Ovarian developmental variation in the primitively eusocial wasp Ropalidia marginata suggests a gateway to worker ontogeny and the evolution of sociality

Many species of social insects such as ants, bees, wasps and termites live in colonies showing high levels of cooperation and division of labour. These features are achieved by the differentiation of colony members into one or a small number of fertile queens (or kings) and a large number of sterile (and therefore defined as altruistic) workers (Wilson, 1971). The evolution by natural selection of such altruistic behaviour on the part of the workers remains a major unsolved problem in insect sociobiology. Inclusive fitness theory (also often referred to as kin selection or Hamilton's rule) provides a powerful theoretical framework for exploring the evolution of altruistic behaviour. Hamilton's rule (rb>c) states that an altruistic trait can spread by natural selection if the benefit of the altruistic act to the recipient (b), devalued by the genetic relatedness between actor and recipient (r), is greater than the cost to the altruist (c) (Hamilton, 1964a; Hamilton, 1964b). Most studies attempting to test Hamilton's rule have measured genetic relatedness between altruists and recipients but relatedness values alone can tell us little about the validity or otherwise of Hamilton's rule (Evans, 1977; Gadagkar, 1990; West-Eberhard, 1978). Attention should therefore now focus on the cost and benefit terms in Hamilton's rule, which so far remain poorly studied (Choe and Crespi, 1997; Gadagkar, 2001; West-Eberhard, 1975). One way in which Hamilton's rule can be relatively easily satisfied is for the altruist to pay only a small cost, and this could happen if the altruist has a lower fertility compared with other individuals in the population. This idea has been proposed repeatedly in such forms as the ‘sub-fertility hypothesis’, ‘parental manipulation’, ‘pre-imaginal caste bias’, etc. (Alexander, 1974; Craig, 1983; Field and Foster, 1999; Gadagkar et al., 1988; Gadagkar et al., 1990; Gadagkar et al., 1991; West-Eberhard, 1975).

These ideas are best tested in so-called primitively eusocial species where many workers can be reproductively totipotent and capable of full ovarian growth, unlike highly eusocial species where workers have permanently reduced or atrophied ovaries even at the time of eclosion (Oertel, 1930; Snodgrass, 1956). Reproductive totipotency makes it possible for workers in primitively eusocial species to transform themselves from sterile to fertile individuals and become replacement queens upon the loss or death of the original queen (Chandrashekara and Gadagkar, 1992; Monnin and Peeters, 1999; Smith et al., 2009; Tibbetts and Huang, 2010). In many social insect species such as honey bees, sweat bees, bumble bees, polistine wasps and vespine wasps, ovarian development is influenced by physiological, social and environmental factors such as body size, larval and adult nutrition, presence or absence of queen pheromones, social interactions and time of year, leading to considerable heterogeneity in the extent of ovarian development among colony members (Backx et al., 2012; Bloch et al., 2002; Duchateau and Velthuis, 1989; Hoover et al., 2006; Kapheim et al., 2012; Motro et al., 1979; Smith et al., 2008; Smith et al., 2009; Tibbetts et al., 2011; Toth et al., 2009; Wheeler, 1986; Wheeler, 1996; Wheeler et al., 1999). In the primitively eusocial wasp Ropalidia marginata, there is a clear pre-imaginal caste bias. Only about 50% of the eclosing individuals build nests and lay eggs when isolated into individual holding boxes in the laboratory and provided with ad libitum food and building material. The other half die without ever laying eggs in spite of living, on average, as long as or longer than the egg layers (Gadagkar et al., 1988; Gadagkar et al., 1990; Gadagkar et al., 1991). The advantage of experimenting with freshly eclosed female wasps maintained in isolation in the laboratory is that we may rule out or minimize the influence of such factors as social interactions, dominance, queen pheromone, etc., and attribute the observed inter-individual variation in reproductive potential to intrinsic properties of the eclosing wasps, including any effects of the larval environment. However, data on ovarian development at different ages could not be collected in the studies mentioned above, in which all females were kept alive until they laid at least one egg or died naturally. In the present study, we therefore measured ovarian development at different ages by similarly maintaining freshly eclosed, virgin female wasps under isolation in the laboratory and found as much or more inter-individual variation in ovarian development as we found in egg-laying ability in the previous studies mentioned above.

To begin with, we provide a brief overview of the natural history of R. marginata in order to help place the results of our laboratory studies in the ecological context of natural colonies. Ropalidia marginata is a tropical, old-world, primitively eusocial paper wasp widely distributed in peninsular India. New colonies may be founded throughout the year, by one or a small group of females. In single-foundress nests, the foundress not only lays eggs but also performs nest building, foraging, brood rearing and all other tasks, until the eclosion of her daughters. In multiple-foundress nests, only one female wasp lays eggs (queen) while all non-reproductive tasks such as nest building, brood care and foraging are performed by the remaining females, who function as sterile workers. However, the queen is replaced from time to time and one of the workers becomes the new queen, and she may in turn be replaced after some time. Because of the tropical climate, the nesting cycle of these wasps is perennial and the nests may be very long lived. The wasps (at least, the females) are also quite long lived. Males stay on their natal nests for about a week, after which they leave, to lead a nomadic life. Female wasps may also leave at various times to found new single- or multiple-foundress nests or to join other newly founded nests, or may stay in their natal nests for their entire life, with or without becoming queens. Thus, there is no real distinction between ‘queens’ and ‘replacement queens’. The mean duration of residence on their natal nests for female wasps is 27 days (s.d. 23 days) (Gadagkar et al., 1982), the mean age at which they become queens (if they do), is 35 days (s.d. 20 days), and they then remain as queens up to a mean age of 103 days (s.d. 65 days). Thus, at least some females are known to remain on their natal nests for over 100 days, take over as queens as late as 78 days after their eclosion and continue to remain as queens for as long as 262 days after eclosion (Gadagkar et al., 1993). Most queens have already mated, although some individuals might mate after assuming the role of queen. Some workers are mated and some are not. Mating is neither necessary nor sufficient for an individual to develop her ovaries, suppress ovarian development in other females, including mated ones, and become the sole egg layer of a colony (Chandrashekara and Gadagkar, 1991). While queens always have well-developed ovaries, and a few workers may have ovaries at an intermediate stage of development, most workers have completely undeveloped ovaries. However, workers can rapidly develop their ovaries and lay eggs within a week or less after they become queens (Gadagkar, 2001).

Thirty-seven nests of Ropalidia marginata (Lepeletier 1836) were collected in and around Bangalore, India, and brought to the laboratory, where they were cleared of all existing adult wasps and larvae. The nests containing the pupae were then maintained in aerated plastic boxes, and monitored daily for eclosing females. Six-hundred and thirty-two freshly eclosed females were removed from their natal nest on the day of their eclosion, immediately transferred to transparent, ventilated plastic boxes (22×11×11 cm), one female per box, and provided with ad libitum food (Corcyra cephalonica larvae, Lepidoptera: Pyralidae), honey, water and a soft wooden block as building material. Each female was randomly allotted to one of the following 22 age classes: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 40, 50, 60, 80, 100, 150 and 200 days. At the stipulated age, the females were collected, dissected and their ovarian measurements (length of largest oocyte, width of largest oocyte, mean length of proximal oocyte, mean width of proximal oocyte, number of mature oocytes, total number of oocytes and number of oocytes with yolk) were subjected to principal components analysis. Principal component 1 from this analysis was used as the ovarian index (OI) for each female. Similarly, 27 different body measurements (see Gadagkar, 2001) were subjected to principal components analysis and the first principal component was used as an index of body size. Adult feeding was determined by noting the number of C. cephalonica larvae consumed during each day of the experiment and was then averaged across all days.

Principal component scores were calculated using StatistiXL, version 1.8 (www.statistixl.com). All statistical analyses were done in R, version 2.11.1 (R Development Core Team, 2010). One-hundred and ninety-three females from all age classes showing at least one oocyte (of any stage of development) were used to fit four functions describing the relationship between age and ovarian index (both variables square-root transformed) to explore trends in the pattern of ovarian development with age: (1) linear increase, (2) saturation (Michaelis–Menten function), (3) unimodal skewed distribution (Ricker function) and (4) quadratic function. An information-theoretic approach (minimum AIC) was used to select the best-fit function, and each function was fitted using lm or nls commands in R, with OI as the response variable and age as the predictor variable. To identify important variables affecting ovarian development, we constructed seven models using unique combinations of age, body size index and feeding. Two-hundred and six females (including those that did not have a single oocyte at any stage of development) at age more than or equal to 20 days were included in this analysis. Individuals below 20 days of age were excluded from this analysis because most of them showed no ovarian development at all. Model selection was performed using information-theoretic criteria (Burnham and Anderson, 2002), viz. second-order Akaike information criterion (AICc) values, Akaike's weights and AIC differences using the AICcmodavg package (Mazerolle, 2010) in R.

A total of 632 freshly eclosed, isolated female wasps were studied by sampling them at predetermined ages ranging from 0 to 200 days and measuring their body size, ovarian development and feeding rate. The wasps showed large variation in ovarian development, even among individuals of similar ages, despite being kept under similar conditions and with access to similar resources (Fig. 1, Fig. 2A, Table 1). All eclosing females started with no ovarian development at day 0, and the number of individuals showing some ovarian development (i.e. with at least one oocyte at any stage of development) increased gradually with age, reached a peak at 100% at 100 days, and decreased slightly thereafter (Fig. 3). The extent of inter-individual variation is dramatically reflected in the fact that the maximum ovarian index in each age class rose very sharply with age, peaking at a value of 8.4 at 80 days, while the mean OI of each age class rose more gradually and peaked at a value of only 4.4 at 100 days (Fig. 2B).

Considering only those wasps that had developed ovaries (196 individuals), we found that ovarian development increased gradually, peaked at a value of 4.8 at 156 days and then decreased slightly, as best described by a Ricker function (Fig. 2C). Thus, a large number of individuals failed to develop their ovaries even up to 80 days of age and a few individuals resorbed their ovaries beyond 100 days of age. Moreover, developed ovaries did not necessarily guarantee egg laying. One-hundred and twenty-seven females that did not lay any eggs had better-developed or similar ovaries compared with the 23 females that laid eggs. Among the 100 day old females, all of which had developed ovaries, eggs layers and non-egg layers did not differ significantly in their ovarian indices (unpaired Wilcoxon rank sum test, W=15, n₁=3, n₂=16, P>0.05). Age, body size and adult feeding rate positively correlated with ovarian development. When taken together, these three variables had higher explanatory power than when they were considered one at a time or in pairs (Tables 2, 3). The age of the wasps studied was varied as part of the experimental design from 0 days (day of eclosion) to 200 days. Addition of other variables to the model, such as the proportion of empty cells, the number of females present on the nests from which these individuals eclosed (before nest collection), the date of eclosion and nest of eclosion, did not add any additional explanatory power (data not shown).

We confirmed that body size was normally distributed (Fig. 4). Feeding rate varied in a highly skewed manner, with most wasps feeding very little. Feeding rate was zero for wasps in the first quartile and was 0.12 C. cephalonica larvae per day for wasps in the third quartile (Fig. 5). It should be emphasized that these feeding rates were observed in spite of the fact that all wasps were provided with ad libitum access to food, although it is not the absolute feeding rates but the highly skewed inter-individual feeding rates that are of interest here. There was a weak but statistically significant positive correlation between body size and adult feeding rate (Fig. 6).

Ropalidia marginata females showed a very high level of variation in the time taken to develop their ovaries and also in the extent of ovarian development, despite having access to similar resources. The average lifespan of R. marginata workers (duration of residence) on their nest is 27±23 days (mean ± s.d.) (Gadagkar et al., 1982). In the present study, about 40% of females failed to show any ovarian development within their lifespan. Because R. marginata nests can be initiated by solitary females, and because mating is not necessary for ovarian development, we consider the possibility that the variation in ovarian development we have observed here may also occur in nature among solitary nest foundresses (which also have no social stimuli, like the isolated wasps in our laboratory experiment). The lack of ovarian development that we observed in 40% or more of the females is thus expected to facilitate the ontogeny of worker behaviour, as the cost of foregoing direct reproduction would be relatively small for many females, i.e. Hamilton's rule (Hamilton, 1964a; Hamilton, 1964b) would be more easily satisfied for some females than for others (Gadagkar, 2001; West-Eberhard, 1975). The fact that it took as many as 100 days for all females to show at least one oocyte suggests considerable delay in reproductive maturation in many females. Such delayed reproductive maturation can facilitate the commonly observed strategy of female wasps to first function as sterile workers and later become future queens of their colonies (Gadagkar, 1991). It is of course possible that variation in ovarian development evolved after the origin of eusociality, although we do not know which came first. Nevertheless, variation in ovarian development makes it easier for females with poor ovarian development to satisfy Hamilton's rule because they would be paying a smaller cost by staying back in their natal nest as sterile workers. This advantage for some females may have facilitated either the origin of eusociality (if variation in ovarian development preceded the origin of eusociality) or the maintenance of eusociality (if the origin of eusociality preceded variation in ovarian development). In short, females with poor potential for ovarian development should find it easier to forego nest founding and become sterile workers. Thus, our study provides further evidence in support of models for the evolution of eusociality discussed in the literature under the theme of ‘sub-fertility’ (Alexander, 1974; Craig, 1983; Gadagkar, 1991; West-Eberhard, 1975) but it should be emphasized that these models in turn show how Hamilton's rule can be satisfied.

Our results also show that some females fail to initiate nests and lay eggs even though their ovaries are as well developed as those of females that do initiate nests and lay eggs. Such females may be incapable of becoming or reluctant to become solitary foundresses but may be capable of becoming replacement queens upon the loss of the original queen or of usurping colonies with weak queens. Thus, our study also provides evidence for direct fitness models for the evolution of sociality discussed in the literature as the ‘sit and wait’ strategy (Nonacs and Reeve, 1993; Starks, 1998). We see no reason why indirect fitness models such as Hamilton's rule and direct fitness models such as ‘sit-and-wait’ should be mutually exclusive; wasps may be expected to gain fitness by any route available to them.

We have shown that age, body size and adult feeding rate are all correlated with an individual's ovarian development. That age is one of the variables important in ovarian development is not surprising. However, age is only one of the variables important in ovarian development, so that, depending on the values of other factors such as body size and feeding rate, individuals of the same age have very different levels of ovarian development and, conversely, individuals of very different ages may have similar ovarian development. We hypothesize that it is such plasticity in ovarian development that allows individuals to adopt different strategies such as nest founding, joining, staying back in the natal nest, usurping, etc., depending on their internal physiological state and the prevailing external environmental conditions. Individuals that resorb their ovaries after a certain age could also contribute to the variable effect of age on ovarian development. Ovarian resorption, also known in ants and bees (Billen, 1982; Duchateau and Velthuis, 1989), could be a strategy to avoid ovipositioning in unfavourable conditions or be aimed at conserving nutrients (Sundberg et al., 2001). As females had ad libitum food available to them throughout their lifetime, starvation may not be the cause of ovarian resorption in our study. However, the metabolic and immunological costs of ageing (Bell and Bohm, 1975; Collier, 1995) may lead to a greater demand by somatic tissue, and survivability may be achieved by compromising reproduction (Ohgushi, 1996). Resorption may also occur because of age and a lack of nesting sites (Bell and Bohm, 1975; Lopez-Guerrero, 1996), and this may be the case here, as many females that developed ovaries failed to initiate nests.

Body size was the second factor that was important in ovarian development. Body size, in turn, is generally influenced by larval nutrition (Edgar, 2006; Karsai and Hunt, 2002; Plowright and Jay, 1977; Richards and Packer, 1996; Roulston and Cane, 2002). Similarly, body size is known to be correlated with food consumption in many insects (Reichle, 1968). As there was a positive correlation between body size and adult feeding rate even when all individuals had access to unlimited food, it may be argued that well-fed larvae developed into larger sized individuals predisposed to consume more food, which then consumed more food and developed better ovaries. Conversely, poorly fed larvae developed into smaller sized individuals predisposed to consume less food, which therefore consumed less food and had poorly developed ovaries (Gadagkar et al., 1988; Gadagkar et al., 1991). Variation in larval nutrition is thus expected to be the starting point of this cascade of events leading to variation in ovarian development that ultimately facilitates the evolution of a worker caste. Such variation in larval nutrition may be brought about by seasonal and other variation in food availability to the foragers, the number of foragers available and perhaps some active differential feeding of larvae to channel them preferentially into queen or worker roles during specific periods in the colony cycle (Gadagkar, 1996; Gadagkar, 2001; Gadagkar et al., 1988; Hunt, 2007; Hunt and Nalepa, 1994; Kapheim et al., 2011).

Adult nutrition was the third important predictor of ovarian development. Oocyte development is essentially a nutrient-driven process requiring large protein and fat resources. Therefore, it is not surprising that females with a higher average caterpillar consumption rate also had better developed ovaries. Adult nutrition has been shown to affect reproductive potential in wasps (Gadagkar et al., 1988; Hunt, 2007; Tibbetts et al., 2011; Toth et al., 2009). Variation in adult nutrition in natural colonies may be brought about by the unequal exchange of food between foragers and intranidal workers as well as being due to unequal trophallaxis among the workers and between workers and larvae. In addition to differential feeding, the actual nutritional status of an individual may depend on its social role in the colony and how much energy is required to perform that role. Energetically expensive behaviours such as foraging may reduce resource allocation to reproduction (Markiewicz and O'Donnell, 2001; West-Eberhard, 1969). It should be noted that it was not possible to measure honey consumption in this study, though sugar consumption is known to affect ovarian development (Nilsen et al., 2011; Page and Amdam, 2007); therefore, our interpretation of nutrition affecting female ovaries is largely based on protein consumption through Corcyra larvae.

We have shown that pre-imaginal factors such as larval nutrition and body size and post-imaginal factors such as age and feeding rate can act together to produce variability in adult reproductive potential, which can in turn facilitate the ontogeny of a worker caste. Our experiments using female wasps in isolation show that the resulting variation in reproductive potential can be very large in R. marginata, thus providing an explanation for how Hamilton's rule could be satisfied even when intra-colony relatedness is quite low (Gadagkar, 2001). Thus, the observed variation in reproductive potential can by itself provide a powerful gateway for worker ontogeny and the evolution of sociality.

R.G. designed the study, S.C. performed the experiments, S.S. analyzed the data, and R.G. and S.S. co-wrote the paper. Our experiments comply with regulations of animal care in India. We thank James Hunt, William Wcislo and an anonymous referee for their helpful comments.