Present address: P. Seppä, Department of Biological and Environmental Sciences, Ecology and Evolutionary Biology Unit, Box 65, University of Helsinki, FIN-90014, Finland
Pekka Pamilo, Department of Biology, PO Box 3000, 90014 University of Oulu, Finland. Tel: +358 8 5531780; fax: +358 8 5531061; e-mail: email@example.com
We studied genetic differentiation between two social forms (M-type: single queen, independent nest founding; P-type: multiple queens, dependent nest founding, building of colonial networks) of the ant Formica truncorum in a locality where the social types characterize two sympatric populations. The genetic results indicate restricted gene flow between the social forms. Female gene flow between the forms appears to be absent as they did not share mitochondrial haplotypes. Significant nuclear differentiation and the distribution of private alleles suggest that male gene flow between the forms is weak. However, the assignment analysis indicates some gene flow with P males mating with M females. The results have potentially important implications concerning social evolution within the forms but they need to be confirmed in other localities before they can be generalized. The colonies in the M-type population have earlier been shown to produce split sex ratios, depending on the mating frequency of the queens. The inferred gene flow from the P to the M type means that the split sex ratio is partly suboptimal, possibly because the P populations are not long-lived enough to influence the behavioural decisions in the M colonies.
A shift from solitary to social life is one of the major transitions in evolution (Szathmary & Maynard Smith, 1995), which in many species has been followed by further development from simple to complex societies. One important change in this later development includes the transition from single-queen societies (monogyny) to societies where many queens coexist and reproduce (polygyny), referred to as sociality evolving a second time (Rosengren & Pamilo, 1983). Such a shift may occasionally have resulted in a pair of related species with alternative social organizations (Wilson, 1971). Similar shifts may also occur sporadically within an existing species, and the result is social polymorphism with some areas being characterized by monogyne and others by polygyne colonies. Polygyny can be further associated with nest proliferation by budding and formation of a network of interconnected nests. Many invasive ant species are characterized by such highly polygyne colonial networks (e.g. the Argentine ant Linepithema humile; Tsutsui et al., 2000; Giraud et al., 2002), but we have only a poor understanding of which factors cause the shift in the social organization, how frequently these shifts take place within a species, and how the two social forms interact within a socially polymorphic species.
Populations of the introduced fire ant (Solenopsis invicta) provide the best-known example of a sympatric mix of two social organizations within a species, with only weak gene flow by males from the monogyne to the polygyne form (Ross & Keller, 1995). These social forms are associated with a genotype difference at one gene, called Gp9, coding for a putative pheromone receptor (Krieger & Ross, 2002) and the queen genotypes are treated differently by workers in the two social forms (Ross & Keller, 1998). Restricted female gene flow has led to very different mitochondrial haplotype frequencies in the monogyne and polygyne forms of S. invicta (Shoemaker & Ross, 1996), demonstrating that the social organization of colonies can strongly influence the spatial structuring and evolution of ant populations. It is essential to examine whether the situation in the fire ant is a special case depending on the specific effects of the Gp9 genotypes, or whether the observed patterns of genetic differentiation can be extended to other ants with similar social variation.
Ants in the genus Formica similarly display two profoundly different types of social organization, which in some cases characterize two closely related species and are sometimes found within a single species. Intraspecific variation of social organization has been shown to be associated with both restricted gene exchange between colonies of different social types (F. exsecta, Seppäet al., 2004) and apparently free exchange (F. selysi, Chapuisat et al., 2004). Formica truncorum provides unique opportunities to study the interaction of two coexisting social forms and the possible impact of such interactions to the fitness of individuals. It is the first species in which split sex ratios were shown in support of worker control of sex allocation (Sundström, 1994). Monogyne colonies with a singly-mated queen produce mainly females and those with multiply-mated queens mainly males, in close agreement with the theoretical predictions under worker control (Boomsma & Grafen, 1990,1991). Male-biased sex ratios in colonies headed by multiply inseminated queens also agree with the queen interests (Pamilo, 1991), and Sundström & Ratnieks (1998) estimated a 37% fitness advantage to multiply mated F. truncorum queens when their colonies produce males and the population wide sex allocation proportion of females is 0.67. These fitness estimates assume that the monogyne form is evolving independently of the polygyne form, as the polygyne colonies normally produce strongly male-biased sex ratios (Pamilo & Rosengren, 1983; Rosengren et al., 1986; Sundström, 1995b) that would influence the optimal sex allocation in the monogyne form if interacting with it.
In this work we study whether the two social forms in F. truncorum are connected by gene flow when sympatric. This is important for understanding the evolution of the nesting strategies and the possible impacts one social form might have on the function of the other, particularly on the optimal sex allocation patterns. We infer the male and female gene flow between the two social forms from the spatial distribution of mitochondrial haplotypes and nuclear microsatellite genotypes in an area where the two social forms occur in close proximity.
Materials and methods
The study population was located in the Tvärminne archipelago, Hanko, SW Finland (Fig. 1). The material included 22 M-type nests inhabiting six islands (sampled in 1995) and 19 nests from a P-type colonial network from the same area (sampled in 1997). Maximum distance between the islands was 3 km and no other good populations of the species are known in the neighbourhood. The smallest distance between the M-type colonies and the P-type network was 300 m, and the colonial network was located within an area of about 2 ha. The social organization of the island nests has earlier been investigated with nuclear genetic markers (Sundström, 1994; Gertsch, 2000). The organization of the P-type colonies was inferred from the genetic data of this study (see below) and from the observations of multiple queens in the nests. The geographical distribution of the mtDNA haplotypes was examined in a small sample of workers from several nests of both social forms in four localities outside the main study area (50–100 km from Tvärminne).
Genetic analyses were performed using both nuclear and mitochondrial markers. Ant DNA for PCR reactions was extracted from head and thorax of adult individuals using either chelex protocol (Walsh et al., 1991) or DNeasy Tissue Kit (Qiagen). For nuclear genetic analysis, seven microsatellite loci, FL12, FL20, FL21, FL29 (Chapuisat, 1996), FE16, FE37 and FE38 (Gyllenstrand et al., 2002) were used. From each nest, five individuals were genotyped. Amplification reactions were carried out in 10 μL volumes containing approximately 10 ng ant DNA, 1× PCR-buffer (10 mm Tris–HCl, pH 8.8, 50 mm KCl, 0.08% Nonidet P40), 1.5 mm MgCl2, 75 μm dNTP, 400 nm of each primer and 0.4 U of Taq polymerase (MBI Fermentas). Samples were PCR amplified under cycling conditions: initial denaturing at 94 °C for 3 min followed by 30 cycles of denaturing at 94 °C for 30 s, annealing at appropriate Ta for 30 s, extension at 72 °C for 30 s followed by a final extension step at 72 °C for 5 min. PCR products were analysed on 6% polyacrylamide gels and visualized using standard silver staining protocol (Bassam et al., 1991) or autoradiography.
Mitochondrial haplotype was determined for one individual per nest. A 1773 bp fragment was PCR amplified using primers CB1 (Crozier & Crozier, 1992) and ND1 (Jermiin & Crozier, 1994) corresponding to bases 11 400–13 168 of the honeybee mitochondrial genome NC 001566 (Crozier & Crozier, 1993). Amplification reactions were carried out in 25 μL volumes containing approximately 10 ng ant DNA, 1× PCR-buffer (10 mm Tris–HCl, pH 8.8, 50 mm KCl, 0.08% Nonidet P40), 2.0 mm MgCl2, 200 μm dNTP, 400 nm of each primer and 2.0 U of Taq polymerase (MBI Fermentas, Vilnius, Lithuania). Samples were PCR amplified under touchdown cycling conditions: initial denaturing at 94 °C for 3 min followed by 10 cycles of denaturing at 94 °C for 30 s, annealing starting at 50 °C for 30 s and lowered 0.5 °C each cycle, extension at 72 °C for 60 s. After the initial 10 cycles of touchdown cycling, an additional 20 cycles with annealing at 45 °C followed. Cycling was ended by a final extension step at 72 °C for 5 min. Five microlitre of the PCR product were digested with 3 U of the restriction enzyme TaqI at 65 °C. Digested products were analysed using the Single Strand Conformation Polymorphism (SSCP) method (Orita et al., 1989). Samples were loaded on 8% nondenaturing PAA gels and electrophoresed at 6 W for 20 h at 4 °C. Banding patterns were subsequently made visible using a standard silver staining protocol (Bassam et al., 1991).
The resulting mtDNA haplotypes were further analysed for nucleotide variation. For each haplotype found, 2–3 individuals were sequenced. PCR products were electrophorezed in 1.5% low melting agarose gels, bands were excised and purified using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany). The purified fragments were ligated into pGEM®-T vector using the pGEM®-T Vector System (Promega, Madison, WI, USA). The recombinant plasmids were used to transform High Efficiency Competent Cells JM109. Recombinant clones were selected by blue/white screening. Minipreparation and purification of plasmid DNA was performed as described in Sambrook et al. (1989). Purified plasmid DNA containing inserts were cycle-sequenced using the Thermo Sequenase kit (Amersham, Little Chalfont, UK) and sequencing primers T7 and SP6 together with internal primers we synthesized, CB76 (5′TAGGCGGTGTAATTGCTTTATT) and ND162 (5′CAAATATTGTCTTATGAGGT). Sequences were electrophorezed in 4.0% Long rangerTM PAA-gels on a LI-COR 4200-2 system. Sequences were base-called and edited using Base ImagIRTM 4.0 (LI-COR) software. Sequences have been deposited in Genbank under the Accession Numbers: AY154884-AY154888.
Social organization was investigated by estimating relatedness following Queller & Goodnight (1989). In the analysis, nests were weighted equally. To obtain standard errors, relatedness estimates were jackknifed over the loci. Due to nonindependence of genotypes in monogyne colonies, these were resampled by drawing one individual randomly from each nest to create 100 datasets. These were used for exact tests of the Hardy–Weinberg equilibrium (software GENEPOP 3.1d, Raymond & Rousset, 1995), estimating genotypic disequilibrium (software FSTAT 2.9.3 by Goudet, 2001), and estimating F-statistics (Weir & Cockerham, 1984; software FSTAT 2.9.3 by Goudet, 2001). Means and standard errors of pairwise estimates were calculated over resamplings. For F-statistics, the M-type population was divided into two parts, the northern islands (a total of 10 nests) and the southern islands (12 nests), as the number of nests on individual islands was too small to allow useful calculations (see Fig. 1). Patterns of nuclear gene flow between populations were studied by assigning paternal haplotypes inferred from the observed worker genotypes. Assignments were done from the observed allele frequencies following Paetkau et al. (1995) by removing the nest in question when calculating the population allele frequencies. When an allele carried by the inferred male was absent from a population, one copy was introduced.
A neighbour-joining tree of the mtDNA haplotypes was constructed using the number of pairwise differences with the software MEGA Version 3.0 (Kumar et al., 2004).
The average relatedness of worker nestmates in the monogyne colonies was r = 0.74 ± 0.02, which is in agreement with previous estimates (r = 0.71 ± 0.03) for this population (Gertsch, 2000). Corresponding estimate within the P-type network was r = 0.01 ± 0.01, confirming polygyny in this population. No linkage disequilibrium was detected in either social form and hence all loci were used for consecutive analysis. No deviations from random mating were detected within the two social forms when analyzed separately.
Genetic variation (Table 1) across loci, measured as heterozygosity and number of alleles, was consistently lower in the polygyne population (one-tailed sign tests: P < 0.01 and 0.05, respectively), and two loci (FL12 and FL29) were monomorphic in the polygyne form. The lower genetic variation in the polygyne form was accompanied by fewer private alleles (4) compared to the monogyne form (19). This indicates a bottleneck during the founding stage of the polygyne colonies. Random samples of 19 nests from the total data set always contained a higher number of alleles (the smallest number in 1000 random samples was 36) than observed in the P population (31 alleles).
Table 1. Number of alleles (A) and expected heterozygosities (He) in monogyne and polygyne populations, respectively.
Genetic structuring within the monogyne form was weak, with FST = 0.014 ± 0.015 (jackknifed over the loci) between the southern and northern groups of islands. Differentiation between the social forms was significantly larger, with FST = 0.11 ± 0.03 (t5 = 3.64, P < 0.05) between the P-population and the northern M-population, and FST = 0.14 ± 0.03 (t5 = 4.41, P < 0.01) between P and southern M.
Assuming strict monogyny of the M-type nests (see Sundström, 1994 and current relatedness estimates), the paternal haplotypes were inferred from the worker genotypes. The inference was not made when there was any uncertainty (e.g. when all workers were identical heterozygotes). Nine of the 31 inferred male haplotypes were more likely to arise from the P population than from the M population. To estimate the reliability of the assignment tests, we generated 1000 random male haplotypes from the observed allele frequencies in the M population. This resulted in a mean of 3.8 misassignments (the male haplotype was assigned to the P population with a higher probability than to M), and the probability for getting seven or more misassignments was lower than 0.05. Thus, some males reproducing in the M population were likely to have immigrated from the P population. We ignore here the possibility that both populations would have received males from outside our study area, as we do not know any strong F. truncorum population in the neighbourhood.
A total of five mtDNA haplotypes were found, one of them outside the main study area. The resulting sequence consisted of 882 bp partial ND1 gene, 750 bp partial cytochrome b gene, 74 bp tRNA-ser and 67 bp noncoding DNA. No indels were detected. A total of 44 polymorphic sites were found. Of these, 25 were situated in the partial ND1 gene, 16 in the partial cytochrome b gene and 3 in the noncoding regions. The transition/transversion ratio was 4.5. The number of substitutions in pairwise comparisons between haplotypes varied from 9 (0.5%) to 30 (1.7%). Three haplotypes (H1, H2 and H3) were found in the monogyne colonies, of these the haplotypes H1 and H3 were found in only one nest each on islands Mellanskär and Sundholmen, respectively. One haplotype (H5) was fixed in the polygyne form and was completely absent from the monogyne colonies. We hence conclude that there was no female gene flow between the M and P forms in this area.
The samples from the localities outside Tvärminne came from one P-type and three M-type populations, as inferred from the worker genotypes. An additional haplotype (H4) was found in a monogyne nest outside the main study area. The other samples had the haplotype H2.
Nucleotide differences between the haplotypes were large and the most divergent haplotype (H3) was found within the monogyne form. The haplotypes characteristic to polygyne and monogyne forms, respectively, did not clearly form separate evolutionary lineages (Fig. 2).
Our results clearly demonstrate that gene flow between the two alternative social forms is strongly restricted in our study area. The populations in Tvärminne did not share any mtDNA haplotype, proving that there has been no female gene flow between them. The level of nuclear differentiation showed that some male gene flow may have taken place, and the assignment of males suggested male gene flow from the P to the M type. As the paternal genotypes could not be inferred in the P population, we could not obtain any estimates of male gene flow in the opposite direction. However, the low genetic diversity in the P population and many unique alleles in M indicate that gene flow from M to P must have been negligible. The result agrees with the earlier findings in Solenopsis invicta (Ross & Shoemaker, 1993) and Formica exsecta (Seppäet al., 2004) in that female dispersal between the social types is very restricted, but we should note that our results come from a comparison of a single pair of populations. Concerning the direction of male gene flow, the results differ from the pattern observed in Solenopsis invicta (Ross & Shoemaker, 1993; Ross, 1997) in which males from monogyne colonies mate with queens from polygyne colonies.
Genetic data have previously suggested limited female dispersal, partial intranidal mating and bottlenecks in polygyne F. truncorum (Sundström, 1993; Elias et al., 2005), and our present results also indicated a bottleneck in the history of the P population. Both behavioural and physiological differences between females of the two forms suggest highly restricted gene flow of females originating in polygyne colonies (Sundström, 1993,1995a). Females of the polygyne type contain less fat and glycogen than young monogyne females and such smaller reserves reduce the probability of successful nest initiation. Furthermore, females from a polygyne colony can be recruited back into the natal colony whereas those from monogyne colonies must disperse. Sundström (1997) also showed that workers of polygyne colonies have a lower rate of rejecting and discriminating females than workers of monogyne colonies. Even though there obviously is a potential for interform female gene flow from monogyne to polygyne colonies, the mtDNA haplotype differences clearly demonstrate that it does not take place. Whether females do not even try to immigrate or are actively discriminated against by workers of polygyne colonies, as is the case in the fire ant (Keller & Ross, 1998; Ross & Keller, 1998), cannot be concluded on the basis of the present data.
Several studies of other ants have shown multiple mtDNA haplotypes coexisting in polygyne colonies (Stille & Stille, 1992; Tay et al., 1997; Goodisman & Ross, 1998; Liautard & Keller, 2001; Seppäet al., 2004), indicating that foreign females can be accepted in them. The unique haplotype in the polygyne form of F. truncorum in Tvärminne suggests that these colonies only recruited new queens originating in the P-type colonies. As the other population with polygyne colonies (about 50 km from our main study site) shared the haplotype H2 with the monogyne form, the polygyne social form does not represent a separate evolutionary line within the species. However, the fixation of the haplotype H5 in the Tvärminne population suggests that either the mitochondrial lineages of the polygyne and monogyne forms in that area have been separated for a long time and H5 has disappeared from the monogyne form or the P population has initially been derived from another source. It seems likely that polydomous colonies can sporadically originate from the monogyne form through a bottleneck (Elias et al., 2005). This results in reduced variation, particularly in mtDNA, and different P-type populations could be fixed for different mtDNA haplotypes. Our results came from a single location and it would be interesting to explore the connections of M and P type populations in a larger geographical area. The observed divergence in the mtDNA sequences suggests that the historical relationships could be readily assessed.
Population bottlenecks have been suggested to induce evolution from less to highly polygyne forms by reducing recognition cue diversity in the invading Argentine ant, Linepithema humile, populations (Tsutsui et al., 2000) and in the Solenopsis richteri clade (Ross et al., 2003). Another possible consequence of a bottleneck is a loss of sex alleles. Such reduction could lead to offspring that are homozygotes at the sex locus and, due to the sex determining system of ants, develop as diploid males that are sterile and impose a genetic load to the colony (Cook & Crozier, 1995). Diploid males have indeed been found in F. truncorum. Gertsch (2000) estimated that the frequency of matched matings, i.e. a male carrying an allele present in his mate, is 14.7% in the monogyne Tvärminne form. We do not have such an estimate from the polygyne form in our study area, but in another nearby P-type population 23% of diploid sexuals were males, indicating a frequency of 46% for matched matings (Pamilo et al., 1994). If the load caused by diploid males is that much bigger in the polygyne social type, it would give a selective advantage to immigrating males carrying new sex alleles, and the genetic load could contribute to the long-term extinction risk of the population. Selection favouring immigrants would homogenize the populations genetically. However, strong nuclear differentiation and distinct allele frequency differences between the social forms in F. truncorum indicate that male gene flow has not been extensive enough to homogenize genetic variation and that the social forms evolve relatively independently even when they occur in sympatry.
The workers in the monogyne colonies benefit by biasing the colonial sex ratio towards the sex which at the population level is underrepresented compared to the optimum of these workers. Such worker control is predicted to lead to split sex ratios, depending on the relatedness asymmetries within colonies (Boomsma & Grafen, 1990). Earlier studies on F. truncorum in Tvärminne have shown that the monogyne colonies headed by a singly-mated queen specialize in raising females and those with multiply-mated queens raise largely males, resulting in the population-wide ratio close to the worker optimum (67% females, Sundström, 1994; Sundström & Ratnieks, 1998). However, the polygyne colonies normally produce highly male-biased sex ratios, and Sundström (1995b)) estimated that the total production of sexuals is roughly equal in monogyne and polygyne F. truncorum nests. The polygyne colonies would thus significantly influence the population-wide sex ratio if sexuals from both type of colonies formed a joint mating pool. In such a case, it would be advantageous for workers in all monogyne colonies to raise females and the currently observed split sex ratios would not represent optimal worker behaviour.
Split sex ratios, as observed in F. truncorum (Sundström, 1994), agree with worker optimum and will also give a large fitness advantage to multiply mating queens (Pamilo, 1991). Consequently, Sundström & Ratnieks (1998) estimated a 37% fitness advantage to multiply inseminated M-type queens of F. truncorum in Tvärminne. This advantage depends on the fact that the population-wide sex ratio is female biased and male production is advantageous to the queens in such a situation. If the sexuals produced by both social forms would form a joint mating pool, that fitness advantage would largely disappear. Including the sexuals produced by the polygyne colonies of Tvärminne would make the population sex ratio close to a 1 : 1 ratio and in that situation all individual sex ratios would be equally good to the queens and there would be no advantage to polyandry.
There are two explanations for this apparent dilemma. Either the dispersal of males from P to M is so low that the effect on the worker behaviour in the M colonies is negligible, or the P-type colony aggregation has existed for such a short time that it has not influenced the behavioural decisions in the M-type colonies. The occurrence of different mtDNA haplotypes in the two P-type populations included in the present study supports the view that these colonies are formed locally from initially monogyne colonies. If they have a restricted life span, they do not have enough time to evolve into separate species; neither do they much influence the adaptive strategies of the sympatric monogyne forms of the species in spite of some gene flow.
We thank P. Gertsch and L. Sundström for supplying DNA samples from the monogyne colonies in Tvärminne. The work has been supported by grants from the Research Councils of Sweden and Finland (77311), and by EU-TMR Network ‘Social evolution’.