We collected 37 V. maculifrons and 13 V. squamosa colonies in and around the city of Atlanta, GA, USA (Fig. 2). Colonies were anesthetized with ether, extracted from the ground, and brought back to the lab. Several workers from each colony were then placed in 95% ethanol for subsequent genetic analysis. Colony collection occurred late in the season (September – November), when usurpation would have been complete, so we did not find mixed colonies containing workers of both species. In addition, most colonies had initiated the reproductive phase of their development and were already producing new queens and males.

We determined the utility of several previously developed microsatellite loci in providing genetic information in V. maculifrons and V. squamosa. Specifically, we investigated if the loci cloned by Thorèn et al. 38 in V. rufa, Hasegawa and Takahashi 39 in Vespa mandarinia, and Daly et al. 40 in V. vulgaris were genetically informative in our study taxa. For these assays, we determined if each locus would PCR amplify in V. maculifrons and V. squamosa. If a locus amplified, we then used agarose gel electrophoresis to determine if the locus displayed size variation in a subset of individuals. In the end, we obtained the genotype of all sampled workers of both species at the eight loci, LIST2003, LIST2004, LIST2013, LIST2019, LIST2020, Rufa 5, VMA-3, and VMA-6, which amplified and were polymorphic in both taxa.

Members of social insect colonies do not represent independent genetic samples because they are related. Consequently, it is necessary to remove the effects of relatedness among colony members to effectively analyze population level data. We avoided the problem of genetic nonindependence of colonymates by inferring the genotypes of the queen and male(s) that produced workers from each colony. Paternity analyses in hymenopteran taxa, such as Vespula, are particularly straightforward because males are haploid and full siblings always display the same multilocus haplotype derived from their father. Our analyses thus resulted in a set of diploid (queen) and haploid (male) individuals that were genetically independent. We used the genotypes of these inferred parental individuals to obtain estimates of genetic variability.

Based on the reconstructed male and queen genotypes, we calculated the allelic richness at the eight loci in V. maculifrons and V. squamosa. In addition, we used the rarefaction method of Petit et al. 41 to correct for the difference in the number of genes sampled in the two species. This method estimates the number of alleles that would be observed in an equal sample of genes from multiple groups based on the number of alleles observed in the actual unequal samples obtained. The use of the standardized allele richness allows for comparisons in the diversity of samples of different sizes. We then calculated the effective number of alleles and gene diversity of each locus to obtain general measures of the overall variability for the markers.

To avoid problems caused by nonindependence of workers sampled from the same colony, we used a resampling technique that yielded unbiased measures of population genetic structure 33. Specifically, we used reduced data sets that included only one individual per colony for population analyses. Briefly, a computer program was written to randomly select a single individual's multilocus genotype from each colony, yielding a new data set with the number of individuals sampled equal to the number of colonies. This procedure was repeated 25 times to produce 25 such data sets. Each of these data sets was then used to calculate the population statistic of interest, and the median of the 25 values was taken as the unbiased estimate. Probability tests implemented by the program GENEPOP 3.4 42 were used in conjunction with this resampling procedure to determine the significance of genotypic disequilibrium between microsatellite markers and deviations from Hardy-Weinberg equilibrium.

We next determined if V. maculifrons or V. squamosa populations showed evidence of isolation by distance over the geographical scale considered. We first calculated estimates of F_ST between all pairs of colonies for each species using GENEPOP. Thus, colonies, which consisted of groups of sampled workers, were considered as 'populations' for these analyses. Significance of the correlation between geographical and genetic distance was assessed with a Mantel test. The relationship was deemed to be significant if the observed correlation was greater than 5% of the 10,000 randomly generated values. Spearman's rank order correlation coefficient (r_S) was used to quantify the association between the genetic distances between nests and the geographical distances that separated them.

We used the molecular genetic data derived from the worker genotypes to investigate how the breeding systems of V. maculifrons and V. squamosa differed. Because annual Vespula colonies are headed by single queens, within colony genetic diversity provides insight into the queen's mating history. We thus used the genetic data to determine the number of males to which queens were mated. We then compared mate number in the two taxa via a Kruskal-Wallis test. We next calculated the effective paternity (k_e3) for workers using the method of Nielsen et al. 43. The metric k_e3 includes information on the number of times a queen mates and the unequal contributions of a queen's male mates to offspring. We also tested if the estimates of k_e3 obtained for V. maculifrons and V. squamosa differed using a Kruskal-Wallis test.

Additionally, we investigated if the distribution of mate number in the two species matched known distributions, as might be expected if particular biological processes (such as constant mate search time or equal probability of mating with particular numbers of males) affected the mating behavior of queens 44. Specifically, we fitted a Gaussian distribution to queen mate number. We expected that mate number would match this distribution if number of mates were a random function of time and mating opportunity. In contrast, deviations of mate number from the Gaussian distribution might indicate the action of selection operating in these species.

We calculated the magnitude of reproductive skew of males mated to queens using the metric B of Nonacs 45. The 95% confidence intervals for each estimate of B were estimated to determine the significance of skew. Values of B were considered to be significant if the 95% confidence intervals failed to overlap 0. Both the skew and significance of skew were calculated using the program Skew Calculator 45. As was the case with our estimates of k_e3, we tested if estimates of B differed between species via a Kruskal-Wallis test.

We calculated the relatedness of workers belonging to the same colony using the program RELATEDNESS 46. We then estimated the relatedness of the queens to their male mates and the relatedness of males mated to single queens to determine if inbreeding occurred or if related males mated the same queen. Standard errors for all estimates were obtained by jackknifing over loci. Finally, we compared our relatedness estimates with similar data collected over two decades ago to determine how the parasitic relationship may have changed mating structure over time.

We sampled 40.56 ± 23.51 (mean ± SD) and 45.62 ± 8.31 workers from each of 37 V. maculifrons and 13 V. squamosa colonies, respectively (Fig. 2). Our total sample size thus consisted of 1501 and 463 workers from the two taxa. The maximum distance between colonies in this study spanned approximately 120 km. Moreover, the sampling locations of V. maculifrons showed considerable overlap with those of V. squamosa (Fig. 2).

We determined if the 43 microsatellite loci developed in related species within the Vespidae were informative in V. maculifrons or V. squamosa. Overall, we found that 16 loci amplified and were variable in both species (Appendix). We used eight of these 16 loci for our study (see Methods). We analyzed reduced data sets consisting of only one individual per colony to determine if loci were in Hardy-Weinberg equilibrium in the two species. We found no evidence of disequilibrium among workers in either Vespula taxon (P > 0.1 for all loci in both species; overall combined P = 0.2388 in V. maculifrons and P = 0.5304 in V. squamosa). In addition, there was no evidence of linkage disequilibrium between any pair of loci (P > 0.3 for all pairs of loci in both species). The eight loci that we ultimately utilized for this study showed substantial variability in the two taxa, with gene diversities (h) often exceeding 0.7 (Table 1). The exceptions were the loci Rufa-5 in V. maculifrons and LIST2019 in V. squamosa, where the variability was somewhat less. Regardless, the variability of the loci was high enough such that the probability of any two males having the same multilocus genotype (nondetection error) was extremely low (<< 0.0001).

The standardized number of alleles (A₅₀, defined as the number of alleles expected to be observed if only 50 genes had been sampled from a given population; Table 1; 41) did not differ significantly between the taxa [12.8 ± 6.3 (mean ± SD) for V. maculifrons and 11.5 ± 4.45 for V. squamosa; paired t-test, t = 0.74, P = 0.48] suggesting that genetic variability was not substantially different between the parasite and its host. In addition, the effective number of alleles did not show strong differences between species (9.2 ± 5.90 for V. maculifrons and 7.0 ± 3.31 for V. squamosa; paired t-test, t = 1.27, P = 0.24). Moreover, the variability at particular loci was similar in the two species. Specifically, the estimates of A₅₀ and expected heterozygosity (h) were significantly correlated across loci in the two taxa (Spearman's correlation coefficient; A₅₀, r_S = 0.738, P = 0.0366; h, r_S = 0.762, P = 0.0280).

The overall estimates of genetic differentiation between colonies in V. maculifrons and V. squamosa were F_ST = 0.1953 and 0.1914, respectively. We found no evidence of genetic isolation by distance among colonies in either species (Fig. 3). The correlation between geographic distance and genetic distance was low and nonsignificant in V. maculifrons (F_ST, r_S = 0.0176, P = 0.4694) and V. squamosa (F_ST, r_S = 0.0036, P = 0.4748). Thus, both of these Vespula species are apparently able to mate at random over the range in which sampling occurred.

As expected, we found that the genotype distributions of workers within colonies always were consistent with the presence of a single queen mated to multiple males. V. maculifrons queens mated with 5.64 ± 1.27 males (mean ± SD; range of 3 – 8, Fig. 4), whereas V. squamosa queens mated with 7.25 ± 1.86 males (range of 5 – 12; Fig. 4). The mate number of queens in the two taxa differed significantly from each other (Kruskal Wallis S = 412.5, P = 0.0037). The estimates of effective mate number (k_e3) in V. maculifrons and V. squamosa were 4.96 ± 1.40 and 5.58 ± 1.66, respectively (range in k_e3 of 2.61 – 8.82 in V. maculifrons and 3.75 – 10.15 in V. squamosa; Fig. 4). In contrast to the actual mate number, the estimates of effective mate number did not differ significantly between species (S = 343, P = 0.2482).

We then turned our attention to the reproductive skew of males mated to the same queen. The mean magnitude of skew was B = 0.0273 ± 0.0409 in V. maculifrons and B = 0.0433 ± 0.0411 in V. squamosa. These values were not significantly different from each other (S = 371, P = 0.0685). However, the estimates of skew were significantly greater than zero in only 13 of the 36 V. maculifrons colonies, whereas the estimates of skew were significant in 10 of the 12 V. squamosa colonies. Thus, the proportions of colonies in which the skew was significant in the two taxa differed (G₁ = 8.553, P = 0.0034). Nevertheless, although paternity skew was statistically significant in many colonies, the overall magnitude of B was low and the actual mate numbers were similar to the effective mate numbers in both species.

We investigated if the distribution of queen mate number in the two species fit known distributions to determine if simple biological processes could explain patterns of queen mating behavior 44. We found that the distribution of mate number in V. maculifrons and V. squamosa differed significantly from that expected under a Gaussian distribution (Shapiro-Wilk test; V. maculifrons, W = 0.920, P = 0.0125; V. squamosa, W = 0.852, P = 0.039). The failure of these distributions to fit the data arose from an observed excess of queens that mated with an intermediate number of males (Fig. 4).

Nestmate V. maculifrons and V. squamosa workers had relatedness values significantly greater than zero (Table 2). However, the relatedness of queens to their male mates did not differ significantly from zero (Table 2), as judged by the fact that the 95% confidence intervals (r ± 1.96 * SEM) overlapped zero. This indicated that inbreeding does not usually occur in either of these species. Furthermore, we found that relatedness of V. maculifrons males mated to single queens was not significantly different from zero (Table 2). But we did find weak evidence that relatedness among V. squamosa males mated to single females was significantly greater than zero. Much of this signal, however, was due to a single colony in which the relatedness among male mates was surprisingly high (r = 0.312).

One of the more striking results arising from our investigation of levels of genetic variation was that measures of allelic diversity were not appreciably different between V. squamosa and V. maculifrons. Population genetics theory suggests that the amount of genetic variation maintained within a population is related to the effective population size 51. Therefore, the shared patterns of diversity among these species suggest that long-term effective population size is not substantially different in the two taxa.

Furthermore, we failed to find evidence for genetic isolation by distance in either Vespula species. The absence of isolation by distance within both taxa suggests that these Vespula wasps exhibit substantial levels of gene flow within the sampling range of this study. This finding is consistent with the high flight capacity of Vespula wasps, which are naturally capable of dispersing over considerable distances 5253. In addition, hibernating Vespula queens may readily be moved via accidental human transportation thereby increasing their effective migratory abilities 5455.

However, our finding of a lack of geographic structure in V. maculifrons and V. squamosa differs somewhat from that found in a previous study of the congener V. germanica in Australia 33. In that investigation, V. germanica was found to display genetic isolation by distance on approximately the same scale investigated in this study. We note, however, that V. germanica is a recent invader to Australia 56. Thus, the population dynamics of V. germanica in Australia likely differ from those of V. maculifrons and V. squamosa in their native range in the United States. In particular, the observed isolation by distance in V. germanica populations in Australia may reflect non-equilibrium conditions that will ultimately fade as gene flow swamps local genetic differentiation 17.

The overall similarity of levels of genetic variation and low levels of structure in V. maculifrons and V. squamosa stand in contrast to expectations derived from other studies of social hymenopteran parasites. For example, ant social parasites often have low population sizes and dispersal capabilities 7. In accord with these predictions, Trontti et al. 57 found that an inquiline parasite species exhibited much greater genetic substructuring than was found in its host. Brandt et al. 58 also found substantial genetic structure in two species of parasitic ants, although the magnitude of structuring did not differ substantially from that of the host species. Thus our findings indicate that social parasites need not show levels of variation distinctly different from their hosts.

We compared the mating system of V. maculifrons and V. squamosa in order to investigate if there were differences between the two species that might be associated with parasitic behaviors. As expected, we found that all colonies of both species were headed by only a single queen, and that all queens of both species were polyandrous 353637. Our estimates of worker relatedness for V. maculifrons and V. squamosa (0.373 and 0.357, respectively; Table 2) did not differ significantly from earlier estimates obtained for these species (0.320 and 0.403; 35). Thus, we found no evidence for temporal variation in the mating systems of these two taxa from samples obtained some 23 years apart. Although only tangential to this study, this result is noteworthy. Few studies have investigated temporal stability of mating structure. Here, we found that structure did not change over a 23 generation time-span. This stability in general mating structure suggests that selective factors affecting mating behavior have remained constant over that time.

We also did not find any convincing evidence of inbreeding between queens and males of both the parasite and its host. In addition, the relatedness of males mated to the same queen tended to be low in both taxa (Table 2). These data are consistent with our understanding of the mating system of Vespula wasps. Namely, although inbreeding is possible under laboratory conditions 59, matings in natural settings tend to take place away from the nest and result in outbreeding 183260. Thus, the parasitic behavior of V. squamosa has not led to a cycle of inbreeding as occurs in some socially parasitic ants 761.

Our primary interest in studying queen mate number was to test hypotheses related to how mate number should evolve in social parasites. We predicted that mate number would be similar in the two taxa, because V. squamosa colonies exist as free-living entities for a substantial part of their life cycle. Therefore, they should retain benefits of multiple mating 37, such as superior task performance of parasite defense associated with increased within-colony genetic diversity, unlike obligate social parasites that are more intimately associated with their hosts 30. Interestingly, we found that queen mate number differed significantly between the species, with parasitic queens tending to have more mates than host queens. This finding may result from a regime of strong selection of parasites on hosts. In this case, hosts of social parasites might be selected to develop less genetically diverse colonies to increase colony uniformity 7 and hence increase parasite recognition. However, this is not likely to be the case in this system, as levels of within-colony genetic diversity, the currency of multiple mating, did not differ substantially between the species. Thus, overall, our results suggest that V. squamosa does indeed bear the benefits of multiple mating likely as a result of the free-living portion of its life cycle 30.

We also found that effective mating frequency of queens was similar between the species. This suggests that unequal use of sperm reduces the effective mate numbers to approximately equal levels in both species, despite the fact that actual queen mate number differed significantly in the two taxa. Although the magnitude of paternity skew was similar between species, such skew was significant more frequently in V. squamosa. Male skew has important implications for the reproductive success of males and, potentially, for how colony members allocate resources to the production of sexuals 62. The underlying mechanisms resulting in the variation of male reproductive skew between the two species remain unclear at this time, but may result from female choice of male sperm or male-male competition occurring among sperm within the female reproductive tract.

Finally, one striking feature of our data was the similarity in the shape of the distribution of mate number in both species. In particular, the distribution of the number of male mates (Fig. 4) indicated that the extremes at both ends of the distribution had been somewhat truncated. This suggests that the number of mates observed does not simply result from random interactions between queens and males. One explanation for the observed pattern is that mate number is under stabilizing selection. Specifically, mating to too few males may be maladaptive 63, whereas mating to too many males may be costly 64. In fact, Goodisman et al. 37 identified a positive correlation between a fitness correlate and number of matings by queens in V. maculifrons, indicating that there is a selective advantage to polyandry. Regardless, the patterns of mating behavior found here suggest that similar factors may influence queen mate choice decisions in V. squamosa. However, queens in each species may achieve the optimal effective mate number via different evolutionary strategies. That is, V. squamosa queens mate with more males, but have greater skew in sperm usage, while V. maculifrons queens mate with fewer males, but have less skew.