Sex-specific division of labor in termites is linked to differential brain architecture and foraging gene expression

Summary

Division of labor in social insects is a key factor contributing to their ecological success, often relying on individual’s age and body size. In termites, it also extends to sexually dimorphic groups; however, the underlying neural and molecular mechanisms remain poorly understood. We investigated the relationship between sex-specific division of labor, brain architecture, and expression of the foraging (for) gene in workers of the termite Syntermes dirus. Behavioral assays revealed a striking sex-based division of labor, with male workers acting as foragers and female workers as builders. Consistent with the cognitive demands of foraging, male workers exhibited a larger brain central complex, a sensory center associated with orientation and navigation. Moreover, relative for expression is higher in male workers than in females and soldiers. Finally, we revisited for gene expression in workers and soldiers of different termite species, suggesting species-dependent effects on foraging. Our findings represent a first step toward elucidating how structural and functional brain changes underlie sex-specific division of labor in termites.

Graphical abstract

Highlights

• •There is a striking sex-based division of labor between Syntermes dirus workers
• •Male workers (foragers) have a larger central complex than females (builders)
• •The for gene, a regulator of foraging activity, is highly expressed in males
• •Differential for expression among termite taxa suggests species-specific effects

Biological sciences; entomology; neurogenetics

Introduction

Social insects are well known for their division of labor, which positively impacts task efficiency, colony productivity, and, consequently, their ecological success.^1,2 The division of labor among workers is usually related to differences in shape or body proportions (size polyethism), where different-sized individuals engage in specific tasks, or the same workers may engage in different tasks over their lifetime (temporal polyethism).^3,4,5

In the order Hymenoptera, eusociality has evolved multiple times, and both size- and age-related division of labor are observed. For instance, in the highly eusocial honeybee, Apis melllifera, nurses and foragers do not differ in their morphology or overall size. Instead, young bees engage in in-hive tasks, while older bees perform riskier tasks, such as foraging and guarding.⁶ In contrast, in the bumblebee Bombus terrestris, the division of labor among workers is based on size, with small workers preferentially engaging in brood care, while larger workers are more likely to forage.⁷ In some ant species, the opposite occurs, with smaller individuals (minors) performing foraging activities, while larger ants are allocated to colony defense.⁸ Notably, all the tasks in the colonies of social Hymenoptera are performed by female individuals, and therefore, no sex-based division of labor is found.⁹

Foraging in insect societies is a complex process that typically involves multiple individuals and various strategies for searching and exploiting food sources.¹⁰ Foraging is often linked to physiological changes in the workers, including glandular activity, hormone levels, circadian rhythms, gene expression, microbial diversity, and morphology of the nervous system.^{11,12,13,14,15,16,17} Concerning the brain, this can involve structural plasticity of specific brain compartments, changes in the concentration of biogenic amines, and variation in gene expression, for instance, the foraging (for) gene, which regulates foraging behavior.^{18,19,20,21,22}

An enriched environment is known to affect the shape of the brain, as shown for different groups of invertebrates and vertebrates,^23,24,25 and in social insects, structural plasticity of the brain has been associated with foraging behavior in bees, wasps, and ants. These changes involve the enlargement of specific neuropils, such as the mushroom bodies (related to learning and memory) and the central complex (related to navigation and orientation).^{18,19,26,27,28} In the case of the mushroom bodies, this enlargement relies on the outgrowth and branching of their intrinsic neurons, the Kenyon cells.^18,29 This is associated with the fact that foraging requires complex navigation, recall of the location of food sources and the colony site, communication among nestmates, and, in essence, the acquisition of complex visual and olfactory information.^30,31

At the molecular level, the foraging (for) gene has emerged as one of the most prominent regulators of foraging behavior both in invertebrates and vertebrates.^32,33 Its role was first identified in divergent foraging patterns in Drosophila melanogaster larvae, underlying the rover-sitter phenotypes.^32,34 The for gene encodes a cGMP-dependent protein kinase (PKG) that regulates downstream targets through phosphorylation. In the social Hymenoptera, for was found to regulate foraging behavior and many other aspects related to the division of labor. Although the pathway linking PKG activity to the actual division of labor is still to be elucidated, it is noticeable that it does not involve a universal mechanism, but is likely to vary across species.^{8,20,35,36,37} In terms of neural localization, for and its protein product are found to be highly expressed in important neuropils for the division of labor, such as the mushroom bodies, and the optic and antennal lobes.^38,39 Furthermore, for expression has also been identified in the corpora allata.³⁸ These glands are the source of juvenile hormone biosynthesis,⁴⁰ a hormone that is strongly associated with division of labor in social Hymenoptera.^16,41 In addition, in both the Chinese honey bee Apis cerana and the stingless bee species Frieseomelitta varia, the levels of the for gene in the abdomen positively correlate with the nurse-to-forager transition, suggesting that PKG activity may also play a role in metabolic regulation, although this role is less well understood than its behavioral effects.^39,42

Termites are considered “eusocial cockroaches,” and their mode of eusociality evolved independently of the Hymenoptera.^43,44 In the termites, all the worker tasks are performed by the immatures that can undergo successive rounds of molts, and the division of labor mostly relies on size and age, often interrelated, but in some cases, also on the sex of the workers.^{14,45,46,47,48} Nonetheless, studies investigating the neural mechanisms underlying division of labor in termites are strongly underrepresented when compared to social Hymenoptera, and the few attempts pointing to structural and functional plasticity of the brain have set the primary focus on the worker-to-soldier molt, and on the reproductive maturation and age.^{49,50,51,52,53,54} No study, as far as we know, has yet attempted to uncover the role of brain plasticity in the context of the worker caste. Similarly, the role of the foraging gene has been addressed only in a single genus, Reticulitermes. In R. flavipes, its expression was analyzed across different castes, photoperiods, and temperatures, which, strikingly, indicated a negative correlation with foraging behavior.⁵⁵ In R. chinensis, however, Xu et al.⁵⁶ found that the knockdown of for greatly impaired the capacity of the termites to follow trail pheromones, suggesting a novel pathway related to the control of foraging communication.

The genus Syntermes (Termitidae) comprises subterranean species whose workers go to the surface at night to forage on the grass- or leaf-litter,⁵⁷ and morphological and morphometrical analyses indicated that Syntermes spp. workers can be divided into male and female lineages.⁵⁶ Thus, while division of labor in Syntermes is suggested to be based on sex and size of the workers, direct observations are scarce due to the termites’ cryptic habits. Within these two lineages, once 3rd instar workers molt into 4th instar ones, they begin to differ in morphology and size, with males being larger and brown-headed, while females are smaller and have a yellowish head. Moreover, only male workers can differentiate into soldiers (Figures 1A and 1B).^57,58

Figure 1.

Caste differentiation and division of labor in Syntermes dirus

(A) Caste developmental pathway adapted from pioneering observations.⁵⁷ Colony tasks are performed by immature individuals, in which younger male and female workers (3rd instar) molt into older ones (4th instar). Soldiers are exclusively males.

(B) 3rd and 4th instar (male and female) workers of S. dirus, highlighting their particular morphological features, such as head sclerotization and size. Scale bars, 1 mm.

(C) Female workers molding moistened soil to build structures around the nest entrance.

(D) Male worker foraging on dead leaves. Figures (C and D) are not on the same scale.

This sex-related pattern of caste differentiation provides an opportunity to investigate how it is reflected in brain morphology and the expression of the for gene, and to do so, we used the Neotropical species Syntermes dirus as a model. In previous field inspections, we had seen that only 4th instar male and female workers of S. dirus go to the surface to perform out-nest tasks, with males often found at foraging sites. Thus, we hypothesized that: (a) division of labor within the worker caste of S. dirus is based on sex/size of the individuals, with males comprising the foraging workforce; (b) males display a higher investment in sensory integration and premotor centers of the brain to cope with foraging tasks; and (c) the for gene, a regulator of foraging activity, is more highly expressed in male workers compared to female workers and soldiers. To test this, we conducted behavioral assays under field and laboratory conditions, analyzed brain structures using immunohistochemistry and volumetry methods, and cloned a partial sequence of the for gene to investigate its spatial expression in S. dirus workers and soldiers. Moreover, we revisited the for gene expression across several termite species, based on publicly available transcriptomes, to unravel its role in the division of labor across different levels of social complexity and nesting habits.

Our results point to a striking division of labor between the workers of the two sexes, with males comprising the foraging workforce. Neuropil reconstruction showed a significant enlargement of the central complex in males, a region involved in orientation and navigation, but not of the mushroom bodies. Furthermore, we also identified higher expression of the for gene in both the brain and abdomen of male workers compared to females and soldiers. Finally, the for transcript levels in workers and soldiers across termite taxa suggest species-dependent effects in the foraging behavior. These findings are discussed in the light of life histories and ecological habits of the different termite groups.

Results

Sexually dimorphic workers engage in different tasks

Male workers were seen leaving the nest around 19:00 (ranging from 18:30 to 19:30). Initially, these workers stood at the nest entrances for a few seconds, moving their antennae, and then started to forage. Since the entrances were located beneath trees, numerous dry leaves accumulated nearby, and male workers did not need to go far from the entrance to collect food (up to 30 cm). Termite workers were not observed on the surface during the day, and several nest entrances used during the previous night were found closed with soil.

There was a striking sexual polyethism related to the behaviors analyzed (Figures 1C and 1D), as male workers were the only individuals engaged in foraging (LRT on Firth’s penalized likelihood: χ²(3) = 63.01, p < 0.001), with a mean of 20.45 ± 15.80 foraging trips per observation period (i.e., cutting pieces of leaves and carrying these back to the nest) (Figure 2A). In a very restricted number of events (six out of the 484 foraging trips), females were seen collecting pieces of leaves close to the nest entrances. In contrast, females were the only individuals involved in building, with a mean of 11.59 ± 11.97 events per observation period (LRT on Firth’s penalized likelihood: χ²(3) = 47.29, p < 0.001) (Figure 2B; Video S1). Soil deposition was performed by both groups, without a statistical difference between males and females (mean of 2.81 ± 8.43 and 5.51 ± 7.58 events per observation period, respectively) (LRT: χ²(1) = 2.29, p = 0.13) (Figure 2C; Video S2).

Figure 2.

Division of labor and leaf consumption between sexually dimorphic workers

(A) Foraging, consisting of cutting/collecting pieces of leaves and carrying these to the nest.

(B) Building, comprising the manipulation and placing of moistened soil around the nest entrance.

(C) Soil deposition, consisting of placing dry soil pellets randomly outside the nest. Data were treated with Firth’s bias-reduced logistic regression (foraging and building) and Type-2 negative binomial GLMM (soil deposition), N = 10 for colonies 2 and 3, N = 7 for colony 1.

(D) Time spent outside the nest to perform respective tasks. Data were treated with GLMM using a gamma family, N = 5 in each colony.

(E and F) Initial and final surface area of the 1,100 mm² dead leaf offered to male- and female-based groups, and treated with a Firth’s bias-reduced logistic regression. N = 5 per caste for colonies 1 and 3. Fw, female workers; Mw, male workers. For (A, B, and C), each dot corresponds to the total of events within a 15-min observation. For (D), each dot represents one individual. For (E), each dot corresponds to a rounded leaf. Values are given in mean ± SD.

The time spent on the surface also varied greatly between the sexes (LRT: χ²(1) = 29.44, p < 0.001), with a mean of 155.13 ± 19.9 s for males and 29.05 ± 7.68 s for females (Figure 2D). This difference in time above ground is explained by the longer time it takes for males to select, cut, and transport leaves, whereas building and depositing soil were performed around the exit hole, and consequently, required a shorter period for the females to return into the nest. These results, followed by those from the leaf consumption assays, corroborated our hypothesis (a) that there is a sex-based division of labor in S. dirus workers.

Leaf consumption is restricted to male workers

Under laboratory conditions, only groups composed of males engaged in cutting leaves; whereas, females never cut the leaves after overnight assays (LRT: χ²(1) = 22.36, p < 0.001). The mean value for the final superficial area of the leaves in male-based groups was 878.01 ± 113.01 mm²; whereas, those in female-based groups did not change compared to the initial ones (Figures 2E and 2F). Together with the field data, this provides evidence that females typically do not engage in foraging, but occasionally may opportunistically collect pieces of leaves next to the entrances.

Male and female workers have different neural architectures

To compare the neural phenotypes of male and female workers, we performed 3D segmentation for the reconstruction of brain neuropils following synapsin immunostaining (Figures 3A–3D). The mushroom bodies (MBs), divided into medial and vertical lobes and calyces, corresponded to 62.59% of the reconstructed neuropils for males and 64.73% for females. The MB lobes were highly developed and shaped as looped structures (Figure 3D). In contrast, the calyces (two units per lobe) were almost fused and encompassed about 16% of the reconstructed brain compartments for males and females. The other measured neuropils, antennal lobes (ALs) and central complex (CX), comprised 28.29% and 9.10% of the neuropils for males and 26.71% and 8.0% for females, respectively. The ALs consist of several glomeruli that were distinguishable by confocal scanning, and it was also possible to observe subdivisions of the CX, such as the central body and the protocerebral bridge (Figures 3B–3D).

Figure 3.

Brain reconstruction in male workers of Syntermes dirus

(A) 3D-brain projection based on serial segmentations of the different brain areas. The mushroom body (MB) lobes (light pink) are well developed and twisted, whereas the MB calyces (green) are reduced and nearly fused. Antennal lobes (ALs) are highlighted in blue, the central complex (CX), including the central body and protocerebral bridge, in purple, and the optic lobes (OLs) in red. As termite workers are blind, the OLs are undeveloped, and their subdivisions cannot be distinguished.

(B–D) Synapsin staining of the different brain regions under confocal microscopy. ML medial lobes, P = peduncle, VL = vertical lobes. Scale bars, 100 μm.

The absolute volume of the sensory and integration regions of the brain (sum of the reconstructed neuropils) did not vary between the sexes (t = 0.59031, df = 14.651, p = 0.56) (Figure 4A). However, the relative values, which consider the head capsule volume, differed significantly, being larger in female workers (t = 14.349, df = 11.72, p < 0.001) (Figure 4B). In other words, female workers have brains that are similar in size to those of males, but they are accommodated in much smaller head capsules (male workers: 15.52 ± 0.603 mm³; female workers: 8.07 ± 0.667 mm³, p < 0.001) (Figure 4C). The absolute values for individual neuropils consistently followed the respective size of the sensory and integration brain region (MB lobes: t = −0.526, p = 0.60; MB calyces: t = 0.630, p = 0.54; LA: t = 1.124, p = 0.28) (Figures 4D–4F), except for the CX, which was larger in the male workers (U = 11, p = 0.003) (Figure 4G). In terms of relative volumes, there was no statistical difference between the groups (MB lobes: U = 24, p = 0.05; MB calyces: t = −0.221, p = 0.828; LA: t = 1.436, p = 0.168) (Figures 4H–4J). The CX, however, was found significantly increased in males, indicating a differential investment in this specific brain area (t = 2.244; p = 0.03) (Figure 4K). The relative enlargement of the CX is in line with our hypothesis (b), that males, as foragers, show enlarged sensory integration and premotor centers. However, the MBs did not differ between the two groups, as further discussed.

Figure 4.

Absolute and relative volume of the brain, head capsule, and neuropils between sexually dimorphic workers

(A) Brain absolute volume.

(B) Brain relative volume.

(C) Head volume. Given the similar absolute brain volume and the discrepant head volume between males and females, relative brain volume is larger in females.

(D–G) Absolute volumes of MB lobes, MB calyces, ALs, and CX, respectively.

(H–K) Relative volumes for the same neuropils. Except for the CX, absolute and relative neuropil volumes did not change between males and females, following Student’s t test or Mann-Whitney U test. N = 11 for males, N = 9 for females, one colony. Fw, female workers; Mw, male workers. Each dot corresponds to one individual, and values are given in mean ± SD.

Sdfor expression is higher in the male worker brain and abdomen

As the behavioral assays strongly indicated a sex-biased division of labor among the workers, we wondered whether this could be related to a sex-specific difference in the expression of the foraging gene. For this, we first identified and cloned a partial sequence of the orthologous gene in S. dirus (Table S1), and then analyzed its expression in the brain and abdominal carcass of the workers (male and female) and soldiers (strictly males).

Consistently, we found that Sdfor expression is significantly influenced by both sex and caste (F value = 6.7690, df = 87, p = 0.0018), being higher in both the brain and the abdomen of male workers compared to those of the female workers and the male soldiers (t = 3.296, df = 85.1, p = 0.0040 and t = 3.038, df = 85.2, p = 0.0088, respectively), suggesting a positive correlation between Sdfor expression and engagement in foraging behavior in this termite species, corroborating our hypothesis (c) (Figures 5A and 5B). Notably, Sdfor expression did not differ between female workers and soldiers in either the brain or the abdominal carcass (t = −0.307, df = 85.4, p = 0.94). Nonetheless, a non-significant trend was found for higher expression of foraging in the brain compared to the abdomen (t = 1.745, df = 85, p = 0.08).

Figure 5.

Relative expression of the gene Sdfor in male and female workers and soldiers (males)

(A) Brain.

(B) Abdominal carcass. Linear mixed models (LMM) using Satterthwaite’s method were employed for comparisons, and the colony was classified as a random factor. Pairwise comparisons were performed using Tukey’s post-hoc test. Different letters correspond to significantly different results. N = 14–15 per caste per tissue for each of three colonies. Fw female workers, Mw male workers, S soldiers. Values are reported as mean ± SD, and letters indicate statistically significant differences in pairwise comparisons.

Role of foraging gene expression varies across termite taxa

Since we identified a positive correlation between the for gene and foraging behavior in S. dirus, we sought to further investigate its role in the division of labor by characterizing its expression across termite species with distinct ecological habits, using publicly available transcriptomes. (Table S2; Figures 5 and 6). In Reticulitermes speratus, a “separate-piece nester” termite (ST), in which workers forage outside the nest, for expression in the head and abdominal tissues of soldiers was significantly higher compared to workers (log₂fc = 0.230, q value = 0.035; log₂fc = 0.344, q value = 0.0007, respectively), (Figure 6A). Notably, no significant difference was found between sexes in both workers (head: log₂fc = −0.16, q value = 0.99; abdomen: log₂fc = 0.044, q value = 0.99) and soldiers (head: log₂fc = 0.06, q value = 0.99; abdomen: log₂fc = 0.002, q value = 0.99), confirming the major effect of caste, but not sex, on foraging gene expression.

Figure 6.

Reanalysis of foraging gene expression levels in publicly available datasets for separate-piece nest (ST) species

(A) Foraging expression in the head (left) and abdomen (right) tissues of male and female workers and soldiers of Reticulitermes speratus. N = 3 transcriptomes per task per tissue.

(B) Foraging expression in the head + thorax tissues of builder and forager female workers of Macrotermes bellicosus. N = 7 transcriptomes per task.

(C) Foraging expression in the head (left) and abdomen (right) tissues of male and female workers and soldiers of Inquilinitermes inquilinus. N = 3 for workers and N = 4 for soldiers per tissue. The y axis shows transcripts per million (TPM); the Wald test was employed to identify differential expression, followed by Benjamini-Hochberg correction; log₂FC and corrected q values are shown for pairwise comparisons. Fs, female soldiers; Fw, female workers; Ms, male soldiers; Mw, male workers; S, soldiers; W, workers. Values are given in mean ± SD.

For Macrotermes bellicosus, there was no apparent difference in for transcript levels between builder and forager female workers (log₂fc = −0.096, q value = 0.38) (Figure 6B). A differential foraging gene expression was, however, observed in Inquilintermes inquilinus, with a higher expression in the head tissue of soldiers compared to workers (log₂fc = 0.162, q value = 0.003), but not in the abdomen (log₂fc = 0.16, q value = 0.621524) (Figure 6C). Within the OPT species (Zootermopsis nevadensis and Neotermes binovatus), no differential expression was observed in the head tissues of workers vs. soldiers (log₂fc = 0.261, q value = 0.99; log₂fc = 0.127, q value = 0.64; log₂fc = 0.029, q value = 0.96, respectively) (Figures 7A and 7B). Details of the transcriptome quantification results are available in Tables S2E–S2K.

Figure 7.

Reanalysis of foraging gene expression levels in publicly available datasets for one-piece nest (OPT) species

(A) Foraging expression in the head tissue of workers and soldiers of Zootermopsis nevadensis. N = 5 individuals per task.

(B) Foraging expression in the head tissue of workers and soldiers of Neotermes binovatus. N = 10 individuals per task. The y axis shows transcripts per million (TPM); the Wald test was employed to identify differential expression, followed by Benjamini-Hochberg correction; log₂FC and corrected q values are shown for pairwise comparisons. S, soldiers; W, workers. Values are given in mean ± SD.

Discussion

Workers of Syntermes dirus display a sex-specific division of labor

Here, we provide solid evidence for a sexually based polyethism among workers of S. dirus, and also highlight neural and molecular mechanisms associated with this so far understudied task allocation setup. Previous studies have already pointed to a sex- and size-based division of labor within Termitidae, with male (larger) individuals engaging in riskier tasks, such as foraging, as shown for fungus-growing termites.^58,59 In Odontotermes distans, for example, they make up nearly 99% of the foragers, whereas 96% of the workers tending the royal pair inside the nest are females.⁴⁵ In some cases, division of labor relies on the size, sex, and age of the workers,⁴⁷ illustrating how complex this phenomenon can be, especially in a hemimetabolous social insect with mixed-sex castes (Figures 1A and 1B).^58,60 Division of labor between S. dirus workers analyzed here resembles that of fungus-growing species, as male workers are foragers and spend longer periods outside the nest, exposed to riskier situations (Figures 2A–2D). Interestingly, male workers become soldiers in Syntermes spp., a caste equally engaged in risky activities,⁵⁷ similar to other closely related termitids performing open-air foraging (i.e., Hospitalitermes spp).⁴⁷

The contribution of females in food acquisition is not supported as follows: first, they rarely manipulate food (at least outside the nest), and second, females do not cut leaves even in the absence of males (Figures 2A–2F), which may also be a consequence of morphological specialization (e.g., mandibular muscle arrangement). In Hospitalitermes medioflavus, both male and female workers engage in foraging, and the latter change from gnawers to carriers as they age.⁴⁷ Although we could not estimate the ages of the sampled termites, we would expect to spot subgroups of females foraging if they engage in this activity, which did not happen during our extensive field observations.

Given their short exposure outside the nest, females, as well as early-instar workers, are likely to be predominantly engaged in intranidal tasks, an assumption confirmed for other termite species with sexually dimorphic workers.^45,48,61 Although such intranidal activities cannot be followed under field conditions, the participation of the females in building activity around the nest entrance, especially in soil deposition, may indicate that groups of intranidal workers are excavating galleries and transporting soil away from the tunnel tips.⁶² Similar to S. dirus, female workers of M. bellicosus are the sole group responsible for building new structures and repairing the nest.⁶¹ Concerning the soil deposition, the overlap of some tasks between worker groups has been already described for other subterranean species. Similarly, the absence of young workers engaging in extranidal tasks is consistent with previous observations.⁴⁸

The neural phenotype is linked to foraging behavior

Reconstruction of the neuropils showed a similar investment in different brain regions between males and females, except for the CX. Studies on caste-biased neuroplasticity in termites are scarce, but, in general, they indicate a higher development of the mushroom bodies in workers compared to soldiers and reproductives.^51,53 This neural investment pattern is expected, given that workers perform a variety of tasks in the colony. In contrast, soldiers are generally restricted to defense, and the royal couple to mating and reproduction.⁶³ Comparisons within a caste are hitherto restricted to Reticulitermes flavipes (Rhinotermitidae), in which there is an absence of sexual dimorphism in the brain anatomy of the workers.⁵⁴ However, it is not clear if and how sex and division of labor are related in this species, contrasting with termitid workers, in which these two parameters are often linked.^{45,47,59,61,64,65}

In contrast to other brain regions, the larger relative volume of the CX in male workers might be associated with their foraging activity (Figure 4K), contributing to the multisensory navigation requirements outside the nest. This neuropil is conserved across Arthropoda and plays a major role in orientation and navigation, serving as a sensory integration and premotor center.⁶⁶ It receives indirect input from the compound eyes and is also associated with visual recognition and sky polarization vision.^67,68,69 Termite workers are, however, blind, indicating that the CX may process mechanosensory and olfactory signals for spatial orientation. Similarly, the small (minim) workers of the ant Atta cephalotes also do not rely on vision when navigating in the dark fungus garden, and they have larger CX compared to the other subcastes, suggesting an important role of this brain center in multisensory navigation.⁷⁰ The well-developed CX of other blind hexapod and non-hexapod arthropods corroborates that the CX is not exclusively involved in visual orientation.^71,72,73,74 S. dirus is a separate-piece nesting termite, whose foraging sites are distant from the nest.⁵⁷ Moreover, male workers forage on the surface, contrasting with workers of most other subterranean termites that follow tunnel directions when foraging.⁷⁵ The necessity to forage above ground and estimate directions, thus, might impose differential sensory demands associated with navigation and orientation in S. dirus male workers.

Unlike studies on Hymenoptera,^19,76,77 we did not find a differential investment in the mushroom bodies between forager and non-forager groups (Figures 4D–4I). This indicates that cognitive demands imposed on male and female workers may be similar, even though they engage in different tasks. Another explanation is that such a division of labor may affect the network organization of the MBs, which does not necessarily result in higher volumes of their compartments.^78,79

The ALs, as the primary olfactory brain region, also did not vary between the two worker groups (Figures 4F–4J). This is not surprising given that termite castes, especially for separate-piece nesting termites, rely on chemical cues to perceive the environment and communicate with nestmates.^12,80 Nevertheless, comparisons between the ALs of workers and other termite castes yielded contrasting results, depending on the species, which likely reflects their distinct ecological habits and cognitive demands.^51,53,54 In this context, comparative analyses for the MBs and ALs between 3rd and 4th instar workers of S. dirus would be valuable to assess the occurrence of possibly age-dependent changes in these neuropils.

Higher for expression matches foraging specialization in S. dirus but not in other termite species

Sdfor gene expression was higher in male workers compared to both females and soldiers, with no significant difference between the latter two (Figures 5A and 5B). Notably, the differences were found in both tissues analyzed, meaning that, despite its well-known behavioral role, the PKG might also play a role in metabolic pathways related to division of labor. Nonetheless, no significant difference was found between female workers and soldiers. This is in agreement with the division of labor observed among sexually dimorphic workers of other termite species and with the fact that soldiers do not forage, although they may scout foraging expeditions.⁴⁸ This result, thus, provides the first evidence that the positive correlation between foraging gene expression (and possibly PKG signaling) and the division of labor of hymenopteran social insects is also extended to a sexually dimorphic system in termites. The roles of the for gene and its protein product, PKG1, in the division of labor in social insects have been extensively studied in the social Hymenoptera, where species-dependent effects on foraging are observed, either by promoting or inhibiting this behavior. In bees, for instance, expression of for gene in the brain of workers may serve as a trigger for foraging in A. mellifera, A. cerana, Frieseomelitta varia, and Bombus terrestris,^{35,36,39,42,81,82} but as an inhibitor in Melipona quadrifasciata and Bombus ignitus.^42,83 Similarly, a dependent effect of for on foraging behavior of ants is also reported.^{84,85,86,87,88} We now extend this interpretation to another major group of social insects, the termites.⁵⁵

The for gene expression among termite species with distinct ecological habits was contrasting. Within ST species, no difference was observed in for expression between M. bellicosus minor workers engaged in building and foraging (females). It is interesting to note that large workers (males) make up 70% of the foragers in this species, whereas minor workers engage mostly in other tasks.⁶¹ Thus, the inclusion of male workers would be valuable in unraveling the role of gene for in the division of labor of M. bellicosus. The lower expression of for in workers compared to soldiers of R. speratus and I. inquilinus suggests that its expression is negatively correlated with foraging, similar to R. flavipes.⁵⁵ While R. speratus colonies employ thousands of workers in food exploitation, foraging through underground galleries,⁸⁹ workers of I. inquilinus do not forage outside the host nest, but consume the dark material located at its bottom.⁹⁰ The higher expression of for in defensive castes of ST species is intriguing, especially considering that higher levels of for and its encoded PKG are associated with defensive behavior in ants.⁸ Thus, a likely role of the for gene in regulating defensive behavior in termite soldiers cannot be ruled out, and further investigations, including pharmacological activation or inhibition of for and its encoded PKG, are necessary to test such an assumption. This is particularly relevant in the context of age-based soldier polyethism,⁹¹ and considering the complex defensive mechanisms and secondary loss of soldiers in some ST termite lineages, with workers engaging in defensive tasks.^80,92,93

No difference was observed between workers and soldiers of the OPT species. These are socially less complex, display flexible caste differentiation, and spend their lifetime in the same piece of wood that serves as both shelter and food source. Therefore, workers and soldiers rarely or, depending on the species, never engage in outside tasks. It could explain the absence of a differential for gene expression triggering foraging behavior, at least in Z. nevadensis and N. binovatus, which live mostly on their food. One could argue that the inquiline condition of I. inquilinus represents an OPT lifestyle. However, this species retains characteristics of ST species, such as constrained caste differentiation.⁹⁴ Although these data are preliminary, it seems that the for gene plays species-dependent effects according to nesting and feeding habits. We advocate for future analyses to address termite species across the social spectrum, including those phylogenetically distant, but with similar open-air foraging like S. dirus (i.e., Hodotermitidae).⁹⁵ Such comparative analyses may shed light on the behavioral effects of for gene expression on castes and social contexts.

In conclusion, our results provide the first approach to unravel the structural and functional brain phenotypes related to a striking division of labor between the sexually dimorphic workers of a termite. The foraging behavior in S. dirus was exclusively performed by male workers, and this was found to reflect a differential architecture of a brain center related to navigation and orientation. At the molecular level, the higher Sdfor expression is possibly associated with the foraging behavior of male workers. These findings, together with the significantly lower expression of for in workers compared to soldiers in other ST species, suggest a species-dependent effect on termite foraging. The non-differential for expression in workers of OPT species is in agreement with their wood-dwelling condition. Further analyses, including species with a wider range of ecological and nesting habits, as well as environmental conditions (e.g., food shortage), should shed light on how morphological differences in the brain, as well as molecular mechanisms such the one exerted by the for gene might regulate behaviors such as foraging and defense across termite taxa.

Limitations of the study

The different assays and methodologies applied in the present study were based on samples from three colonies of Syntermes dirus, except for brain anatomy and volumetry (one colony) and the leaf-cutter bioassay (two colonies). Replication across a larger number of colonies is essential to evaluate potential colony effects and to assess the consistency of the data at the population level. Although these limitations do not appear to have affected the leaf-cutting assays, given the marked behavioral differences between male and female workers of S. dirus, we advocate for broader sampling in future neuroanatomical studies, in order to robustly generalize brain morphological patterns across colonies.

Resource availability

Lead contact

Further information and requests should be directed to and will be fulfilled by the lead contact, Iago Bueno da Silva (buenoiago2@gmail.com).

Materials availability

This study did not generate new, unique reagents. The reagents are commercially available and are detailed in the key resources table.

Data and code availability

• •Data supporting the results are shown in the Supplementary material (Tables S1 and S2, and Videos S1 and S2) and deposited on Zenodo: https://doi.org/10.5281/zenodo.18310697. Partial sequences of Sdforaging and SdRpl32 are deposited in NCBI, GenBank: PV254826, PV254827.
• •This study did not generate new codes. The codes employed for the analyses of public transcriptomes are available on GitHub: https://github.com/aroquej/foraging_project_termites.
• •Any additional information is available from the lead contact upon request.

Acknowledgments

The authors are grateful for the financial support from Fundação de Amparo à Pesquisa do Estado de São Paulo (I.B.S.: 24/02356-1, F.S.N.: 21/05598-8, A.R.J.: 24/06229-4), Conselho Nacional de Desenvolvimento Científico e Tecnológico (K.H.: CNPq 302209/092843), and Fundação Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (A.R.J. e J.E.B.: finance code 001). I.B.S. holds a Fyssen Postdoctoral Fellowship (Fyssen Foundation, France). We also thank Prof. Reginaldo Constantino for the help in identifying the species, the Laboratório Multiusuário de Microscopia Confocal (LMMC), Elizabete Rosa Milani for the technical assistance with the confocal analysis, Prof. Judith Korb for making the M. bellicosus genome available for our analysis, and Prof. Carlos Sarmiento for fruitful comments on a prior version of the manuscript.

Author contributions

Conceptualization, I.B.S., A.R.J., and F.S.N.; data curation, I.B.S., A.R.J., J.E.B., and L.V.; formal analysis, I.B.S., A.R.J., and L.V.; funding acquisition, K.H. and F.S.N.; investigation, I.B.S., A.R.J., K.H., and F.S.N.; methodology, I.B.S., A.R.J., J.E.B., and L.V.; project administration, F.S.N.; software, I.B.S., A.R.J., and L.V.; resources, K.H. and F.S.d.N.; supervision, K.H. and F.S.N.; validation, I.B.S., A.R.J., K.H., and F.S.N.; visualization, I.B.S., A.R.J., L.V., H.K., and F.S.N.; writing – original draft, I.B.S. and A.R.J.; writing – review and editing, I.B.S., A.R.J., L.V., K.H., and F.S.N.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse monoclonal anti-Synapsin (α-Synorf1)	Developmental Studies Hybridoma Bank (DSHB)	Cat#3C11; RRID:AB_528479
Goat anti-mouse IgG, Alexa Fluor 594	Thermo Fisher Scientific	Cat#A-11032; RRID:AB_2534091
Bacterial and virus strains
Escherichia coli dh5α	Thermo Fisher Scientific	Cat#18265017
Software and algorithms
Amira (v5.3.3)	FEI/Thermo Fisher Scientific	RRID:SCR_007353
R v4.4.1	R Core Team	RRID:SCR_001905
ImageJ (v. 1.53e)	NIH	RRID: SCR_003070
Oligo Explorer (v.1)	Gene Explorer TM
Kallisto	Pachter Lab	RRID: SCR_016582
Sleuth	Pachter Lab	RRID:SCR_016883
BBMap	SourceForge	RRID: SCR_016965
BUSCO (Benchmarking Universal Single-Copy Orthologs)	Evgeny Zdobnov Lab	RRID:SCR_015008
Chemicals, peptides, and recombinant proteins
Triton X-100	Sigma-Aldrich	Cat#T8787
Paraformaldehyde (PFA)	Sigma-Aldrich	Cat#P6148
Bovine serum albumin (BSA)	Sigma-Aldrich	Cat#A9647
Sodium azide (NaN3)	Sigma-Aldrich	Cat#S2002
Ethyl alcohol, 200 proof	Merck	Cat#1.00983
2-Propanol Alcohol	Sigma-Aldrich	109634
Chloroform	Sigma-Aldrich	102445
Methyl salicylate	Sigma-Aldrich	Cat#M6752
DNase I (10×) + Buffer	Thermo Fisher Scientific/Invitrogen	AM2222/AM8170G
RNase A	Sigma-Aldrich	10109142001
Phosphate-buffered saline (PBS)	Thermo Fisher Scientific	Cat#10010023
Critical commercial assays
Illustra GFX PCR DNA and Gel Band Purification Kit	Cytiva	Cat#28903470
Ultrascript cDNA Synthesis Kit	PCR Biosystems	PB30.11–10
qPCRBIO SyGreen Mix Separate-Rox	PCR Biosystems	PB20.14–05
Deposited data
Supporting data	This paper	Zenodo: https://doi.org/10.5281/zenodo.18310697
Sdforaging partial sequence	This paper	GenBank: PV254826
SdRpl32 partial sequence	This paper	GenBank: PV254827
Oligonucleotides
Primers for Sdfor clonning	This paper	N/A
Primers for SdRpl32 clonning	This paper	N/A
Primers for Sdfor qPCR	This paper	N/A
Primers SdRpl32 qPCR	This paper
Primers NADH-dh qPCR	Hojo et al.96
Recombinant DNA
pGEM - T Easy vector	Promega	Cat#A1380

Experimental model and study participant details

Termite species and sampling

Three colonies of Syntermes dirus (Burmeister) found in the campus of the University of São Paulo, Ribeirão Preto, Brazil (FFCLRP-USP) were chosen for this study. Workers of S. dirus are divided into two lineages, males and females, comprising 3rd and 4th-instar individuals (Figures 1A and 1B). We selected males and females from the 4th instar, as well as soldiers, based on morphological features.⁵⁷ We did not include 3rd instar workers in our study because they are rarely observed at nest entrances. Thus, we hereafter refer to male and female workers as those of the 4th instar.

Method details

Division of labor under field conditions

The three colonies were monitored daily for about one month to investigate which tasks were performed by male and female workers. Behavioral observations were conducted over 15 min at 27 different nest entrances (7 for colony 1 and 10 for each of colonies 2 and 3), totaling 27 inspections and 405 min of observation. For each inspection, the number of the behavioral events and the caste involved were quantified as follows: foraging (i.e., cutting pieces of leaves and carrying them to the nest), building (i.e., carrying moistened soil and modeling it at the nest entrance through head and mouthpart movements), and soil deposition (i.e., leaving the nest carrying dry soil pellets with the mandibles and randomly placing them farther away from the entrance).

Additionally, we aimed to quantify the duration males and females spent outside the nest while performing their respective tasks (see the results section for the behavioral repertoire). To do so, we randomly selected 15 male and female workers from each of the three nests and tracked them individually from the moment they left the nest to perform their tasks until they returned inside. The time spent outside the nest by each individual was quantified using a chronometer.

Leaf-cutting bioassay

As we noted a few cases where females were carrying small pieces of leaves (see the results section, Figure 2A), we aimed to test whether these individuals contributed actively to the foraging process by selecting and cutting leaves, or whether they were opportunistically collecting small pieces of leaves disposed next to the nest entrance. For this purpose, ten male- and female-based groups (five from two different colonies) were kept in 9 cm Petri dishes filled with moistened soil and a rounded piece of leaf. Each group comprised five individuals sampled during their above-ground activities. The leaves were collected around nest entrances and cut with scissors into standardized circles of 1,100 mm². The groups of termites were left overnight in the dark at 25 °C and inspected the next morning. The leaf cutting was determined by comparing the initial and final areas of the rounded leaves. Before and after the experiment, the leaves were individually photographed with a scale bar and imported into the software ImageJ (v. 1.53e). After calibrating the scales, the surface of each leaf was selected using the “Color threshold” and “Wand” tools. The respective area was then calculated using the “ROI manager” tool.

Brain dissection, fixation, and incubation

For brain volumetry, individuals (males and females) were sampled from colony 1 during their foraging and building activities, respectively, and kept under laboratory conditions until brain dissection. Termites were briefly cooled on ice, their heads were removed with a blade, and photographed in a Petri dish against a white background using a Leica M125 C stereomicroscope with a 2 mm scale bar. The following measurements were extracted: maximum head width, head length, and head depth (estimated by ½ head width). Head volume was calculated by applying the formula for an ellipsoid: ¾ ∗π ∗ ½ head width ∗ ½ head height ∗ ½ head depth.^51,53 The heads were then kept in PBS solution (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO_4, pH 7.4) and opened with forceps. Each brain was carefully isolated and individually fixed in PFA solution (4% paraformaldehyde, pH 7.4) for 2 h at room temperature (RT). The samples were washed on a shaker in PBST 0.2% (PBS +0.2% Triton X-100) eight times for 30 min each.

Brains were immersed overnight in an antibody-blocking solution (10% PBS 0.1M, 2% bovine serum albumin, and 0.004% sodium azide (NaN₃) diluted in PBST 0.2%) at RT on a shaker. Next, they were incubated in 1:30 primary antibody against synapsin (α-Synorf1, DSHB, USA) for 3 days at RT. After washing in PBST 0.2% eight times for 30 min each, they were incubated in a 1:300-diluted fluorescence-labeled secondary antibody (Invitrogen Alexa Fluor 594, Fragment of Goat Anti-Mouse IgG, Thermo-Fisher, USA) for 3 days at RT. Finally, the brains were dehydrated in a graded ethanol series (50, 70, 80, 90, 2x 98, and 100%) for 30 min each, transferred to glass vials containing methyl salicylate, and kept for at least one week, aiming to improve sample clearing.

Confocal laser scanning microscopy and volumetry

A total of 11 brains from males and nine from females were individually analyzed. Images of whole brains were acquired on a Leica SP5 confocal microscope at 1024 x 1024 pixels, using 10× dry objectives on sequential slices of 1.3 μm optical thickness. The files were imported into the software Amira (v. 5.3.3), and the neuropil volume was measured by individually outlining the following compartments with the brush tool: mushroom bodies (MB), divided into calyces and lobes, antennal lobes (AL), and the central complex (CX). The sum of these neuropils was used as a proxy for the investment of the “sensory and integration brain region”.⁵³ Since male and female workers differed in their head volumes, comparisons were performed using both absolute and relative measures. Relative brain volume was estimated as the ratio between the sensory and integration brain area and head volume, and relative neuropil volume as the ratio between individual neuropil volume and the sensory and integration brain area.

Cloning and partial sequencing of the Sdfor and Sdrpl32 genes

Since the S. dirus genome has not been sequenced so far, to identify the putative orthologs in S. dirus, we blasted (BLASTN) the Apis mellifera sequences of the genes Ribosomal protein L32 (Rpl32 isoform X1: XM_006564315.3) and foraging (cGMP-dependent protein kinase for: XM_026444521.1) against the nucleotide sequences available for the order Blattodea. We retrieved Rpl32 orthologs from Cryptotermes secundus (XM_023846691.2), Coptotermes formosanus (BLKM01005267.1), and Zootermopsis nevadensis (XM_022059022.1), and orthologs of the foraging gene from Reticulitermes flavipes (MK764266.1 and MK764267.1), Reticulitermis chinesis (OM831127.1), and C. secundus (XM_023857366.2 and XM_023857367.2). These sequences were aligned, and degenerate primers for cloning the respective S. dirus sequences (Table S1) were designed using the Oligo Explorer tool v. 1.5 (Gene Explorer TM).

For the cloning process, conventional Polymerase Chain Reactions (PCR) were performed using the following protocol: 95 °C for 2 min, followed by 40 cycles at 95 °C for 30 s and 40 °C for 60 s, and a final extension of 10 min at 40 °C. The PCR products were loaded onto 1.5% agarose gels for electrophoresis. The respective bands were purified using the Illustra GFX PCR DNA and Gel Band Purification Kit (Cytiva) following the manufacturer’s instructions. The purified products were then ligated into a pGEM T-easy vector (Promega) for transformation of competent DH5ɑ E. coli grown overnight at 37 °C in solid LB medium with 100 μg ⨉ mL⁻¹ Ampicillin (Sigma-Aldrich). Positive clones were transferred to liquid LB medium and left to grow overnight at 37 °C at 200 rpm, followed by DNA extraction. Briefly, 3 mL of LB medium was centrifuged at 10,000 × g for 10 min, and the supernatant was discarded. The bacteria pellet was subsequently resuspended in 200 μL buffer solution (50 mM glucose, 25 mM Tris-HCl pH 8, 10 mM EDTA pH 8), followed by addition of 200 μL of a lysis solution (NaOH 0.2 N, SDS 10%), and lastly 200 μL of ammonium acetate (10 M, pH 7.8). After 10 min of incubation on ice, the samples were centrifuged at 10,000 × g for 30 min. The supernatant was collected, and 480 μL of isopropanol was added for DNA precipitation, followed by another centrifugation at 10,000 × g for 10 min. The precipitated DNA was washed by adding 500 μL ice-cooled ethanol 70%, followed by centrifugation at 10,000 × g for 2 min. The pellet was dried and resuspended in DEPC water. The purified DNA was treated with RNase A (Sigma-Aldrich) for 10 min at 37 °C to eliminate residual RNA. The concentration was assessed in a Nanovue (GE Healthcare) spectrophotometer.

The sizes of the respective DNA products were confirmed by electrophoresis in a 1.5% agarose gel, and the identity of the sequences were confirmed via Sanger sequencing using M13 primers. Sequence orthology was confirmed by comparing the obtained fragments against known sequences in the NCBI database using BLASTN. The sequences were deposited in the NCBI database (Genbank accession codes Sdfor: PV254826; Sdrpl32: PV254827).

For the gene expression analyses, specific primers were designed using the software Oligo Explorer v.1 (Gene Explorer TM), based on the sequenced fragments obtained. To determine primer amplification efficiencies, separate pooled cDNA mixes were prepared for each tissue, with each pool containing samples from males, females, and soldiers. Primer efficiencies were assessed across these pools, and only the ones showing acceptable and comparable efficiencies among tissues were selected for downstream analyses (Table S1). Except for NADH-dh obtained from Hojo et al.,⁹⁶ all primers were specifically designed for this study.

Tissue dissection

To identify the putative role of the foraging gene in the division of labor in S. dirus, we sampled termite workers (male and female) and soldiers from the same colonies used for behavioral assays. All sampling procedures took place between 19:00 and 22:00. Male workers were collected during foraging trips, while female workers were sampled while depositing soil pellets at the nest entrances. Soldiers were sampled while standing at the nest entrances or patrolling the foraging workers. After sampling, the specimens were snap frozen in liquid nitrogen and stored at −80 °C. Their brains and abdomens, respectively representing potential behavioral and metabolic sites where for might act regulating division of labor, were then dissected in ice-cold Insect Ringer solution (128 mM NaCl, 17 mM Na₂HPO₄, 20 mM KH₂PO_4, pH 7.4). From the abdomens, we removed the gut and used only the carcass (mainly fat body, epidermis, and abdominal ganglia) to avoid contamination of the samples with non-termite RNA. After dissection, the respective tissues were transferred to 1.5 mL Eppendorf tubes containing 250 μL TRIzol reagent (Invitrogen), and then stored at −80 °C until RNA extraction.

RNA extraction and cDNA synthesis

For RNA extraction, we followed the manufacturer’s protocol (Invitrogen) with minor adaptations. Briefly, the dissected tissues of the termite workers (male and female) and soldiers kept at −80 °C were homogenized in TRIzol and kept under RT for 5 min before adding 50 or 100 μL chloroform (Synth). After 3 min at RT, they were centrifuged at 13,000 × g at 4 °C for 15 min. The aqueous phase was collected, transferred to a new microtube, mixed with 125 or 250 μL of ice-cooled isopropanol, and stored overnight at −20 °C. At this step, 0.25 μL of glycogen (20 μg/μL, Thermo Scientific) was added to the samples to improve RNA yield. After centrifugation at 13,000 × g for 30 min at 4 °C, the supernatant phase was removed, and 250 or 500 μL of ice-cooled 75% ethanol was added. Following centrifugation at 4,000 × g at 4 °C for 10 min, the supernatant was removed, and pellets were left to air dry at RT. The RNA pellets were solubilized in DEPC-treated deionized water, and 8 μL of the RNA samples was treated with 2 μL of RNA-free DNase (Ambion, Austin TX, USA) at 37 °C for 30 min. Then, 1 μL of EDTA (50 μM) was added and samples were incubated at 75 °C for 10 min. The concentration of the samples was normalized to 0.1 μg/μL, and 0.5 μg of RNA was used as a template for cDNA synthesis (cDNA Synthesis Kit, PCR Biosystems, London, UK) following the manufacturer’s protocol, diluted 10×, and kept at −20 °C for gene expression.

RT-qPCR analysis

The RT-qPCR assays were performed using the qPCR SyGreen Mix (PCRBiosystems). The reactions consisted of 5 μL of SybrGreen, 1.25 pmol of each primer, 1 μL of 10× diluted cDNA, and ultrapure water for a final reaction volume of 10 μL. The assays were carried out in a StepOnePlus Real-Time PCR system (Applied Biosystems, Waltham, MA, USA) under the following conditions: 95 °C for 2 min, followed by 40 cycles at 95 °C for 5 s and 60 °C for 22 s. The threshold cycle (Ct) values were used to calculate the relative expression of the genes of interest following the 2^−ΔΔCt methodology.⁹⁷ SdRpl32 and NADH-dh were used as reference genes, as they have been suggested to be stable for other termite caste comparisons.^96,98,99,100

Reanalysis of publicly available transcriptomes

To further investigate the role of the foraging gene in the division of labor across termite species, publicly available datasets were retrieved from NCBI repositories (for accession codes and associated publications, see Table S2A). We selected representative species of different termite families, including Zootermopsis nevadensis (Archotermopsidae), Neotermes binovatus (Kalotermitidae), Reticulitermes speratus (Heterotermitidae), Macrotermes bellicosus, and Inquilinitermes inquilinus (Termitidae), which also cover the diversity of ecological habits of termites: one-piece (OPT) and separate-piece (ST) nesters.^{14,101,102,103} OPT species (including Z. nevadensis and N. binovatus) are less socially complex, display flexible caste differentiation, and spend the whole lifetime in a piece of log that serves as shelter and food source. On the other hand, ST species (including S. dirus, R. speratus, M. bellicosus, and I. inquilinus) are more socially complex, displaying constrained developmental flexibility and central-place foraging.¹⁰⁴

We first employed Trimmomatic¹⁰⁵ to remove sequencing adapters and low-quality ends with default parameters. Next, the software BBMap¹⁰⁶ was used to index the coding sequences (CDS) of the reference genome and subsequently remove reads that did not map. This step was essential to improve sample quality and read mapping by removing symbiont and contaminant reads that were highly enriched in some libraries, particularly those generated from whole bodies or entire abdomens (Table S2B). Next, the software Kallisto¹⁰⁷ was employed for the indexing of the reference CDS with a k-mer of 31, and subsequent quantification and normalization of the reads based on pseudoalignments for each transcriptome with a 100 boostrapping. Finally, the quantification of the reads was compared and subjected to statistical analysis using the “Sleuth” package.¹⁰⁸ We employed the Wald test to compare data and check for statistically significant differences in gene expression, followed by the Benjamini-Hochberg adjustment test for multiple comparisons.

For species whose reference genomes were not available, we selected and merged representative libraries from each comparison and subsequently generated de novo transcriptome assemblies using the RNA Spades tool, with default settings.¹⁰⁹ Next, we removed symbiont and contaminant RNA by blasting the generated transcriptome assemblies against a local database consisting of mRNA sequences from Blattodea (Taxonomy ID: 85823) downloaded from NCBI, including cockroaches and termites. We established a minimum E-value of 1e⁻⁵⁰, and sequences that lacked a significant hit were removed. Next, we utilized the CD-hit-est tool¹¹⁰ to cluster isoforms and reduce redundant sequences. Finally, we performed a Benchmarking Universal Single-Copy Orthologs (BUSCO) analysis using BUSCO v5.8.2¹¹¹ to assess the completeness and redundancy of the newly-generated reference transcriptomes, using the Insecta lineage dataset (insecta_odb12) as query (see Table S2C for BUSCO results and selected libraries). The quantification and statistical analyses were performed as described for species with available genomes (see Table S2D for annotation of foraging transcript for each species analyzed).

Quantification and statistical analyses

All statistical analyses were carried out in R software (v. 4.4.1).¹¹² Different methods were employed to compare the behaviors of the workers observed in the field. To compare foraging and building activity between groups, represented as separated or nearly perfectly separated data, we employed a Firth’s bias-reduced logistic regression, available under the package “logistf”,¹¹³ with predictor significance evaluated using a likelihood ratio test. For this purpose, we assumed each behavior as dependent variable, and sex and colony as fixed effects. The soil deposition behavior and time spent on the surface were individually compared by a generalized linear mixed model (GLMM, “glmmTMB” package¹¹⁴) using Type-2 negative binomial and gamma distributions, respectively. We applied different models due to the presence of zeros and data overdispersion concerning the soil deposition behavior. In these cases, the behavior and time were assumed as dependent variables, and sex and colony as fixed and random effects, respectively. The significance was evaluated using a likelihood ratio test (LRT). For leaf consumption assays, we transformed data into binary variables (cut leaves = 1; do not cut leaves = 0). Due to the perfect data separation between males and females (see the results section), we also applied a Firth’s bias-reduced logistic regression followed by LRT. In this case, leaf cutting was set as a binomial variable, while sex and nest were assumed as fixed effects.

The neuropil volumes between male and female workers were compared using a Student’s t test or a Mann-Whitney U test according to data distribution. For gene expression analysis, data distribution and homoscedasticity of variance were assessed through Shapiro-Wilk and Levene’s tests, respectively. The lmer function of the “lme4” package¹¹⁵ was employed to create general linear models, and emmeans was used for pairwise comparisons. For these models, gene expression was assumed as dependent variable, tissue and sex as fixed effects, and colony was considered a random effect, and the residuals were further checked using the “DHARMa” package.¹¹⁶

Published: March 7, 2026