Our study species, Formica fusca
(L), is a common soil-dwelling ant species which builds small colonies (500–2000 workers) in semi-open habitat (peat bogs, recently logged areas, etc.). The species is facultatively polygyne (i.e. colonies contain multiple queens), with previous estimates of effective queen number between two and five (Hannonen et al. 2004
). Queens commence egg laying in spring, they continue laying for 2–3 weeks and this first offspring cohort comprises all sexual and most worker offspring to be produced during each breeding season. New workers are produced each year in discrete cohorts, and have an average lifespan of ca
1 year, i.e. workers born in July tend sexual brood the next spring and perish during their second summer. The study population, comprising around 60 colonies, is located on a 3
ha island, Lilla Träskön, off the coast of southern Finland. The population consists of both monogyne and polygyne colonies, with a yearly colony mortality of polygyne colonies of around 40% and an average rate of within-colony queen turnover close to 35% (Bargum et al. in press
To obtain polygyne colonies, we collected adult workers from 56 colonies of unknown social structure. From as many colonies as possible, worker pupae (n=53), female sexual (gyne) pupae (n=32) and male pupae (n=29) were also collected on the same sampling occasion. As in other species of Formica, this species produces split sex ratios, with 59% of sexual-producing colonies producing only males or only gynes. In addition, we collected two to six established queens per colony from 13 polygyne colonies at the same site.
To determine the social type (monogyne or polygyne) of the colonies we had collected, we genotyped eight adult workers per colony at six polymorphic DNA-microsatellite loci (FL12, FL20: Chapuisat 1996
; FE13, FE17, FE19, FE21: Gyllenstrand et al. 2002
) following the protocol described in Hannonen et al. (2004)
. The colony-specific genotype distributions were inspected manually and the colonies were assigned as monogyne (n
=18) if the genotypes were consistent with a single mother, or polygyne (n
=38) if they were not. From 12 of the colonies determined as polygyne, sufficient numbers (n
>10) of gyne and worker pupae were available for analysis. From these, we genotyped eight additional adult workers, up to 16 worker pupae, and up to 16 gyne pupae using the above protocol. The number of male-producing polygyne colonies with an adequate sample size was too small (n
=5) to allow robust conclusions. Therefore, we restricted our analysis to colonies producing exclusively gynes. Finally, we dissected the established queens and genotyped them, as well as the contents of their spermatheca, again following the protocol described in Hannonen et al. (2004)
. Based on the genotype data thus obtained, we estimated relatedness with the program Relatedness
v. 5.0.4 (Queller & Goodnight 1989
). Average values were calculated by weighting nests equally, and standard errors were estimated by jackknifing over colonies.
We compared and quantified the number and identity of queens contributing to gyne and worker offspring in four steps. First, we compared the relatedness among gynes with that among worker brood within each colony. If the number of mothers and their reproductive apportionments are similar for both gynes and workers, the relatedness values should converge between the two types of offspring. If one caste is produced by fewer queens or skew is higher, we expect relatedness values to be higher within that caste.
Second, we quantified and compared the effective number of queens producing worker and gyne offspring using the formula
is the average genetic relatedness among female offspring of a single queen (taking into account the frequency of multiple mating and the relatedness among the male mates of the queen); rQ
is the average relatedness among nest mate queens; rm
is the average relatedness among the males mated with different queens of the same colony (Queller 1993
; Ross 1993
; Seppä 1994
); and rc
is the estimated relatedness among offspring of one caste (e.g. Queller 1993
; Ross 1993
). To calculate rs
, we estimated queen-mating frequency from the spermathecal contents obtained from the old queens. As we did not have access to offspring of these queens to estimate paternity skew and consequently effective mating frequency, we used the observed mating frequency. For the same reason, we could not estimate the relatedness between the male mates of queens and hence assumed rm
to be zero, i.e. that the males were unrelated. This is a reasonable assumption as rm
was shown to be zero in another population of the same age and species (Hannonen et al. 2004
). These assumptions give a maximum estimate of effective queen number, which should be unbiased for the two offspring groups. RQ
was estimated based on the genotypes of the established queens.
Third, to quantify the degree to which apportionment differs between worker and gyne brood, we estimated breeder shifting between castes (i.e. the degree to which the same queens contributed to gyne and worker brood, respectively) by applying the formula (Pedersen & Boomsma 1999
is the estimated relatedness among gyne offspring; rw
is the relatedness among worker offspring; and rg,w
is the relatedness between castes. This formula was originally developed to measure queen turnover, i.e. breeder shifting between temporally separated offspring clutches, but may equally well be used to describe differences in queen apportionment between any two groups. In our case, this represents a shift in breeder identity (breeder shifting) between different types of brood. The estimate can take values from 0, signifying no shift in the number and identity of queens producing gynes and workers, and 1, which signifies a complete absence of overlap in breeder identity between gyne and worker offspring. We also used this approach to compare changes in queen apportionment between the adult workers present in the colony and the gyne and worker brood they were raising. In this way, we can assess whether the same maternal queens that had produced the adult workers also produced the gynes and the new workers. To test for a difference in turnover between adult workers to gyne brood and adult workers to worker brood, we obtained confidence intervals for the turnover values by jackknifing over colonies.
Finally, we investigated whether the two offspring groups are genetically differentiated by partitioning the observed allelic diversity into variation between colonies, variation between castes within the same colony, and within-colony variation between individuals of the same caste, using the software Arlequin
v. 2.0 (Schneider et al. 2000
). The F
-value of interest here is the one reflecting differentiation between castes within colonies. A significant deviation from zero would indicate that the castes are genetically differentiated. Such genetic differentiation may arise if the number of reproductive individuals producing each caste differs considerably, or if different individuals produce each caste. Since two colonies had missing data for one locus each, we calculated the F
-values for two datasets: all 12 colonies using four loci and 10 colonies using all six loci. We also quantified the number of unique alleles, i.e. alleles found in one caste but not the other, within each colony across all loci.
To assess whether workers gain inclusive fitness benefits through a shift in the identity of queens producing worker and gyne brood, we estimated the relatedness of adult workers to gyne versus worker brood. We obtained the relatedness estimates by implementing the asymmetric relatedness algorithm (px
) in Relatedness
v. 5.0.4 (Queller & Goodnight 1989
), assigning adult workers as px
and either worker or gyne brood as py
. We then conducted a pairwise comparison on the respective relatedness values to assess whether adult workers were more closely related to gynes than to worker brood.