Behav Ecol Sociobiol (1993) 33:345-354 Behvioal Ecology and Sociobiology (j Springer-Verlag1993

Genetic populationstructure and sociogeneticorganisation in Formicatruncorum (; Formicidae) LiselotteSundstrom Department of Zoology, PO Box 17 00014, University of Helsinki, Finland

Received March 26, 1993/Acceptedafter revision August 14, 1993

Summary.The genetic population structureand the so- Introduction ciogenetic organization of the red wood truncorumwere compared in two populations with Hamilton's(1964) principle of inclusivefitness is a central monogynouscolonies and two populationswith polygy- concept in explanatorymodels of the evolution of euso- nous colonies. The genetic population structure was cialityamong .The model incorporatesgenetic re- analysed by measuring allele frequency differences latednessas a key factor in the generalcost-benefit for- among local subsets of the main study populations.The mula for helpingbehaviour and is basedon the probabil- analysis of sociogeneticorganisation included estimates ity that two individualscarry genes that originatefrom of nestmatequeen and nestmateworker relatedness, ef- the same ancestor.The principlehas been used most ex- fectivenumber of queens,effective number of matingsper tensively to understand the evolution of queen,relatedness among male matesof nestmatequeens among Hymenopteraand in developingtheories about and relatednessbetween queens and their male mates. social evolution within this group (Hamilton1964; Triv- The monogynouspopulations showed no differentiation ers and Hare 1976;Rosengren and Pamilo 1983;Nonacs between subpopulations,whereas there were significant 1988;Ratnieks 1988; Pamilo 1991b,c). OriginallyHamil- allele frequencydifferences among the subpopulationsin ton (1964)used the eusocialHymenoptera as an example the polygynous population.Workers, queens and males of his theory because the haplodiploidsex determining showed the same geneticalpopulation structure. The re- system of this group can cause higher relatednessbe- latednessamong nestmateworkers and among nestmate tweensisters than betweensisters and brothersor mother queenswas identicalin the polygynoussocieties. In three and offspring.Females may thereforemaximise their in- of the four populations there was a significant het- clusive fitness by raisingfemale siblings instead of their erozygoteexcess among queens.The queenswere related own offspring.This asymmetryin relatednessalso causes to their male mates in the polygynous population potential conflictsbetween workers and queens (Hamil- analysed,but not in the monogynousones. The data sug- ton 1964; Triversand Hare 1976).This conflict has be- gest limiteddispersal and partialintranidal mating in the come importantfor evaluatingkin selection arguments populations with polygynous colonies and outbreeding (Triversand Hare 1976; Boomsma and Grafen 1990, in the populations having monogynous colonies. 1991; Pamilo 1990a, 1991b, c). Genetic relatednessis Polyandrywas commonin both populationtypes; about thereforea key parameterfor understandingthe conse- 50% of the femaleshad mated at least twice. The males quencesof inclusivefitness for social evolution. contributedunequally to the progeny,one male fathering The argumentspresented above often assumethat so- on average75% of the offspringwith double matingand cial insectcolonies have one singlymated queen.Howev- 45-80% with three or more matings. er, the social structureand consequentlythe geneticrela- tionships among colony members in social Key words:Population structure - Genetic relatedness- colonies may vary both inter- and intraspecifically Sociogenetic organisation - Queen number - Social in- (Holldoblerand Wilson 1990,pp. 209-21 1).Ant colonies sects may have several reproducingqueens (polygyny),the queens may mate several times (polyandry),the queens may contribute unequally to the brood and the males may contributeunequally to the brood producedby their mates(Elmes 1980; Pamilo 1982a,1993; Ross 1988;Stille et al. 1991;Heinze et al. 1992).Changing the geneticrela- Correspondence to: L. Sundstr6m tionships among workers within a colony may change 346 the relatednessasymmetries within the colony and hence sociogenetic organization (Seppi 1993) includes the effects the nature of conflict between workers and queens. of these factorsand is hereused to describethe genealog- Therefore,the social organisationof colonies is a critical ical relationshipsamong nestmatesrelative to the whole factorwhen sex allocationunder worker control is evalu- breedingpopulation. ated (Triversand Hare 1976;Boomsma and Grafen1990, The genetic structurewithin ant populationscan be 1991;Pamilo 1990a,199 1b,c) and when analysingpredic- dividedinto two parts,that due to subdivisionof popula- tions about the evolution of colony characteristics tions and that due to relatednessamong individuals with- (Pamilo 1991c). in colonies. The populationstructure reflects patterns of Despite the disadvantagesdecreased nestmate relat- gene flow betweensubpopulations and can be measured edness causes to workers,both polygyny and polyandry by allele frequencydifferences. The distributionof indi- can be favoured by colony-level selection if genotypic viduals within a subpopulationis reflectedby the geno- diversityenhances efficiency (Crozier and Consul 1976; type distributionswithin groups in relation to that be- Moritz and Southwick1987; Page et al. 1989).Polygyny tweengroups. Population viscosity, as definedby the ten- may also be favouredif dispersalrisks are high (Pamilo dency to reproducein the natal patch, also affects the 1991c)or if queensare shortlivedand brood development relatednessestimates, increasing them in geneticallysub- takes a long time (Nonacs 1988; Seppi 1993;Rosengren dividedpopulations. Genetic subdivisionof populations et al. 1993).Kin selectionarguments have been shown to is oftenconsidered favorable for evolutionarychange, be- support the evolution of polygyny both from the work- cause it facilitates genetic change (Wright 1931, 1940). ers' and the queens'point of view under specificcondi- Social behaviourmay thus be a resultof selectionin vis- tions (Pamilo 1991c).The relatednessamong coexisting cous populations(Hamilton 1964; Wilson 1980; Good- queens reflectsthe way colonies become polygynous,by night 1992;Taylor 1992).Secondary polygyny, the coex- recruitingdaughters or by recruitingrandomly from the istence of several reproductive queens, is frequently population. Holldobler and Wilson (1990, pp. 187-188) linked to limited dispersalby sexuals and colony repro- cite only two relatednessrecords for coexisting queens duction by budding (Cherixet al. 1991; Forteliuset al. and Rosengrenet al. (1993)cite eight. The studies avail- 1993).Such limited dispersalis expectedto be expressed able on nestmatequeen relatednesshave shown that co- also in the genetic populationstructure. existing queens in moderatelypolygynous often are Both the populationstructure and the matingpattern related (Craig and Crozier 1979; Pamilo 1981, 1982b; among the individualswithin that populationcan be de- Pearson 1982; Pamilo and Rosengren1984; Douwes et terminedby allozymeanalysis. In this paper I will focus al. 1987;Seppa 1993).In highlypolygynous , such on the genetical population structureand its relation- as Formica aquilonia and Solenopsis invicta (Pamilo ships to the sociogenetic organisation of colonies by 1982b;Rosengren et al. 1993; Ross 1993)queen related- comparingmonogynous and polygynous societies of a ness is close to zero. single species. I will present data on genealogicalrela- Polyandry may be favourable for workers under a tionshipsamong workers,males, queens and theirmates, numberof conditions;increased genetic diversity may be includingthe degreeof polyandryamong queensand the adaptivefor colonies (Crozierand Page 1985; Hamilton occurrenceof unequalsperm contribution of the queen's 1987;Sherman et al. 1988)and it can be facilitatedby the mates.This procedureallows a comparisonof the mating sex allocation conflict (Starr1984; Moritz 1985; but see structure in monogynous and polygynous societies, Woyciechowskiand Lomnicki 1987; Ratnieks 1990).In henceproducing information necessary to analysefactors monogynouscolonies polyandrymay be eitherfavoured affectingthe evolution of polyandryand to test the argu- or selectedagainst, depending on the colony life history ments for kin selectionin the evolution of polygyny. patternand the net effectthe productionof diploidmales has on colony mortality (Page 1980; Crozier and Page 1985; Ratnieks 1990). The occurrenceof diploid males Methods cause high mortalityof monogynouscolonies in Solenop- sis invicta(Ross and Fletcher1986). If a smallproportion Populationand colony samples.Formica truncorum is generallycon- sideredas a species of Formicas.str., although it differsecologically, of diploid males in the brood cause a high mortalityrisk morphologicallyand geneticallyfrom the group of mound-building for a colony, polyandryshould be selectedagainst. How- red wood-ants of the F. rufa group (Yarrow 1955; Collingwood ever, if the queen-workerconflict is a major component 1979; Pamilo et al. 1979). The colony structure is very flexible, of selection,and the occurrenceof diploid males is not, ranging from clearly monodomous to highly polydomous (Rosen- then polyandryis expectedto evolve in monogynousso- gren et al. 1985, 1986). Moreover, the monodomous colonies are cieties (Nonacs 1992). In contrast polyandryis not ex- generally monogynous, whereas the polydomous ones tend to be polygynous (Sundstr6m 1989). The two differentcolony types are pected in polygynoussocieties unless it conferssome se- polymorphicat the same loci and show no systematicallele frequen- lective advantagefor individualqueens. cy differencesbetween major populations. Estimates of genetic relatednessare also important Ants were collected from four differentlocations in south-west when testingkin-selection arguments for the evolutionof Finland and east Jutland (Mols), Denmark (Fig. l). All of these polygyny.The effectivenumbers of coexistingqueens and main populations except Espoo consisted of several sites separated the effectivenumbers of matings by either unsuitable terrain (Denmark) or water (Inkoo, Hanko). derivedfrom relatedness The averagedistances between sites were much greaterin Denmark estimates are, however, significantlyaffected by the ge- than in the Finnish populations. Table 1 gives the number of netic relationshipsamong queens,among the male mates colonies sampled in each population and the type of material col- of queens and betweena queen and her mate. The term lected (workers,reproductive queens, males). Sundstrom(1989) has 347

100 km 100 km

ESPOO

3 HANKO 2 3 INKOO

3 '02,4 2 MOLS

6 7 @ 5 Z 1 s g g s 4 l g f2 | Fig. 1. The sampling sites in Denmark 2ZN < > ?v: ffi / / and Finland. The numbersin the insets - | correspond t } | | zsJ /of l { } 2q,4 to the differentsites in each 1 km population. Insets show the details of 9 km- | | 1!2 ___ 1 km l } i a 0 the Danish population in Mols, the 5 km 1 Hanko population and the Inkoo pop- ulation

shown that the colonies in Inkoo are polygynous and that the ones population (FiJ can be decomposed into effects due to inbreeding in Hanko are monogynous. The other populations have not previ- within subpopulations (Fis)and due to allele frequencydifferences ously been analysed. among subpopulations (F,,) where (1-Fit) = (1-FSt)(1-FjS)(Wright 1943).All these parameterswere calculated weighting the colonies Laboratorycolonies. To estimate the number of matings by queens equally.The reason for this is that sample sizes should not affectthe and the allele frequencies of male mates, the progeny of single contribution each colony makes to the population average. queens collected from polygynous colonies in Inkoo were analyzed. Sample sizes for statistical testing must in the case of social The colonies sampled frequentlycontained well over 20 wingless, insects be defined using the effective population size. For worker mated and functional queens. The queens were housed in ceramic data this is the number of reproductive individuals each colony flowerpots(see Pamilo 1982a)with approximately100 workers and represents,corrected by the numberof colonies sampled.For queen fed maple syrup and standard diet (Keller et al. 1989) and kept at data the number of individuals sampled can be used, because in low but constant humidity. Of 163 colonies 95 produced worker their case every individualpotentially contributesto the gene pool. progeny and 10 produced both male and female sexuals.The corre- The test variable and the effective sample sizes for testing the sponding data were obtained for the monogynous colonies by de- F-values were thus determinedas follows (Li and Horwitz 1953): ducing the genotype of the queen and her mate(s)using both worker Workers x2 = 3NF N = (0.75/rn),where n = number of and male genotype information. Fst colonies analysed and r = estimated worker nestmate relatedness Electrophoresis.Generally 10-30 workers or worker pupae, 10 x2 = NF2 N = males and 10 old queens collected from the natural colonies (Table FiS (0.75/rn) Queens x2 = 2NF N = number of individuals 1), and all or a maximum of 30 pupae from the laboratorycolonies, Fst analysed = NF2 were analysed by horizontal starch gel electrophoresis. If adult FiSX2 workerswere used, the abdomenswere removedand the ants rinsed The product NF is multipliedby 3 for workers because they repre- in water before deep-freezing(-70?C). This procedurewas not nec- sent at least three independent haploid genomes (two from the essary for males or old queens, but the abdomens of the old queens mother and one from the father) from the previous reproductive were removed for dissections. The electrophoreticprocedure is de- generation (Pamilo 1983). scribedin Seppa (1992).Three loci were studied:isocitrate dehydro- The relatednesswas measured from allele frequencydata as a genase (Idh,two alleles),phosphoglycerate kinase (Pgk, two alleles) regression coefficient (Pamilo and Crozier 1982; Pamilo 1989, and aconitase (Aco, three alleles).All stainings used buffersystem II 1990b;Queller and Goodnight 1989)using the algorithmof Pamilo described in Seppa (1992). (1989) and the weighting scheme of Pamilo (1990b). The standard errors of the means were obtained by jackknifingover the colonies. Geneticrelatedness and matingstructure. Genotype frequencydata The standard errors obtained correspond to the standard error of were used to test both the relatedness (r) among nestmates and the mean. The relatednessestimates thus obtained were tested for subdivision of hierarchicalpopulations. The population structure deviationsfrom zero, 0.75 or 0.25 using Student'st-distribution. The was analysed using F-statistics (Wright 1951). A deviation from relatednessestimates are affectedby both genetic differentiationof panmixis is determined by comparing the observed (Hobs) and ex- subpopulations and by local inbreeding (Pamilo 1989). Therefore pected (Hexp) amounts of heterozygosity over all loci and is de- the relatedness values used for estimating the effective number of scribed by the inbreeding coefficient F = 1-(Hobs/Hexp). Subdivi- reproductivefemales per colony were corrected for inbreeding as sion of the population is determinedas Ft = (Ht-Hs)/H,, where Hs suggested by Pamilo (1990b).In the Inkoo population this correc- is the mean expected amount of heterozygosity in the subpopula- tion yielded negative relatedness estimates both for queens and tions and Ht is total expectedamount of heterozygosityin the whole workers because the differentiationamong sites was considerable. population (Nei 1987, p. 190). The inbreeding coefficient within a Thereforethe relatednesswas estimated separatelyfor each site in 348

Inkoo and the mean of those estimates is used. This is in contrast to Results the procedure used by Seppa (1993) where inbreedingwithin local sites boosted the relatednessestimates. Diploid males, detected by Genetic relatedness among nestmate workers, heterozygous genotypes in any of the three loci, were found in the Inkoo population and omitted from all analyses. queens and males If the queens or their mates are related to each other or do not contributeequally to the brood the relatednessestimates alone will The worker relatednessestimates from naturalcolonies not give the correct estimate of the number of reproducingindivid- and laboratorycultures are shown in Table 1. The varia- uals within colonies. Therefore,the effective number of queens is tion in relatednessamong populationsis very high and estimated using the formula of Ross (1993) all the populations have a relatednessestimate lower Ne= (4rs-rq-2rmi)/(4rf-rq-2rmi) (1) than 0.75, indicatingeither multiple mating by queens, where rs is the average relatednessof progeny of a single queen, rq polygyny or both. In the populationswith high related- is the averagerelatedness of nestmatequeens, rfis the averagerelat- ness values each colony usuallyhas only one or two off- edness of nestmate workers and rml is the average relatedness spring genotypesat each locus and in the latter case in among the male mates of coexisting queens in single polygynous roughly equal proportions.If there are more than two colonies (Ross 1993). This estimate corresponds to the harmonic workergenotypes they can all be explained mean number of queens per nest if all nestmate queens reproduce by multiple equally. The present data allow the estimation of all these parame- mating.This fits the assumptionof monogynyand occa- ters. sional polyandry.In the populationswith low relatedness Similarily,because all males may not contribute equally to the values single colonies frequentlyshow severalor all pos- progeny of their mates and there may be queen-to-queenvariation sible workerand male genotypes,which can only be ex- in the number of matings, the effectivenumber of mates of a single plainedby the presenceof severalmatrilines. The related- queen is: ness estimates in the monogynous colonies correspond Me= (2rf-0.5-rm2)/(2rs-0.5-rm2) (2) well to those from the laboratorycultures, while the esti- where rfs is the average relatedness of progeny belonging to the matesfrom the polygynouscolonies are significantlylow- same patriline (0.75), r, is the average relatedness of single queen er (t = 9.7, P < 0.0001) and hence cannot be explained by progeny and rm2 is the average relatednessamong mates of single polyandryalone. queens (Ross 1993).Again all the parametersneeded can be estimat- The average relatednessestimates for both workers ed from the currentdata. This estimate correspondsto the harmon- and queensin the Inkoo ic mean number of mates per queen if these mates always share populationare equal,but boost- equally in paternity (Pamilo 1982a). ed by differentiationamong islands and give negative The relatedness between nestmate workers and males and the relatednesseswhen correctedfor differentiationamong relatednessbetween queens and their mates were calculated as the sites (islands).When each site is analysed separately, regressionbetween the two groups (Pamilo 1990b)and expressedas worker, queen and male relatednessesare positive but the regressionof workers or queens on males. For the progeny of a lower than the ones in the pooled data (Table 1). The singly mated queen the expected regressionof workers on males is effectivenumbers of 0.25, but 0.5 for the regressionof males on workers. Workersthus queens are calculatedbased on the have 25% of their genes in common with their brothers, whereas latter values and are presentedin Table 1. Because the males have 50%/Oof their genes in common with their sisters. The monogynous populations in Denmark and Hanko do genotypes of the male mates of queens were deduced by combining not show any such subdivision of populations the in- the informationfrom worker and male and queen genotypes. breedingcorrection can be applied to the pooled data. The observed number of mates of each queen was determined All queensare winglessand most of them are functional, from the progeny by combining the genotype informationfrom all with filled three loci for each individual. In the laboratory colonies the queen spermatheca(105 filled, 14 empty)and corpora genotype was also available. The proportion of offspringfathered lutea scarson the ovaries(108 positive, 11 negative)indi- by each male in double and triple matings was determined using cating egg laying. formulas 1 and 2 from Pamilo (1993).In the formulathe proportion of offspring fathered by the ith male is pi and the relatedness g = 0.25 + 1/21pi2. I use the parameterg to distinguish this mea- Genetic population structure sure of pedigree relatednessfrom the estimated average relatedness r among colonies. The paternity contributions of each male were Table 2 shows the inbreedingcoefficients relative to the corrected for a finite sample size giving Xpi2 = (N Xpi2-1)/(N-1), where N is the number of individualsanalysed within each colony. total population(Fit) and decomposedinto effectsdue to In this way relatedness was estimated separately for each colony inbreedingwithin subpopulations(Fi,)and due to genetic with one multiply mated queen. From these data the means and subdivision of populations (F,t). where (1-Fit) = (1- variancesover all colonies were calculated.Corrections for inbreed- Fst)(1-Fis).When populationdifferentiation (F,t) increas- ing and allele frequency differenceswere not made because there es the Fit value will also increase. An excess of het- was no significantinbreeding in the colonies included and no allele erozygotes(negative may mask the effectsof frequencydifferences between mates. The relatednessfor progenies Fj1) genetic of singly mated queens was calculatedthe same way as the average subdivisionof populations.This has consequencesfor the intracolonyrelatednesses above (Pamilo 1989, 1990b)using a subset relatednessestimates as they are calculated assuming of colonies determinedas having a singly mated queen on the basis that populationsare in Hardy-Weinbergequilibrium, as of genotype distributions.Some double (or triple) matings are not will be explainedbelow. detected by this method because multiple male mates of single The negativeFis values indicatea significantexcess of queens may have identical genotypes. This was corrected by calcu- heterozygosityin lating the probabilitythat two males have the same genotype using two of the three loci among queens in the allele frequenciesin the local population. Thus, if the proportion both monogynous populations (Denmark and Hanko) of males with an identical genotype is p and the observed propor- (x2= 13.6, P<0.001, x2 = 3.9, P <0.05). Neither the tion of double matings D0, then the correctedproportion of double queens nor the workersshow any genetic subdivisionof matings is: Dc = Do/(l-p). eithermonogynous population. There is also a slight but 349

Table 1. Within-casterelatedness estimates of workers,reproductive queens and males and effectivenumber of queens in differentpopulations of Formicatruncorum

Population N r SE r* SE NE

Monogynous populations Denmark Workers 91 0.60*** 0.03 0.60*** 0.03 1 Males 33 0.37 ** 0.04 Hanko Workers 23 0.60*** 0.05 0.59 ** 0.05 1 Males 18 0.36* 0.05

Polygynous populations Inkoo Workers 89 0.25*** 0.02 0.10*** 0.06 16 Queens 76 0.24*** 0.02 0.07*** 0.03 Males 52 0.14*** 0.02 0.07** 0.03 Espoo Workers 21 0.11 ** 0.04 - 5.6 Queens 11 0.00 0.04 Laboratory colony Workers 95 0.60*** 0.05 0.59*** 0.05

N = numberof colonies analysed,r = relatedness,r * relatednesscorrected for inbreeding(Denmark, Hanko, Espoo, laboratorycolony) and average relatednesson islands (Inkoo) NE= effectivenumber of queens, determinedusing the correctedestimates (r*) Other asterisksindicate significantdeviations from zero (reproductivequeens, workers and males from polygynous colonies), 0.75 (workers from monogynous colonies and laboratory colonies) or 0.5 (males from monogynous colonies): * P<0.05; ** P<0.01; *** P<0.001

Table 2. Genetic population structureand inbreedingcoefficients ance and increasesthe intracolonydiversity causing low- er relatednessestimates. The genotypedistributions in all N Fit Fst F i loci correspondto those expectedfor brother-sisterpairs, and also for one female(the queen)producing them. The Monogynous populations worker-maleregression is slightly lower in the polygy- Denmark Workers 91 0.08 0.03 -0.12 nous colonies, but this estimate is probablyboosted by Denmark Queens 85 -0.33*** 0.05 -0.40*** Hanko Workers 24 0.06 0.03 -0.09 the allele frequencydifferences among islands (positive Hanko Queens 23 -0.34 0.05 -0.41 * Fi value). The Tt values describingthe genetic subdivisionof Polygynous populations the Inkoo populationare similarfor queensand workers Inkoo Workers 89 0.21*** 0.19*** 0.02 and indicate considerable allele frequency differences Inkoo Queens 76 0.08 0.19*** -0.13*** among subpopulations.The Fi, value is effectivelyzero Espoo Workers 21 0.05 for the workersand shows that workerallele and geno- Espoo Queens 11 0.09 type frequencies are in Hardy-Weinbergequilibrium N =number of colonies analysed within subpopulations.Therefore the positive Fit value Asterisks indicate significant deviations from zero * P < 0.05, among workersonly reflectspopulation subdivision, not ** P<0.0l, *** P<0.001 inbreeding.The allele frequenciesdo not differbetween queens and their mates (X2< 1.8, P >0.05 for all loci and all populations).The Espoo population does not show significant heterozygote excess among the queens in the any inbreeding. polygynous Inkoo population (X2= 13.2, P <0.001), but not among the queens in the Espoo population (Fit posi- tive) (Table 2). Number of matings The estimates of relatedness between workers and nestmate males are lower in the monogynous colonies The results from the laboratorycultures show multiple than expected for a brother-sister relationship. This may mating of the queens (Table 3). The effectivenumber of be caused by the high excess of heterozygotes among matings(Me) is 1.43in all populationsassuming that the queens in two of the three loci (Denmark: Pgk: F = - male mates of single queensare unrelated.The estimated 0.2+0.1, Aco: F = -0.18+0.08, Idh: F = -0.03+0.1; relatednessamong male mates of single queens (rm2) is Hanko: Pgk: F = -0.44+0.11, Aco: F = -0.30+0.13, zero in all cases,whereas the averagerelatedness of male Idh: F = -0.02 + 0.03). The corresponding regression co- mates of nestmate queens is higher than zero in the efficients of nestmate workers on males is for the Danish polygynous Inkoo population (Table 4). This difference population 0.10 + 0.06, 0. 11 + 0.05 and 0.25 + 0.08 and for occursbecause relatedness estimates use the background the Hanko population 0.17+0.1, 0.20+0.06 and population as a referencefor the degree of similarity 1.00 +1.00 respectively. In Hanko Idh is monomorphic within groups (Pamilo 1990b).The average estimate of for the colonies that produced males. An excess of het- relatednessamong male mates of nestmate queens uses erozygotes among queens reduces the between-nest vari- the othercolonies in the populationas reference,whereas 350

Table 3. The number of matings and paternity contribution of males as detected from offspringgenotypes

Population N % Number of mates

1 2 27p2 > 3 Lp

Corr Obs Corr Obs Obs

Inkoo 95 12 57 58 13 12 0.77 + 0.05 25 0.42 +0.09 Denmark 93 27 70 75 20 15 0.53 +0.05 3 0.37 + 0.15 Hanko 23 19 11 13 10 8 0.56 + 0.05 2 0.65 + 0.25

The observed values (obs) are corrected (corr) for undetected matings by males with identical genotypes. The paternity contribution is expressed as the variance in the proportion of offspringfathered by each male (_7p2') with a correctionfor sample sizes, % = percentage of male mates with identical genotypes calculated using the populationwise allele frequencies

Table 4. Relatedness(r ?SE) of queens to their male mates, among the male mates of nestmate and single queens and between nestmate workers and males

Population N n Queen to Among the male mates of Workersto males male mate Nestmate queens Single queens

Inkoo 10 95 0.16+0.04*** 0.1 1+ 0.04 * 0.01+0.03 0.08 + 0.02 (23)*** Denmark 5 85 0.02+0.02 0.002+0.02 0.02+0.07 0.14 + 0.03 (33)*** Hanko 5 23 0.10+0.06 0.005+0.11 0.03+0.13 0.18+0.05 (18)

The number of male-producingcolonies analysed is shown in parentheses N representsthe number of colonies analysed in Inkoo and the numberof sites in the monogynous populations from Denmark and Hanko (the queen genotypes within each site have been pooled), n representsthe number of queens Significancelevels (t-tests) are shown for deviations of r from zero (queen to male mate, among male mates of queens) or deviations from 0.25 (workers to males): *** P<0.001 the estimateof male mates of single queens also includes the male mates the other queensin the same colony. Be- cause the male mates of nestmatequeens on averageare D 0.8o- relatedthis will mask any relatednessamong male mates of single queens. Assumingthat the relatednessof male 0.6-4 mates of single queens is approximatelythe same as the a: average relatedness among male mates of nestmate 0.4- queens in a polygynous colony (r = 0.11, Table 4), the 0.2- effectivenumber of mates rises only slightly (Me = 1.5). The pedigreerelatedness (g) was estimatedfor proge- 0- nies of multiplymated single queens.The expectedrelat- 0 2 4 6 edness among progeniesof doubly mated queensis 0.5 if paternitycontribution is equal, but this estimatewill be Numberof mates affectedby how males contributeto the offspring.The Fig. 2. Pedigreerelatedness estimates (g) and 95% confidenceinter- vals for progenies from laboratory colonies, with a queen mated varianceof paternitycontribution (Xp12) (Table 3) shows one, two, three,four and over four times. The value of g is estimated that paternityis unequal,corresponding to one male fa- separatelyfor each colony using information on the proportion of thering60-90% of the offspringwith double matingand offspringfathered by each male. The relatednessestimate for proge- 40-80% with three or more matings.The g values were nies of singly mated queens correspondsto the average relatedness tested separately on the progeny from the laboratory (r).The numberof coloniesincluded in the analyses and significance levels = ** = *** = colonies for 2, 3, 4 and 5 mates using t = where are shown (* P<0.05, P<0.01, P<0.001). (x-,t)/sx, The solid line shows the expected relatednessat equal contribution [t is the expectedrelatedness with equal contribution,x is of males the observedmean value and sx is the standarderror of the mean(Fig. 2). Only data fromthe laboratorycolonies were used for this purpose.They are more reliableboth higher than zero in the polygynouspopulation (t = 4.3, because the queen genotypes were known and because n = 10, P < 0.01),but not in the monogynousones (Den- the expected proportion of male mates with identical mark: t = 2, n = 5, P>0.05; Hanko: t = 1, n = 5, genotypeswas low. The observedvalues are in the tested P > 0.05) (Table 4). The relatedness among the male cases significantlyhigher than the ones expectedat equal mates of nestmate queens also differsfrom zero in the contribution(Fig. 2). polygynous colonies (t = 2.8, n = 10, P<0.05), but not The relatednessbetween queens and their mates is in the monogynous ones (t = 0.1, n = 5, P >0.05). The 351 queen to male mate regressionis higher than that for tions are in Hardy-Weinbergequilibrium. Excess het- workersto males,but the differenceis not significant.The erozygositywill decreasethe intercolonygenetic variance two meansfall withinthe 95% confidenceintervals of the and increasethe intracolonyvariance, resulting in lower means (Table4). relatednessestimates. Males are particularilyaffected by this, because they inherit their entire genome from the queen.On the basis of workerand male genotypedistri- Discussion bution the males are likely to be producedby the queen. Worker reproductioncannot be completely ruled out, Genetic population structure and intracolony relatedness but is not likely to occur as this would, rather,increase the relatednessbetween workers and males. The results show considerablevariation in estimatesof The parameterF includesall factorsthat reflectgenet- genetic relatednessamong populations.This variationin ic subdivision of populations - both increased ho- relatednesspatterns correspond to differencesin the ge- mozygosity due to inbreedingand the Wahlundeffect netic structureof the populations.Low intracolonyrelat- (Wahlund 1928) resulting from allele frequencydiffer- edness due to polygyny correlateswith considerablege- ences among subpopulations.There was a significantly netic differentiationbetween subsets of main popula- positive Fit value only for workers in the polygynous tions, whereas no such differentiationwas detected in Inkoo population consisting of several subpopulations. those showing high intracolony relatedness.This sug- Because the effectivepopulation size is very large the gests limited dispersaland partial intranidalmating of polygynouspopulations, inbreeding is expectedto be low sexualsin the polygynouspopulation and outbreedingin and thereforethe declinein heterozygosityis expectedto the monogynouspopulations. be slow. Decomposingthe Fit value into its components The smallvariation in the relatednessestimates within shows that it consists almost entirelyof allele frequency each populationmoreover indicates sociogenetically rel- differencesamong subpopulations.This suggestslimited atively uniform colonies within populations. The de- dispersalin the polygynouspopulation. In the monogy- crease of relatednessbelow 0.75 can be explained by nous populationsthe lack of allele frequencydifferences polyandryin the populationsfrom Denmark and Hanko, betweensubpopulations suggests either continuous gene which is confirmedby the genotypedistributions among flow betweenthem, or that those populationsare young known workerand male offspringof single queens.This and representseveral independent colonizations. is in correspondancewith earlierresults for this species Similar correlations between sociogenetic organisa- (Sundstrom 1989). The Inkoo and Espoo populations tion and populationstructure have been shown in Formi- show a clearlylower within-nestrelatedness estimate and ca exsecta (Pamilo and Rosengren 1984) and Formica the workerand male genotypescan in most cases not be sanguinea (Pamilo 1981; Pamilo and Varvio-Aho1979), explainedas being producedby a single multiplymated althoughthe patternis different.F. exsecta shows signs of queen. The expectedrelatedness among nestmatework- inbreedingin monogynouspopulations. In F. sanguinea ers decreasesto 0.25 with increasedpolyandry. All the the variationin queen numberis continuousand higher empiricalestimates for polygynous nests are well below numbersof queenscorrelate with higherpopulation vis- this value, when corrected for genetic differentiation cosity (Pamiloand Varvio-Aho1979). among subpopulations.Therefore two differentcolony Environmentalfactors such as habitat structurehave types, monogynous and polygynous, can be distin- been proposedas agentsaffecting the evolutionof polyg- guished. The effectivenumber of queens per nest in the yny (Holldoblerand Wilson 1977;Rosengren et al. 1985, Inkoo populationvaries between 2 and 16, assumingei- 1986).In the presentcase the archipelagoislands repre- ther zero or the observed(0.1) relatedness among queens. sent a very uniform habitat structure both in size, However, most colonies sampled for old queens con- longevityand quality,but still containboth monogynous tainedmore than 20 winglessqueens, most of whichwere and polygynouspopulations. Such differencesin colony inseminated and egg-laying. The effective number of structuremay reflect differentages of the populations queensis very sensitiveto smalldifferences in the related- studied. Monogyny is likely to preceed polygyny in ness estimateamong queenswhen the workerrelatedness Formica ants (Rosengren and Pamilo 1983) and the is close to zero. Thereforeestimates of the effectivenum- polygynouscolonies may originallybe foundedby sever- ber of queensshould be treatedwith cautionwhen work- al independentcolonizations. Thus the now polygynous er relatednessis low. The relatednessestimates may be populations may have had much the same population further decreasedby queens being supersededand re- structureas the monogynousones have now. This is part- placed. Queens of monogynous Formicaspecies appear ly supported by the results for the Espoo population, to live on average 20 years (Pamilo 1991a),but no esti- where the relatednessamong queens is lower than that mates on averagelife span of polygynousFormica queens among queensin the Inkoo population.The Espoo pop- are available. ulation may representa transition period when young The relatednessestimates for nestmate males in the femalesdisperse, but join existingcolonies yieldingnon- monogynous populations are significantlylower than related queens, but related workers if the number of 0.5. Similarilythe worker-maleregression coefficients are queens is relativelylow. Subsequentintranidal mating lower thanl0.25. This is largely due to the excess het- and adoption of daughterqueens will increasethe relat- erozygosity among the queens in these populations,as edness among queens,as is observedin the Inkoo popu- the expectationsof 0.5 and 0.25 assumethat the popula- lation. 352

The similarityin geneticpopulation structure and ge- tory colonies. However,with three or more matingsthe netic relatednessestimates between workersand queens proportion of undetectedmatings increases.Therefore in polygynouscolonies suggestthat daughterqueens are the estimatesof proportionof offspringfathered by each recruitedback to the colony. Queenswere related to their male are less reliablewhen the numberof matingsis high. mates in the polygynous societies,but not in the mono- The proportion of offspringfathered by each male is, gynous ones. This also suggests intranidalmating and however,clearly unequal also when there are only two limiteddispersal in the polygynouspopulations, whereas mates.The relatednessestimates obtained for two species nuptialflights and outbreedingis the rulein the monogy- of social wasps (Vespula squamosa and Paravespula mac- nous populations. This connection between polygyny ulifrons)were considerablylower than the ones obtained and intranidalmating and monogyny and dispersalhas in this study. The femaleshad mated with two to seven been suggested by behavioural studies on the present males and relatednessamong the progenieswere 0.40 and populationsand severalother Formica species(Fortelius 0.32, spermbeing used randomlyfor fertilizations(Ross 1987; Forteliuset al. 1993; Rosengrenet al. 1993).How- 1986). ever, the slightly lower relatedness estimates among Polyandryis considereda key factorin increasingge- queens than among workers in the Espoo populations netic diversity in monogynous societies (Crozier and indicaterelatively few but unrelatedqueens or reproduc- Page 1985)emphasizing the adaptivesignificance of di- tive competitionamong queens. versity.It may also be selectedagainst if multiplemating Although there were no consistent allele frequency leads to a larger fraction of colonies producingdiploid differencesamong workers,queens and maleswithin sub- males and any proportionof diploid males causes a load sets of the polygynous societies or among monogynous to the colony (Page 1980; Crozierand Page 1985).This populationsthere was an excess of heterozygotesamong predictionthus dependson the effectdiploid males have queens,but not among workers.Such a patternhas earli- on monogynouscolonies, while polygynousones should er been observedin Formica exsecta (Pamiloand Rosen- not show any distinction.In Solenopsis invicta the occur- gren 1984).High levelsof heterozygositymay be associat- rence of diploid males caused high mortalityof incipient ed with heterosiseffects and may give a selectiveadvan- monogynous monandrouscolonies (Ross and Fletcher tage during the nestfoundingstage or yield faster larval 1986).The monogynous and polygynous colonies of F. developmentrates selectinghighly heterozygousfemales truncorumare equallypolyandrous so neitherscenario is to develop into queens. unequivocallysupported. With the evolution of polygy- ny, mating frequenciesmay also be expected to decline (Nonacs 1992),but this is not supportedby the present Mating structure results.However, if the averagemating frequencyis less than two, as is the case in this species,the populationis Polyandryis common, with about 50% of the females expected to share more characteristicswith a monan- being at least twice mated both in monogynous and drous society than with a polyandroussociety (Nonacs polygynous colonies. The frequencyof triple matings is 1992). higher in the polygynous societies than in the monogy- Although polyandryhas a minor effecton increasing nous ones. This may be due to a higher availabilityof diversityin polygynoussocieties, it may be advantageous males becauseof the male-biasedsex ratio (Rosengrenet for individualqueens. If the frequencyof diploidmales in al. 1986; own observations).It may also be a methodo- the population is high, females may benefit by mating logical artifact because the queen genotypes in the multiply to ensure reception of viable sperm. A larger monogynouscolonies were only indirectlyinferred from geneticheterogeneity among the offspringmay in itselfbe the workerand male genotypesand are thereforenot as adaptive and may also increase the chance of having reliable as for the laboratory colonies. The number of some highly heterozygous individuals among the off- matings is variable in other Formica ants; some are spring.Considering the observedexcess of heterozygotes polyandrouswhile others are largely monandrousirre- among queens this may have important fitness conse- spectiveof the numberof queens(Pamilo 1993; Pamilo et quencesfor the reproducingindividuals. al. submitted).In general,earlier genetical studies have shown monandryor low levels of polyandryin ants (Page Acknowledgements.I wish to thank P. Pamilo, R. Rosengren, J.J. 1986; Ross et al. 1988; van der Have et al. 1987; Seppa Boomsma, P. Seppa, W Fortelius, P. Punttila, P. Nonacs and two 1993),although high numbersof matingshave been ob- anonymous refereesfor constructivecomments on this manuscript. servedin behaviouralstudies (Page 1986). C. Olsson, L. Walin,and S. Bjorkholmprovided help in the field and Multiplematings caused a smallerdecrease in the re- the laboratory. The work was financed by the Finnish Natural latedness estimates of single queen progenies,than ex- Sciences Research Council, the Entomological Society of Helsinki and Oskar Oflunds Stiftelse. pected if all males contributeequally to the progeny.The data show that the males contributeunequally. This is the rule also among other polyandrous Formica ants (Pamilo 1993).A problemhere is that seeminglyunequal References contributionsof the males may in fact be due to two males havingthe same genotype.This numberis likely Boomsma JJ, Grafen A (1990) Intraspecificvariation in ant sex to ratios and the Trivers-Harehypothesis. Evolution 44:1026-1034 be quite small becausethe expectedproportion of males Boomsma JJ, GrafenA (1991)Colony-level sex ratio selectionin the havingidentical genotypes is less than 12%in the labora- eusocial Hymenoptera.J Evol Biol 3:383-407 353

Cherix D, Chautems D, Fletcher DJC, Fortelius W, Gris G, Keller Pamilo P (1981) Genetic organizationof Formicasanguinea popula- L, PasseraL, RosengrenR, Vargo EL, WaltherF (1991)Alterna- tions. Behav Ecol Sociobiol 9:45-50 tive reproductive strategies in Formica lugubris Zett. (Hy- Pamilo P (1982a)Multiple mating in Formicaants. Hereditas97:37- menoptera: Formicidae).Ethol Ecol Evol 1:61-66 45 Collingwood CA (1979) The Formicidae (Hymenoptera) of Pamilo P (1982b) Genetic population structure in polygynous Fennoscandiaand Denmark. Fauna Entomol Scand 8:1-174 Formicaants. Heredity 48:95-106 Craig R, Crozier RH (1979) Relatedness in the polygynous ant Pamilo P (1983) Genetic differentiationwithin subdividedpopula- Myrmeciapilosulat. Evolution 33:335-341 tions of Formicaants. Evolution 37:1010-1022 Crozier RH, Consul PC (1976) Conditions for genetic polymor- Pamilo P (1989) Estimating relatedness in social groups. Trends phism in social Hymenopteraunder selection at the colony level. Ecol Evol 4:353-355 Theor Pop Biol 10: 1-9 Pamilo P (1990a) Sex allocation and queen-worker conflict in CrozierRH, Page RE Jr (1985) On being the right size: male contri- polygynous ants. Behav Ecol Sociobiol 27:31-36 butions and multiplemating in social Hymenoptera.Behav Ecol Pamilo P (1990b) Comparison of relatednessestimators. Evolution Sociobiol 18: 105--15 44:1378-1382 Douwes P, Sivusaari L, Niklasson M, Stille B (1987) Relatedness Pamilo, P (1991a) Life span of queens in the ant Formicaexsecta. among queens in polygynous nests of the ant Leptothoraxacer- Insectes Soc 38:111-120 vorum.Genetica 75:23-29 Pamilo P (1991b) Evolution of colony characteristicsin social in- Elmes GW (1980) Queen numbers in colonies of the ants of the sects 1. Sex allocation. Am Nat 137:83-107 Myrmica.Insectes Soc 27:43-60 Pamilo P (1991c) Evolution of colony characteristicsin social in- Fortelius W (1987) Differentpatterns of female behaviourin mono- sects 2. Number of reproductiveindividuals. Am Nat 138:412- and polydomous Formicapopulations. In: Eder J, Remboldt H 433 (eds) Chemistry and biology of social insects. Peperny, Pamilo P (1993) Polyandryand allele frequencydifferences between Miinchen, pp 293-294 the sexes in the ant Formicaaquilonia. Heredity 70:472-480 Fortelius W, Rosengren, R, Cherix, D Chautems, D (1993) Queen Pamilo P, CrozierR (1982)Measuring genetic relatednessin natural recruitment in a highly polygynous supercolony of Formica populations: methodology. Theor Pop Biol 21:171-193 lugubris(Hymenoptera; Formicidae). Oikos 67:193-200 Pamilo P, Rosengren R (1984) Evolution of nesting strategies of Goodnight K (1992) The effect of stochastic variation on kin selec- ants: genetic evidence from differentpopulation types of Formi- tion in a budding viscous population. Am Nat 6:1028-1040 ca ants. Biol J Linn Soc 21:331-348 Hamilton WD (1964) The genetical evolution of social behaviour.J Pamilo P, Varvio-AhoS (1979) Genetic structureof nests in the ant Theor Biol 7:1-52 Formicasanguinea. Behav Ecol Sociobiol 6:91-98 Hamilton WD (1987) Kinship, recognition, disease, and intelli- Pamilo P, VepsalainenK, RosengrenR, Varvio-AhoS-L, PisarskiB gence: constraintsof social evolution. In: It6 Y, Brown JL, Kik- (1979)Population genetics of Formicaants II. Genetic differenti- kava J (eds) societies: theoriesand facts. Japan Scientific ation between species. Ann Entomol Fenn 45:65-76 Societies, Tokyo, pp 81-102 Pearson B (1982) Relatedness of normal queens (macrogynes)in Have T van der, Boomsma JJ, Menken SBJ (1988) Sex investment nests of the polygynous ant Myrmicarubra Latreille. Behav Ecol ratios and relatednessin the monogynous ant Lasius niger (L.). Sociobiol 12:1-4 Evolution 42:160-170 Queller DC, Goodnight KF (1989) Estimating relatedness using Heinze J, Lipski N, H6lldobler B (1992) Reproductivecompetition genetic markers.Evolution 43:258-275 in colonies of the ant Leptothoraxgredleri. Ethology 90:265-278 Ratnieks FLW (1988) Reproductiveharmony via mutual policing Holldobler B, Wilson EO (1977) The number of queens: an impor- by workers in eusocial Hymenoptera.Am Nat 132:217-236 tant trait in ant evolution. Naturwissenschaften65:8-15 Ratnieks FLW (1990) The evolution of polyandry by queens in Holldobler B, Wilson EO (1990) The ants. Springer,Berlin Heidel- social Hymenoptera: the significance of timing of removal of berg New York diploid males. Behav Ecol Sociobiol 26:343-348 Keller L, Cherix D, Ulloa-Chacon P (1989) Description of a new RosengrenR, Pamilo P (1983) Sex ratio strategyas relatedto queen artificial diet for rearing ant colonies as Iridomyrmexhumilis, number,dispersal behaviour and habitat quality in Formicaants Monomoriumpharaonis and Wasmanniaauropunctata (Hy- (Hymenoptera:Formicidae). Entomol Gen 11: 139-151 menoptera: Formicidae).Insectes Soc 36:348-352 Rosengren R, Cherix D, Pamilo P (1985) Insularecology of the red Li CC Horwitz DG (1953) Some methods of estimatingthe inbreed- wood ant Formicatruncorum Fabr. 1. Polydomous nesting, pop- ing coefficient.A. J Hum Genet 5:107-117 ulation size and foraging. Mitt Schweiz Ges 58:147-175 Moritz RFA (1985) The effects of multiple mating on the worker- Rosengren R, Cherix D, Pamilo P (1986) Insularecology of the red queen conflict in Apis melliferaL. Behav Ecol Sociobiol 16:375- wood ant Formicatruncorum Fabr. 2. Distribution,reproductive 377 strategy and competition. Mitt Schweiz Ges 59:63-94 Moritz RFA, Southwick EE (1987)Phenotype interactions in group Rosengren R, Sundstrom L, Fortelius W (1993) Monogyny and behaviour of honey bee workers (Apis melliferaL.). Behav Ecol polygyny of Formicaants - a result of differentdispersal tactics? Sociobiol 21:53-57 In: Keller L (ed) Queen number and sociality in insects. Oxford Nei M (1987) Molecularevolutionary genetics. Columbia Universi- University Press, Oxford ty Press, New York Ross KG (1986) and the problem of sperm utilization Nonacs P (1988) Queen number in colonies of social Hymenoptera in social insects. Nature 323:798-800 as a kin-selected adaptation. Evolution 42:566-580 Ross KG (1988) Differential reproduction in multiple-queen Nonacs P (1992) Male parentageand sexual deception in the social colonies of the fire ant Solenopsisinvicta (Hymenoptera; Formi- Hymenoptera.In: WrenschDL, EbbertMA (eds) Evolution and cidae). Behav Ecol Sociobiol 23:341-355 diversity of sex ratio in insects and mites. Chapman and Hall, Ross KG (1993) The breeding system of the fire ant Solenopsis New York invicta:effects on colony genetic structure.Am Nat 141:554-576 Page RE Jr (1980) The evolution of multiple mating behaviour by Ross KG, Fletcher DJC (1986) Diploid male production- a signif- honey bee queens (Apis melliferaL.). Genetics 96:263-273 icant colony mortality factor in the fire ant Solenopsis invicta Page RE Jr (1986) Sperm utilization in social insects. Annu Rev (Hymenoptera:Formicidae). Behav Ecol Sociobiol 19:283-291 Entomol 31:297--320 Ross KG, Vargo,EL, FletcherDJC (1988) Colony genetic structure Page RE Jr, Robinson GE, Calderone NW, Rothenbuhler WC and queen mating frequency in fire ants of the subgenus So- (1989) Genetic structure,division of labor and the evolution of lenopsis (Hymenoptera:Formicidae). Biol J Linn Soc 34:105- insect societies. In: Page RE, Breed MD (eds) The genetics of 117 social insects. Westview,Boulder, pp 15-29 354

Seppa P (1992)Genetic relatednessof workernestmates in Myrmica Taylor P (1992) Altruism in viscous populations - an inclusive fit- ruginodis(Hymenoptera: Formicidae) populations. Behav Ecol ness model. Evol Ecol 6:352-356 Sociobiol 30:253-260 TriversRL, Hare H (1976) and the evolution of the Seppi P (1993) Sociogenetic organizationof Myrmicaruginodis and social insects. Science 191:249-263 Myrmica lobicornis (Hymenoptera: Formicidae) colonies and WahlundS (1928)Zusammensetzung von Populationenund Korre- populations: number, relatednessand longevity of reproducing lationserscheinungenvom Standpunktder Vererbungslehreaus- individuals.J Evol Biol, in press betrachtet.Hereditas 11:65-106 Sherman PW, Seeley TD, Reeve HK (1988) Parasites, pathogens, Wilson DS (1980)The naturalselection of populations and commu- and polyandry in social Hymenoptera.Am Nat 131:602-610 nities. Benjamin/Cummings,Menlo Park Starr CK (1984) Sperm competition, kinship and sociality in the Woyciechovski M, Lomnicki A (1987) Multiple mating of queens aculeate Hymenoptera. In: Smith RL (ed) Sperm competition and the sterility of workers among eusocial Hymenoptera. J and the evolution of animal mating systems. Academic, New Theor Biol 128:317-327 York, pp 427-464 WrightS (1931) Evolution in Mendelianpopulations. Genetics 97- Stille M, Stille B, Douwes P, (1991) Polygyny, relatednessand nest 159 founding in the polygynous myrmicineant Leptothoraxacervo- Wright S (1940) Breeding structure of populations in relation to rum(Hymenoptera; Formicidae). Behav Ecol Sociobiol 28:91- speciation. Am Nat 74:232-248 96 Wright S (1943) Isolation by distance. Genetics 28:114-138 Sundstrom L (1989) Genetic relatednessand population structure WrightS (1951)The genetical structureof populations.Ann Eugen- in Formicatruncorum Fabr. (Hymenoptera:Formicidae). Actes ics 15:323-354 Coll Ins Soc 5:93-100 YarrowIHH (1955)The Britishants allied to Formicarufa L. (Hym; Formicidae).Trans Soc Br Entomol 12:1-48