Genetic diversity and colony structure of Tapinoma melanocephalum on the islands and mainland of South China

Abstract Aim Tapinoma melanocephalum is listed as one of the most important invasive pest species in China. Information regarding the patterns of invasion and effects of geographic isolation on the population genetics of this species is largely lacking. Location South China. Methods To address this problem, we genotyped 39 colonies (two colonies were collapsed due to genetic similarity) using microsatellite markers and mitochondrial DNA sequencing to compare colony genetic structure of T. melanocephalum on the mainland and islands of South China. Results An analysis of the colony genotypes showed that the genetic diversity of the mainland population was slightly higher than that of the island populations but not significantly so. However, the observed heterozygosity on Shangchuan Island (SCD) was significantly lower than that of the other colonies. We also found six haplotypes in 111 mitochondrial DNA COI sequences. The relatedness (r) value between colonies of SCD was 0.410, higher than that of the other populations. The genetic clusters among colonies were not related to geographic locations and exhibited admixture likely due to frequent human‐mediated dispersal associated with trade between the mainland population and the islands. Pairwise F STs between populations showed differentiation among mainland populations, while SCD displayed high levels of divergence (FST > 0.15) from most mainland populations. There was no significant isolation by distance among colonies. Most populations showed signs of a bottleneck effect. Main conclusions Our study suggests that there was no significant difference in the genetic diversity among the islands and the mainland; however, the lower genetic diversity, the higher degree of genetic divergence from other colonies, and the higher relatedness among nestmates made the SCD population stand out from all the others.

Tapinoma melanocephalum (Fabricius) (Hymenoptera: Formicidae), the ghost ant, is a ubiquitous invasive species and widely distributed in the tropical and subtropical latitudes worldwide, and even invades temperate zones through commercial activity and human trade (Espadaler & Espejo, 2002;Wilson, 1967). T. melanocephalum is not only a household-infesting ant (Nickerson, Bloomcamp, & Fasulo, 2004), but also an agriculture pest associated with hemipteran species, such as mealybugs and scales (Fowler, Bernardi, Delabie, Forti, & Pereira-da-Silva, 1990;Venkataramaiah & Rehman, 1989). It was first known to invade China in ca. 1929. Currently, it is frequently found in disturbed habitats where it nests in old branches, dry grass clumps, moist grass, and plant stems. It is also likely to be found in rotten wood, soil, and decayed parts of trees or beneath bark as well as in the walls of houses and in plant pots. It has successfully spread through several provinces in South China and some areas of North China through human assistance (Wetterer, 2009).
In their introduced ranges, changes in colony structure have facilitated success of many invasive social insects, such as the ant Anoplolepis gracilipes and Linepithema humile (Corin, Abbott, Ritchie, & Lester, 2007;Thomas, Becker, Abbott, & Feldhaar, 2010). Invasive social insects often form supercolonies containing in some cases billions of workers and thousands of queens spread over hundreds of square kilometers (Holway, Lach, Suarez, Tsutsui, & Case, 2002;Holway & Suarez, 2004). Like other invasive ants, colonies of T. melanocephalum are polygynous, usually unicolonial; mating is intranidal; colonies are mostly founded by budding; and swarming is generally lacking (Smith & Whitman, 1992). Workers within unicolonial societies, or supercolonies, can move freely among different nests without aggression. Different supercolonies can exist near each other and exhibit distinct colony boundaries (Waldman, Frumhoff, & Sherman, 1988). On Christmas Island, the invasive ant Anoplolepis gracilipes established two supercolonies with an unusual reproductive strategy resulting in workers that were heterozygous at all microsatellite loci examined (Thomas et al., 2010).
There have been several hypotheses to explain unicoloniality of invasive ants from a genetic perspective. Tsutsui, Suarez, Holway, and Case (2000) reported that unicoloniality of invasive ants in their introduced regions may result from loss of genetic diversity. Invasive populations showed lower genetic diversity due to a genetic bottleneck. Lower genetic variability may result in a change in the nestmate recognition system, leading to decreased aggression (Bourke & Franks, 1995;Tsutsui et al., 2000;Vásquez & Silverman, 2008).
Another mechanism for unicolonial formation has been proposed for Linepithema humile by "genetic cleansing." Giraud, Pedersen, and Keller (2002) hypothesized that colonies could share common recognition alleles. These recognition signals might gradually mix with colonies or supercolonies due to reduced interpopulation aggression and increased encounter rates between colonies.
Genetic diversity has played an important role in the evolutionary changes of species in response to changing environmental stress (Frankel & Soulé, 1981), and it is one of the three levels of biodiversity promoting the long-term survival of a species (Frankham, Briscoe, & Ballou, 2002;Hedrick & Kalinowski, 2000;Morley et al., 2004). The ability of invasive ants to resist environmental stress and adapt to changes in the environment is dependent on genetic diversity (Beardmore, Mair, & Lewis, 1997). Therefore, studies on genetic diversity would be helpful for understanding the evolution and adaptation of invasive ants in their introduced environments (Sakai et al., 2001). Genetic constraints potentially influence initial colonization and will also influence the rate of spread (Antonovics, 1976;Sakai et al., 2001). Islands are isolated from nearby mainland by large water barriers and are usually smaller in area than the mainland (Losos & Ricklefs, 2009a). Geographic isolation of invasive species from the mainland results in experiencing different environmental conditions, food resources, and biological competitors and predators. In addition, the water surrounding islands acts as a geographic barrier to dispersal that limits gene flow both between island populations and between island and mainland populations (Barnett et al., 2013;Duffie, Glenn, Vargas, & Parker, 2009). Previous bottlenecks or continued isolation from the mainland can also decrease genetic diversity of invasive species in islands (Losos & Ricklefs, 2009b;Pimm, Diamond, Reed, Russell, & Verner, 1993).
Here, we explore the genetic and environmental (geographic isolation) factors in introduced areas to determine and compare the population genetic structure and colony characteristics of T. melanocephalum between island and mainland regions using six microsatellite loci and COI sequence data, and examine the relationship between two levels of population structure. The aim of this study is to characterize the breeding structure of the colonies and to investigate the relationships among populations and the mechanism of population spread.  Table S1).

| Sampling and study sites
After collection, the samples were preserved in 90% ethanol and stored at 4°C until DNA extraction.
The polymerase chain reaction (PCR) amplification was performed in 25 μl containing 1.5 μl of DNA extract (25 ng/μl), 1 μl of each primer (10 μmol/L; synthesized by Sangon Biotech (Shanghai) Co., Ltd., Shanghai, China), 12.5 μl Premix rTaq (TaKaRa, Japan), and 9 μl of ddH 2 O. The cycling conditions for the amplification were as follows: 3 min at 94°C followed by 35 cycles of 30 s at 94°C, 30 s at annealing temperature (see Table 1 for annealing temperature for each primer), 45 s at 72°C, and an extension of 5 min at 72°C. The PCR products were diluted 1:10 with 1× dilution buffer (5 × 930 DNA Dilution Buffer) and separated by a Fragment Analyzer™ automated capillary electrophoresis system (Advanced Analytical Technologies, Inc., USA). The alleles were scored using the software PROSize 2.0 (Advanced Analytical Technologies, Inc., USA).
The partial cytochrome oxidase subunit I (COI) gene of the mitochondrial DNA was amplified and sequenced from 111 individuals from all colonies using the primers LCO1490 (5′-GGT CAACAAATCATAAAGATATTGG-3′) and HCO2198 (5′-TAAA CTTCAGGGTGACCAAAAAATCA-3′) (modified from Hebert, Ratnasingham, & de Waard, 2003; GenBank accession number: JQ913600). The PCR amplification reaction and cycling followed the method described by Chiotis, Jermiin, and Crozier (2000), and the PCR products were sent for sequencing to Sangon Biotech Co., Ltd (Shanghai, China). All sequences were edited in DNAStar. If necessary, errors were manually corrected and then aligned using MEGA 6.0 software after blasting in NCBI. Based on a 610-bp fragment, the genetic diversity of the COI gene was estimated.

| Colony boundary test
To determine whether workers collected from different points could be attributed to the same colony, genotypic differentiation with pairs of samples collected from the different geographic points from each site was performed using the log-likelihood G test in GENEPOP version 4.5.1 (Raymond & Rousset, 1995). The p values were detected using Fisher's probability with the standard Bonferroni correction.
Two sampling locations were considered to be different colonies if the genotypic differentiation of workers from them was shown to be significantly different (Deheer & Vargo, 2004;Dronnet, Chapuisat, Vargo, & Bagnères, 2005).

| Genetic diversity testing
The Hardy-Weinberg equilibrium for each locus and linkage disequilibrium (LD) for each pair of loci were tested by GENEPOP version 4.5.1 (Raymond & Rousset, 1995). To avoid high relatedness within colonies, we randomly selected one individual from each colony to test. For linkage disequilibrium testing, we performed 10 repetitive calculations by taking 10 different individuals from each colony. The p values were tested using the Bonferroni correction for multiple tests (Rice, 1989 The variability at five microsatellite loci in Tapinoma melanocephalum was tested from mainland populations and island populations. The number of alleles (A) and allele richness (AR) were estimated among all samples. PHWE shows the p value for the test of Hardy-Weinberg equilibrium. Data with lowercase or uppercase letter indicate no significant difference between island populations and mainland populations using an independent t test (p > .05).
allele for each locus and colony with the exception maximization (EM) algorithm (Chapuis & Estoup, 2007;Dempster, Laird, & Rubin, 1977). The number of alleles and the allele richness in each colony and population were analyzed by FSTAT 2.9.3 (Goudet, 1995). The observed heterozygosity and the Shannon index of each colony or population were estimated by GENALEX 6.5 software (Peakall & Smouse, 2006). The Shannon index is a measure of the genetic diversity of a population that is estimated based on the allele number and frequency, and the resulting values ranged from 0 (low diversity) to 4.6 (high diversity; Colwell, Mao, & Chang, 2004). When we compared the genetic diversity among all island and mainland populations, our results were estimated at the population levels of the island and mainland, instead of at the colony level. Significant differences were determined by SPSS 14.0 using an independent t test.
The observed heterozygosity of all populations did not have similar variances and was compared using a Kruskal-Wallis test, and if the differences were significant, multiple comparisons of means were performed with the Mann-Whitney test. In addition, the haplotype diversity of each colony was detected by DnaSP (Rozas, Sánchez-DelBarrio, Messeguer, & Rozas, 2003).

| Genetic structure of colonies and populations
To determine the genetic relationships among populations, each colony was considered a subpopulation in each geographic site (similar to that reported by Vargo, 2003) and was subjected to an independent simulation run by STRUCTURE software based on the admixture model and an allele frequency-correlated model, with different genotypes of individuals allocated to different genetic clusters (Evanno, Regnaut, & Goudet, 2005;Pritchard, Wen, & Falush, 2010). Here, we ran all individuals genotyped at the population level, taking the first individual genotyped per colony. We performed 10 runs for each value of K with the simulated number of geographic populations from 1 to 11 with 5,000 runs of MCMC after a burn-in period of 10,000 runs that was discarded after burning. After running, we uploaded the results file into STRUCTURE HARVESTER (http://taylor0.biology.ucla.edu/structureHarvester/#) to determine the most likely number of genetic clusters (K) using the method of Evanno et al. (2005).
The coefficient of relatedness among workers was estimated within all 37 colonies sampled and worker nestmates of colonies from 11 geographic populations for genetic analysis using FSTAT 2.9.3, following Queller and Goodnight (1989) using all individual genotypes as the reference population. The index of relatedness R weights each allele inversely by its frequency in the population so that rare alleles are given a relatively higher weight.

| Isolation by distance
To estimate the degree of genetic differentiation at the population level at this scale, we calculated the pairwise F ST s for 11 geographic populations with one individual from each colony using GENEPOP version 4.5.1. The differentiation index was interpreted in a standard manner as follows: a low degree of genetic differentiation (0 ≤ F ST < 0.05); a medium degree of genetic differen- and the natural log of the geographic distance (km) were run using GENEPOP version 4.5.1 with Mantel tests (Raymond & Rousset, 1995).

| F-statistics analysis
To assess genetic variation among the island and mainland populations, we performed a three-level analysis of F-statistics (Weir & Cockerham, 1984). In this part, only five loci (locus 434, locus 592, locus 911, locus 1178, and locus 1315) were used in the F-statistics analysis. We followed the approach of Thorne et al. (1999)

| Genetic bottleneck test
The program BOTTLENECK 1.2.02 was used to detect whether the populations recently experienced bottlenecks (Piry, Luikart, & Cornuet, 1999

| Colony boundaries and breeding structure
The results of the tests of genetic differentiation between collec- and therefore considered separate colonies (Table S1). An analysis of the number of alleles per colony indicated that all of 37 colonies of mainland and island populations had more than four alleles and was consistent with their polygyne social structure that has been previously reported (Wetterer, 2009; Table 2).

| Testing for Hardy-Weinberg equilibrium and linkage disequilibrium
Four of the six microsatellite loci exhibited significant deviations from Hardy-Weinberg equilibrium (see Table 1). Across all populations,

| Genetic diversity of populations
Overall, 778 individuals were successfully genotyped at six microsatellite loci. The results showed that the mainland populations had higher average numbers of alleles (A) and allelic richness (AR) than the island populations (Table 1), but there was no significant difference (A: t = 0.323, df = 8, p = .755; AR: t = 0.263, df = 8, p = .799). In addition, we also analyzed the genetic diversity of each colony using microsatellite and mitochondrial DNA data. The number of alleles for each colony ranged from 4.4 to 9.0, and the allelic richness varied from 4.3 to 8.8 (see Table 2). Range of observed heterozygosity among all colonies was from 0.33 (COLSCD2) to 0.67 (COLDAD2).
We found that Shannon's index varied from 0.92 (COLSCD2) to 1.79 (COLZJ2), indicating that all populations and colonies exhibited a modest level of genetic diversity. At the population level, the number of alleles in all populations varied from 5.30 (SJ) to 7.67 (ZJ), allelic richness ranged from 4.8615 (SCD) to 7.49 (ZJ), and the observed heterozygosity ranged from 0.36 (SCD) to 0.57 (GZ). The results indicated that the SCD populations exhibited the lowest genetic variation. The average observed heterozygosity of the SCD points was slightly lower than that of several mainland populations (Kruskal-Wallis test, χ 2 = 19.24, df = 10, p = .037). Therefore, with the exception of the SJ and BH populations, the genetic diversity of the SCD population was significantly lower than that of the mainland populations (p < .05). Regarding the mitochondrial DNA data (Table 2)

| Genetic structure of populations
The genetic clustering of the T. melanocephalum populations by Bayesian clustering analysis was grouped into two branches (K = 2 and 4; Figure S1) by STRUCTURE using the delta-K method, with 20 runs for K ranging from 1 to 11 (Figure 3). Interestingly, the two genetic clusters showed no relationship with the geographic location of colonies, and individuals from one site were assigned to different genetic clusters. Most of the colonies of the islands were mixed with colonies from the mainland. Approximately 50% membership of BH2 and HBD2 colonies were not uniquely assigned to either of the two genetic clusters (Figure 4). In addition, when K increased to 4, the population structure from Naozhou Island and Shangchuan Island showed more clearly approximately >75% assignment probability. Individuals F I G U R E 3 Delta-K values based on 10 runs of K ranging from 1 to 11 using STRUCTURE F I G U R E 4 Colony genetic structure of Tapinoma melanocephalum based on genotypes for K = 2 and K = 4 using the program STRUCTURE. Each individual is represented by a vertical bar. The same color stands for individuals that are assigned to the same cluster from other colonies were clustered into different genetic clusters and mixed with colonies from mainland and island areas (Figure 4).
The relatedness R among workers within colonies over five loci varied among populations, ranging from 0.331 to 0.373, with an average of 0.337 ± 0.07, while pairwise comparisons between colonies of each population showed that relatedness varied from 0.065 to 0.410 (mean R = 0.355 ± 0.035) (see Table 2). The average genetic relatedness between the two colonies of the SCD population was 0.410, which is greater than zero. The higher relatedness between the SCD colonies is consistent with their lower allelic richness compared with the other more genetically diverse populations. As explained by Boomsma, Fjerdingstad, and Frydenberg (1999), relatedness among workers from the same colony headed by a once-mated queen will average 0.75. In this study, all colonies had much lower relatedness values and were considered polygyne, with all queens probably descending from a single foundress based on the mtDNA data.

| Molecular variance
Hierarchical F-statistics were estimated for all colonies. Three levels of hierarchical AMOVA showed that genetic differentiation was higher within colonies (F IND-COLONY = 0.44133, p < .001) than within the populations (F REGION-TOTAL = 0.20965, p < .001) and within region (F COLONY-REGION = 0.26028, p < .001; Table 4).
For the mainland and island regions, the degree of genetic

| Bottleneck effect
We examined signatures of a genetic bottleneck using BOTTLENECK, with the two-mutation model. The results revealed a significant heterozygosity excess (p < .05) in 11 geographic populations, except for the BH, NZD, and SCD populations under the Wilcoxon signed-rank test of the infinite allele model (Table 5).
There was no significant heterozygosity excess in any of the Our results suggest that the populations of SJ, ZJ, MZ, WZD, and DAD experienced a recent genetic bottleneck under the Wilcoxon test and sign test of two-mutation models (Table 5).

| Genetic diversity of populations
Study of genetic diversity and potential for rapid evolution is helpful for understanding colonization dynamics and spread of invasive social insects (Sakai et al., 2001). Geographic characteristics can affect dispersal range and gene flow of invasive species and further influence genetic diversity of introduced populations (Barrett & Husband, 1990;Moody & Mack, 1988). For islands, a natural barrier has the potential to influence species diversity or population size (Jensen et al., 2013). In our study, the genetic diversity of the mainland populations of T. melanocephalum was slightly higher than that of the island populations, but there were no significant differences.
However, colonies in the SCD population had lower allelic richness and number of alleles than mainland colonies; the observed heterozygosity of the SCD population was dramatically lower than that of several mainland populations. Multiple introductions of invasive populations could promote higher genetic diversity of alien species compared with a single introduction (Sakai et al., 2001;Wilson, Naish, & Boulding, 1999). In our study, microsatellite data showed that the genetic diversity of island populations was close to that of the mainland populations. This was also supported by mitochondrial DNA data. One possible reason is that populations from the mainland and the islands are multiply introduced with movement of human beings and frequent commercial trade.
However, invasive species are usually associated with population bottlenecks in the introduced regions and reduction in diversity may TA B L E 4 Population differentiation estimated by F-statistics for 11 geographic regions Asterisks represent significant differences from zero: *p < .05; **p < .01, and ***p < .001. occur due to genetic drift during colonization (Barrett & Husband, 1990). Low levels of heterozygosity have been taken as evidence of bottlenecks in populations in the recent past (less than four N e generations ago) (Luikart & Cornuet, 1998). Interestingly, Argentine ants in the southeastern United States (introduced populations) experienced bottlenecks and exhibit higher levels of genetic diversity than other introduced populations (Buczkowski, Vargo, & Silverman, 2004). Some populations may display lower genetic diversity while no bottleneck effect is detected. This may occur if the population has been present for long periods of time during which they recover from the signatures of a bottleneck (Green, 1990). In our study, most island populations of T. melanocephalum showed similar allele richness with that of the mainland populations, and three island populations experienced no detectable bottleneck.

| Population genetic structure
Using the STRUCTURE program, we obtained two genetic clusters for all T. melanocephalum populations, which did not correspond to distinct geographic locations and admixtures of island and mainland colonies. T. melanocephalum has been described as an intranidal mating species and was shown to form unicolonial populations by colony budding (Hölldobler & Wilson, 1977). Gene flow has shown to be limited between different colonies or supercolonies because of intranidal mating (Espadaler & Rey, 2001;Helanterä, Strassmann, Carrillo, & Queller, 2009). STRUCTURE analysis showed some colonies could not be assigned to either of the two clusters. This result suggests that the clusters are not completely distinct. It may be attributed to long-distance dispersal of the ghost ant that frequently occurs due to human activities (Seifert, 2010). Therefore, newly introduced populations may mix with earlier arrivals to form a single more genetically diverse unicolonial society, thus promoting gene flow at large scales among island and mainland populations. On the other hand, based on our results showing each colony had a single mtDNA haplotype, it appears that each of our study colonies descended from a single queen. It also suggests colonies do not fuse the way Argentine ant colonies do (Thomas et al., 2010). Notably, colonies of most of populations we tested were not uniquely assigned to either of the two genetic clusters by STRUCTURE analysis using microsatellite data. In addition, most workers in our study were heterozygous at microsatellite loci. Taken together, this may indicate a special breeding structure in this species in which mating occurs in or near the nest and a majority of males come from external nests.
Obviously, the reproductive habits of this species in China feature need further study.
Among the paired mainland populations, there was no differentiation. However, when the SCD populations were paired with most of the mainland populations, we found a high level of genetic differentiation, and F ST s ranged from 0.0007 to 0.325. In particular, the SCD populations and the ZH populations may have different origins leading to the observed strong genetic differentiation. This is supported by one of the ZH colonies, ZH1, having a different mtDNA haplotype than the SCD colonies. There was no correlation between the genetic differentiation and geographic distance among all colonies, which suggests that gene flow among colonies is not related to distance, and isolation by sea has had minimal influence on gene flow. Five islands were separated by 16-50 km from the nearest mainland.
Hierarchical AMOVA suggests that the genetic divergence of T. melanocephalum was mainly due to differences within colonies instead of between colonies in regions or sampling regions. Due to founder events, colonies achieved rapid differentiation. The changes in allele frequency were affected by differential survival of individuals during the initial bottleneck (Chakraborty & Nei, 1977;Tarr, Conant, & Fleischer, 1998).
Our results indicated low relatedness within all individuals of colonies. But, in the SCD population, colonies exhibited the highest r value (0.410) and the lowest allelic richness; these colonies all clustered together. Similar to a previous report, invasive species in introduced regions tend to form supercolonies or colonies (Holway et al., 2002) with reduced genetic variability and increased nestmate recognition (Payne, Tillberg, & Suarez, 2004;Tsutsui et al., 2000). This may also be attributable to the possibility that the SCD population was derived from only one introduction. Strong hurricanes and fierce floods may also affect the population distribution and expansion and could also lead to genetic loss. In addition, mtDNA data were uniform within a colony, suggesting that each colony came from a single matriline.
Many reports of invasive species in different geographic regions suggest that social behavior is related to colony breeding structure.
In its native range, the Argentine ant forms small supercolonies that are effectively unicolonial by intranidal mating and kin recognition (Pedersen, Krieger, Vogel, Giraud, & Keller, 2006). Intranidal mating has also been reported in Monomorium pharaonis (Wilson, 1971) and Lasius sakagamii ant species (Yamauchi, Kinomura, & Miyake, 1981), as well as T. melanocephalum, the Argentine ant, and Anoplolepis gracilipes (Bustos, Cherix, Lachaud, & Fourcassié, 1998;Drescher, Blüthgen, Schmitt, Bühler, & Feldhaar, 2010). A previous report revealed a negative correlation between relatedness and aggression in Anoplolepis gracilipes, which suggested kinship as one factor causing the formation of a supercolony. Kinship also facilitates colonization (Drescher, Blüthgen, & Feldhaar, 2007). Supercolonies of A. gracilipes undergo random mating; intranidal mating system in this ant can be observed due to genetic cohesion in these two supercolonies (Thomas et al., 2010). In the introduced range, the invasive red imported fire ant exhibits a higher degree of polygyny than in native regions (Ross & Keller, 1995). Polygynous colonies in native areas have fewer queens who are close relatives compared to the many unrelated queens found in introduced populations (Ross, Vargo, & Keller, 1996), leading to the observed high queen numbers and decreased relatedness for introduced ranges.
Closely related colony members in polygynous nests may arise from inbreeding, reproductive dominance of one queen, and/or high relatedness among queens (Giraud, Blatrix, Poteaux, Solignac, & Jaisson, 2001). Nestmate queens in polygynous species are generally related, but to a degree that varies greatly between species, some species having nestmate queens that are almost unrelated (Keller, 1995). Therefore, the queen number and relatedness among nestmate queens of T. melanocephalum should also be a focus of future studies as well the relationship between colonies in China and those in other parts of the world.

ACK N OWLED G M ENTS
This study was supported by the Science Foundation for the Excellent Youth Scholars of Guangdong Province (No. Yq2013031).

CO N FLI C T O F I NTE R E S T
None declared.