Shift from independent to dependent colony foundation and evolution of ‘multi-purpose’ ergatoid queens in Mystrium ants (subfamily Amblyoponinae)bij_1257 198..207

Division of labour improves fitness in animal societies. In ants, queens reproduce, whereas workers perform all other tasks. However, during independent colony founding, queens live as solitary insects and must be totipotent, especially in species where they need to forage. In many ants, solitary founding has been replaced by dependent founding, where queens are continuously helped by nestmate workers. Little is known about the details of this evolutionary transition. Mystrium rogeri from Madagascar and Mystrium camillae from Southeast Asia (subfamily Amblyoponinae) have winged queens, but three congeneric species from Madagascar reproduce with permanently wingless queens instead. We show that this ‘ergatoid’ caste has distinct body proportions in all three species, expressing a mixture of both queen and worker traits. Ergatoid queens have functional ovaries and spermatheca, and tiny wing rudiments. They can be as numerous as workers within a colony, but only a few mate and reproduce, whereas most behave as sterile helpers. The shape of their mandibles makes them unsuited for hunting and, together with a lack of metabolic reserves (i.e. in the form of wing muscles), this means that ergatoid queens cannot be solitary foundresses. In comparison with winged queens, ergatoid queens are less costly per capita and they experience lower mortality. They remain in their natal colonies where they can either reproduce or function as helpers, making them a ‘multi-purpose’ caste. Within the Amblyoponinae, ergatoid queens replace winged queens in Onychomyrmex as well. However, in this genus, ergatoid queens are ‘sole-purpose’, few are produced each year and they reproduce but do not work. Hence, different types of ergatoid queens evolved to replace winged queens in ants. © 2009 The Linnean Society of London, Biological Journal of the Linnean Society, 2009, 98, 198–207.


INTRODUCTION
Division of labour is a key feature of all animal societies. Individuals that specialize on distinct tasks perform more efficiently; hence, the fitness of the society is increased (e.g. honeybees; Seeley, 1982). Although, in many social species, all individuals retain the ability to switch between tasks during their life, in ants and some other eusocial insects, selection has shaped adult phenotypes that are dedicated to specific tasks. The most widespread examples are queen and worker castes, which are morphologically specialized for reproductive and nonreproductive tasks, respectively (Oster & Wilson, 1978). Although either caste cannot function alone (e.g. workers usually cannot mate), together, they allow for improved functioning of the colonial unit.
In many social insects, colonies are founded by solitary individuals (independent colony foundation; ICF) and division of labour is impossible during this initial phase. A foundress needs to carry out all tasks, including reproduction, nest building, foraging, and nest defence (e.g. Polistes wasps; Reeve, 1991). Accordingly, strong selective pressures act on the phenotype of such foundresses so that they remain totipotent. In many ants, winged queens disperse by flight and found new colonies alone (Peeters & Molet, 2009). In basal lineages, founding queens must even forage outside the nest before their first brood emerges (nonclaustral ICF), although, in the majority of higher ants, queens are provisioned with large metabolic reserves that are sufficient to feed the first brood (claustral ICF). Mortality is high during ICF, and has been selected against in numerous unrelated ant genera. Instead, dependent colony foundation (DCF) has evolved, whereby established colonies divide on a regular basis. This strategy eliminates the solitary founding stage and reproductives cannot succeed without sterile nestmates (Peeters & Ito, 2001). Among social wasps and bees, DCF involves both winged queens and winged workers that fly together to new nests (Ross & Matthews, 1991;Visscher, 2007). In ants, however, workers are always wingless and the queen(s) is forced to disperse on foot. Consequently, the ancestral winged queens have disappeared in many species (Heinze & Tsuji, 1995;Peeters & Ito, 2001). In 200-300 species, they are replaced by sexually reproducing workers ('gamergates'). However, in most ants, workers cannot mate and store sperm; hence, wingless reproductives have evolved that are morphologically distinct from the worker caste. They have a functional spermatheca but their thorax is reduced and resembles that of the workers'; hence, they are termed 'ergatoid' queens (Wheeler, 1910;Peeters, 1991).
Ergatoid queens occur in a large number of unrelated ant taxa, but relatively few studies have described their morphological characteristics and the pattern of investment in sexuals. The worldwide subfamily Amblyoponinae is basal among ants (Saux, Fisher & Spicer, 2004;Brady et al., 2006) and exhibits a variety of colonial reproductive strategies and associated queen phenotypes. Thus, it comprises an ideal taxon to assess the consequences of the shift from ICF to DCF. For example, in Onychomyrmex, winged queens do not exist and colonies possess a single ergatoid queen that is physogastric and highly fertile (Brown, 1960). This corresponds to the general pattern found in ants with ergatoid queens. However, a recent study in Mystrium 'red' revealed that each colony has numerous ergatoid queens that are smaller than the workers (Molet, Peeters & Fisher, 2007a). There are no winged queens, unlike other species in this genus. Whether this is an isolated example is unknown. Accordingly, we compared colony demography, caste morphometry, and reproductive division of labour in two other species from Madagascar (Mystrium mysticum and Mystrium oberthueri) as well as in two species with winged queens (Mystrium camillae from Southeast Asia and Mystrium rogeri from Madagascar). We review the repeated shifts to obligate DCF in the subfamily Amblyoponinae, where the ancestral winged queens have been replaced by different types of wingless reproductives, including gamergates. We find two contrasting patterns (i.e. 'sole-purpose' versus 'multipurpose' ergatoid queens), indicating that recurrent selection against winged queens can be solved in different ways. We suggest that wingless reproductives can readily evolve in ants because two castes (winged queens and wingless workers) already exist, and larval development can be modified to create intermediate adult phenotypes that are cheaper than winged queens. Adults and brood were counted in the field, and some cocoons were opened a few days after collection to identify the pupae inside (worker, gyne or male). In addition, ten colonies of M. camillae were collected from secondary forests in Java (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003). Specimens from M. camillae colonies are in F. Ito's collection; specimens from the four other species are deposited at the California Academy of Sciences and imaged on http://www.antweb.org.

MORPHOMETRY
We used a Leica MZ6 microscope to take dorsal and lateral photographs of all adults from one colony per species: 23 ergatoid queens and 61 workers of M. 'red'; 22 ergatoid queens and 23 workers of M. mysticum; 21 ergatoid queens and 30 workers of M. oberthueri; one dealate queen, 19 winged gynes, 24 workers, and one intercaste of M. rogeri; and one dealate queen and 24 workers of M. camillae. We measured mandible length, maximum head width, thorax volume, and cross-sectional area of the first gaster segment with ImageJ (http://rsb.info.nih.gov/ij) sensu Molet et al. (2007a). To compare mandible shape among workers and ergatoid queens of M. 'red', we measured crosssectional area at the base, mid-way, and tip in ten REPLACEMENT BY ERGATOID QUEENS 199 individuals of each caste. In addition, the thorax of various adults was studied by scanning electron microscopy. A few pupae were photographed to study thorax structure.

DISSECTIONS
We dissected a proportion of queens (N = 681) and workers (N = 223) from 52 colonies to check for ovarian development and spermatheca. Individuals were classified according to mating status (full or empty spermatheca) and the type of oocytes in their ovaries. Yolky oocytes were divided into two categories: 'mature' (i.e. the size of an egg) and 'immature'. Non-yolky oocytes were not counted.

STATISTICAL ANALYSIS
We performed allometry analyses between thorax volume and gaster area in all species and castes. The one exception was M. camillae, where only one queen was available for measurement. We used (S)MATR, version 1.0 (Warton & Weber, 2002) (available at: http://www.bio.mq.edu.au/ecology/SMATR/) to compute allometry coefficients (i.e. the slope of the regression lines) using standardized major axis regression, and to test for isometry and compare allometries between groups (Molet et al., 2007a). All other statistical analyses were performed with STATISTICA, version 7.1.
By contrast, M. mysticum and M. oberthueri ergatoid queens have a smaller thorax volume than conspecific workers (Fig. 3A), similar to M. 'red' (Molet et al., 2007a). The thorax of ergatoid queens has an appearance similar to that of workers because the pronotum is relatively long (Fig. 4). Importantly, however, ergatoid queens retain rudiments of both anterior and posterior wings. Unpigmented pupae have large wing pouches (Fig. 4) similar to winged queen pupae, but young adult ergatoid queens only retain tiny wing rudiments at the exact location of the wings in M. rogeri and M. camillae queens. In addition, the pleurite and notum of ergatoid queens, unlike those of workers, are incompletely fused in the region of the wing rudiments (Fig. 4).
In M. camillae and M. rogeri, both winged queens (based on the dissection of N = 13 and N = 3 queens, respectively) and workers (N = 40 and N = 105) have ovaries with eight ovarioles (except in one colony of M. rogeri; see below). However, these queens have a much larger gaster than conspecific workers (Fig. 3B). Active ovaries in M. rogeri queens are longer than in fertile ergatoid queens (Molet et al., 2007a). However, in contrast with workers from ergatoid species, M. rogeri workers have functional ovaries and can lay eggs following the death of the queen (after 3 months in the laboratory; M. Molet & C. Peeters, unpubl. data). One of the M. rogeri intercastes had the same gaster area as queens, but the other four had smaller gasters. These four intercastes were dissected and had ten, 11, eight, and eight ovarioles, whereas the queen of the same colony had 13. Mystrium rogeri colony BLF10994 lacked a queen, and workers had more ovarioles (i.e. 8-16) than in other colonies, with 14 workers having mature oocytes.
Mystrium rogeri and M. camillae workers show a higher variability in their body proportions than workers of ergatoid species (Fig. 5). For example, their mandible coefficient of variation (CV) is much higher than that of M. mysticum, M. oberthueri, and M. 'red' workers (12.5% and 20.1%, respectively, versus 7.6%, 4.2%, and 10.0%). Head and mandible sizes of M. rogeri and M. camillae workers range from being similar to ergatoid queens to being similar to workers of ergatoid queen species (Fig. 5). Hence the workers in species with ergatoid queens exhibit very little size polymorphism compared to workers of queenright species.
All ergatoid queens of M. 'red' have a soft reddish cuticle throughout their adult life, unlike the black ergatoid queens in M. mysticum and M. oberthueri. Having an unsclerotized cuticule is possible in ergatoid queens because they do not forage outside the nest and thus are less subject to dehydration. The ergatoid queens of M. 'red' are also less numerous in each colony, and individually more fecund.

ERGATOID QUEENS ARE A CASTE DISTINCT FROM WORKERS
In all three Mystrium species lacking winged queens, reproduction is carried out by ergatoid queens that have distinct morphology and body proportions relative to conspecific workers. They are smaller than workers and, given their tiny mandibles (Fig. 2), they are unlikely to be hunters (confirmed in the laboratory for M. oberthueri; Molet et al., 2007b). Together with a lack of wing muscles serving as metabolic reserves, this makes ergatoid queens unfit to be solitary foundresses. This is in Figure 5. Proportions between body parts in queens and workers of the three species with ergatoid queens compared to workers of the two species with winged queens. Workers of the latter two species have shapes that overlap those of ergatoid queens and conspecific workers. Each point corresponds to one individual.
contrast to the winged queens of M. rogeri that are larger than workers and have strong mandibles (Fig. 2) and, thus, are able to hunt during ICF. We can conclude that these ergatoid queens are not totipotent, and DCF is obligate (i.e. there is uninterrupted cooperation between queens and nestmate workers). These three species with ergatoid queens are closely related, and are derived relative to M. rogeri and M. camillae (DNA sequence data from seven nuclear genes, B. Noonan and B. Fisher, unpubl. data). Because Mystrium workers apparently lack a functional spermatheca, gamergates do not exist in this genus. Accordingly, the only possible response to selection against winged queens in Mystrium was to evolve a new wingless caste that is capable of sexual reproduction: the ergatoid queen caste [termed the 'intermorph caste' in Molet et al. (2007a) but corresponding fully to the ergatoid queens found in other ant taxa].
Ergatoid queens were found in large numbers in all colonies of M. oberthueri, M. mysticum and M. 'red'. However, only a few mated and reproduced, whereas most remained virgin and infertile, and the latter performed brood care (e.g. M. oberthueri; Molet et al., 2007b). The distinctive mandibles of Mystrium workers (Fig. 2) allow specialized predation on large centipedes, as also observed in three species of Amblyopone (Gotwald & Lévieux, 1972;Ito, 1993b;Masuko, 1993). In M. rogeri, workers are continuously polymorphic (Fig. 5) and their highly variable mandibles are suitable for both brood care and hunting (Molet et al., 2007b). In M. camillae, only large individuals were seen in the foraging area of laboratory nests (F. Ito, unpubl. data). In Mystrium species with ergatoid queens, the workers have large mandibles suited for hunting but not for brood care. By contrast, ergatoid queens have reduced mandibles and are unfit to hunt, but they can care for the brood. Such atypical division of labour may explain the reduction in size polymorphism of workers relative to M. rogeri and M. camillae; all but the larger workers have been eliminated in species with ergatoid queens, and ergatoid queens have taken over the tasks of small workers.
The lack of mated ergatoid queens in some colonies may be the result of field collection occurring just before the mating season, so mated queens may have died during the year or left the nest during a colonyfounding event. Some virgin ergatoid queens could have mated a few days later (e.g. we observed males flying towards the colonies during collection, and in the laboratory virgin ergatoid queens of M. oberthueri performed sexual calling and mated with foreign males). The absence of dealate queens in some M. rogeri colonies is due to either natural mortality or escape during collection.

EVOLUTION OF 'MULTI-PURPOSE' ERGATOID QUEENS
Mystrium ergatoid queens do not disperse individually from their natal colony, and they presumably mate near the entrance. They can be as numerous as the workers in a colony, although it is not known whether they are produced year-round. The majority of ergatoid queens are infertile and care for the brood. They contribute to the success of DCF, an endeavour that relies on a large number of nonreproductive helpers (Macevicz, 1979). Such 'multi-purpose' ergatoid queens are also found in other subfamilies; for example, Myrmicinae (all species of Ocymyrmex: Bolton & Marsh, 1989;Forder & Marsh, 1989;Eutetramorium mocquerysi: Heinze, Hölldobler & Alpert, 1999;Myrmecina nipponica: Ohkawara, Ito & Higashi, 1993) and Ponerinae (Leptogenys diminuta: Ito & Ohkawara, 2000), and they correspond to the 'reproductive intercastes' listed by Peeters (1991). In Mystrium, interspecific comparisons reveal that ergatoid queens are considerably cheaper per capita relative to winged queens (fresh weights of 5.95 mg for M. 'red' ergatoid queens versus 14.51 mg for M. rogeri queens; Molet et al., 2007a). Moreover, winged queens are generally reared in large numbers once a year (i.e. more than 100 winged gynes in two colonies of M. rogeri), although only few are likely to succeed in establishing a new colony. Hence, this large reproductive investment may be dramatically reduced by producing multi-purpose ergatoid queens, where individuals that do not reproduce remain in the colony as labourers.

IN AMBLYOPONINAE
The worldwide subfamily Amblyoponinae comprises ten genera with distinctive morphology and specialized predatory behaviour. This subfamily is placed in the poneroid group, which comprises a heterogeneous assemblage of taxa at the base of the ant tree (Brady et al., 2006). In Amblyopone (Gotwald & Lévieux, 1972), Apomyrma stygia (Brown, Gotwald & Lévieux, 1970), and Prionopelta (Hölldobler & Wilson, 1986;Ito & Billen, 1998), the degree of dimorphism between winged queens and workers is rather limited, meaning that a lone foundress needs to raise workers that are almost as big as herself. Accordingly foundresses must hunt (i.e. ICF is nonclaustral, as documented in Amblyopone australis; Haskins & Haskins, 1951). Species that have a poor success rate during nonclaustral foundation have two evolutionary choices: either rear more expensive queens (i.e. larger reserves to reduce the frequency of foraging trips) or shift to DCF. The first scenario is seen in Myopopone castanea, where winged queens are much bigger than workers and have more ovarioles (32 in queen, 6-22 REPLACEMENT BY ERGATOID QUEENS 205 in workers: Ito & Ohkawara, 1994, F. Ito unpubl. data); the details of ICF remain unknown. The second scenario is more widespread. In species of the Amblyopone reclinata group, winged queens have been replaced by gamergates (Ito, 1993a). In Adetomyrma, both winged and ergatoid queens have been found in undescribed species from Madagascar (B. L. Fisher, unpubl. data). In all species of Onychomyrmex, winged queens have been replaced by ergatoid queens (Brown, 1960). The gaster of these queens is much larger than that of workers and they are often physogastric (Wheeler, 1916). Very few (N = 1-6) ergatoid gynes are reared annually, virgin infertile ergatoids are not present year-round in colonies, and colonies are strictly monogynous (e.g. Onychomyrmex hedleyi; Miyata et al., 2003). The ergatoid queens in Onychomyrmex can be described as 'sole-purpose' because reproduction is their exclusive function. This syndrome is true in the majority of ants studied [i.e. only one ergatoid queen (also called 'dichthadiiform' in species where they have a huge gaster) occurs in each colony] (Peeters & Ito, 2001). This is in sharp contrast to species with multi-purpose ergatoid queens, where many are produced per colony and can either reproduce or function as labourers, as described in the present study for Mystrium. Sole-purpose and multipurpose ergatoid queens can be found in various other ant taxa, where they convergently evolved to replace winged queens.

PHENOTYPES IN ANTS
Ergatoid queens are novel phenotypes that express a mixture of queen and worker traits. Evidence for this is provided by the occurrence of tiny wing rudiments in the ergatoid queens of M. mysticum, M. oberthueri, and M. 'red', in addition to functional ovaries and spermatheca. Although wings sometimes start developing from imaginal wing discs in ant workers, they only occur in pupae and have disappeared at the time of emergence (Sameshima, Miura & Matsumoto, 2004). In the ergatoid queens of Mystrium, wing rudiments persist in adults. These structures have no function, although their persistence suggests a recent origin of winglessness in the three species with ergatoid queens. Wing pouches in pupae may be related to the hormonal mechanisms that activate the development of queen-like ovaries.
We found a few intercastes in M. rogeri, which are erratically-produced queen-worker intermediates (e.g. in Temnothorax nylanderi; Plateaux, 1970). Importantly, they were first to start laying eggs upon the death of the founding queen in colony BLF 14779 (M. Molet & C. Peeters, unpubl. data). These intercastes varied in thorax morphology, and were all larger than conspecific workers. They are thus unlike the regularly produced ergatoid queens described in the present study, but provide insights into the evolutionary origins of the latter. In environments where DCF is a better reproductive strategy than ICF, the regular production of cheaper intercastes that can mate and reproduce (e.g. as found in Pachycondyla obscuricornis ;Düssmann, Peeters & Hölldobler, 1996) could be selected for, resulting in the evolution of an ergatoid queen caste with fixed body proportions (e.g. Myrmecina nipponica; Miyazaki et al., 2006). In the ants, the shift from ICF to DCF leads to modifications at the level of both the colony (i.e. reproductive investment and division of tasks) and the individual (i.e. novel phenotypes), resulting in better colonial economy.