Molecular phylogenetic analysis and morphological reassessments of thief ants identify a new potential case of biological invasions

Species delimitation offered by DNA-based approaches can provide important insights into the natural history and diversity of species, but the cogency of such processes is limited without multigene phylogenies. Recent attempts to barcode various Solenopsidini ant taxa (Hymenoptera: Formicidae: Myrmicinae), including the thief ant Solenopsis saudiensis Sharaf & Aldawood, 2011 described from the Kingdom of Saudi Arabia (KSA), were precipitated by the unexpected existence of a closely related species, the Nearctic S. abdita Thompson, 1989 within the S. molesta species complex native to Florida. This finding left the species status of the former uncertain. Here, we investigated the taxonomy and phylogeny of these two species to determine whether or not S. abdita represents a new global tramp species. We inferred a phylogeny of the two species using DNA sequence data from four nuclear genes (Abd-A, EF1α-F1, EF1α-F2, and Wingless) and one mitochondrial gene (COI) sampled from populations in Florida, Guatemala, Hawaii, and Saudi Arabia. Both species clustered into one distinct and robust clade. The taxonomy of S. saudiensis was re‐examined using morphometrics. A reassessment of the morphological characters used to diagnose the worker and queen castes were consistent with molecular evidence. Based on combined morphological and molecular evidences S. saudiensis is declared as a junior synonym of S. abdita (syn. nov.). In addition, our findings indicate that S. abdita is a novel global tramp species which has a far wider distribution than previously thought and has established itself in many new habitats and different geographic realms.


Measurements and indices.
Measurements and indices were performed as previously described 3,29,58 . All measurements are in millimeters. Molecular data generation. The phylogenetic relationships among our samples were inferred using molecular data from Shreve et al. 56 , who sequenced four nuclear genes (see below) and COI to estimate a global phylogeny of Solenopsis. Their data was subsampled to include representative New World species discussed by Rasool et al. 55 as well as Old World species. In addition, we generated two new S. saudiensis COI barcodes from Riyadh, which were identical. Finally, all S. saudiensis, S. abdita, and S. cf. abdita samples used by Shreve et al. 56 were re-extracted and sequenced in a different laboratory to prevent contamination and ensure that no samples were mixed up. Molecular methods follow Brady et al. 59 and Moreau et al. 60 Briefly, total genomic DNA was isolated from whole single workers with the Qiagen DNeasy Blood and Tissue kit (Qiagen Inc., Valencia, CA, USA). Only a single individual from each collection event was used to avoid subsampling colonies. DNA sequence data were generated from four nuclear protein-coding genes (Abdominal-A (Abd-A), elongation factor 1-alpha F1 (EF1α-F1), elongation factor 1-alpha F2 (EF1α-F2), and Wingless (Wg)), and the mitochondrial protein-coding gene cytochrome c oxidase I (COI). Primer sequences, PCR amplification, and Sanger sequencing protocols are given in Brady et al. 59 and Moreau et al. 60 We only deviated from the given protocols by adding BSA (0.08 mg/mL final concentration) to the final PCR reaction mix and using a touchdown PCR procedure to increase specificity, which started 5 °C above the published annealing temperatures and decreasing by 0.4 °C/cycle for 12 cycles. PCR amplicons were Sanger sequenced in both directions using PCR primers and the BigDye Terminator 3.1 kit on an ABI 3730xl capillary sequencer (Applied Biosystems, Carlsbad, CA, USA). Sequence traces were assembled in Geneious Prime 2020.05 (https ://www.genei ous.com) and deposited in GenBank (GenBank accession numbers MT550038-MT550618; see Supplementary Table S2). For comparison, Rasool et al. 55 COI sequence data from S. saudiensis collected from the Riyadh region, KSA, were included (GenBank accession numbers KR916796-KR916802; see Supplementary Table S2).

Molecular data analysis.
We assembled two molecular datasets: a multilocus dataset derived from Shreve et al. 56 consisting of four nuclear loci, and COI, to better place S. saudiensis within a global biogeographic framework. We also assembled a COI barcoding dataset to compare against the S. saudiensis haploytype described by Rasool et al 55 .
Each locus was globally aligned using the global iterative refinement method (G-INSI-i) implemented in MAFFT 7.402 (Katoh and Standley 61,62 ;-globalpair-maxiterate 1000-retree 100). The concatenated multilocus dataset produced a 2,546 bp alignment, of which 457 nucleotides were variable and 338 were parsimony informative. Use of other alignment algorithms did not impact phylogenetic tree estimation. For each dataset, www.nature.com/scientificreports/ we estimated maximum likelihood trees using IQTREE 1.6.12 63 , simultaneously estimating the optimal model of nucleotide substitution using ModelFinder 64 (multilocus: SYM + R3, COI: TIM2 + F + I + G4) on an unpartitioned dataset. We estimated branch support using ultrafast bootstraps 65 55 (Fig. 1B), which includes the Rasool et al. 55 data, confirm the overall biogeographic pattern recovered with the multilocus dataset. Importantly, it also shows that all Saudi Arabian S. saudiensis samples and the single Hawaiian S. cf. abdita sample all share an identical COI haplotype.
The principal component analysis of the COI dataset shows three clusters (Fig. 2). The first principal component, which explains 24.37% of the variation and more than twice that of the second principal component, clearly separates the S. saudiensis and S. abdita samples from the other Solenopsis species. The S. saudiensis and S. abdita samples are poorly separated, and their differentiation is mainly derived by the first principal component.
According to the present work, S. elhawagryi is clearly associated with other Eurasian species and quite distinct from S. saudiensis, which falls out in a clade of Nearctic species. The molecular results are supported by our morphological analyses. The two species are easily separated based on the possession of the presence/absence of postpetiolar teeth and polymorphy/monomorphy of the worker castes. The former species has a postpetiole process in all castes and is polymorphic, whereas the latter lacks a postpetiole process and is clearly monomorphic.
Morphological reassessments/new synonymy. The genus Solenopsis, comprised of some of the smallest ants in the subfamily Myrmicinae, includes numerous minute species with minor workers less than 2 mm length. The material of S. abdita and S. saudiensis are ideally studied with high magnification microscopes, the Leica M205 C Stereomicroscope with a magnification zoom range of 20.5 × to examine and detect diagnostic characters that demonstrate clear morphological similarities between the two species. These similarities can be summarized in the following diagnosis ( Fig. S1A-F, Fig. S2A-F) (Fig. 3): monomorphic species.
Head. Eyes minute with two to five ommatidia only seen with higher magnification, more frequently two; funicular segments 3-8 about twice as broad as long; anterior clypeal margin with a central pair of stout projecting teeth and a lateral pair of short, broad, basal, blunt teeth. Mesosoma. Dorsum of mesosoma smoothly curved, not flattened before and after the metanotal groove; metanotal groove acutely impressed in profile. Postpetiole. Postpetiole about 1.3 × broader than long in dorsal view; nearly hexagonal in profile with a distinctly convex dorsal surface. Pilosity. Relatively abundant and long hairs sparse on mesosoma, petiole, postpetiole, and gaster; more than 10 erect hairs on the dorsum of promesonotum; posterior tibial hairs mostly appressed. Sculpture.    www.nature.com/scientificreports/ mesosomal dorsum, distinctly larger (HL 0.72-0.84, HW 0.64-0.78, EL 0.24), a broad petiolar node rounded in profile, and postpetiole with a distinct subpetiolar pair of teeth. Comparing S. abdita/saudiensis to S. pergandei (Forel, 1901), the three species are uniform yellow with eyes consisting of two ommatidia and present a distinct acute metanotal groove. However, S. abdita/saudiensis can be easily separated by the distinctly longer head when seen in full-face view (CI 75-87); abundant, short, subdecumbent or appressed body pilosity; and well-developed anterior central and lateral pairs of clypeal teeth. By contrast, S. pergandei has a nearly quadrate head (CI 89-93); profuse, suberect, and longer body pilosity; and a blunt central pair of anterior clypeal teeth while the lateral pair is absent.
The analysis of Rasool et al. 55 shows a sister cluster of two unrelated species, S. mameti Donisthorpe, 1946 from Mauritius and S. saevissima (Smith, 1855) from Brazil. Morphologically, S. abdita/saudiensis and S. mameti are clearly distinct, as the former is uniformly yellow, with a shallow metanotal groove and less abundant, short, scattered body pilosity, whereas the latter is unicolorous dark brown with a deep metanotal groove and abundant, long body pilosity. Solenopsis saevissima is completely different from S. abdita / saudiensis and easily separated by numerous sets of characters including brown color, strong polymorphism in any nest series, profuse, stiff, and long body pilosity, conspicuously large eyes that contains about 12 ommatidia in the longest row, and an emarginated posterior margin of head seen in full-face view.
Solenopsis abdita/S. carolinensis. Among the Nearctic species, S. abdita can be confused with S. carolinensis Forel, 1901 and Thompson 36 was not able to present a practical differential diagnosis between the two species. However, Pacheco and Mackay 3 successfully recognized the former species by the shorter scape, the broader petiole, and the appressed hairs on the tibiae whereas S. carolinensis has the tibiae with suberect hairs. In addition, the queen caste can be useful in the identification where the queen of S. abdita is dark brown and has smaller eyes, while the queen of S. carolinensis is yellow.

Discussion
The field study conducted by Rasool et al. 55 in the Riyadh region did not turn up evidence of Solenopsis species other than S. saudiensis. However, the sampling methods and efforts deployed in their study are insufficient to conclude that other Solenopsis are absent, as many localities, habitats, and microhabitats in the province, which has a high diversity of natural and agricultural habitats, were left unexplored.
Our molecular results are clearly consistent with the Arabian revision of the Solenopsis fauna 29 and the morphological traits used in species recognition. Based on molecular data as well as a morphological reevaluation of both S. abdita and S. saudiensis, our results indicate that S. saudiensis, described in 2011, represents a junior synonym of S. abdita. Thompson 36 states that types of S. abdita were deposited at the Museum of Comparative Zoology (MCZ), the Florida State Arthropod Collection in Gainesville (FSCA), and the Natural History Museum of Los Angeles County (LACM), but extensive searches in these museums were unable to locate the type materials. The absence of S. abdita types has been observed before by Pacheco and Mackay 3 . In many Solenopsis species, however, the morphological distinction of species on the basis of the worker caste is arduous (e.g.  -group 3,11 ), therefore, the study of the sexual castes including queens and males represent a useful addition for species delimitation 11 . Here, the comparison of the reproductive female caste of S. saudiensis described by Sharaf et al. 72 with the original description of the queen of S. abdita (Thompson, 1989) revealed that most of their taxonomic characters match, including body size, sculpture, pilosity and reflected a straightforward synonymy. The few minor exceptions include body color, which is dark brown in the former species and reddish brown to almost black in the latter species. However, coloration in ants presents wide variation within and between populations 3,11 . www.nature.com/scientificreports/ Ecological similarities are also found in the nesting habits of the two species, since both species were encountered nesting in palm logs (family Arecaceae) 3,36 . Solenopsis saudiensis has been collected in or near date palm plantations, Phoenix dactylifera L., on the Arabian Peninsula 28 and S. abdita has been reported to be commonly found in rotten palm logs in the USA 36 . The nesting preference of S. abdita includes a broad range of habitats that are either moist or mesic niches in Florida including sandhill, swamp forest, grass tussocks of seasonal ponds, bases of pines in flatwoods, hammocks, rotten wood and palm logs 36,73 , or bases of date palm trees in the KSA 28 where nests are built near the soil surface 73 .
These results also demonstrate the non-native status of the populations of S. abdita within KSA and represent the first known record of this species in the Old World. Introduced populations are also characterized by a reduced genetic pool as a consequence of a bottleneck effect following their introduction; which we observed in the form of populations from Saudi Arabia and Hawaii presenting identical COI sequences. The presence of this species in two regions outside its native range (the Arabian Peninsula and Hawaii), coupled with particular morphological and ecological traits such as small body size, polygyny 36 , lestobiotic lifestyle, and association with disturbed environments, supports the tramp status of this species 74 . Indeed, individuals of S. abdita in KSA were commonly encountered in date palm groves 28 and highly disturbed urban habitats (one of the two type series was found nesting under a discarded carpet next to a human settlement 28 ) but also in more natural habitats such as nature reserves (e.g. Rawdhat Khorim 75 ). However, nature reserves are not necessarily disturbance-free and sometimes even the most pristine reserve can have exotic species along roads or buildings. Together, these results contrast with the conclusion of Rasool et al. 55 of S. abdita being strongly associated with and specialized to colonize date palm groves following an adaptive process involving a large and strong gene pool.
While limited by the extent of the sampling used in our study, the results tend to indicate a New World origin for S. abdita potentially spanning the Nearctic and Neotropical realms. Given that the two samples from the www.nature.com/scientificreports/ Nearctic and the Neotropical regions (Florida and Guatemala, respectively) are genetically distinct and the species falls out in the New World clade, it seems likely that the native range is also somewhere in the New World (possibly circum-Caribbean). Currently, S. abdita is predominantly recorded from Florida and surrounding states ( Fig. 4A; based on data from AntMaps 76 ), which may entirely be an artifact of the geographic focus of the species keys used to identify thief ants. If species of Solenopsis known as fire ants from the geminata (e.g. S. aurea, S. geminata, and S. xyloni) and saevissima (e.g. S. invicta, S. richteri, and saevissima) groups are notorious invaders in tropical to warm temperate climates regions 3,11,14,77 , this contrasts with the few successful introductions of the thief ants (previously referred as Diplorhoptrum) that have been recorded and their limited distribution within the introduced range (e.g. S. globularia, S. papuana, and potentially S. terricola to be confirmed as introductions in Florida). While an argument might be made for a potential candidate synonymy of these species with S. abdita, this possibility can be readily excluded by morphological examination. Solenopsis globularia (Smith, F., 1858) is easily distinguished by the greatly dilated/enlarged postpetiole seen in dorsal view, S. papuana Emery, 1900 has larger eyes that consist of three ommatidia plus a high profile of the propodeal dorsum, and S. terricola Menozzi, 1931 is a uniform dark brown species.
The identification of S. abdita as a new introduction within two distinct regions of the world [the Arabian Peninsula and Hawaii (Fig. 4B)] raises several questions. First, what is the extent of the native range of this species? And are populations from Guatemala part of the native or exotic range of this species? Second, due to the challenges of identifying of S. abdita and other thief ants, how many unidentified records of these species exist that potentially demonstrate a wider introduced range? Our study represents a case that could be expanded to more parts of the world to identify both Solenopsis specimens and other challenging ant taxa known to include major tramp species (e.g. Cardiocondyla 78 , Pheidole 79 , Tetramorium 80 ).
Several of these taxa are widespread tramp species frequently involved in human-mediated dispersal. Invasive and tramp species tend to have far-reaching geographical distributions and share life history traits including foraging behavior, nest structure, and queen number 9,16 .
Our phylogenetic and morphometric results indicate that invasive characteristics evolved within monomorphic S. abdita, such as its small size, lestobiotic lifeway, and phenotypic plasticity, could potentially confound taxonomists. Increased phylogenetic taxon sampling and improved species-level taxonomy using ultrastructural tools will be necessary to explore the issue of invasive origins in further detail.

Data availability
The specimens used in this study have been databased and the data are freely accessible on AntWeb (https ://www. antwe b.org). Main data needed to evaluate the conclusions in the paper are present in the paper. Additional data that support the findings of this study are available from the corresponding authors upon reasonable request.