tarsi brown, calypter margin and fringe brownish yellow, anepisternum with only 1/3 or less brown. Lonsdale (2011) mentioned some color variations that he observed in his specimens, including variable brown markings on the anepimeron and the anteroventral corner of the anepisternum, and the anepisternum is sometimes predominantly brown along the ventral margin. He also observed that specimens from western North America are sometimes darker with only the dorsal 1/4 of the anepisternum (similar to that seen in our analyzed specimens from Brazil), meron, and katepisternum yellow. Specimens reared from Jacaranda Juss. have also a total lack of pigment, and other specimens rarely have the lateral yellow stripes on scutum connected along the posterior Table 6. Range, mean, and standard deviation of COII interspecific sequence divergences (uncorrected) expressed as percentages for L. huidobrensis and L. sativae

Among-Crop Genetic Structure
For both species, no genetic structuring was found among crops sampled. For Brazilian populations of L. huidobrensis from Gypsophila and Chrysantemum, a single haplotype was found for each molecular marker. The Brazilian haplotype of COI-3′ sampled herein from L. huidobrensis was the same found previously by Parish et al. (2017) sampled from six individuals from lettuce, cucumber, and bean from Minas Gerais and São Paulo states. Similarly, based on other molecular markers of L. huidobrensis, no genetic structuring among different crops has previously been found by Scheffer (2000) based on COII, where a single haplotype from W. Java, Sri Lanka, Israel, and Guatemala, was found from celery, cabbage, potato, fava bean, mustard, snow pea, leek, and Chrysantemum. Additionally, a single haplotype of COII from L. huidobrensis was sampled from nine different host plants throughout Yunnan Province in China (He et al. 2002).
Likewise, although more genetic variation was found among samples of Brazilian L. sativae in both markers as compared to L. huidobrensis, no evidence of genetic structuring was found among crops. As an example, the most sampled haplotype of L. sativae in Brazil of COI-5′ (Fig. 13, SA.1) was collected from all four crops studied herein (tomato, melon, watermelon, and onion) in addition to bean (Ferreira et al. 2017) from Espírito

Genetically Homogeneous Liriomyza huidobrensis
Different haplotypes found for L. huidobrensis from around the world show very few differences, with pairwise intraspecific distances up to 1.1% (COI-5′), 2.0% (COI-3′), and 0.9% (COII). Considering the COII marker, there is more genetic structuring within South America (three Brazil and Ecuador haplotypes separated by 5 substitutions) when compared to the other genetically homogeneous regions of the world (three haplotypes separated by two substitutions). Previous authors have suggested that the European and recently introduced populations (e.g., Middle East and Asian) of L. huidobrensis originated from somewhere in South America (de Goffau 1991, van der Linden 1993, Scheffer 2000, which could be true given the higher genetic variability seen among COII South American populations studied herein. Nevertheless, a larger sampling of South American populations is needed to corroborate this hypothesis. In disagreement with our findings, Scheffer (2000) reported very high pairwise intraspecific distances up to 5.33% for COI-3′+COII, distances calculated from individuals belonging to two monophyletic clades, one consisting of samples from California and Hawaii and the other from Central and South America. However, Scheffer et al. (2014) elaborated a new multiplex PCR method and found that those L. huidobrensis reported from California and Hawaii are truly L. langei and that probably L. huidobrensis does not occur in California and Hawaii (Scheffer 2000, Lonsdale 2011 andS. Scheffer personal communication). In the present analysis, these mistaken sequence identifications were corrected to L. langei.

Liriomyza sativae: Adding Evidence for a Species Complex
Within L. sativae, high intraspecific pairwise genetic divergences for both fragments of COI, with 0-6.9% (Table 3) for COI-5′, and 0-9.6% (Table 3) for COI-3′ were observed. These findings are in agreement with previous results, which demonstrated high distances between COI clades within L. sativae ranging from 2.5-8.8%. High divergence values were also observed in another related species, L. trifolii, with uncorrected pairwise distances between major clades ranging from 4.7-5.7% (Scheffer and Lewis 2006). Nevertheless, according to Scheffer and Lewis (2005), the mitochondrial variation found within L. sativae exhibits some geographic structure. The clade 'sativae-A' was found only in Florida, Guatemala, Honduras, and Mexico (recently found by Pérez-Alquicira et al. (2018)); clade 'sativae-L' was found only in Colorado on wild lupines; and clade 'sativae-W' was more widespread, being found in Florida, California, Arizona, Mexico (recently found by Pérez-Alquicira et al. 2018), and all invasive Old World populations. Our present phylogenetic results from COI-5′ (Fig. 2) recovered three clades with maximum posterior probability values: L. sativae-W (formed by Australia, Bangladesh, China, French Polynesia, India, Japan, Pakistan, Papua New Guinea, and Sri Lanka populations) and sister to a clade composed of L. sativae-A (Mexico) + L. sativae-B (Brazil), while the COI-3′ analysis (Fig. 3) recovered four clades also with maximum posterior probability values: L. sativae-A (also including Guatemala, USA: Florida and Honduras) + L. sativae-B (Brazil) as sister to the clade composed by L. sativae-L (USA: Colorado) + L. sativae-W (formed by Israel, Japan, Malaysia, Papua New Guinea, Philippines, Saudi Arabia, Sri Lanka, USA, Venezuela, and Vietnam). These clades had previously been recovered by Scheffer and Lewis (2005) and Parish et al. (2017), while the confirmation of the Brazilian clade with the COI-5′ marker is unprecedented. Scheffer and Lewis (2005) suggested that more genetic diversity may be found in the Americas. Two studies on L. sativae populations from Brazil have been previosly conducted (Parish et al. 2017, Ferreira et al. 2017) and although both used the COI gene, each sequenced a different region of it. Thus, we have decided to work with both fragments herein, to be able to contrast our results.  Southeastern Brazil (Espírito Santo State), but they did not compare their sequences to other available L. sativae sequences. They do show a maximum likelihood tree, including their sequences and a single other L. sativae sequence (KF962593) from Bangladesh, that shows a very long branch separating them, but unfortunately, the tree has no scale bar and no sequence divergence is reported between these. In any case, all of L. sativae COI-5′ sequences generated herein group with Ferreira et al.'s sequences and also newly corroborate the 'sativae-B' clade with this marker (Fig. 2). Thus, both markers indicate a single unique lineage of L. sativae in Brazil, sister to the 'sativae-A' clade, which shows sequence divergences of 2.4-4.6% in COI-5′ and 1.8-2.6% in COI-3′ to members of the 'sativae-A' clade.
Deep mitochondrial divergences and phylogenetic clade structure found within species of L. cicerina, L. sativae, and L. trifolii made previous authors suggest the presence of cryptic species (Scheffer andLewis 2005, 2006;Carapelli et al. 2018). Considering this, we are following Scheffer and Lewis (2005) in the assumption that the Brazilian endemic population also represents another possible cryptic species in this complex. Apparently, these cryptic species may show particular geographic restrictions and/or host preferences, such as the case of lupin-associated L. sativae in Colorado (Scheffer and Lewis 2005) and host-associated cryptic species of L. cicerina in Tunisia (Carapelli et al. 2018). Although the Brazilian endemic populations reported herein do not have any particular host preference, these haplotypes have not yet been found outside of Brazil. Thus, more sampling of L. sativae in other areas of South America is still needed to corroborate or refute the geographical restriction of this supposedly cryptic species. For example, Scheffer and Lewis (2006) suggested that within the L. trifolii-W clade, a shallowly diverged, but phylogenetically distinct subclade restricted to peppers (from Mexico, Hoduras, Florida, and California) provided substantial evidence for another separate species in this complex that seemed to be host-associated. However, the distinction of this supposedly pepper-associated clade was refuted by Pérez-Alquicilla et al. (2018) in a study involving Mexican L. trifolii populations of peppers, tomatillos, and onions.

Conclusions
Brazilian populations of both Liriomyza leafminer species studied have shown no or little intraspecific genetic divergences in all mtDNA markers studied, even though specimens were collected in different crops and geographic regions of the country. In the case of L. huidobrensis, only a single haplotype was found in each of the three markers, and in the case of COI (HA.1 and HB.1) haplotypes, they were also shared with Old World populations. However, in the case of the COII the Brazilian haplotype found was restricted to Brazil, and was slightly more divergent from Ecuadorian populations than other Old World haplotypes. This apparent higher genetic variability suggests that future mtDNA studies of L. huidobrensis populations in South America should perhaps focus on this genetic region. For L. sativae, our study added support to the previous suggestion that this species may be in fact a complex of cryptic species, and that further sampling of American populations is needed to test this hypothesis and to investigate the origin of introduced Old World populations. The present study also confirmed the presence of a single divergent L. sativae lineage in Brazil based on two Further investigations on leafminers focusing on morphology, biology, host plants, and integrated molecular data for different countries are strongly recommended. This will provide more information to establish the pest status, invasion processes, gene flow, and speciation.

Supplementary Data
Supplementary data are available at Annals of the Entomological Society of America online.