Caste differentiation in social insects is a representative case of polyphenism, in which multiple discrete phenotypes are seen in a single species (Nijhout, 1999, 2003). Castes of termites are primarily divided into two types of individuals: fertile (reproductives) and sterile individuals (neuters) (Thorne, 1996; Roisin, 2000). The sterile castes include workers, presoldiers and soldiers, while the fertile ones are alates, primary reproductives and secondary reproductives (Figure 1). Alates with imaginal characteristics (i.e., wings, testes, ovaries) that develop during the nymphal stages perform nuptial flights, shed their wings, mate, found new colonies, and then become primary reproductives (i.e., kings and queens). When primary reproductives die or become scenescent, or when colonies expand, some individuals differentiate into the secondary reproductives.

In contrast, soldiers and workers are sterile castes not engaged in reproduction. Late-stage larvae that behave as workers are called as “pseudergates” in the relatively basal termite lineages (Thorne, 1996). In this review, we will expediently use the term “worker” including pseudergate sensu lato (Korb and Hartfelder, 2008; Roisin and Korb, 2011). A part of workers differentiate into soldiers with drastic morphological modifications. A soldier possesses a pair of extremely elongated mandibles and/or a frontal projection with an exocrine gland from which defensive substances are discharged (Noirot, 1969; Weesner, 1969).

Depending on the species, the developmental pathways are roughly divided into two patterns, i.e., a linear pathway or a bifurcated pathway (Roisin, 2000; Bourguignon et al., 2009; Figure 1). Basically, all larvae hatched from eggs have the potential to develop into any castes. Classically, assuming that all egg contains the similar or identical genetic information, caste developmental fates are determined depending on various environmental factors experienced during postembryonic development, as well as sex and egg factors (Noirot, 1990, 1991). Major environmental stimuli include nutritional and seasonal factors such as temperature and humidity (Noirot, 1991; Fei and Henderson, 2002; Liu et al., 2005; Scharf et al., 2007). Termites can receive these stimuli by themselves, and in some cases, this environmental information can be transferred through social interactions such as allogrooming and/or trophallaxis, the latter of which is the exchange of nutritious food by stomodeal (i.e., oral) or proctodeal (i.e., anal) transfer (Noirot, 1991; Tarver et al., 2011). However, little is known about the relationships between environmental factors and patterns of social interactions (e.g., Machida et al., 2001).

Except for environmental stimuli, sexes are also important factors for the caste differentiation (Noirot, 1990). Termites show the XY sex determination system (Matsuura, 2011), and neuter castes are specialized in or biased toward either sex in some species (Roisin, 2000). For instance, in the genus Nasutitermes only male workers differentiated into soldiers (Hojo et al., 2004). In addition, a candidate X-chromosome-linked gene, worker, is reported to affect the caste differentiation in Reticulitermes speratus (Hayashi et al., 2007). The worker is hypothesized to have two alleles (wk^A and wk^B). They reported that male (XY) larvae possessing wk^A were destined to be workers, while in female (XX), heterozygous in worker locus determined the differentiation into workers. Larvae with other genotypes basically develop into nymphs although only the genotype for females with homozygous wk^B is thought to be lethal. This is the first report showing a genetic influence in reproductive division of labor (Kitade et al., 2011). However, it should also be noted that the caste fates are also affected by coexisting castes (Hayashi et al., 2007), suggesting that, except for environmental cues, social interaction among castes (or phenotypes) is the major factor affecting the caste composition in termites. The abnormal caste ratios set by artificial manipulations were gradually returned to the normal caste ratios just by keeping those colonies under laboratory conditions (Lefeuve and Bordereau, 1984; Mao et al., 2005; Park and Raina, 2005). These experiments were performed under laboratory conditions where environmental conditions were stable, so that the changes of social interactions are suggested to be the important factors adjusting the caste ratio.

Historically, studies on the caste ratio regulations in termites started with the reports on seasonal transitions of caste composition in some species, such as Odontotermes redemanni (Banerjee, 1966) and Trinervitermes ebenerianus (Sands, 1965). Howard and Haverty (1981) found that, in Reticulitermes flavipes, the caste proportion in a colony changed with seasons. They reported that the soldier ratios increased concurrently with temperature and peaked when alate differentiation was about to occur. Furthermore, the ratios of soldiers and reproductives were kept at a low level (2 and 0.1%, respectively, on average during a year). Generally, as reviewed by Haverty (1977) on 112 termite species, the proportion of soldiers in a colony is low in all the investigated species.

In social hymenopterans, the low proportion of reproductive castes was often discussed in terms of conflicts among colony members (Bourke and Franks, 1961). Conflict among reproductives is also suggested to exist in termite sociality since the artificial addition of extra reproductives resulted in cannibalism among reproductives (Lüscher, 1961). In contrast, the adaptive meaning of low soldier proportion has been discussed based on the results of experimental manipulations. Basically, a colony needs soldiers because soldier is the only caste that can intercept attacks by enemies (mainly predators) (Noirot and Darlington, 2000). However, the possession of soldiers in colonies should incur some costs, since soldiers cannot feed by themselves but require help from workers (Noirot and Darlington, 2000; Korb and Hartfelder, 2008). Actually, an excess of soldiers caused high mortality of nestmates in Coptotermes formosanus (Haverty, 1979). Thus, maintaining the appropriate proportion of soldiers is necessary for termite colonies. However, only a few studies on the adaptive significance of caste proportion have been reported in termites.

The most famous discoveries on the regulatory mechanisms of reproductive caste differentiation in termites were a series of studies on the differentiation of secondary reproductives of Kalotermes flavicollis performed by Lüscher (1952, 1961) in the middle of 20th century. He clarified the developmental pathway of the focal species and the fact that workers had the potential to differentiate into both secondary reproductives and soldiers. Moreover, his subsequent studies revealed that workers differentiated into secondary reproductives or soldiers after the removal of primary reproductives or soldiers, respectively, from the colony (Lüscher, 1960, 1961, 1974). These results suggested that the caste differentiation was repressively regulated by other colony members (castes) probably through pheromones. Furthermore, the differentiation of secondary reproductives was not inhibited when the anus of primary reproductives was sealed, indicating that some substances were transferred through proctodeal trophallaxis (Lüscher, 1955, 1974). In K. flavicollis, fewer alates were differentiated when nymphs were isolated with the royal pair than in groups of orphan nymphs (Springhetti, 1969, 1971, 1972). This inhibitory effect occurred only when the king and queen had directly contacted with nymphs, suggesting that the direct communications such as stomodeal and/or proctodeal trophallaxis, or grooming would be responsible for it.

In contrast to the inhibitory effects of primary reproductives, the alate production was shown to be accelerated by the presence of soldiers in Zootermopsis nevadensis (Lüscher, 1975). Similarly, the differentiation of secondary reproductives of K. flavicollis was accelerated by soldiers (Springhetti, 1969). Both cases indicate that the differentiation to reproductives seemed to be induced by the presence of soldiers (Figure 2).

The inhibition of soldier differentiation by soldiers was first reported in incipient colonies of Z. angusticollis. Most of the incipient colonies possessed only one soldier for a long time. Castle (1934) discovered that, when a soldier was introduced into another incipient colony before the first soldier was differentiated, soldier differentiation was suppressed. When the soldier was removed from an incipient colony, another workers started to differentiate into a new soldiers. By using individuals collected from mature colonies, the inhibitory effects of soldiers on the additional soldier differentiation were demonstrated in K. flavicollis (Figure 2) (Springhetti, 1969, 1985). Lefeuve and Bordereau (1984) firstly reported the chemical characteristics of soldier pheromone using Nasutitermes lujae. They revealed that not only living soldiers but also chemical components extracted from soldiers using organic solvent inhibited the soldier differentiation. These inhibitory effects did not occur unless workers directly contacted with the soldier extracts, suggesting that the pheromone suppressing additional soldier differentiation was a non-volatile, non-polar substance. Furthermore, Okot-Kotber et al. (1991) reported that soldier extracts of Reticulitermes flavipes inhibited soldier differentiation in termite species belonging to a different family, suggesting that the inhibitory mechanisms are shared among termite lineages.

On the other hand, in N. lujae and Cubitermes fungifaber, the soldier differentiation occurred more frequently in the groups of workers reared with primary reproductives than those without reproductives (Bordereau and Han, 1986). Similar effects of reproductives on the additional soldier differentiation were found in K. flavicollis (Springhetti, 1969, 1970) and Z. nevadensis (Maekawa et al., 2012). Generally in termites, the soldier differentiation is stimulated by the presence of primary reproductives and suppressed by other soldiers (Figure 2). The presence of secondary reproductives is thought to stimulate the soldier differentiation as well as primary reproductives, although the direct evidences have not been reported. In general, the caste differentiation appears to be inhibited by the same caste while induced by a different caste, probably through pheromone actions (Bordereau, 1985; Noirot, 1991).

Early studies demonstrated that the transplantation of corpora allata (CA; juvenile hormone-producing endocrine glands) induced caste differentiation of recipient workers in K. flavicollis (Lüscher, 1952, 1960). CA from soldiers, reproductives and late-stage nymphs induced the soldier differentiation, whereas CA from workers and early-stage nymphs induced the differentiation into reproductives. These results suggest that juvenile hormone (JH) plays important roles in the caste differentiation, and the JH synthesis activity in CA differs among castes. In addition to the CA transplantation, the applications of ectopic JH analogs (JHAs) to workers also induced the soldier differentiation in many termite species, indicating that an increase in JH titer stimulated differentiation into soldiers (Howard and Haverty, 1979). Thus, JH has been recognized as the most important endocrine factor for caste differentiation, and a model of caste differentiation by JH action was proposed, in which high JH titer induces soldier differentiation whereas low JH titer causes differentiation into alates (Nijhout and Wheeler, 1982; Figure 3).

Although classical studies suggested that the patterns of JH-titer transitions during intermolt determined the future caste differentiation, precise measurements of JH titer were not performed until the establishment of JH quantification methods (Zera, 2007). So far, three major methods for JH quantification have been reported: radioimmunoassay (Greenberg and Tobe, 1985; Goodman et al., 1990), gas chromatography-mass spectrometry (GC-MS) (Shu et al., 1997), liquid chromatography-mass spectrometry (LC-MS) (Westerlund and Hoffmann, 2004). The JH quantification by GC-MS was developed earlier than those using LC-MS. For the JH quantification to be a routine lab procedure, however, GC-MS analysis has some disadvantages due to the requirement of many samples and mass consumption of harmful organic solvents for sample preparation (Bergot et al., 1981; Rembold and Lackner, 1985). Recently, a rapid and accurate method for JH quantification by LC-MS was developed to quantify insect JHs (e.g., JHI, II, and III) and their metabolites (Westerlund and Hoffmann, 2004). By applying this method, Cornette et al. (2008) and Gotoh et al. (2008) established the quantification of JH titers in a termite Hodotermopsis sjostedti. Cornette et al. (2008) further analyzed the patterns of JH titers during differentiation into reproductives and soldiers in H. sjostedti, indicating that JH titers of pseudergates prior to soldier differentiation were constantly high, whereas, those of nymphs before molts into alates remained low (Figure 3). These two JH transitions leading to the soldier and alate differentiations are consistent with the patterns proposed by Nijhout and Wheeler (1982), although there are minor differences in the patterns leading to the stationary molt (repeating the same instar) to pseudergates (Figure 3). It was also demonstrated that JH titers of nymphs before alate differentiation were low in Cryptotermes secundus (Korb et al., 2009a) and Reticulitermes speratus (Maekawa et al., 2010).

As for the regulations of the hemolymphatic JH titer in insects, it is known that JH titer is not only regulated by the activity of JH synthesis in CA and secretion from CA, but also by other biological processes such as sequestration and degradation (Gilbert et al., 2000; Nation, 2008). Actually, the transcription levels of hexamerins (a factor related to JH sequestration) were shown to be up-regulated during soldier differentiation (Scharf et al., 2005, 2007; Cornette et al., 2013), so that the sequestration process is suggested to modify JH titer during the caste differentiation.

Based on the fact that JH is the most important factor involved in the caste differentiation in termites, it was considered that the JH actions mediate the effects of interactions among colony members on the caste differentiation (Henderson, 1998). By applying the technique of GC-MS analysis in Coptotermes formosanus, it was shown that the isolation of workers from soldiers increased the worker JH titers, suggesting that the presence of soldiers may repress the JH titer of nestmates (Mao et al., 2005; Park and Raina, 2005). Since these studies did not employ the application of JH or JH analogs, the observed effects were suggested to occur under natural conditions.

The JH quantification method by LC-MS that was performed in H. sjostedti (Cornette et al., 2003) was also applied to analyze precise effects of the soldier presence on worker JH titers during the soldier differentiation in R. speratus (Watanabe et al., 2011; Figure 4). The results of this study in which JH titers of JH-treated workers reared with/without soldiers were quantified (Figure 4A), revealed that, under the absence of soldiers, the JH application rapidly increased the worker JH titers (Figure 4B), resulting in the elevation of presoldier differentiation ratio from workers (Figure 4C). On the other hand, under the presence of soldiers, the rapid increase in worker JH titers by JH application was suppressed (Figure 4B), leading to the low induction rates (Figure 4C). These results suggested that the soldier differentiation triggered by the JH application was suppressed by the soldier presence, probably by lowering the worker JH titer. This suppression by the soldier presence occurred very rapidly (within a few days), suggesting the rapid and accurate regulation of soldier-caste ratio in termites.

Extremely elongated mandibles used to bite or flick enemies (mechanical defense), or frontal glands discharging aversive and/or toxic substances (chemical defense) toward enemies (Deligne et al., 1981; Prestwich, 1984) were formed through the distinctive morphogenetic events during the process of soldier differentiation (Noirot, 1969; Weesner, 1969). This process can be induced artificially by using JH or JHA, which initiates the soldier differentiation (Miura et al., 2003; Šobotník et al., 2010).

In order to examine the relationship between JH titer and mandibular length, various concentrations of JH were applied to workers in Reticulitermes speratus, in which soldiers possess elongated mandibles (Tsuchiya et al., 2008; Figures 5A,B). The results showed that mandibular lengths of induced presoldiers depended on the concentrations of applied JH. Namely, higher concentrations of JH induced presoldiers with longer mandibles while lower concentrations induced presoldiers with shorter mandibles. Interestingly, the resultant mandibular lengths were also affected by the presence of soldiers even under the same JH concentration; workers coexisted with soldiers molted to presoldiers with shorter mandibles while those without soldiers molted to presoldiers with longer mandibles. These results strongly suggest that the presence of soldiers affects the developmental processes in the soldier differentiation by controlling JH titer (Tsuchiya et al., 2008). Actually, a subsequent work quantifying the endogenous JH titer in the same species supported this idea (Watanabe et al., 2011).

In some species of rhinotermitids, the frontal gland seen in soldiers (Figure 5C) is also formed during the soldier differentiation induced by JH applications (Lelis and Everaerts, 1993; Watanabe and Maekawa, 2008; Šobotník et al., 2010). In R. speratus, Watanabe and Maekawa (2012) showed that the degree of frontal-gland formation depended on the concentrations of applied JH; higher JH concentration induced well-developed glands whereas lower one did not. It was also suggested that the soldier presence suppressed the frontal-gland formation via the regulation of JH titer (Figure 5D). These studies focusing on the development of soldier characters indicated that interactions among individuals did not only affect the caste differentiation but also the caste-specific morphogenesis. In these cases, JH titers in workers might be suppressed to lower levels by the presence of soldiers (Figure 5D). Since the social interactions are suggested to affect the expression patterns of genes that were shown to regulate the soldier-specific morphogenesis (Koshikawa et al., 2002, 2003, 2005, 2010; Hattori et al., 2013; Toga et al., 2012, 2013), the relationships between those gene expression patterns and social interactions will be clarified sooner or later.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.