Genetic analyses
Genetic analyses were used to infer the social structure (i.e., monogyne or polygyne) and genetic relatedness within and between each fire ant mound analyzed (Fig 1). For each mound, eight workers were randomly selected for DNA extraction. DNA was extracted from individual workers following a modified Gentra-PureGene protocol (Gentra Systems, Inc. Minneapolis, MN, USA).
To determine the social form of each mound, we pooled the DNA extracted from the eight individual workers per mound and screened this pooled sample for the presence of the Gp-9b allele, exclusively present in polygyne colonies (Krieger & Ross, 2002; Ross & Keller, 1998). A PCR reaction was performed on each pooled sample using the specific primer pair 24bS and 25bAS (Valles & Porter, 2003). This primer pair amplifies a 423 bp amplicon, and a successful amplification denotes the presence of the Gp-9b allele, thereby characterizing the workers as polygyne. Amplifications were performed according to the protocol described in Valles and Porter (2003) and visualized on a 1% agarose gel. In all, we identified 38 monogyne and 35 polygyne mounds across all six fire ant sites using theGp-9 method, of which 11 monogyne and 8 polygyne mounds were treated with the isotope tracer.
In addition, five microsatellite markers previously developed forS. invicta (Sol11, Sol20 , Sol42 ,Sol49 and Sol55 ; Krieger & Keller, 1997) were amplified for each of the eight individual workers per mound. The allelic polymorphism of these five microsatellites was previously shown to be suitable to delimit colonies of S. invicta and infer their colony structure. The microsatellites were genotyped using the M13-tailed primer method (Boutin-Ganache, Raposo, Raymond, & Deschepper, 2001), consisting of 5’-fluorescently labeled tails with 6-FAM, VIC, PET or NED dyes to facilitate multiplexing. DNA amplifications were performed in a volume of 15 µL including 0.25-1.0 U of MyTaq™ HS DNA polymerase (Bioline), 2 µL of MyTaq™ 5x reaction buffer (Bioline), 0.08 µL of each primers, 0.08 of each M13 dye and 1 µL of the DNA template. PCR reactions were carried out using a Bio-Rad thermocycler T100 (Bio-Rad, Pleasanton, CA, USA). PCR products were sized against LIZ500 internal standard on an ABI 3500 genetic analyzer (Applied Biosystems, Foster City, CA, USA). Allele calling was performed using Geneious software v.9.1 (Kearse et al., 2012).
For every mound, the social structure result obtained with theGp-9b method was confirmed using microsatellite markers, inferring whether all workers from a mound could be assigned to a single queen (carrying one of the two alleles of the mother queen at each microsatellite marker studied). Polygyny was deduced when more than one worker per colony could not be unambiguously assigned to a single queen (see Appendix S2 for results). In addition, we compared the relatedness coefficients (r ) between monogyne and polygyne mounds (as identified using Gp-9 ) using analysis of variance (ANOVA) to verify that relatedness coefficients were significantly lower in polygyne versus monogyne mounds (i.e., suggesting the reproduction of several unrelated queens) and to determine any differences by site. We also used t-tests to establish if relatedness coefficients were significantly different from 0 for polygyne mounds (i.e., multiple unrelated queens producing workers within a single mound) and 0.75 for monogyne mounds (i.e., one singly-mated queen producing workers within a mound). Relatedness coefficients were calculated using the program COANCESTRY v.1.0 (Wang, 2011), according to the algorithm described by Queller and Goodnight (1989). Relatedness coefficients were weighted equally and standard errors (SE) were obtained by jackknifing over colonies.
Colony spatial structure was investigated for the six sites to determine whether distinct mounds of S. invicta, especially those collected within 5 m of each other, consisted of a single colony (i.e., polydomy) or separate colonies. To answer this question, genotypic frequencies at all mounds were compared using a log-likelihood (G)-based test of differentiation using GENEPOP ON THE WEB (Rousset, 2008). Bonferroni’s correction was applied to account for multiple comparisons of all pairs (adjusted P -value < 0.0008). Significance was determined using a Fisher’s combined probability test.
Colony clustering was visualized for each site by plotting individuals on a principal component analysis (PCA) using the adegenet R package (Jombart, 2008). The clustering of mounds into distinct colonies was also represented by Bayesian assignments of individuals into genetic clusters (i.e., colonies; K) using STRUCTURE v.2.3.4 (Pritchard, Stephens, & Donnelly, 2000). For each site, STRUCTURE simulations were run with values of K from 1 to the total number of mounds encountered in each site and repeated 10 times for each value of K. Each run included a 5 × 104 burn‐in period followed by 1 × 105 iterations of the MCMC. The most likely number of groupings was evaluated using the ΔK method (Evanno, Regnaut, & Goudet, 2005) implemented in Structure Harvester v.0.6.8 (Earl, 2012). Additional details and results for clustering analysis can be found in Appendix S2 and S3.