Termite soldier mandibles are elongated by dachshund under hormonal and Hox gene controls

Although the morphological design of organisms is determined by genetic information, morphologies can also be altered by environmental stimuli, known as phenotypic plasticity or polyphenism (Nijhout, 2003). Insects are a major group of arthropods and share a basic body plan in serially homologous body segments; however, the allometry or proportions of body parts is diversified, particularly the size of appendages, including the mouthparts (Angelini et al., 2012a; Boxshall, 2004). As mouthpart modifications in insects played important evolutionary roles in adaptations to various environments, they exhibit considerable variation (Grimaldi and Engel, 2005; Koch, 2001; von Lieven, 2000). In some cases, however, the function of mouthparts is not related to feeding. Enlargement of mandibles is a striking example, as seen in male stag beetles fighting over mates (Emlen, 2008) and in termite soldiers defending their colonies against predators (Deligne et al., 1981).

In termites, the soldier caste has a unique morphology and function, and is a synapomorphy specific to the termite lineage, as it is present in all termite species with a few exceptions (Noirot, 1969). Soldiers can be differentiated from workers (or pseudergates, i.e. false workers or working larvae) through two molting events: the first molt leads to the presoldier stage; the second to the soldier stage (Fig. 1A; Roisin, 2000). During the differentiation, the anterior body parts are largely modified through solider-specific morphogenesis (Koshikawa et al., 2002; Miura and Matsumoto, 2000; Watanabe and Maekawa, 2008). Although soldier morphology is diversified among termite species, the biting type is thought to be the most primitive (Prestwich, 1984). In these species, soldiers possess elongated and heavily sclerotized mandibles (Koshikawa et al., 2002; Weesner, 1969).

The differentiation of termites into soldiers is regulated by a key endocrine factor: juvenile hormone (JH) (Cornette et al., 2008; Miura, 2019; Nijhout and Wheeler, 1982). In many termite species, the application of JH or its analogue (JHA) to workers induces differentiation into soldiers and morphological changes (Fig. 1C; Howard and Haverty, 1979). Although the mechanisms that control the effects of JH are unclear, knockdown of the JH receptor gene Methoprene-tolerant (Met) or hexamerin genes inhibit soldier differentiation (Masuoka et al., 2015; Zhou et al., 2006). Therefore, the downstream gene network upregulated by JH likely plays an important role in soldier-specific morphogenesis (Cornette et al., 2013; Miura, 2001, 2005; Miura and Scharf, 2011). The insulin/insulin-like growth factor signaling (IIS) pathway is also upregulated in mandibles before the presoldier molt and is required for soldier morphogenesis (Hattori et al., 2013).

The detailed morphogenetic processes of soldier differentiation have been studied in the damp-wood termite Hodotermopsis sjostedti (Archotermopsidae), a basal termite species (Koshikawa et al., 2002; Koshikawa et al., 2003; Fig. 1A). In soldier differentiation, the apical region of the mandible is more elongated than the proximal region (Fig. 1B; Koshikawa et al., 2002). Prior to the molt into presoldiers (PS), when mandibles are dramatically elongated, new folded cuticles and epidermal layers grow under the mandibular cuticle of pseudergate (PE) and expand at the presoldier molt (Koshikawa et al., 2003; Sugime et al., 2015). Therefore, epithelial cells of mandibles are suggested to proliferate before the PS molt. However, the mechanisms by which endocrine signals regulate mandible-specific allometric development are unclear, although several molecular studies on downstream factors have been reported (e.g. Ishikawa et al., 2010; Koshikawa et al., 2005, 2010; Miura et al., 1999; Toga et al., 2013). Thus, because the mandible grows allometrically during soldier differentiation, we hypothesized that patterning factors provide spatial information to the mandibular epithelia.

Here, we have investigated the expression patterns of morphogenetic regulatory genes that are widely conserved and play important roles in body-part identity and the proximodistal patterning of insect appendages (Carroll, 1995; Kojima, 2004). In this study, using H. sjostedti, the expression patterns of 18 candidate genes were analyzed in mandibles during soldier differentiation. These included Hox genes, which provide segment identity [labial (lab), proboscipedia (pb), Deformed (Dfd), Sex combs reduced (Scr) and Antennapedia (Antp)]; segment polarity genes, which regulate the anteroposterior polarity of each segment [decapentaplegic (dpp), wingless (wg), engrailed (en) and hedgehog (hh)]; and appendage-patterning genes, which determine the proximodistal patterns and joint formation in appendages [Distal-less (Dll), dachshund (dac), homothorax (hth), extradenticle (exd), aristaless (al), Serrate (Ser), spitz (spi), paired (prd), and cap'n’collar (cnc)] (Angelini et al., 2012b; Coulcher and Telford, 2012; Kojima, 2004; Table S1). The temporal and spatial expression patterns were analyzed at seven developmental stages (Fig. 1C) and in five appendages. Results showed that dachshund (dac), which identifies the intermediate domain along the proximodistal appendage axis, was upregulated in the mandible-specific manner, and functional analyses using RNA interference (RNAi) showed that dac was required for the mandibular elongation. Furthermore, to clarify the epistatic relationships among differentially expressed toolkit genes (dac), hormone-receptor genes (Met and InR) and Dfd, the expression levels of these factors were analyzed after RNAi knockdown of each.

For expression analyses, sequences of toolkit genes were obtained by RNA sequencing of transcriptomes from eggs, alates, soldiers and neotenics in the focal termite species (see Materials and Methods for details). To screen for genes that were highly and specifically expressed in mandibles before the presoldier molt, the expression levels of target genes in mandibles were first compared at different developmental stages (Fig. 1C): pseudergate (PE), 1 week after JHA application (JHA1w), 2 weeks after JHA application (JHA2w), JHA-induced presoldier (JHAPS), natural presoldier (NPS), soldier (Sol) and stationary molt (SM). In the qPCR examinations (Fig. 2), genes that showed significantly higher expression during the course of the soldier differentiation (JHA1w and JHA2w) compared with pseudergate (PE) and stationary molt (SM) were chosen as ‘primary candidates’. Among the Hox genes (lab, pb, Dfd, Scr, and Antp), pb and Antp were more highly expressed in JHA1w than in the other developmental stages (Fig. 2, Tukey's test; P<0.05). By contrast, lab, Dfd and Scr were not highly expressed in JHA1w and JHA2w. Among the segment polarity genes (hh, dpp, wg and en), hh was more highly expressed in JHA1w and JHA2w than in PE, but was not significantly different from that in SM. The expression of dpp showed no significant difference between JHA1w and PE, but was lower in JHA2w. Although the expression levels of wg and en were significantly higher in JHA1w than in PE, that of en did not significantly differ between JHA1w and NPS. Of the leg-patterning genes (Dll, dac, hth, al, exd, prd, spi and Ser), only dac showed significantly higher expression in JHA1w than at the other developmental stages. Dll was not upregulated during the differentiation. hth and al were more highly expressed in JHA1w than in PE, but its expression was not significantly different from that in SM. prd and spi expression did not show significant differences between the different developmental stages, and Ser expression was significantly higher in JHA1w and JHA2w than in PE, and was highest in SM. Therefore, the four genes (pb, Antp, dac and wg) showing the highest upregulation in JHA1w were chosen as the ‘primary candidates’.

In addition to the four genes, six genes (Dfd, en, hh, Dll, Ser and al) were chosen as ‘secondary candidates’, as these genes showed interesting expression patterns (upregulation at JHA1w/2w) and/or are known to possess important functions, but did not show significant differences from either PE or SM.

The expression levels of all the 18 genes at JHA1w were then compared among antennae (Ant), mandibles (Md), maxillae (Mx), labium (Lab) and forelegs (FL) to determine whether they were upregulated in a mandible-specific manner (Fig. 3). Only the expression of dac was significantly (twofold) higher in mandibles than in the other appendages among the primary candidate genes (Fig. 3, Tukey's test; P<0.05). Among the secondary candidate genes, Dfd was expressed specifically in mandibles although at higher levels in maxillae. It was also shown that en was expressed at high levels in mandibles, but the difference was not as large as for dac. Furthermore, prd clearly showed a mandible-specific expression pattern, although significant upregulation was not detected during the soldier differentiation (Fig. 2). Therefore, only dac showed the most distinctive expression patterns (i.e. soldier differentiation-specific and mandible-specific patterns), so further functional analyses focused on dac.

To determine the detailed localization of dac inside a mandible, qRT-PCR and immunological staining were performed. Results of qRT-PCR comparing distal and proximal parts of right/left mandibles, by cutting a mandible into two parts, clearly showed that the expression levels in distal parts were higher (more than two folds) than those in proximal parts in both right and left mandibles (Fig. 4A). Additionally, other genes that are known to provide axis information in insect appendages, i.e. dpp, wg, hh, en, hth, al and Dll, were also examined, of which only en showed significantly higher expression in the distal part (Fig. S2). On the other hand, hth showed higher expression levels in the proximal part, although a significant difference was detected only in left mandibles. The expression of Dll was undetectable.

In addition to the qPCR examinations, immunological staining was also carried out using two different monoclonal antibodies against Drosophila Dac protein. One of the two examined antibodies (mAbdac1-1) clearly showed the localization in the focal termite mandibles. Strong Dac signals were detected in epithelial tissues at the apical tip of mandibles and in the area between the first and second tooth in the distal half of the mandibles (Fig. 4B). In the proximal part, Dac signals were also detected in the marginal area of mandibles (basal teeth area), although the signals were weaker than the apical area. Dac signals were not detected at the outer margin of mandibles.

To elucidate the function of dac in mandibular morphogenesis during soldier differentiation, RNAi was performed to JHA-treated pseudergate (PE) individuals (Fig. 5). dac RNAi resulted in presoldiers (PS) with markedly smaller mandibles than those of normal presoldiers and the GFP control (Fig. 5B,C, Fig. S3). The effects of dac RNAi on morphogenesis of 13 other body parts showed that the mandible length (ML) and the length between the apical and first teeth (AL) were smaller in dac^RNAi PSs than in normal PS (Fig. 5C, Fig. S3, Tukey's test; P<0.05). However, most of the other parameters did not significantly differ between normal and dac^RNAi PS. Moreover, the mesonotum width (MsW) and the metanotum width (MtW) of the dsRNA-injected PS (dac^RNAi, GFP^RNAi) were larger than those of normal PS. In principal component analysis (PCA) using the 13 parameters (Fig. S3), the distinction between PE and PS was represented by PC1; thus, PC1 reflects total body size as well as allometric differences. In addition, the distinction between dac^RNAi PS and control PS (RNAi untreated and GFP^RNAi) was represented by PC2. Thus, the mandibular morphology in dac^RNAi was intermediate between that of PE and PS (Fig. 5B,C, Fig. S3).

Furthermore, geometric morphometric analysis was conducted using 20 landmarks along the outline of the left mandibles in dac^RNAi individuals (Table S4, Fig. S4). In this analysis, Dll^RNAi was also added to confirm whether Dll did not affect the mandibular elongation in the focal termite species. Principal component analysis showed that the shape of dac^RNAi mandibles (PC1: −0.1) reduced the degree of elongation, particularly at the apical part, and showed a highly curved (i.e. round-shaped) outer edge (Fig. S4). Thus, dac-RNAi disrupted the soldier-specific mandibular morphogenesis during the soldier differentiation.

To determine the epistatic relationships among hormonal factors, Hox genes and dac, RNAi knockdown of Met, InR, Dfd and dac was performed, followed by quantification of gene expression levels of those genes (Fig. 6). dac expression was reduced in Dfd^RNAi, Met^RNAi and InR^RNAi presoldiers (Fig. 6A, Tukey's test; P<0.05). Dfd expression was reduced in Met^RNAi and InR^RNAi, but not in dac^RNAi PS (Fig. 6B, Tukey's test). Met and InR expression was, respectively, reduced by InR^RNAi and Met^RNAi, which indicates reciprocal regulation (Fig. 6C,D, Tukey's test; P<0.05). By contrast, dac^RNAi did not affect the expression of Dfd, Met and InR (Fig. 6B-D, Tukey's test).

Our data strongly suggest that dac is a key factor in mandibular elongation during soldier differentiation, and that its expression is regulated by Dfd downstream of JH and insulin signaling. Four genes (i.e. pb, Antp, wg and dac) were significantly upregulated in mandibles at 1 week post-JHA application (JHA1w, Fig. 2). Because the folded mandibular tissue is prepared and almost ready to expand in JHA2w (Koshikawa et al., 2003), the patterning information must be provided by regulatory genes before JHA2w. Therefore, upregulation of the four regulatory genes in JHA1w is plausible if these genes are responsible for the mandibular morphogenesis.

Among the Hox genes that determine the identity of body parts along the anterior-posterior axis, Dfd is known to provide the mandibular identity (Carroll, 1995; Hughes and Kaufman, 2002; Hughes and Kaufman, 2000). In addition, Dfd reportedly contributes to mandibular remodeling during the soldier differentiation (Toga et al., 2013). In H. sjostedti, Dfd expression was upregulated in JHA1w, although it was not significantly different from that in PE (Fig. 2). The comparison of Dfd expressions among appendages showed that Dfd was expressed not only in mandibles but also in maxillae (Fig. 3). This pattern of Dfd expression exactly corresponds, to our knowledge, to that revealed in other insects (Hughes and Kaufman, 2000; Rogers and Kaufman, 1997). Although the identity of maxillae also requires expression of pb and/or Scr, Dfd seems to be the only required factor for the mandibular identity, which is also the case in the termite soldier differentiation.

By contrast, pb and Antp were highly expressed in JHA1w, although expression was not mandible specific (Fig. 3). The spatial expression patterns of pb and Antp, i.e. upregulation in more-posterior appendages, also corresponded to those in other insects (Hughes and Kaufman, 2000; Rogers and Kaufman, 1997). Expression of these Hox genes could also be necessary, because the posterior appendages are mildly modified in the soldier differentiation (Koshikawa et al., 2002).

As the mandible-specific morphogenesis occurs in response to high JH titer, spatial information for the body-part identity should be provided by the upregulation of Hox genes downstream of the activity of JH, or by continuous expression throughout postembryonic development. The expression levels of some Hox genes were high even at the PE stage, and some were upregulated by JH, although the patterns differed among Hox genes (Fig. 2). This suggests that the Hox genes need to be expressed even during the postembryonic development, probably for postembryonic modifications.

The antero-posterior appendage patterning genes, en and hh were upregulated in mandibles, respectively at JHA1w and JHA2w, but both were upregulated also in SM (Fig. 2). Hence, these genes are likely expressed at each molt, as are hth, al and Ser (Fig. 2). By contrast, dpp expression in mandibles was decreased by JHA (Fig. 2). This pattern was different from wing-disc formation in Drosophila melanogaster, in which en and hh expression induces dpp for antero-posterior axis formation (Estella et al., 2012; Kojima, 2004). Ser was slightly upregulated during soldier differentiation but the expression level was higher in SM. In Drosophila wing formation, Ser is expressed at the D/V boundary of dorsal cells (Kim et al., 1995). Such axis formation patterning would be also required for patterning at every molt. In insect embryos, dpp expression is known to localize at the distal tip (Angelini and Kaufman, 2005a). However, during the soldier differentiation in termites, dpp was downregulated in mandibles, suggesting that the mandibular modification in postembryonic development is different from the general appendage formation.

Mandibular enlargement during soldier differentiation occurs along the proximo-distal axis (Koshikawa et al., 2002, 2003); thus, the above result (Fig. 2) was likely due to the restriction of antero-posterior growth of mandibles (from teeth to the outer edge) during soldier differentiation. The antagonistic effects of wg and dpp within the imaginal disc provide patterning information along the proximo-distal axis of appendages (Brook and Cohen, 1996; Jiang and Struhl, 1996; Penton and Hoffmann, 1996; Struhl and Basler, 1993). In the imaginal discs of Drosophila, high levels of wg and dpp signals reportedly induce Dll expression, leading to the formation of distal appendages (Lecuit and Cohen, 1997). During the mandibular elongation of termite soldiers, the expression pattern of wg corresponded to the pattern in Drosophila, but those of dpp and Dll did not (Fig. 2). Thus, Dll expression was reduced simultaneously with the downregulation of dpp in mandibles after JHA treatment. Upregulation of wg and dac at JHA1w suggests an interaction between the two factors (Fig. 2), but wg expression was not mandible specific (Fig. 3).

During the formation of insect appendages, Dll makes the major contribution to the patterning of the distal region (Panganiban and Rubenstein, 2002). The mandible is thought to be a gnathobasal appendage that lost the distal appendage region, where Dll is expressed (Angelini et al., 2012a; Boxshall, 2004; Panganiban et al., 1994; Popadić et al., 1998; Snodgrass, 1935). However, the apical region (from the first teeth to the tip) of the mandible is considerably enlarged during soldier differentiation (Koshikawa et al., 2003). Although some genes for the patterning of apical mandibular region were predicted, our data (Fig. 2, Fig. S4) suggest that Dll is not involved in or required for the mandibular enlargement during soldier differentiation, as reported in other insects (Angelini and Kaufman, 2004, 2005b; Beermann et al., 2001; Coulcher and Telford, 2013; Panganiban et al., 1994; Popadić et al., 1998; Simonnet and Moczek, 2011). This situation is very similar to the mandibular elongation in male stag beetles (Gotoh et al., 2017).

Among 18 candidate genes, only dac showed characteristic expression patterns in the mandibles during soldier differentiation; i.e. mandible-specific upregulation prior to molt into presoldiers (Figs 2 and 3), and knockdown of dac expression restricted mandibular elongation during the molt into presoldiers (Fig. 5, Figs S3 and S4). The dac expression was higher in the distal part of the mandibles (Fig. 4A), which is similar to the of embryonic mandibles in T. castaneum (Coulcher and Telford, 2013). The detailed expression localization of dac (Fig. 4B) corresponded well to the areas of epithelial tissues where extensive cell proliferation takes place along the inner edges of mandibles before the molt into presoldiers (Koshikawa et al., 2002). However, the dac localization at the inner edge did not explain the cell proliferation at the outer (lateral) edges of mandibles. Interestingly, this situation is consistent with the phenotypes of dac-RNAi individuals.

Knockdown of dac affected enlargement of the mandibles, but not that of other appendages (Fig. 5, Fig. S3). The mandibular elongation in dac-RNAi individuals was significantly suppressed compared with those of GFP- and Dll-RNAi individuals (Fig. S4). The function of dac in mandibular development corresponds to that in other insects (Angelini and Kaufman, 2004; Coulcher and Telford, 2013; Giorgianni and Mann, 2011; Gotoh et al., 2017; Popadić et al., 1998; Simonnet and Moczek, 2011). In H. sjostedti, the number of mandibular teeth does not change during soldier differentiation, but the proportion changes owing to the increase in tooth distance (Koshikawa et al., 2003). dac^RNAi PS exhibited restricted enlargement of the apical region of the mandibles with highly curved or rounded outer edges (Fig. 5, Fig. S4). Considering that the dac expression was restricted to the inner edges (Fig. 4B), the RNAi knockdown of dac should inhibit the cell proliferation at the inner edge, so that the epithelial cell proliferation would only occur at the outer edge, resulting in the mandibular shape with curved outer edge. It is therefore suggested that there are other factors responsible for the cell proliferation at the outer edge.

In Drosophila leg discs, dac expression is directly regulated by Dll (Giorgianni and Mann, 2011). Because Dll was not upregulated in mandibles, this regulatory mechanism was not applicable to soldier differentiation. In Tribolium, however, dac is also expressed in mandibles and the proximal region of legs, which are not affected by Dll (Prpic et al., 2001). So this pattern of dac expression in Tribolium likely corresponds to that in termite mandibles (Angelini et al., 2012a). Unlike the mandibular development at metamorphosis in holometabolous insects, in the case of termite soldier differentiation, pre-existing mandibles are just modified, so that the loss of function of dac does not result in complete or partial loss of mandibles but in the allometric changes observed in this study.

To assess morphogenesis during postembryonic development in insects, clarification of the links between hormonal signals that circulate throughout the body and patterning genes that mediate tissue-specific responsiveness is needed (Gotoh et al., 2015; Lavine et al., 2015; Truman and Riddiford, 2007; Villarreal et al., 2015). Therefore, we also examined the epistatic relationships of dac with putative upstream factors (i.e. JH and insulin), which provide temporal physiological information, and with Hox genes, which provide spatial information (Fig. 6). dac and Dfd expression in JHA1w was regulated by Met and InR, and Dfd expression was necessary for that of dac (Fig. 6E). These findings suggest that JH and IIS have their effects upstream of morphogenetic factors during soldier differentiation. In T. cataneum larvae, the expression of hh, which is important for the proliferation of imaginal cells, is inhibited by JH (Villarreal et al., 2015). In H. sjostedti, however, hh was upregulated in mandibles under the high-JH condition, although it was not significantly different from that during SM (Fig. 2). The difference in the response of hh expression to JH between H. sjostedti and T. castaneum was possibly due to alternation of the cis and/or trans regulatory elements. Some other factors, such as the Fat/Hippo signaling pathway, have been suggested to mediate between endocrine and morphogenetic factors (Gotoh et al., 2015; Lavine et al., 2015). Thus, tissue-specific cell proliferation leading to the mandible-specific elongation during termite soldier differentiation is coordinated by the interactions between endocrine signals and patterning genes (Fig. 6E). These interactions would also be applied to the differentiation of other castes, such as alates. The detailed link between endocrine signals and patterning genes is still unclear, so that further studies will enhance understanding of the regulatory mechanisms underlying morphological diversity.

Colonies of H. sjostedti Holmgren (family Archotermopsidae) inhabiting rotten wood were collected on Yakushima Island (Kagoshima Prefecture, Japan) in May 2011 and May 2014. Colonies were maintained in plastic containers with nest logs at ∼25°C under constant darkness, and were occasionally fed moistened pinewood. Caste categories were identified based on previous studies of the focal species (Miura et al., 2000; 2004): larval instars, pseudergates (PE), nymphs, alates, neotenics, presoldiers (PS) and soldiers (Sol) (Fig. 1A).

To identify major toolkit genes in the focal termite species, RNA-sequencing was performed (see supplementary Materials and Methods). Total RNAs were extracted from eggs, alates, soldiers and neotenics of H. sjostedti, yielding 181,084,900 paired-end reads. After filtering, 140,958,127 paired reads and 18,567,603 single reads, the paired-end read counterparts of which were filtered out, were retained for further analyses. De novo transcriptome assembly of the sequence reads generated 164,434 contigs with a total of 161,836,366 bases and a N50 of 2075 bp. The predicted amino acid sequences of Tribolium castaneum toolkit genes were used for BLASTp searches to identify the termite orthologs from transcriptome data of H. sjostedti (Boxshall, 2004; Richards et al., 2008). The top-hit sequences from the termite transcriptome database were defined as putative orthologs of the toolkit genes. We performed phylogenetic analyses that included genes from other animals to confirm the orthologs of putative toolkit genes (Table S1, Fig. S1A-J, see supplementary Materials and Methods for details).

To generate individuals undergoing soldier differentiation, a JH analog, pyriproxyfen, was applied (Ogino et al., 1993). As previously described (Sugime et al., 2015), 10 pseudergates (PE) were placed in a Petri dish (ø 70 mm) lined with filter paper containing 10 µg pyriproxyfen (Sigma-Aldrich) and moistened with distilled water. Petri dishes were monitored daily and maintained at 25°C under constant darkness. The presoldier molt typically occurred ∼14 days after JHA application. During soldier differentiation, samples were obtained 1 week after JHA application (JHA1w), JHA2w and PS induced by JHA (JHAPS) stages. In addition to these JHA-treated termites, pseudergates (PE), natural PS (NPS), Sol and PE in the stages prior to the stationary molt (SM) were sampled from the stock colonies. NPS were included for comparison with JHAPS to evaluate artificial effects that might be induced by the application of an unnatural compound, i.e. JHA. To obtain the SM, PE that exhibited whitish abdomens due to gut purging were collected from the stock colonies (Koshikawa et al., 2005; Sugime et al., 2015).

To quantify expression levels in mandibles during the soldier differentiation, total RNA was extracted from mandibles along a time course, i.e., PE, JHA1w, JHA2w, JHAPS, NPS, Sol and SM, and real-time qPCR was performed for the selected toolkit genes. Total RNAs were also extracted from different appendages, including mandibles (Md), maxillae (Mx), labium (Lab), antennae (Ant) and forelegs (Fl) dissected from 20 individuals 1 week after the JHA application (JHA1w) from three different colonies. The samples were frozen in liquid nitrogen and preserved at −80°C until RNA extraction. Methods for RNA extraction, reverse-transcription, real-time qPCR, selection of reference genes and statistical analysis were performed as previously described (Cornette et al., 2013; Hattori et al., 2013; Ishikawa et al., 2010; Koshikawa et al., 2005, 2010). Details are provided in the supplementary Materials and Methods.

To analyze localization of the dac expression in mandibles, qRT-PCR and immunostaining were carried out. First, to roughly determine in which part of mandibles dac is expressed, qRT-PCR comparisons between distal and proximal mandibles were carried out. At 1 week after the JHA application (JHA1w), right and left mandibles were dissected, each of which was further dissected into two parts (distal and proximal parts) as shown in Fig. 3. RNA extraction and qPCR were performed as mentioned above. To further determine the detailed spatial patterns of dac expression, immunological staining was performed using the monoclonal antibodies against the Drosophila Dachshund protein (mAbdac1-1 and mAbdac1-2; Mardon et al., 1994; purchased from Developmental Studies Hybridoma Bank, University of Iowa) and a confocal laser microscope (OLYMPUS FV3000). See supplementary Materials and Methods for details.

To evaluate its function, dac was knocked down by RNAi using a method established to assess soldier differentiation of the focal species (Hattori et al., 2013; Koshikawa et al., 2010). Dll, which is responsible for the elongation of distal appendages, was also knocked down by RNAi (Panganiban and Rubenstein, 2002). Fragments of the target genes were amplified by PCR using specific primers and subcloned, and the sequences confirmed. dsRNA was then synthesized and diluted to a final concentration of 3 µg/µl (dac, Dll and GFP) or 1 µg/µl (InR), and injected to PE at day 3 after JHA application. See supplementary Materials and Methods for details. The termites were maintained in Petri dishes at 25°C for 2-3 weeks. Only individuals that molted within 3 weeks were used for the morphometric study. Phenotypes for RNAi-treated individuals were evaluated by morphometric analyses. To evaluate the effects of dac RNAi on soldier differentiation, we conducted principal component analysis based on the measurements of 13 body parts. To further investigate the effects of RNAi on detailed mandibular shapes (untreated, GFP^RNAi, Dll^RNAi and dac^RNAi), morphological changes were assessed, based on 20 landmarks along the outline of the mandibles, using geometric morphometric techniques (see supplementary Materials and Methods for details, Table S3).

To evaluate the epistatic relationships among dac, JH signaling, insulin signaling and Dfd expression, qPCR was performed after reciprocal RNAi treatments. dsRNAs for InR and dac, and siRNAs for Met and Dfd were injected into termites. Detailed methods for the preparations of dsRNA and siRNA are described in the supplementary Materials and Methods. After RNAi of dac, Met (JH receptor), InR (insulin receptor) or Dfd (Hox), the transcript levels of these genes in mandibles were quantified 7 days after JHA application (i.e. 4 days after RNAi) by qPCR. RNA extraction, reverse transcription and qPCR were performed as described above.

We express our gratitude to Dai Watanabe for valuable comments on the analyses. Tadao Matsumoto provided us the first opportunity to work on this fascinating insect. We also thank two anonymous reviewers who gave us constructive comments on the former version of manuscript. Computational resources were provided by the Data Integration and Analysis Facility, NIBB.