Caste Determination

Note: The following text is from Holldobler and Wilson's The Ants, Chapter 8. It's a bit dated but is still a good introduction. A few more-modern references are included under References.

A wealth of studies on caste determination have followed the pioneering experiments by Michael V. Brian (1951a), who first suggested the key steps on the divergence of queens and workers in the genus Myrmica. All of the research has come together in an overriding principle: the female castes of ants are differentiated by physiological rather than genetic factors. In other words, allelic differences almost never separate the female castes of a colony, even to the extent of slightly biasing individuals to develop into one caste as opposed to another. The single exception known applies to alternative forms of reproductive females among colonies and not to the ordinary complement of castes within single colonies. In the European slavemaking ant Harpagoxenus sublaevis, ergatomorphic reproductives (wingless and resembling workers) differ from fully winged (gynomorphic) queens by a single recessive allele. The ergatomorph intercastes possess the genotypes EE, Ee, or ee, while winged queens are always ee (Buschinger, 1975a; Buschinger and Winter, 1975; see Figure 8-49). The ratios of the workers to queens produced are also affected by the alleles (Winter and Buschinger, 1986). The control in this case is not absolute. Larvae of all three genotypes develop into workers under stringent feeding regimes. When the larva makes the physiological commitment to become a reproductive, it adheres strictly to the type of reproductive prescribed by its genotype, in other words an intercaste or a gynomorph, but the likelihood that it will make this decision in the first place is affected by the E and e alleles. When a larva has an E allele, its development is slowed. It is also more sensitive to inhibition by adult gynomorphs and hence is more likely to become a worker instead of an ergatomorph. Conversely, ee larvae grow faster and are less sensitive to the presence of gynomorphs. They are consequently more likely to turn into gynomorphs themselves (Winter and Buschinger, 1986).

Given that caste determination is almost entirely environmental in ants as a whole, the observed caste systems can be usefully interpreted as the consequence of variations among species on growth transformation during larval development (Oster and Wilson, 1978). With the aid of this concept, evolutionary scenarios can be constructed that are subject to experimental testing. Consider first the production of a small class of majors by skewing of the adult worker size-frequency distribution, the lowest grade of worker polymorphism known in the ants. As suggested in the schema of Figure 8-50, newly hatched larvae vary in body weight, probably according to a normal distribution. It is likely that they also vary in other qualities, such as the amount of yolk available to them during embryonic development, the temperature at which they developed, and so forth. Suppose that some of these factors influence the final body size attained by the larvae and hence by the adult ants. If this growth transformation remained constant with increases in initial size, the result would be a set of workers whose size-frequency distribution is approximately normal. Such distributions are typically accompanied by isometry (non-allometric growth), and consequently the entire worker caste will be monomorphic. Suppose, however, that the transformation is such that the larger the initial size of the larva (or the greater the quantity of other caste-biasing factors present), the more rapid its growth, so that individuals starting large finish proportionately larger. Such a relationship might be described as follows: y1 = F[y0] where y0 and y1 are the body size (or quantity of other caste-biasing factors) at the start and finish of larval growth. If the transformation is linear, the result will be monomorphism, in conformity with the conditions just described. Where the transformation is exponential, so that the subsequent rate of growth increases as a function of the starting point 0, the result will be a size-frequency distribution skewed toward the upper size classes. The magnitude of the exponential constant represents the sensitivity of larval growth to the initial conditions encountered by the young larva. Alternatively, the constant can depend on conditions encountered later in larval life, with similar final results. By relatively small adjustments in the transformation function, the caste system of a species can be conspicuously altered.

The next simplest conceivable step in the elaboration of size-frequency distributions is the introduction of decision points during development, as illustrated in Figures 8-51 and 8-52. The decision point, or control point as it is sometimes called, is a time in development at which one or the other of two sets of growth constants are acquired by the immature form. Thereafter the individual will proceed in its development toward one subcaste or another, with little chance of deviation. Decision points are efficient devices for sorting colony members into two or more independent populations. The rules for shunting individuals into one direction as opposed to the other can be adjusted so as to regulate the relative sizes of the two populations. The examples given in Figure 8-55 utilize threshold size at a critical developmental time as the shunting rule; this is the situation in queen-worker determination in the ant Myrmica ruginodis (Brian, 1955). Larvae that attain a certain size by a critical time continue rapid development toward a still larger ultimate size; those that fall short proceed at a distinctly slower rate and are even more behind at the completion of adult development. By setting the threshold size low or high relative to the usual size-frequency distribution of larvae at the critical development time, species can arrange a lower or higher proportion of the larger caste.

Shunting rules are based on a variety of token stimuli in addition to size. No fewer than six classes of such stimuli have been identified in various ant species as influencing the determination of individual females to the worker caste as opposed to the queen caste. Two of these are also known to be potent in minor-major worker determination; the latter process is still relatively unexplored and will probably be shown to be subject to other influences as well (see reviews in Wilson, 1971; Schmidt, 1974a,b; Brian, 1979a, 1983, 1985; Nijhout and Wheeler, 1982; de Wilde and Beetsma, 1982; Passera, 1984; Vargo and Fletcher, 1986a,b; D. Wheeler, 1986a).

(1) Larval nutrition. Competition among larvae alone might produce the bimodal size-frequency curves that underlie most queen-worker and minor-major distinctions in ant species. Larvae that attain a threshold size by a critical developmental time are shunted toward the larger caste.

(2) Winter chilling. Intraovarian eggs of Formica and larvae of Myrmica that have been chilled have a greater tendency to develop into queens, an apparent device for timing the emergence of queens in the spring or early summer. Other responses to temperature, humidity, or photoperiod could produce crops of queens (or even worker subcastes) at other times. These would depend on local climatic conditions and the idiosyncratic features of the colony life cycle in each species.

(3) Temperature. The larvae of Formica and Myrmica tend to develop into queens more readily if reared at temperatures close to the optimal for larval growth.

(4) Caste self-inhibition. The presence of a mother queen inhibits production of new queens in Myrmica, Monomorium, and Solenopsis; likewise the presence of soldiers inhibits the production of soldiers in Pheidole. This negative feedback loop could obviously serve to stabilize caste ratios when there is a need to fix the ratio instead of making it flexible in response to short-term needs.

(5) Egg size. In Formica, Myrmica, and Pheidole, the larger the egg, the more yolk, and the more likely the larva is to develop into a queen as opposed to a worker. No information is available on the relation of egg size to minor-major determination.

(6) Age of queen. Young queens of Myrmica tend to produce more workers; the queen's age could, of course, be reflected in the size of the eggs she lays. Smaller egg size in the batch produced by a nest-founding queen might further explain the occurrence of nanitic workers in the first brood; in other words, egg size could evolve to insure the production of nanitics at this stage in the colony life cycle.

Some of these factors are merely biasing in their effects, rendering an individual more likely to take one direction as opposed to the other upon reaching the point of bifurcation. Others exert their influence directly at the decision points themselves. Often one factor can override another in the following manner: if condition a prevails earlier instead of a', then the larva can respond to either b or b', but a' forces the same response regardless of whether b or b' is present. For example, mature larvae of Myrmica ruginodis subjected to winter chilling have the capacity to develop into either queens or workers. But only those that subsequently reach a weight of 3.5 mg within approximately eight days after the start of posthibernation development actually become queens. Larvae not exposed to chilling always develop into workers, regardless of their size.

The scheduling of receptiveness to caste-biasing stimuli almost certainly represents an idiosyncratic genetic adaptation on the part of each individual species. The later the decision point, the more flexible is the system, in the sense that it permits the colony to make rapid adjustments in the caste ratios. This would seem a priori to be of special advantage to species that possess a soldier caste subject to occasional heavy mortality. The older the larvae are when shunted to the soldier developmental pathway, the shorter the time required to fill gaps created by casualties. It is consequently of interest that in Pheidole the point of soldier-vs-minor worker determination is in the final instar (Passera, 1974a; D. Wheeler and Nijhout, 1984). It is also true, as stressed by D. Wheeler (1986a), that a wide separation of the queen-worker and major-minor decision points minimizes confusion of the two and a more precise regulation of caste ratios.

The timing of queen-worker bifurcation in fact varies greatly among species. In Myrmica ruginodis it is very late --about a week prior to the cessation of larval growth. But in Formica polyctena the opposite is true: a larva is determined to queen or worker within 72 hours after hatching from the egg (Bier, 1958a,b; Schmidt, 1974a,b; Weaver, 1957). A similarly wide variation in timing has been observed among species of bumblebees (Röseler, 1974; Röseler et al., 1984). The full range of timing in decision points is illustrated in Figure 8-53. The relation of these differences to the ecology of the species awaits investigation.

Queen determination in Myrmica. The most thoroughly studied system of caste determination in ants is that of the genus Myrmica, pursued by M. V. Brian and his associates in England through various ramifications from 1950 to the present time. The species employed has been primarily Myrmica ruginodis, with supplementary studies made on Myrmica rubra. Both species are polygynous. In the earlier work Brian sometimes referred to Myrmica rubra as Myrmica laevinodis and to Myrmica ruginodis variously as Myrmica rubra, Myrmica rubra macrogyna, or just Myrmica. Fortunately, the two species appear to be similar in key aspects of caste determination.

Myrmica ruginodis and Myrmica rubra, like most north temperate myrmicines, overwinter with their brood in the larval stage. In most mature colonies the following spring, some of the female larvae develop into queens and others into workers. The annual cycle is illustrated in simplified form in Figure 8-54. As the “slow” larvae, hatched in early summer, reach a certain point in their development in the late summer or fall, they become dormant. This crucial juncture is in the terminal (third) instar. More precisely, the larvae are halted at a stage of early pupal development when the brain has migrated halfway back into the larval prothorax. Many larvae do not proceed beyond this 0.5 point until the following spring, a condition referred to as “primary diapause.” Others are able to go on until the brain is 0.8 into the prothorax, but they always halt there in a condition of “secondary diapause.”  Still other, younger larvae do not reach the 0.5 stage at all before cold weather; these pass the winter in a nondiapause state. In any case, all the brood overwinters in the larval stage, and it is from some of these vernalized larvae that the yearly crop of queens is matured in the spring.

Diapause, as pointed out by Brian (1956b), has the effect of holding over until winter all of the late brood in the larval stage. The dormant state can be broken by sustained high temperatures and handling by spring workers. Both of course are normally encountered by larvae in wild colonies in the spring. The important point is that chilling in winter temperatures confers on the larvae the capacity to sustain a high growth rate in critical periods of the final instar and, as a result, to metamorphose into queens. Dormancy itself is not a prerequisite. Some small larvae, as already mentioned, are immobilized by winter cold before they enter stage 0.5 dormancy; yet the following spring they too have the capacity to transform into queens.

Final caste determination occurs in the spring larvae late in their terminal instar. The critical periods were revealed by experiments in which sets of larvae were starved at different periods in pupal development. Of course, larvae are not normally steered into one path of development or another by any such regimen of sudden starvation or overfeeding. In order to learn about the natural course of determination, it was necessary to follow the development of many larvae being reared individually under relatively undisturbed conditions. When this was done, a clearer picture of the role of the growth rates was obtained. Brian learned first that the time required for spring larvae to develop into worker pupae did not differ from the time required for development into queen pupae. In both cases duration of development ranged from 9 to 21 days and averaged about 13 days. But the final weight attained by larvae transforming into the two castes differed greatly, averaging about 4.5 mg in the worker and 8 mg in the queen. Clearly the larvae destined to be queens either must grow at a faster rate or else start at a higher weight. And either can be the case. In general, queens come from spring larvae that either start relatively large and maintain a moderate to high growth rate or else start at a medium weight and maintain a consistently high growth rate. For the most part larvae that have failed to reach a weight of about 3.5 mg by stage 0.8 are destined to become workers.

An important finding from Brian's studies is that caste determination in Myrmica ruginodis is essentially worker determination, or the failure of this event to occur. This means that the worker is an individual diverted from a normal female (that is, queen) course of development by having part of its adult system shut down. For convenience, the imaginal discs of a larva can be divided into a dorsal set, containing wing buds, gonad rudiments, and ocellar buds, and a ventral set, containing leg buds, mouthpart buds, and central nervous system (Brian, 1957a, 1965a). In the case of queen development leading to a more or less typical hymenopteran female, both sets maintain full growth and development at the onset of pupal development. But in the case of larvae destined to become workers, the dorsal set stops growth and development for the most part, and only the ventral set continues on. The abruptness of the shutdown is quite striking. It is interesting that the dorsal organs do not always shut down together. One of the anomalous “castes” described by W. M. Wheeler (1905) was the pterergates, or otherwise normal workers bearing external vestiges of wings. One worker of Myrmica scabrinodis figured by Wheeler had wings as long as the thorax itself.

Growth studies, including the starvation experiments, disclosed the important role of chilling in the development of queen potential in larvae, and of nutrition in leading Myrmica larvae to growth beyond the thresholds required to sustain development as queens. Subsequent experimentation has revealed the existence of at least four additional factors influencing caste determination in the closely related species Myrmica rubra (Brian, 1963, 1979a, 1983). First, an increase in temperature from 22°C to 24°C, that is, from the optimal temperature for larval survival to slightly above, results in an eightfold increase in the proportion of larvae metamorphosing into workers. Second, the presence of nest queens results in a fourfold increase in the proportion of workers. The latter effect is caused at least in part by a change in the behavior of the workers evoked by the perception of queen pheromones. When two sizes of larvae are presented to adult workers in the presence of a queen, the small larvae are fed more and the larger larvae less than in the case of control groups lacking a queen. Also, the larger larvae are bitten and licked more in the presence of the queen, presumably lessening their chances of survival (Brian and Hibble, 1963; Brian, 1983; Elmes and Wardlaw, 1983). Similar changes are obtained when dead queens are presented daily to queenless colonies (Carr, 1962). Workers make sharper discriminations if they are exposed to queens during the first several weeks after they emerge as adults than if they are kept away from this caste (Brian and Evesham, 1982; Evesham, 1984b). The presence of a primer pheromone is suggested by the fact that when the sterol fraction of extracts of the heads of queens are fed to larvae or applied topically, larval growth is inhibited (Brian and Blum, 1969).

A third influence is blastogenic. Small eggs laid during periods of most rapid oviposition yield higher numbers of workers. When queens are allowed to emerge from hibernation at 20°C, the rate of oviposition rises during the first three weeks to a maximum that persists through the following three to four weeks. Then the rate gradually declines toward zero, reaching a very low level after about the sixteenth week. Simultaneously the size of the eggs changes, declining rapidly in the first three weeks and then remaining approximately constant. Eggs laid during the first three weeks show a greater capacity for transforming to queens than do eggs laid after three weeks when both kinds are cultured under identical conditions.

Finally, a most interesting additional blastogenic effect discovered by Brian and Hibble (1964) is that eggs laid by different queens differ markedly in their tendency to produce queens or workers. Most of the variation appears to be the result of age; younger queens have a higher tendency to lay worker-biased eggs. This effect was foreshadowed by the results of early experiments on Pheidole pallidula by Goetsch (1937), who introduced eggs from colony-founding queens into larger colonies in which normal-sized workers were being reared. Despite the richer environment, dwarf-sized minor workers typical of first brood were obtained. If confirmed, this appears to be the first demonstration of some form of ovarian, or truly blastogenic influence on the determination of worker subcastes.

To summarize, there are at least six factors operating that determine whether a Myrmica female will become a worker or a queen: larval nutrition, winter chilling, posthibernation temperature, queen influence, egg size, and queen age. The next question should logically be, what is their relative importance in nature? The clearest way to view the entire caste-determining system is to regard it metaphorically as a series of checkpoints arranged more or less in sequence. An egg “aspires” to develop into an adult queen. This ambition is “approved” by the colony providing the following two checkpoints are passed. First, has the larva been through diapause and chilled to resume full development? Second, has the larva reached the requisite size by the onset of adult development in the final larval instar? In addition, are the mother queens nearby and potent, and is the colony young? If so, borderline cases are more likely to fail the queen test and be consigned to workerhood. Taken together, the caste-biasing factors make it more likely that the Myrmica colony will produce new queens in the spring and also when it is large and robust--the conditions under which it can most profitably invest in reproduction.

Queen determination in Formica. Work on the Formica rufa group, paralleling in many respects that on Myrmica, was published by Gösswald and Bier (1953-1957) and Schmidt (1974a). As in Myrmica, there was originally much confusion in the taxonomy of this difficult species group, but it has been largely straightened out by Yarrow (1955), Lange (1958), and Betrem (1960). Two species were used in the Gösswald-Bier study: Formica nigricans (= rufa pratensis) and Formica polyctena (= rufa rufopratensis minor) . The following account applies primarily to Formica polyctena.

These formicine ants hibernate without brood. In the spring, when the nest temperature rises to 13°C or above, the queens migrate to the warmest part of the nest near the surface, lay batches of eggs (referred to as the “winter” eggs), and afterward retreat to cooler parts of the nest. Eggs laid at temperatures under 19.5°C remain unfertilized, and as a consequence those first produced in the spring, when the nest temperatures are between 13° and 19.5°C, are male. Also, the smaller the colony, the poorer its thermoregulation, and, hence, the higher the proportion of males produced (Gösswald and Bier, 1955). Later eggs in this first “winter” batch are fertilized and capable of producing either queens or workers. Eggs laid still later--the “summer” eggs--are capable of producing only workers.

The winter and summer eggs differ strikingly from each other in several ways. Viewed in ovarian preparations, the winter eggs contain more RNA around their nuclear membranes and have a much more voluminous polar plasm than is the case in summer eggs; in addition, the nurse cells have larger nuclei (Bier, 1953, 1954a). It has been postulated, quite reasonably, that these cytological differences are in some way intimately connected with the later biasing of larval growth, but the relation has not been proved experimentally.

It is clear in any case that chilling of the eggs in the ovaries makes them bipotent with respect to caste. Final determination, however, occurs during about the first 72 hours of larval life. This was proven in experiments in which Gösswald and Bier (1957 and contained references) introduced eggs of Formica polyctena into colonies of Formica nigricans in order to permit the tracing of individual development. When 30 or more host nigricans workers, deprived of their own queen, were given small numbers of polyctena eggs, they reared queens. Groups of less than 30 workers reared workers. At 27°C the young larvae remained plastic for 72 hours.

Two other factors were discovered that match those in Myrmica. The presence of Formica nigricans queens inhibits development of the Formica polyctena eggs into queens. If a queen is present with a large group of workers, the winter eggs transform into either queens or workers, but, if no queen is present, the eggs always transform into queens. Part of the difference is due to the propensity of workers to direct the flow of food to queens differentially. Workers just emerged from hibernation, as well as young workers, increase the tendency of larvae hatched from winter eggs to develop into queens.

In summary, although caste determination in Formica differs from that in Myrmica in several important details, there is a close resemblance in general pattern. Multiple controls exist; most of the six factors of Myrmica also occur in Formica. A close interplay of responses to hibernation and nutrition characterizes both. In Formica, as in Myrmica, the relative weight and precise degree of interaction of the factors under natural conditions are still unknown.

Queen determination in other ants. The queen-determining systems of other ant species have so far proved to be variations on the themes exemplified by Myrmica and Formica. The differences discovered are in the timing of the determination during development of the egg and larva, and the relative importance of the six environmental factors.

The formicine Plagiolepis pygmaea resembles Myrmica ruginodis in the key events of queen determination. The spring larvae that have overwintered in the final instar are bipotent. If they are exposed to inhibitory queen pheromones during the first nine days after diapause is broken, they develop into workers. When the queen is absent or less potent, the better nourished larvae become queens while those less well fed turn into workers. The queen secretes the pheromone primarily or entirely from her head (Passera, 1974b, 1980b, 1984; Suzzoni et al., 1980). The Myrmica pattern is also repeated in Leptothorax nylanderi. Larvae that overwinter and emerge from diapause are bipotent; in the absence of the queen and with sufficient nourishment, they develop into queens. The critical period, like that of Myrmica and Plagiolepis, is in the third and final instar (Plateaux, 1971).

A more Formica-like determination has been recorded in the Mediterranean myrmicine Pheidole pallidula. Bias toward queen development originates during the egg stage. Eggs laid by fat, heavy queens after hibernation are strongly prone to yield queens (Passera, 1980b, 1984).

Monomorium pharaonis, a cosmopolitan house ant of tropical Old World origin, has the most aberrant queen-determining system yet discovered in ants. The colonies are unusual in being potentially immortal; as old queens die, they are replaced by young queens that mate in the nest. The typical longevity of the queens is no more than 200 days, the shortest known in the Formicidae as a whole. As Petersen-Braun (1975, 1977) demonstrated, inhibition by the queen pheromone is decisive, and the brevity of queen life regulates bursts of new queens at intervals of three to four months due to what Petersen-Braun has called a state of “physiological queenlessness.” Queens pass through three periods in their 28-week life span:  a juvenile phase which lasts about a month, a fertile phase of two to three months, and a final senile phase. During both the juvenile and senile periods, the queen lays eggs destined to become workers regardless of larval nutrition. During the lengthy intervening fertile phase, as the queens approach senescence, their eggs are bipotent. If the larvae hatched from these eggs are given sufficient food they become queens, and if not, they turn into workers. As a consequence, a crop of new queens arrives about the time the old queens die off. In the intermediate period, spanning 4-5 developmental periods from egg to adult, an “avalanche” of new workers is eclosed. These ants had been initiated during the juxtaposed senile and juvenile periods of the two queen generations. The inhibitory pheromones are secretions known to be picked up from the cuticle of the queens by the workers, but their ultimate glandular source remains unknown (Berndt and Nitschmann, 1979).

Still incomplete information indicates that queen inhibition is a very widespread if not universal phenomenon, at least in ant species possessing a queen caste in the first place (Vargo and Fletcher, 1986a,b, 1987). In the ponerine Odontomachus “haematodes” troglodytes (= O. “haematodes”) larvae are sensitive to queens during their final one or two instars. Once determination has occurred by this means, it does not appear to be reversible by changes in nutrition (Colombel, 1978). In colonies of the pseudomyrmecine Tetraponera anthracina, the sensitive period is very late in larval life. Worker-destined larvae commence metamorphosis soon after this point, but those determined as queens continue growing for a while (Terron, 1977).

Soldier determination in Pheidole. The mode of regulation of ratios of minor and major workers is now relatively well understood in the genus Pheidole. When Passera (1974a) altered the percentages of Pheidole pallidula majors (or “soldiers”) from the usual 3-5 percent, most returned to this level within 75 days through the differential production of minors and majors. Even colonies containing 50 or 100 percent majors at the outset dropped to an average level of 20-25 percent and were still declining at the end of the experiment. A similar regulation has been demonstrated in Pheidole morrisii by Gregg (1942), P. vinelandica by D. Wheeler and Nijhout (1981, 1984), and Pheidole dentata by Johnston and Wilson (1985). These results show that major/minor ratios are not set entirely by an automatic shunting of larvae at the critical developmental time consistent with the model of simple determination depicted in Figure 8-56. Feedback controls have been added that can temporarily increase or decrease the proportion directed to development as majors.

The nature of the controls was elucidated in Pheidole vinelandica by D. E. Wheeler and H. F. Nijhout (1981, 1983, 1984). Worker larvae have four instars. They molt to the final instar when they are about 0.6 mm in length, and soldier determination takes place when the larvae grow to 0.9-1.2 mm in length. Minor worker larvae usually begin metamorphosis at 1.3 mm, while major worker larvae continue to grow beyond that point. Methoprene, a juvenile hormone analog, delays metamorphosis (as expected) and yields an abnormally large crop of soldiers. Conversely, the presence of an excess number of soldiers reduces the production of new soldiers, an effect evidently due to a contact inhibitory pheromone. The model suggested by Wheeler and Nijhout to explain these results incorporates a feedback loop from the colony to the individual. When the colony is growing and foraging workers bring in sufficient food, individual larvae are well fed, and as a consequence their juvenile hormone titer is elevated in some fashion, predisposing them toward soldier determination in the final instar. However, as the proportion of soldiers in the colony rises, so does the amount of inhibitory pheromone to which the still-growing larvae are exposed. The inhibitory pheromone has the effect of making the larvae less sensitive to juvenile hormone, so even though many have an elevated titer of the hormone, it is no longer sufficient to trigger soldier development--and the larva transforms into a minor worker. In essence, a balance is struck between the proportion of soldiers in the colony and the sensitivity of the developing larvae so that the proportion of soldiers remains more or less constant.

We noted earlier that each species of Pheidole has a characteristic norm in its major/minor ratio, and that at least in the case of Pheidole dentata, differences among colonies of the same species appear to have a partial genetic basis. The Wheeler-Nijhout model explains how much variation can arise during evolution. All that would be required are mutations affecting production of the juvenile hormone, production of the soldier inhibitor pheromone, or sensitivity to either of these substances.