Our experimental work significantly improves our understanding of worker reproduction in the invasive A. gracilipes. We could attribute male production to workers in 25 captive colonies out of 66 male-producing colonies. In our queen transfer experiment, the absence of a queen triggered an increase in physogastric workers suggesting the existence of queen control over worker reproduction. Dissections of physogastric workers revealed that their ovaries were more likely to contain yolky oocytes. We also found that physogastric workers were less aggressive and less likely to forage than normal workers, which indicates that their presence may be costly to colony foraging capacity and defence. The head width and wing length of worker-produced males were slightly larger than males for which we could not rule out a queen or worker origin. Finally, 5.5% of worker-produced males and 14.3% males that could have been produced by dealate queens or workers were heterozygous. We found that most queens were homozygous while most workers were heterozygous for at least one locus. This indicates that the reproduction of A. gracilipes is unusual and may involve diploid males and/or gynandromorphs, consistent with previous suggestions [23,29].To the best of our knowledge, this study is the first to test for the potential cost of worker reproduction in an invasive ant species. Given the cost of worker reproduction, male-production by workers is unexpected in a highly successful invasive species such as A. gracilipes, which may indicate that worker reproduction has benefits.
Frequency of worker reproduction and evidence of queen control
Male production by workers occurred in 25 colonies but not in another 70 colonies that were always queenless. Nor did we find male production by workers within any of the colonies from the queen transfer experiment, including six queenless colonies that we monitored past the end of the experiment and that had been queenless for 116 to 186 days. The absence of male production in these colonies could be due to the low worker count, as we found from our observations of 233 captive colonies that worker number was positively associated with male production. In Taiwan, three queenless colony fragments out of nine produced male brood after being kept for four months in the laboratory, and adult males were observed in one of these fragments two months later [21]. This matches our observation that it took 54.1 ± 10.2 days (mean ± SD) for males to develop from eggs to adult. Our observations on male production are all based on laboratory-kept colonies. Caution must be taken when extrapolating to field colonies, which are not as likely to be queenless, although queenless aggregations of A. gracilipes workers and brood are frequently observed in the field (personal observation).
Additional results from our colony observations suggest that queens may limit worker reproduction, though we cannot rule out worker policing. In our queen transfer experiment, removing queens triggered an increase in physogastric workers, and moving the queen back after 60 days led to a decrease in physogastric workers. Observations from Lee et al. (2017) [21] suggests that physogastric workers may switch from producing trophic eggs in queenright conditions to producing viable male eggs in queenless conditions. Workers of several ant species switch from trophic egg production to male egg production when the queen dies or disappears (e.g. Aphaenogaster senilis [36]; Aphaenogaster cockerelli [37]; Prolasius advena [38]; N. apicalis [10]; O. longinoda and O. smaragdina [17], but to the best of our knowledge, A. gracilipes is the only invasive ant species that has been found to do so. Social insect queens can inhibit worker reproduction with queen pheromones, i.e. chemical signals indicating the reproductive status of the queen [2,39]. Several experiments with ants, wasps, and some bees have shown that applying synthetic queen pheromones to queenless colonies inhibits worker reproduction by preventing workers from activating their ovaries and by causing secondary oocyte resorption [12,39–41]. Worker reproduction could also be controlled through the policing of reproductive workers [5,8]. For example, queens and workers could behave aggressively towards egg layers or destroy worker-laid eggs [9].
Physogastric workers dissected as part of the dissections of entire colonies and following the aggression tests were more likely to have yolky oocytes than normal workers. Yolky oocytes indicate the presence of fertile or trophic eggs [13,31]. There was no difference in yolky oocyte presence at the end of the queen transfer experiment, probably because colonies in the queen transfer experiment had been queenless for a relatively short period (60-120 days vs 108 and 143 days for dissections of worker ovaries of entire colonies and 102-212 days for aggression tests). We only observed yellow bodies in physogastric workers (7.8-12.9%). Yellow bodies can indicate active oviposition of viable eggs, although they are sometimes observed in trophic egg-layers [13,32–34]. Physogastric A. gracilipes workers originating from Taiwan also had a higher reproductive potential than normal workers [21]. They had more well-developed ovaries and more yolky oocytes than normal workers [21]. Yellow bodies were also only observed in physogastric workers (13%) [21]. Histological sections of the abdomen of physogastric workers indicated that fat bodies were more abundant in physogastric than in normal workers [21]. The distended abdomen of physogastric workers could therefore be due to the presence of fat bodies and well-developed ovaries. The reproductive or trophic egg-layer status of individual workers can only be determined by ovary dissections or by observations of egg-laying. Physogastric workers are more likely to be reproductive than normal workers, but some non-reproductive workers may have a temporarily distended abdomen from feeding extensively on liquids. Additional research is needed to determine an objective way to non-destructively distinguish reproductive from non-reproductive workers.
Costs of worker reproduction
We found that behavioural differences between physogastric and normal workers may decrease the competitive ability of the colony. Physogastric workers in queenright and queenless colonies were infrequently observed in the foraging area and were mostly observed inside the nesting tubes during the queen transfer experiment. These observations suggest that physogastric workers do not contribute to foraging as much as normal workers and may spend more time tending to the brood than contributing to foraging activities. Physogastry may also affect the ability of workers to defend the colony during interspecific conflicts. We found that physogastric workers were less aggressive towards O. smaragdina workers and were less likely to engage in a fight than normal workers, which would reduce the potential of A. gracilipes colonies with a large proportion of physogastric workers (such as queenless colony fragments) to become behaviourally dominant. In queenright colonies, the queen may limit the proportion of physogastric workers and thus minimize the costs associated with worker reproduction, such as a decrease in foraging and defence activities.
The production of males by workers also generates costs for other ant species [18]. For example, in Neoponera obscuricornis colonies, two costs are associated with worker reproduction following queen removal: an increase in energetic cost associated with aggressive interactions between workers for egg-laying and a decrease in colony labour due to reproductive workers spending less time working for the colony [19]. Costly worker conflicts about which workers become reproductive and which workers continue to contribute to colony labour also take place in Aphaenogaster senilis [36]. We have never observed aggressive interactions among workers so it is unlikely conflicts take place among A. gracilipes workers as to which will become physogastric and which physogastric workers will produce males. Adult males do not appear to originate from a single dominant physogastric worker in queenless A. gracilipes colonies, as our genetic results indicated that males originated from more than one worker in at least one of our queenless colonies. This result is consistent with male genetic data for A. gracilipes in Taiwan, which showed four different alleles at one locus (Ano10) in one queenless colony fragment [21].
Without a queen, A. gracilipes colonies are doomed because reproductive workers are unable to lay worker eggs due to their lack of spermatheca [21]. The only chance of survival for a queenless colony would be to merge with a queenright colony and/or adopt a queen. Our workers in queenless colonies readily accepted their original queen back in the nest after being separated for 60 days. In the Northern Territory (Australia), laboratory-kept A. gracilipes queenless colonies were successfully merged with queenright colonies from a different source colony [42]. Orphaned colonies may therefore merge with other colonies and/or adopt a queen from a different colony in the field. However, the increase in proportion of physogastric workers, which have less competitive ability and do not contribute to foraging as much as normal workers, following queen death could precipitate the demise of orphaned colonies before such opportunity arises.
Potential benefits of worker reproduction
Despite the costs associated with their lack of contribution to foraging and defence, the role of physogastric workers as trophic-egg layers in queenright colonies may be significant [21]. Colony observations have shown that trophic eggs may represent a major part of the larval diet in A. gracilipes [21]. We did not observe trophic eggs during the queen transfer experiment, but any trophic eggs produced by physogastric workers would likely have been fed to the queen and brood immediately after being laid, as observed in queenright A. gracilipes colonies by Lee et al. (2017) [21].
Worker reproduction may also increase the fitness of deceased A. gracilipes queens and orphaned workers because it is their last opportunity to contribute to the gene pool. In Taiwan, the seminal vesicles of A. gracilipes worker-produced males contained viable sperm suggesting that they are able to mate [21]. Although the reproductive mode of A. gracilipes is unresolved, genetic data and laboratory observations suggest that intranidal mating is the main mode of reproduction for this species [27,43]. In eight of our captive colonies in which males were present, we observed alate queens with no prior exposure to males lose their wings before observing eggs in the colony. If queen brood or virgin queens were present in the colony at the time of the queen’s death and did not inhibit the production of males by workers, intranidal mating between worker-produced males and virgin queens could occur. Such a strategy could prolong the life of a colony after the queen’s death.
Size and genotypes of males and implications for A. gracilipes reproduction
The head width and wing length of worker-produced males were significantly larger (4.0-4.3%) than for queen or worker-produced males and Weber’s length tended to be larger, which may provide worker-produced males with a competitive advantage [26]. We do not know whether A. gracilipes queens select the males they mate with, whether this selection involves male sizes, and whether larger males have a competitive advantage over smaller ones. Larger males of some Pogonomyrmex harvester ants are more successful at mating than smaller males because they can be more successful at gaining access to a mate and transfer a greater proportion of their sperm [25,26]. It would be informative to test whether A. gracilipes queens choose larger males, potentially selecting worker-produced over queen-produced males.
We also found that 4.9% of worker-produced males and 21.1% of males that could have been produced by dealate queens or workers were heterozygous. Our findings are different to those of Lee et al. (2017) [21] who found all 14 A. gracilipes worker-produced males from a single queenless colony fragment to be hemizygous, and most of the 20 males from one queenright colony to be heterozygous. Elsewhere they have been genotyped, field-collected heterozygous A. gracilipes males were found to be common (Borneo [23], Christmas Island [27], Arnhem land Australia [29]). For example, about 50% of males collected in Borneo were heterozygous [23]. A heterozygous genotype in males would typically indicate diploidy.
In ant populations, when a queen mates with a male sharing the same genotype at the sex determination locus (or loci, i.e. match mating), half of the diploid offspring produced by the queen will be homozygous at the sex determination locus (or loci) and develop into diploid males instead of workers [4,24]. Diploid male production is especially common in ant populations that have low genetic diversity (such as invasive populations), and hence low sex determining allele diversity [4,24]. Intranidal mating may be common in A. gracilipes [27,43] which would increase the chance of mating between related individuals and increase diploid male production.
Heterozygous A. gracilipes males that were produced by queens can be diploid, but it is unlikely that heterozygous males produced by workers are diploid. Anoplolepis gracilipes workers do not possess a spermatheca and are unable to mate [21] so their male offspring cannot be diploid through match mating. Instead, heterozygous males may be produced as a result of genetic mosaicism in which an individual possesses two distinct genotypes i.e. two sets of cells that are genetically different and spread across the body [44,45]. In the case of A. gracilipes worker-produced males, heterozygous individuals could be hemizygous but combine the two genomes of a single worker, which would explain why they possess two different alleles at some loci.
Some heterozygous males produced by queens may also not be diploid but genetic mosaics. Diploid males tend to be sterile [3,4,24], but in some ant species, a low proportion of diploid males produce sperm and can father triploid progeny [46,47]. In A. gracilipes, dissections of the seminal vesicles of 16 putative diploid males revealed that all of them possessed motile sperm, which suggests that they are not sterile [21]. Given that heterozygous males (putatively diploid) are apparently common for this species [this study, 23,27,29], we would expect a high prevalence of triploid workers resulting from successful mating between a queen and a diploid male. However, evidence of triploid A. gracilipes individuals has never been reported [21,23,27,29,48]. Heterozygous males may therefore be genetic mosaics with both maternal and paternal cells (i.e. gynandromorphs) [29]. Gynandromorphs, can occur in Hymenoptera and may combine the morphological features of males and females [49–51]. Some gynandromorphs can have bilateral symmetry with one side female and the other male, while other gynandromorphs are mosaics with male and female tissues spread across the body [45,52]. The four males which had one eye that was oversized compared to the other eye (Fig. 2) may be sex mosaics with a conspicuous phenotype. In ants, sex mosaics sometimes present an enlarged eye (female) on one side of the head and a smaller eye (male) on the other side [50,52].
In addition to gynandromorphs, A. gracilipes reproduction may also involve a caste determination system. We found that most genotyped workers were heterozygous for at least one locus, and that most queens were homozygous. This genetic pattern is typical of A. gracilipes populations and suggests that female castes are determined by a genetic component for this species [21,23,27,29,48]. A potential caste determination system could be linked to gynandromorphy in males. Queen-produced gynandromorph males could produce sperm from their inherited paternal or maternal cells, and female castes could be determined by a combination of male and female alleles [29]. The reproductive mode of A. gracilipes may contribute to the ecological dominance of this ant by maintaining a high number of heterozygous workers that may be better adapted to human-modified environments, as has been suggested for another invasive ant species, Wasmannia auropunctata (the little fire ant or electric ant) [53,54].