Head and mandible shapes are highly integrated yet represent two distinct modules within and among worker subcastes of the ant genus Pheidole

Abstract Ants use their mandibles for a wide variety of tasks related to substrate manipulation, brood transport, food processing, and colony defense. Due to constraints involved in colony upkeep, ants evolved a remarkable diversity of mandibular forms, often related to specific roles such as specialized hunting and seed milling. Considering these varied functional demands, we focused on understanding how the mandible and head shape vary within and between Pheidole subcastes. Using x‐ray microtomography and 3D geometric morphometrics, we tested whether these structures are integrated and modular, and how ecological predictors influenced these features. Our results showed that mandible and head shape of majors and minor workers tend to vary from robust to slender, with some more complex changes related to the mandibular base. Additionally, we found that head and mandible shapes are characterized by a high degree of integration, but with little correlation with feeding and nesting habits. Our results suggest that a combination of structural (allometric) constraints and the behavioral flexibility conferred by subcaste dimorphism might largely buffer selective pressures that would otherwise lead to a fine‐tuning between ecological conditions and morphological adaptation.

In seed harvester ants, seeds are collected by foraging workers (i.e., slender body and mandibles). These workers then bring the seeds into the nest where they can be stored. Later, larger workers with specialized phenotypes (i.e., robust body and mandibles) can process them. Such species often inhabit desert and savanna ecosystems (Brown et al., 1979) but can also be found in tropical forests (Wilson, 2003). Among the several ant genera that exhibit this behavior, Pheidole Westwood (Myrmicinae: Attini) can be considered an ideal study system due to its inordinate diversity, wide distribution, and the presence of a specialized subcaste.
Pheidole is the most diverse genus among ants, with 1,167 extant species (Bolton, 2020). However, estimates suggest the existence of at least 1,500 Pheidole species (Wilson, 2003) and possibly well over 2,000 . One of the most remarkable peculiarities of Pheidole is the conspicuous dimorphism between its workers. This dimorphism is characterized by the division between "minor workers" and more robust and macrocephalic workers, known as "major workers" or "soldiers" (Wilson, 2003). Both subcastes also have morphological modifications involving the reduction of the sting apparatus and the absence of functional ovaries. The sting apparatus in Pheidole is quite atrophied (Kugler, 1978) and is likely to be nonfunctional. Along with sting simplification, the absence of functional ovaries (Hölldobler & Wilson, 2009) can result in a lower body volume, decreasing the energy cost involved in worker production for the colony, which allows for a large number of individuals coexisting in the nest (Hölldobler & Wilson, 2009), and the potential caste specialization related with sterility (Oster & Wilson, 1978).
Pheidole majors are morphologically and behaviorally adapted in the nest and resources defense, food transportation and processing, as well as food storage (Mertl et al., 2010;Mertl & Traniello, 2009;Sempo & Detrain, 2004;Wilson, 1984Wilson, , 2003. Pie and Traniello (2007) showed that morphological variation in majors and minor workers can be attributed mainly to allometric changes, with little dissociation between morphological features. Additionally, the authors demonstrated a lower degree of morphological integration in major workers than in minor workers, which was also found by Friedman et al., (2019) using different approaches for both measurements and analysis.
Although most Pheidole species are associated with generalist and scavenger habits, seed predation and harvesting account for a significant portion of their diet and have evolved independently multiple times in New and Old World species (Economo et al., 2015;Moreau, 2008). Macrocephaly is often correlated with seed processing in ants, and therefore with the demand for considerable muscle strength (Ferster et al., 2006). However, recent studies have questioned the relationship between worker head size and granivory in Pheidole. Holley et al., (2016) studied this relationship, predicting that majors in seed-harvesting species should have the musculature optimized to open the seed, which would produce wider heads compared with those that do not consume seeds. Nevertheless, the authors found that Pheidole species did not exhibit a relationship between cephalic size and seed-milling behavior. In contrast, Holley et al., (2016) reported that there is a greater difference in head size between major and minor workers in the granivorous species when compared to nongranivorous ones. However, these previous studies are based on 2D or linear morphometrics. The recently developed 3D approaches made it possible to capture almost all the shape variation, consisting of the most powerful tool to test for these issues.
These approaches are especially useful when considering structures that are difficult to quantify, given their irregular surfaces (e.g., insect mandibles and mesosoma).
Considering the morphological complexity in Pheidole, we used x-ray microtomography and 3D geometric morphometrics to investigate how mandible and head shape vary within and between its subcastes. Also, we tested whether these parts are integrated and modular, how much the allometric effect is responsible for differences in shape, and how ecological predictors influenced these features.
Due to the worker's conspicuous dimorphism, mainly related to size, as already pointed out by Pie and Traniello (2007), we predicted a strongly allometric effect on the head and mandible shape for majors and minors, following the pattern found for the genus. Because of the important role of the head in bearing the muscles associated with mandibles, we hypothesized that these structures (head and mandibles) are strongly integrated and should be part of the same module, with less covariation between modules than within them.
In this scenario, the shape of head and mandible would covariate jointly, as the mandibular demands would be met by the shape of the head, thus optimizing its functions. However, some alternatives are possible, for instance, the lack of integration and modularity between head and mandible, as well as the indication of strong integration with most covariation occurring between modules instead of within them.

The lack of head and mandible integration and modularity in
Pheidole would be possible if the mandibular demands do not depend directly on the head shape for the muscular volume and disposition. Thus, the mandible could have evolved as a structure semi-independent from the rest of the head, which would result in a great plasticity and diversity of shapes. Additionally, it can indicate that generalized head shapes would allow different mandible forms to meet their demands. As demonstrated by Holley et al. (2016), the massive head volume in Pheidole major workers is not directly associated with pressures involving the requirement for a strong capacity for process seeds (i.e., mechanical function), thus, suggesting the possible dissociation between mandible and head shape. The second possibility is that these structures are strongly integrated, however, that they are not part of the same module. Considering the integration without modularity (Roseman et al., 2009), the genetic effects would not be strongly grouped as expected for structures that display significant modularity; instead, they may be spatially restricted but continuous, even overlapped (Zelditch et al., 2012). This would allow mandible and head to vary their shape similarly, without being affected as a single unit in its developmental and evolutionary process.
Several studies have pointed out that lifestyles (e.g., fossorial, marine, and parasitic) and feeding ecologies may be the main predictors for diversification and specialization of forms in different organisms (Da Silva et al., 2018;Olsen, 2017). Indeed, insects have mouthparts with characteristic adaptations that evolved in a context mainly related to the variety of exploitable food sources, which resulted in feeding specialization for better functional performance in many lineages (Krenn, 2019). To optimize mandible and head shape to explore their environment while considering the potential impact that ecological predictors play in morphological changes, we expect that feeding and nesting preferences should affect morphological patterns of these structures in Pheidole workers. However, an alternative possibility is that workers' dimorphism, as well as the behavioral flexibility they provide for the worker force, can reduce ecological pressures on head and mandible shape, which could enable these structures to vary independently of feeding and nesting habits.
We envision three contrasting scenarios of morphospace occupation for minor and major workers. Because majors tend to be behaviorally specialized for food processing and defense, we expect that head and mandible shape in these individuals would be strongly constrained to a set of optimal morphological optima when compared to minors. This would lead to a more restricted morphospace into which major workers could diversify, whereas minors would be freer to evolve different forms. Alternatively, given that minors are responsible for most quotidian colony functions, such as brood care, foraging, and nest maintenance, they would be more constrained by the need to perform simultaneously such a variety of tasks, whereas major workers could be freer to evolve into more specialized morphologies. Finally, major and minor morphologies could be so intimately linked in their developmental pathways that they would constrain one another in their evolutionary possibilities, such that their morphospace occupation would essentially mirror one another.

| MATERIAL AND ME THODS
We compiled a dataset of 3D models of major and minor workers (one specimen per subcaste) of a total of 27 Pheidole species (N = 54 specimens) (Table S1). We assumed that the morphological variation between individuals of the same subcastes within the same species is relatively small when compared to the variation between different subcastes and different species. We selected our focal species based on the conspicuous morphological variation among them, as well as the availability of material and the reliability of the associated biological data. However, as no molecular data are available for most of the selected species, explicit phylogenetic comparative analyses were not performed.
Our dataset represents the main groups recognized in the New World, where they cover a broad geographical distribution from the South of the United States to the South of Brazil. Additionally, they represent the main feeding and nesting habits found in Pheidole. We categorized their feeding behavior as granivorous and nongranivorous, and their nesting habits as soil, twigs, and plants. We chose to include species with missing biological data so that we could build a morphospace encompassing the largest possible morphological variation in the genus.
A micro-CT/μCT scans ZEISS Xradia 510 Versa (Carl Zeiss AG) generated the scans used to construct those 3D models. Scan settings were selected according to yield optimum scan quality: 4x objective, exposure times between 1 and 5 s, source-filter "Air," voltage between 30 and 50 keV, power between 4 and 5W, and field mode "normal." The combination of voltage, power, and exposure time was set to yield intensity levels between 15,000 and 17,000 across the whole specimen. Scan times varied from 27 to 50 min, depending on exposure times. Full 360-degree rotations were done with a number of 801 projections. The resulting scans have resolutions of 1013 × 992 × 999 (H × W × D) pixels and voxel sizes range between 2.25 and 5.39 μm. The processing and postprocessing of raw DICOM data were performed with Itk-snap 3.8.0 (Yushkevich et al., 2006). Desired volume renderings were generated by adjusting the range of color space to a minimum so that the outer surface of specimens remains visible at the highest available quality. The 3D models were rotated and manipulated to allow a complete virtual examination of scanned specimens. Semilandmarks were divided into three curves and one patch, such that one curve started in the L7 and ended in L1; second, it started in L7 and ended in L6; the third started in L6 and ended in the L5 ( Figure 1d). Lastly, the patch was placed connecting the four anchor points in L1, L5, L6, and L7 ( Figure 1d). We opted to apply the measurements to a single half of the head due to bilateral symmetry. We measured mandible shape by digitizing 12 landmarks and 41 semilandmarks using the same software. The landmarks corre- All geometric morphometrics analyses were performed in R (R Development Core Team, version 3.6.2. 2019) using the geomorph package v. 3.3.2 . Landmark and semilandmark coordinates were aligned using Generalized Procrustes Analysis (GPA), in which the specimens were translated to a common location, scaled them to unit centroid size, and optimally rotated them using the least-squares criterion (Rohlf & Slice, 1990). To test the influence of allometry, we used a Procrustes regression (Goodall, 1991;Adams & Collyer, 2018) of independent contrasts of shape (Procrustes coordinates) on independent contrasts of size (centroid size). Allometry was tested within both subcastes combined in a single dataset and for each subcaste separately. Statistical significance was assessed using a residual randomization permutation procedure (1,000 permutations). Additionally, we performed a Procrustes regression to access variation in allometric slopes between subcastes, testing common or unique allometries. To visualize shape changes associated with the allometric variation, we estimated the average allometric trends within groups (i.e., subcastes) using the common allometric component (CAC; Mitteroecker et al. 2004) approach.
Principal component analyses (PCAs) were performed to visualize the distribution of species shape along different axes in tangent space for majors and minor workers combined, as well as for each subcaste separately. Our discussion will focus on the axes that together explain at least 70% of the total variation. Thin-plate spline deformation grids (Bookstein, 1991) were employed to visually describe the shape differences. To test the morphological integration between head and mandible, we ran a two-block partial least squares analysis for Procrustes shape variables with 1,000 permutations (Adams & Collyer, 2016;Rohlf & Corti, 2000). For the modularity test, we combined the mandible and head dataset with normalize centroid size  and considered the degree of modularity in two hypothesized modules (head and mandible) of Procrustes shape variables, comparing this to a null hypothesis of random assignment of variables, with 1,000 permutations (Adams, 2016).
We assessed the influence of the ecological predictors on the shape patterns using linear models of shape variation. The analyses were performed in R (R Development Core Team, v. 4.0.3, 2020) using the geomorph v. 3.3.2 and rrpp v. 0.6.2 packages . We performed a Procrustes regression considering a residual randomization permutation procedure (Collyer et al. 2015) and contrasting the linear models, using the simple allometry as a null model. For this analysis, we employed three biological models with the null model nested with all the others. Additionally, we ran a model comparison based on log-likelihoods for size and shape,  including an adjusted tolerance for the shape models due to the number of variables exceeds the number of observations.

| RE SULTS
We found a strong allometric effect for both head (p = .001) and mandible shape (p = .001) using the combined dataset (major and minor workers), as well as for major and minor heads (p = .003 and p = .049, respectively), and minor mandibles (p = .002) when analyzed separately. Allometric trajectories of major and minor worker heads exhibit significant differences, with size accounting for 20% of the total shape variation (p = .001), subcastes explaining 15% (p = .001), and the interaction between size and subcaste describing 5% of the remaining variation (p = .006) ( Table 1). Mandibles also show significant differences, in which size explains 28% (p = .001), subcastes represent 43% of shape variation (p = .001), whereas their interaction was not significant (Table 1). The homogeneity of slopes test found a significant (p = .006) difference between allometric slopes of the head, suggesting nonparallel slopes between majors and minors (Table 1). Additionally, this test indicated no significant (p = .067) difference between allometric slopes for the mandible, which implies parallel slopes (Table 1).
The CAC plot for the head shows two separated slopes (ma- The reconstructed morphospaces (Figures 5 and 6) indicated an overlap in morphology among species with different food preferences, as well as nesting habits, with no statistically significant differences among these ecological predictor categories. This pattern suggests that variations in head and mandible for majors and minors could be relatively independent of ecological pressures.
Procrustes analysis of variance showed that size is more important for mandible and head shape of majors (Z = 1.05 and 1.78, respectively; Table 2) and minor workers (Z = 2.43 and 1.44; Table 2) than the ecological predictors. However, size was the only statistically significant for major heads (p = .05; Table 2) and minor mandibles (p = .01; Table 2). Considering the effect of food preference and nesting habits on size, our results suggested no statistical significance (Table 2). But higher Z values were related to nesting for

F I G U R E 4 Principal component analysis of the head (a and b) and mandible (c and d) shape of Pheidole workers (a and c) with and (b and d) without allometric effects. Deformation models indicate extreme shapes along PC
major mandible size (0.05; Table 2) and food preference for head size (0.71; Table 2); while minors presented higher values associated with food preference considering mandible (0.88; Table 2) and head sizes (0.80; Table 2).
The linear models show results congruent with those found in the morphospaces. Our results showed that the model including only size provided the best fit to our data (Tables 3 and 4). Additionally, for models involving ecological effects on size (Table 4), our results suggest that models including food preference has the best fit to our data. The only exception is for the minor mandible size, in which the model that includes only nesting habit has the best fit (Table 4).
Conversely, majors showed a high Z value in the model that considered the influence of food preference on mandible (1.30; Table 3) and head (0.12;  Table 3) and head (0.48; Table 3).

Regarding morphological integration, our results indicated significant scores with high r-PLS values for all combinations between and
within majors and minor workers, and between their mandible and head (Figure 7). This result implies that morphological changes in the mandible were accompanied by corresponding changes in the head.

| D ISCUSS I ON
Changes related to size and shape, their evolutionary history, and the effect of ecological predictors driving those changes are a cornerstone of macroecology and macroevolution (Bright et al., 2016;Dumont et al., 2011;Olsen, 2017). In ants, workers of about 13% of the species are subdivided into morphologically variable and ecologically adapted subcastes (Oster & Wilson, 1978), so understanding this pattern and how they are influenced by biotic and abiotic predictors is a key concern in myrmecology. Several studies have explored general morphological consequences of this division into subcastes and their morphological adaptations (Holley et al., 2016;Powell, 2008Powell, , 2009Powell & Franks, 2005Tschá & Pie, 2019). However, with new sources of information, such as mi-croCT, new questions can be postulated considering body parts that were not previously explored, due to constraints in the quantification of their size and/or complexity.
Here, we used a comparative approach to explore how the head and mandible shape of workers vary and are affected by allometry, modularity, and integration in Pheidole, while also describing their morphospace. Mouthparts are often viewed as one of the targets for natural selection, which is also expected considering the ecological demands associated with the division of labour between Pheidole majors and minor workers. This is also true for the head, as this body region holds the muscles and the appendages related to the feeding and colony maintenance, as well as the sense organs. Contrary to the belief about ecological pressures over mandible and head shape, we find that, in Pheidole, morphological variation of these structures is characterized by a strong degree of independence between them coupled to high integration and a pronounced allometric effect, with little correlation with food preferences and nesting habits.
Our integration results indicate that there is greater variation in the head shape of major workers, which would not be directly associated with the mandibular demands (Figure 7). The strong degree F I G U R E 6 Principal component analysis of the head (a and b) and mandible (c and d) shape of Pheidole minor workers considering (a and c) food preference and (b and d) nesting habitat. Deformation models indicate extreme shapes along PC of independence between head and mandible for majors and minors suggests that these structures display significant modularity when compared to the null hypothesis of no modular association (Adams, 2016). Strong integration and significant modularity could represent a key factor in shaping the morphology and diversity of heads and mandibles in Pheidole workers, thus constraining the possible morphological patterns that these structures can assume, in spite of ecological pressures.
We found a strong relationship between head and mandible with size, both for minor workers and head shape in majors. This suggests that size is an effective mechanism by which Pheidole workers may adjust to biological demands (e.g., changes in food preference and nesting habits). Our analysis also showed that, although there is a significant change in shape related to size within the worker caste, allometry does not explain all shape variation (Table 1). Differences between subcastes had a considerably high effect on the head shape, but with size explaining most of the variation, whereas for the mandible shape, the allometric effect was significantly lower than the subcaste (Table 1). These results suggest that majors and minor workers are more different than expected only due to size variation, which implies that morphological changes may be largely related to different tasks that each subcaste performs within the colony. These results corroborate Pie and Traniello (2007), who argued that worker shape in Pheidole varies predominantly through changes in size, even considering that this genus inhabits diverse ecological niches. Analysis of common allometric component (CAC) showed that size-related shape changes are mostly concentrated on the elongation of the head and thickness of the mandible. These results described a contrast between small individuals with relatively square-shape heads and slender mandibles, while large individuals TA B L E 2 Results of Procrustes analysis of variance considering the head and mandible shape and size of Pheidole majors and minor workers Note: In the analysis of shape, the effect of size, food preference, and nesting were considered for head and mandible. In the analysis of size, only the effect of food preference and nesting were considered. Bold values indicate significant results.

TA B L E 3
Results of model comparison using analysis of variance with randomization of residuals in a permutation procedure considering the head and mandible shape of Pheidole majors and minor workers

F I G U R E 7
Results of integration (r-PLS) and modularity (CR) tests for mandible and head of Pheidole workers; (a) major and (b) minor workers. The arrows indicate the modules and its respective values influence the head shape of seed-harvesting species, nor its size, as previously shown by Holley et al., (2016). Wheeler (1910) already hypothesized the absence of a direct relationship between head shape and granivory, arguing that morphological modifications related to the seed processing would also be needed to tear exoskeletons from other insects commonly collected by these ants. Indeed, our results indicate that potentially not only head shape is not affected by these requirements, including nesting habit, but also that this seems to be the case for mandible shape, in which distinct morphologies that occupy morphospace would compensate for pressures related to these ecological predictors. An alternative possibility might be related to behavioral or physiological aspects, which in this case would facilitate the exploration of different niches, regardless of the head and mandible shapes.
These results should be taken with the caveat that we did not consider these correlations in a phylogenetic context, and the sample size of species was relatively small. However, the main risk of an unmeasured phylogenetic effect would be a spurious ecologymorphology correlation, not a lack of correlation, so we do not think it likely that consideration of phylogeny will qualitatively change our results. To test this expectation, a more comprehensive study including a more comprehensive number of species from different biogeographical regions and with reliable biological and molecular data is necessary.
Aside from the morphological variation in the mandible shape for majors and minor workers regarding thickness, and continuous variation from robust mandibles to slender ones, and continuous variation in the curvature, our results showed an important variation related to the mandibular base. This region, which is often seen as more conserved, exhibited a set of key modifications such as its angle to the internal margin, the size, and shape of the atala (Richter et al., 2020), as well as the internal articulation margin. These changes are related to the muscle insertions (i.e., atala with the opening muscles) and points of articulation with the head. Variations in those regions would be directly associated with changes in the way these ants use their mandibles to interact with the environment; however, more data considering behavior and biomechanical properties are necessary to elucidate their functionality. Additionally, major workers exhibited more variation in the elongation of the mandible, as well changes in its thickness (PC1 - Figure 7a, b), when compared to the minors (PC1 - Figure 7c, d).
Our results also suggest that variation is much stronger in major In African cichlid fishes, the explosive evolution in several species with morphological highly specialized oral jaws was mainly caused by the presence of a pharyngeal jaw (Albertson et al., 2005;Burress et al., 2019;Kocher et al., 2004). For these fishes, the pharyngeal jaw, as it plays a role in the basic chewing functions, enabled the mandible to diversify into several different morphological patterns (Albertson et al., 2005;Burress et al., 2019;Kocher et al., 2004;Liem, 1973;Liem & Osse, 1975). The same occurs in the process of evolution of new genes, in which gene duplication, seen as one of the most important contributors to this process, allows new copies to mutate, as their functions will not be lost due to its duplication (Kaessman, 2010;Long et al., 2003Long et al., , 2013Taylor & Raes, 2004).
Considering head morphology, contrary to that found for mandible variation, minor workers head diversified along different shape axes (Figure 6c, d), with PC1 being more important for minors than for majors (Figure 6a, b). Regarding minor workers, most of the variation in head shape was related to elongation and thickening. This enables a significant increase in muscle adhesion surface, as well its size, consequently promoting growth in its volume. Additionally, head diversification in minor workers is slightly constrained in its morphospace, when compared with majors, following a similar pattern as that suggested for mandible diversification in our results. As discussed, this higher volume may not be related to specific food preferences, either nesting habit. However, it could enable a considerable advantage for defense, food loading, as well as the processing of a variety of food items (e.g., Wheeler, 1910). Wilson (2003) has postulated that Pheidole's hyperdiversity may result from the dimorphism between its workers, the majors being specialized individuals related to specific tasks such as defense, and food transport and processing, as well as assuming distinct functions in the absence of minors. On the other hand, Mertl el al. (2010) proposed that the occupation of niches in Pheidole would have as its main cause behavior-related adaptations, and morphological modifications would be less involved in its speciation.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data for this paper can be accessed on Dryad at https://doi. org/10.5061/dryad.1rn8p k0sz.