Conditional Use of Sex and Parthenogenesis for Worker and Queen Production in Ants

See allHide authors and affiliations

Science  03 Dec 2004:
Vol. 306, Issue 5702, pp. 1780-1783
DOI: 10.1126/science.1105453


The near-ubiquity of sexual reproduction in animal species has long been considered a paradox because sexually reproducing individuals transmit only half of their genome to their progeny. Here, we show that the ant Cataglyphis cursor circumvents this cost by using alternative modes of reproduction for the production of reproductive and nonreproductive offspring. New queens are almost exclusively produced by parthenogenesis, whereas workers are produced by normal sexual reproduction. By selectively using sex for somatic growth and parthenogenesis for germline production, C. cursor has taken advantage of the ant caste system to benefit from the advantages of both sexual and asexual reproduction.

The main advantage of asexual reproduction is that it confers a twofold advantage over sexuality by allowing, generation by generation, the transmission of twice the number of genes to offspring (1, 2). However, asexual reproduction is also associated with both short-term and long-term disadvantages, including a lower genetic diversity of offspring and a reduced rate of adaptive evolution of species (3, 4). The nature and the degree of the cost associated with asexual reproduction is expected to vary across taxa, depending on the biology of the species and the type of environment in which they live (13).

In ants, as in other Hymenoptera, females are usually produced by sexual reproduction and are diploid, whereas males develop from unfertilized eggs and are haploid (3). The diploid fertilized eggs can develop into either new queens (gynes) or workers, with the developmental switch generally under environmental control (5). In the Cape honey bee and five ant species, however, unmated workers may reproduce by thelytokous parthenogenesis (611); that is, they may produce female offspring from unfertilized eggs. While conducting a population genetic study of one of these species, the ant Cataglyphis cursor, we discovered that not only unmated workers but also mated queens can use thelytokous parthenogenesis. Pedigree analyses indicated that queens use automictic parthenogenesis with central fusion where two of the four products of meiosis merge. Unlike workers, queens use this mode of reproduction specifically to produce new queens.

Cataglyphis cursor is a common ant in the dry forests of Europe. Colonies usually contain a single queen and up to 3000 workers. Only few colonies produce gynes, and the number of gynes produced per colony is small. This is because C. cursor has an unusual mating system whereby gynes mate near the parental nest before leaving the colony with adult workers to initiate new colonies 3.2 to 11.3 m away (12). Previous studies also indicate that C. cursor workers can produce both gynes and workers parthenogenetically in colonies that have lost the mother queen (7).

We collected 38 large colonies in Southern France and genotyped 532 workers at four highly polymorphic microsatellite loci (expected heterozygosities, 0.833 to 0.944) (13). The genotypes indicated that 35 of these colonies contained a single reproductive queen (monogyny), whereas three colonies contained offspring from at least two queens. Analysis of lab-raised worker progeny (n = 437 freshly eclosed workers) from 12 queens showed that they had mated with an average of 5.6 ± 1.3 males (range, 4 to 8).

A detailed analysis of the 35 monogynous colonies showed that most of the workers in these colonies could only have been produced by sexual reproduction. Overall, 476 of the 489 workers (97.3%) genotyped in the 35 colonies harbored, at one or several loci, alleles that were not present in the mother queen and came from one of the queen's mates. It is impossible to determine whether the 13 workers harboring only alleles identical to those of their mother were fathered by a male that had no allele distinct from those of the queen or whether they had been parthenogenetically produced. Because the four microsatellite loci were highly polymorphic, the probability of mating with a male harboring no diagnostic allele at any of the four loci was low, ranging from 0.0001 to 0.013 across colonies according to the queen's genotype. Thus, of the 476 workers, only one was expected to have no diagnostic alleles if they were all sexually produced. Hence, it is likely that some or all of the 13 workers with no diagnostic paternal allele may indeed have been asexually produced (the estimated proportion of asexually produced workers is 2.5% when corrected for the probability of nondetection of paternal alleles).

A total of 56 gynes were produced by 10 of the 35 monogynous colonies. In contrast to workers, most of these gynes (54 of 56) had alleles at the four loci that could all be attributed to the queen (Fig. 1A), hence these gynes had been produced by parthenogenesis. The alternative explanation, that these 54 gynes had been fathered by a male having no diagnostic alleles, can be ruled out. Queens and males came from the same gene pool, as indicated by a lack of significant difference in allele frequencies for the four loci (Fisher exact test, all P > 0.05) and the workers' Fis value (an index of observed versus expected homozygosity), which was not significantly different from zero (Fis = 0.011 ± 0.015, n = 35 colonies; two-tailed t test, t = 0.691, P = 0.494). This, together with the high allelic diversity, resulted in a very high probability to detect a male's genetic contribution. However, none of the gynes produced in nine of the 10 colonies had any diagnostic allele, even though the likelihood of such a matched mating was lower than 0.013 in each of the nine colonies (range, 0.0001 to 0.013). Overall, the probability that all the fathers of the gynes produced in the nine colonies had no diagnostic alleles was P < 10–28. Indeed, the genotypes of workers in these nine colonies confirmed that all or most of the males that mated with the queens had diagnostic alleles at one or more loci (Fig. 1B). The outcome of the vast majority of gynes being produced by parthenogenesis was that the relatedness between queens and gynes was very high (r = 0.864 ± 0.046, n = 56 gynes) and significantly greater (P < 0.001) than the theoretical value of 0.50 expected under sexual reproduction.

Fig. 1.

Respective proportion of gynes (A) and workers (B) harboring maternal alleles only, and therefore interpreted as parthenogenetic daughters, in each of 10 colonies (colony numbers are laboratory designations). The sample size for each colony is indicated above the bars.

Most of the 54 parthenogenetic gynes were neither genetically identical to each other within a colony nor genetically identical to their mother queen. The discrepancies resulted from gynes being homozygous at some loci where the mother queen was heterozygous. In all cases, the gynes were homozygous for one of the two maternal alleles. This is the expected pattern under automictic parthenogenesis with central fusion. Because two of the four products of meiosis merge, the offspring have the same genotype as their mother for the loci that did not cross over, whereas the offspring is homozygous for one of the two maternal alleles if crossing-over did occur (14, 15). The frequency of transition from heterozygosity is expected to vary across loci depending on their distance to the centromere (15). Consistent with this prediction, the frequency of transition from heterozygosity to homozygosity varied significantly across the four loci, presumably reflecting differences in the distance between each locus and the centromere (Table 1; χ2 = 25.53, P < 0.0001).

Table 1.

Proportion of gynes homozygous for a given locus when the mother was heterozygous at that locus. The sample size for each locus is indicated.

Locus Sample size Percentage of gynes homozygous
Ccur11 53 5.7
Ccur46 47 46.8
Ccur58 47 34.0
Ccur63b 41 17.1

The expected outcome of automictic parthenogenesis is a gradual increase in homozygosity over time (16). Indeed, the overall level of homozygosity was significantly higher in gynes than in workers (Fig. 2; Fisher's exact test on the number of homozygous versus heterozygous loci in gynes and workers: P < 0.0001). Accordingly, F statistics revealed a significant excess of homozygosity in gynes (F = 0.396 ± 0.12, P < 0.001) and queens (F = 0.255 ± 0.051, P < 0.001) but not in workers from the same colonies (F = 0.002 ± 0.016, P = 0.45). By increasing the levels of homozygosity, parthenogenesis should result in reduced queen survival and fitness, much like inbreeding does. However, the fitness effect might be limited for ant queens because they stay in the protected environment of the nest, except during colony founding. Even at this stage, the intensity of this cost should vary according to the mode of colony founding, with selection against more homozygous queens being higher in species where queens start a new colony on their own and lower in species, such as C. cursor, where queens do not go through a stage of independent colony founding (12).

Fig. 2.

Overall homozygosity detected in gynes (white) and workers (black) at all four loci, for each of 10 colonies. The sample size for each colony is indicated above the bars.

In addition, two processes appear to counteract the process of genetic homogenization induced by automictic parthenogenesis. The first is the occasional production of gynes by sexual reproduction. The overall production of such gynes was 3.6% (2 of 56) in the 10 colonies studied. The second process is the occasional queen production by worker parthenogenesis. Because workers are usually produced by sexual reproduction, their contribution to gyne production will contribute to the maintenance of heterozygosity in gynes and queens, just as under queen sexual reproduction.

Although C. cursor queens do not require mating to produce diploid offspring, they have retained sexual reproduction to produce workers, which suggests that sexual reproduction has important benefits for colony function. The observed mating frequencies in this species lie on the high end of the continuum of mating frequencies reported in ants (17). A possible explanation is that genetic input from an increased number of mates compensates for the negative effect of high queen homozygosity on colony genetic diversity. Parthenogenetic production of workers at the level observed for gynes would lower colony genetic diversity, which could lead to reduced defense against parasites, less efficient division of labor, and a decreased range of environmental conditions that a colony can tolerate (1820). These costs are akin to those thought to lead to the instability of parthenogenetic reproduction in nonsocial organisms (2). Multiple mating lowered the overall relatedness of nestmate workers to r = 0.42 (SEjackknife = 0.02, n = 35), a value well within the range of values reported in other ants (21). Thus, the high queen mating frequency may cancel out reduced genetic diversity at the colony level stemming from the relatively high queen homozygosity.

Using alternative modes of reproduction for the queen and worker castes may also enhance cooperation within the social group by aligning the interests of queens and workers. Parthenogenetic production of gynes by queens reduces conflict with workers because, just like queens, workers are significantly more closely related (t = 2.31, df = 43, P = 0.03) to the parthenogenetic gynes (r = 0.59, SEjackknife = 0.07, n = 10) than they would be to sexually produced gynes or to gynes produced parthenogenetically by other workers (these two values are identical to the relatedness between workers, r = 0.42). As a result, workers should police the reproduction of other workers (22). The almost complete lack of worker-produced gynes in colonies containing a queen is consistent with this idea.

Conditional use of parthenogenesis for queen production might also occur in other ants, yet it may remain unnoticed because it primarily occurs in dependent-founding species where it is most difficult to detect. In ants there is a strong association between the mode of colony founding and the number of queens, with dependent colony founding being almost exclusively restricted to species with high numbers of queens per nest (23, 24). The likelihood of detecting parthenogenesis with genetic markers is low in such species because it is very difficult to determine the maternity of female offspring. As a result, only a handful of studies in highly polygynous ants are sufficiently detailed to have enabled the detection of parthenogenesis.

This study shows that by taking advantage of the social caste system, C. cursor colonies can benefit from the advantages of both sexual and asexual reproduction. By using alternative modes of reproduction for the queen and worker castes, queens can increase the transmission rate of their genes to their reproductive female offspring while maintaining genetic diversity and social cohesion in the worker population. These findings, together with those of other recent genetic studies (2529), indicate greater flexibility of the ant reproductive and social systems, thus providing an ideal ground to test various evolutionary predictions.

Supporting Online Material

Materials and Methods


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article