Nearly all of the ant specimens used in this study were collected by participants—most of them school children—in the School of Ants citizen science project (SOA, www.schoolofants.org), between 2011 and 2013 (Lucky et al., 2014). SOA was designed to facilitate public engagement in scientific research; one goal was to use participant-collected samples of widespread ants for population genetic studies. Other goals included increasing interest and trust in science through public participation in scientific research (Vitone et al., 2016), facilitating monitoring of ant distributions in human-modified environments (Lucky et al., 2014), and encouraging the study of urban biodiversity.

All specimens were collected following the standardized SOA protocol (Figure 2, Supplementary Material 1), in which participants used cookie crumbs to attract ants, then collected ants in zip-top bags, froze them overnight, and mailed them to the SOA team. Upon receipt, ant specimens were identified to genus and species (wherever possible) and vouchers were stored at −18°C in 95% EtOH.

From hundreds of samples containing Tetramorium workers, 90 samples from 26 states were included in genetic analyses, representing the broadest possible US geographic range (Figure 2, Supplementary Material 2). To minimize unintentional sampling of ants from the same colony, specimens were collected from localities at least 1 km from any other sample, far beyond the estimated territory size of 42 m² in a single colony of T. immigrans (Brian and Elmes, 1974). DNA was extracted destructively using the Qiagen DNeasy blood and tissue kit (QIAGEN, Valencia, CA), according to standard protocol with the only change being the final elution volume of 75 μL DNA for each sample. A 658 bp fragment of the mitochondrial COI gene, commonly used as a “barcoding” gene, was amplified using primers LepF1 (5′-ATT CAA CCA ATC ATA AAG ATA TTG G-3′) and LepR1 (5′-TAA ACT TCT GGA TGT CCA AAA AAT CA-3′) using standard PCR conditions (Folmer et al., 1994).

To contextualize North American T. immigrans among native range samples, we also took advantage of additional, although not fully overlapping (348 bp), COI sequences of European T. immigrans from previous study (Schlick-Steiner et al., 2006). We built a TCS haplotype network (Clement et al., 2000) using the combined European and our own COI data in the program POPART v1.7 (Leigh and Bryant, 2015).

Double digest Restriction-site Associated DNA sequencing (ddRADseq) library preparation followed (Peterson et al., 2012) using the 78 samples that successfully generated COI sequences. In addition, to compare inter- vs. intra-colonial levels of genetic diversity and test for inbreeding, we included 16 workers from a single colony in North Carolina. In brief, the rare EcoRI and common MseI restriction sites were used for restriction enzyme digestion. Illumina sequencing adaptors and custom barcodes between 8 and 14 bases in length, which varied by at least four bases (Parchman et al., 2012), were ligated onto the sticky ends and PCR-amplified using primers for Illumina sequencing adaptors. Amplified digestion-ligation products were pooled into equal volumes for sequencing. Gel-based size selection for 200–500 base pair fragments and purification of the sequencing library using Agencourt AMPure XP beads (Beckman Coulter) were performed at the UF ICBR core-sequencing laboratory (http://www.biotech.ufl.edu), and sequenced using Illumina NextSeq500 high-output flow cell for a shared 150-cycle sequencing run.

Illumina adaptor sequences were automatically trimmed from all reads by Illumina's BaseSpace software. All subsequent bioinformatics steps were conducted using Stacks v. 2.3e (Catchen et al., 2013). Raw Illumina reads were demultiplexed using process_radtags, quality controlled (-c -q -r) and trimmed to 100 base pairs (-t). The Stacks program denovo_map.pl was used for de novo assembly, locus discovery, and genotyping. A minimum of three mismatches was allowed between stacks during loci assembly (-M 3), and an allowance of two mismatches was permitted between stacks (-n 2). Finally, the Stacks populations program was used to identify loci violating Hardy-Weinberg equilibrium (-hwe), which were subsequently removed via the blacklist option (-B). The remaining loci were then filtered to include only loci present in all four populations (-p 4), and in 90% of the individuals (-r 0.9), for downstream analyses.

Basic population and summary statistics were calculated using Stacks populations sumstats.summary output, the R packages adegenet v2.1.1 (Jombart and Ahmed, 2011) and hierfstat v0.04-22 (Goudet, 2005). To infer population structure the Bayesian-based clustering analysis, STRUCTURE v2.3.4, was used (Pritchard et al., 2000). We tested the support for 1 through 5 genetic clusters (K = 1–5) and ran 10 replicates per K with 100,000 generations burn-in followed by 500,000 MCMC repetitions, using default settings. To infer the most likely number of unique genetic clusters we used the ΔK method (Evanno et al., 2005), implemented in STRUCTURE Harvester (Earl, 2012). The STRUCTURE plots were visualized using pophelper v2.2.9 (Francis, 2017). We also analyzed the SNPs data using the discriminant analysis of principle components (DAPC), which has fewer marked assumptions than STRUCTURE, using adegenet. The most probable number of clusters for DAPC was determined using sequential k-means clustering using the find.clusters function and the associated Bayesian Information Criterion (BIC). DAPC plots were made for individuals under the most probable cluster membership and labeled by U.S. geographic origins (Northeast, Mid-Atlantic, Midwest, and Western). Hierarchical analysis of molecular variance (AMOVA) was conducted to detect differences among populations using the R package poppr (Kamvar et al., 2014).

Out of the 90 specimens that were tested, 14 failed to amplify; this was presumed to be due to DNA degradation during shipping. We were able to verify that all specimens morphologically identified as T. immigrans in this study were indeed this species using DNA barcode authentication of samples. No other cryptic species in this complex were found among these samples and very little genetic difference was found within the US. Comparison of US and European populations using a short fragment of COI recovered all US haplotypes as identical to the most commonly collected haplotype from Europe (Figure 3).

The success rate of sequencing citizen-collected samples was high (98% success) with ddRADseq, on average, there were 970,000 ± 36,000 SE reads per individual. After stringent filtering 2,255 unlinked SNPs were retained for downstream analyses. The North American population of T. immigrans has little population structure: across the entire sampled range overall F_ST was low (F_ST = 0.0150, Table 1A). Within the one North Carolina colony for which 16 nestmates were sampled, we found near-identical genotypes (F_ST = 0.001, Table 1A). The inbreeding coefficient (F_IS) for the North Carolina colony was −0.4706, while the continent-wide F_IS was 0.0376 (Table 1A). Pairwise F_ST values between the four geographic regions were also low, ranging from 0.0056 to 0.0297 (Table 1B). The majority (94.82%) of genetic variation was found among individuals, rather than between regions as indicated by AMOVA (Table 2).

The optimal number of populations supported by STRUCTURE analysis was K = 2 (Figure 4), with Midwest and Mid-Atlantic samples grouped together while Western and Northeast populations had greater similarity to each other. For K = 3, additional structure was detected with two Northeastern individuals from Vermont having shared ancestry assigned primarily to this third cluster (Figure 4). The optimal number of clusters detected using k-means clustering was also two, corresponding to the lowest BIC value (Figure 4). DAPC was subsequently carried out using the optimal two clusters and retained three discriminant eigenvalues, which separated Midwest population from the other three regions (Figure 5).

The most common, widespread pavement ant in North America is unequivocally T. immigrans, as evidenced by COI and ddRADseq data sampled from colonies sampled across North America. The near absence of genetic diversity (~0.0%) revealed in our study is consistent with previous, small-scale studies that found 0.0–1.2% COI sequence divergence for T. immigrans in its native range (Schär et al., 2018). We found no evidence of other T. caespitum complex species beyond the known presence of T. tsushimae (Steiner et al., 2006), which was readily distinguishable from T. immigrans based on morphology and geographic distribution.

Despite having been established in North America since at least the 1800s, little genetic divergence exists among populations of T. immigrans within the non-native range. Whether this is a result of selection or reflects the limited diversity present in the population upon establishment is unclear. We recovered only a few genetic variants from across a vast geographic range, and thousands of sites across the genome. COI sequences, for example, were nearly identical across samples; among three haplotypes identified, each differed from the others by only a single SNP. When compared with other European samples, the USA specimens clustered with the most common haplotype found throughout different regions of Europe (Figure 3).

Genomic data also indicate that pavement ants comprise a single, largely undifferentiated population across North America, with close examination revealing differences in structure that varied depending upon the analytical method used. In both STRUCTURE and DAPC, SNP data provided support for the existence of two weakly differentiated T. immigrans clusters (Figures 4, 5). In the K = 2 scenario, most samples were assigned to one cluster, Cluster 1, that was commonly encountered in all four geographic regions, with Cluster 2 represented only in a small subset of samples from the Northeast and West. Additional, minor, genetic sub-clusters were identified when K = 3. By contrast, DAPC analyses identified the composition of Midwestern samples as distinct from the other three geographic regions (Figure 5). This overall lack of strong genetic structure was also confirmed by AMOVA, as genetic variation among individuals far surpassed variation between assigned geographical regions (Table 2). Surprisingly, this low diversity appears not to result from inbreeding, as indicated by a low inbreeding coefficient (F_IS) among the 16 nestmates from a single North Carolina colony (Table 1A). While the patterns of nestmate variability could be underestimated by only examining a single colony, when combined with the lack of strong population structure across the US, they suggest that the entire North American population is likely descended from either a single introduction or multiple introductions from a single, genetically homogeneous lineage.

Our data do not allow us to distinguish between a single-colony and a multiple-colonies-from-a-single-lineage introduction scenario, but the presence of a rare social parasite, Tetramorium atratulum (Schenck) in the non-native range may provide an informative clue. This workerless ant is an obligate parasite of T. immigrans and is found at low densities in both the native and non-native range of its host (Dash and Sanchez, 2009). Tetramorium atratulum is fed and cared for exclusively by T. immigrans workers, which do not recognize this interloper as a threat. It seems unlikely that a viable population of this parasite could have become established in North America as a result of a single-colony host introduction. Instead, its presence suggests that a pulse of introduction and establishment of closely related pavement ant colonies may have been more likely. Future studies on the population genetics of T. atratulum could prove to be especially illuminating as to the introduction history of both this parasite and host species.

Despite the fact that the establishment of the European pavement ant in North America precedes that of other invasive ants by a century, the low genetic diversity in this species is nonetheless similar to that of other invasive ant species (Tsutsui and Suarez, 2003; Ascunce et al., 2011; Eyer et al., 2018). The combination of outbreeding and the genetic bottleneck experienced during founding resulted in a widespread population today that largely lacks genetic structure. The continent-wide range expansion from the initial site of establishment was likely facilitated by human-assisted dispersal, as we found no evidence of geographic isolation across the non-native range.

The genetic uniformity of T. immigrans across North America is surprising, considering its reproductive biology, which differs from that of other widespread and ecologically dominant non-native ant species. Low genetic diversity in this outbreeding species appears to be a result of the severely limited genetic diversity of the initial population upon establishment rather than inbreeding or clonal reproduction. Invasive ant success is often associated with unicoloniality, where colonies reach massive size and density as a result of multi-queen (polygynous) and multi-nest (polydomous) colonies that display no intercolonial aggression (Tsutsui and Suarez, 2003). Unicolonial populations often possess extremely low genetic diversity as a result of introduction bottlenecks and inbreeding (Tsutsui et al., 2003; Ascunce et al., 2011; Eyer et al., 2018) and in some cases clonal reproduction depresses diversity even further (Foucaud et al., 2009; Pearcy et al., 2011).

Tetramorium immigrans is widespread in paved habitats throughout most of temperate North America and maintains large and often competitively dominant colonies; however, each colony possesses only a single, outbreeding queen and has distinct territorial boundaries that workers aggressively defend against non-nestmates (Bruder and Gupta, 1972). By contrast, T. tsushimae, the other US introduced pavement ant in the T. caespitum complex, is polygynous. This difference in reproductive biology likely contributes to the competitive dominance of T. tsushimae where both species co-occur (Reuther, 2009; Trager, pers. comm.); whether this more recently introduced pavement ant will replace T. immigrans entirely remains to be seen.

Little is known about where, specifically, T. immigrans first became established in the US and precisely when this occurred. Brown (1957) suggests an introduction in or before the early 1800s. Given this timeline, it is surprising that stronger population structure does not exist in the non-native range. The yellow crazy ant, Anoplolepis gracilipes Smith, became structured and differentiated in its Australian exotic range within a matter of decades (Gruber et al., 2013). Reproductive biology may offer some insight into these differences. A. gracilipes colony foundation is dependent, with new colonies “budding” by fission from a parent colony; resulting in restricted gene flow and outbreeding distance. In contrast, a single mated T. immigrans queen independently establishes a new nest without attendant workers, a process known as independent colony foundation (Bruder and Gupta, 1972). Outbreeding and the mating flights of T. immigrans sets them apart from many other well-studied invasive ants and may explain why minimal population structure exists within the US.

The results of this study show a higher genetic diversity among colonies than within a single colony (Table 1A), confirming that, at least in the region of this single-colony sample (North Carolina), T. immigrans does not form supercolonies in the exotic range. If not unicoloniality, what, then, can account for the success of T. immigrans? It seems likely that the combination of habitat suitability and the almost complete lack of competition from other North American ant species in the human-built environment (that is, open, gravelly, and well-drained sites), has aided in the expansion of T. immigrans (Cordonnier et al., 2019). The development of roads and railways, as well as increasing urbanization, have effectively rolled out the red carpet for this species. The generalist feeding habit of T. immigrans, combined with a preference for habitats sharing the qualities of paved environments in temperate climates, created a suitable and largely competitor-free niche for T. immigrans after its introduction into early 1800s North America. Urban T. immigrans have been found to have carbon-13 levels similar to those in human foods, to a greater extent than co-occurring ant species, suggesting that T. immigrans is exquisitely capable of taking advantage of human-produced (and discarded) food resources (Penick et al., 2015).

Although T. immigrans is most often associated with urban areas and paved surfaces, populations of this species have also become established in undisturbed, unpaved habitats in the western US, in California, Nevada, Utah, and Idaho, where the incursion of this species into intact ecosystems is associated with reduction of local ant species richness (Ward, pers. comm.). As the pavement ant is not strictly limited to the built environment, research on this species' ability to compete with native ants would be of interest. Further studies comparing the intraspecific genetic variations and behavior of this species in both its native and non-native ranges would shed more light on the mechanisms leading to the success of pavement ants, and indeed other synanthropic species, in the Anthropocene.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.