Simulated biomechanical performance of morphologically disparate ant mandibles under bite loading

Insects evolved various modifications to their mouthparts, allowing for a broad exploration of feeding modes. In ants, workers perform non-reproductive tasks like excavation, food processing, and juvenile care, relying heavily on their mandibles. Given the importance of biting for ant workers and the significant mandible morphological diversity across species, it is essential to understand how mandible shape influences its mechanical responses to bite loading. We employed Finite Element Analysis to simulate biting scenarios on mandible volumetric models from 25 ant species classified in different feeding habits. We hypothesize that mandibles of predatory ants, especially trap-jaw ants, would perform better than mandibles of omnivorous species due to their necessity to subdue living prey. We defined simulations to allow only variation in mandible morphology between specimens. Our results demonstrated interspecific differences in mandible mechanical responses to biting loading. However, we found no evident differences in biting performance between the predatory and the remaining ants, and trap-jaw mandibles did not show lower stress levels than other mandibles under bite loading. These results suggest that ant feeding habit is not a robust predictor of mandible biting performance, a possible consequence of mandibles being employed as versatile tools to perform several tasks.

mechanical demands need to be withstood by their mandibles.Additionally, we predicted that sturdier mandibles, such as those of major workers in polymorphic species like Pheidole spp., would perform better than slender mandibles in simulations of pressure biting, regardless of the main feeding habit of the species, due to their more robust constitution that allows for a better concentration of stresses around thicker regions of the mandibles 33 .By conducting this comparative study across a wide range of ant species, we aimed to improve our understanding of the relationship between mandible morphology and bite performance and shed light on the functional implications of mandible variation in ants.

Colour maps
In strike simulations, most mandibles of predatory and omnivorous species concentrate relatively higher stresses around the mandibular articulations with the head, while the stresses were comparatively lower along the mandible blade.Some species, however, showed a large surface area with intermediate to high stress levels along the mandible blade.Among predatory species, Acantognathus brevicornis, the major worker of Carebara sp.01, Lophomyrmex sp.01, Platythyrea cribrinodis, and Parasyscia sp. had high-stress concentrations around the mandibular articulations, whereas Bothroponera fugax, the minor worker of Carebara sp.01, the media worker of Eciton burchellii, Ectatomma edentatum, Myrmica ruginodis, and Octostruma petiolata also showed high levels of stress along the mandible blade.Similarly, omnivorous species such as Cephalotes pusillus, Formica fusca, Heteroponera dentinodis, Lasius niger, Odontomachus chelifer, the major worker of Pheidole aper, and Wasmannia affinis concentrated high levels of stress around the mandible articulations with the head, while Azteca sp., Camponotus zenon, Dorymyrmex sp., the minor worker of Pheidole aper and Solenopsis sp.04 exhibited relatively high stresses along the mandible blade too.Finally, the leaf-cutting species Acromyrmex aspersus showed intermediate to high stress levels along the mandible blade (Fig. 1a).
In terms of pressure biting, several predator species, including the major worker of Carebara sp.01, Ectatomma edentatum, Lophomyrmex sp.01, Octostruma petiolata, and Parasyscia sp., showed proportionally lower levels of stress along the mandible blade, a pattern found only in the major worker of Pheidole aper among omnivorous species (Fig. 1b).Meanwhile, some omnivorous species, such as Solenopsis sp.04 and the minor worker of Pheidole aper showed a large surface area in their mandible blade with proportionally high stress levels (Fig. 1b,d).The leaf-cutting ant Acromyrmex aspersus was among the species with lower relative stress levels along the mandible blade.Although most species showed relatively higher stress levels covering a larger surface area in pressure than strike biting, there were some important exceptions to this pattern, such as the mandibles of Acromyrmex aspersus, Ectatomma edentatum, Octostruma petiolata, Holcoponera striatula, and the major worker of Eciton burchellii, which exhibited only subtle differences in stress distribution between biting conditions (Fig. 1).
For simulations using only the apical tooth, we excluded the mandibles of Acantognathus brevicornis, Eciton burchellii, Odontomachus chelifer, and Strumigenys denticulata due to their specialized morphologies, in which the independent use of the apical tooth is improbable.In general, when comparing the use of the entire masticatory margin to employing only the apical tooth in strike (Fig. 1a,c) and pressure (Fig. 1b,d) simulations, we found no relevant differences.However, Holcoponera striatula and Lenomyrmex foveolatus represented two exceptions, where using the apical tooth only resulted in a more spread-out distribution of higher stress levels, representing a worse stress pattern (Fig. 1c,d).
Ectatomma edentatum and Holcoponera striatula, which are closely related and exhibit similarities in their mandible morphologies, showed clear distinctions in stress distribution in strike and pressure biting.Specifically, Holcoponera striatula showed a larger surface area of the mandible with relatively higher stresses than Ectatomma edentatum (Fig. 1).This finding suggests that Ectatomma edentatum exhibits superior biting performance compared to Holcoponera striatula.

Intervals method
To further explore the interspecific differences in mandibular stress distribution, we conducted PCAs for each biting scenario using the percentage of mandibular volume covered by 15 stress intervals as input variables.In strike simulations with the entire masticatory margin, the first two components of the PCA explained 85% of the variance.PC1, which accounted for 54% of the variance, represented stress intervals ranging from low to high, except for the highest stress interval, which was more closely associated with PC2.An increase in PC1 indicated essentially a larger mandibular volume covered by intervals of low stress (Fig. 2a).The negative range of PC1 included Eciton burchellii workers, Odontomachus chelifer, Strumigenys denticulata and Acantognathus brevicornis, all of which had a larger mandibular volume filled with higher stress intervals.In contrast, the positive range of PC1 featured species such as Cephalotes pusillus and the major workers of Carebara sp.01 and Pheidole aper, which had a larger mandibular volume filled with the lowest stress intervals (Fig. 2a).PC2 explained 30% of the variance and mainly distinguished mandibles with a higher coverage of the highest stress levels (interval 15) from the remaining species in its negative range.This pattern was observed in the major worker of Eciton burchellii, Strumigenys denticulata, and Acantognathus brevicornis.The positive range of PC2 was associated with an increase of intervals 5-9, but no species or group of species was distinctly isolated in this region of the PCA (Fig. 2a).
For simulations of pressure with the entire masticatory margin, the first two components of the PCA accounted for 87% of the variance.PC1 explained 46% of the variance and primarily distinguished species with a proportionally larger mandibular volume displaying higher stress (such as the major worker of Eciton burchellii and Strumigenys denticulata) from species more associated with intervals 2-5 (such as Holcoponera striatula) (Fig. 2b).PC2 explained 41% of the variance and was positively associated with intervals 8-12.It mainly differentiated Octostruma petiolata and Acantognathus brevicornis from other species in its negative range, as these species had lower proportions of their mandibles filled with such intervals (Fig. 2b).
In strike simulations using only the apical tooth, PC1 explained 68% of the variance and showed a positive association with intervals 2-7 while being negatively associated with intervals 13-14.Along the negative range of PC1, Lenomyrmex foveolatus and the median worker of Eciton burchellii displayed a larger mandibular volume with higher stress levels.In contrast, on the PC1 positive range, Cephalotes pusillus, the major worker of Carebara sp.01 and Pheidole aper stood out for having a larger mandibular volume with the lowest stress interval (Fig. 2c).PC2 explained 22% of the variance and was associated with a larger mandibular volume filled with intervals 9-11.This axis mainly differentiated the major worker of Pheidole aper and the median worker of Eciton burchellii from other species in its negative range, given their lower proportion of mandibular volume filled with such stress intervals (Fig. 2c).
In simulations of pressure using only the apical tooth, the first two components accounted for 87% of the variance.PC1 explained 52% of the variance and showed a positive association with intervals 2-4 while being negatively associated with intervals 12-14.This axis mainly differentiated Azteca sp., Cephalotes pusillus, Parasyscia sp., and Ectatomma edentatum from other species in its positive range (Fig. 2d).PC2 explained 36% of the variance and was positively associated with intervals 7-10.This axis mainly isolates Octostruma petiolata from other species in its negative range, as it had a lower proportion of mandibular volume covered by such intervals (Fig. 2d).

Discussion
In this study, we aimed to investigate how mandible morphological variation influences stress patterns during biting and whether predatory ants exhibit enhanced biting performance compared to omnivorous species.We simulated the mechanical behavior of mandibles from 25 ant species belonging to different feeding habits, including predatory, omnivorous, and leaf-cutting species, and tested four biting scenarios.Bite simulations revealed no apparent specialization in biting performance between predatory and omnivorous species, and differences in stress patterns are more pronounced at the interspecific level than between feeding habits.Similar stress patterns were observed among species of all feeding habits, contrary to our initial expectation that predatory ants would exhibit superior performance compared to omnivorous species.This general pattern was reflected in the distribution of stress intervals along the mandibular volume, where groups of predatory and omnivorous species were not observed on the PCAs of all biting scenarios.The only exception to this general pattern was observed in the specialized predators (trap-jaw ants), which were generally isolated from other species on the PCAs.These ants typically had a larger volume of their mandibles filled with intervals of intermediate to high stresses, contradicting our hypothesis that trap-jaw ants would show a better mechanical performance in strike biting.The more robust mandibles of Pheidole aper and Carebara sp.01 majors showed relatively lower stresses and were associated with the lowest stress intervals in all biting conditions, partially differing from our hypothesis that such mandibles would perform better only in pressure biting.Although there were no evident differences in stress patterns when comparing the differential use of the masticatory margin (entire or apical tooth only), we observed differences between pressure and strike biting.Pressure biting tended to generate higher relative stress levels along a larger mandibular surface area than strike bite, suggesting that pressing with mandibles is more mechanically demanding than striking.However, pressure biting tends to generate lower values of non-normalized stress than striking bite.Therefore, the mechanical demands of pressure biting are more relevant in long-lasting biting, where tiny failures can accumulate along the cuticle than in an acute high-stress peak, which is characteristic of strike biting.In some species, such as Acromyrmex aspersus, Ectatomma edentatum, Octostruma petiolata, Holcoponera striatula, and the major worker of Eciton burchellii, there were only subtle differences in stress distribution between pressure and strike biting, indicating a lack of specialization towards a specific biting behavior.
The form-function relationship of mandibles has been explored with FEA in other insect groups, such as Odonata 8 and beetles [47][48][49][50]60 . Regrding ants, the relationship between mandible morphology and feeding habit is supported by data from fossil 40,61 and current lineages 34,35,62,63 .Studies focused on ant functional groups have found relationships between mandible morphology and species feeding habits [64][65][66] .However, not all studies have found clear correlations. A recent study usng a geometric morphometric approach with 3D data of the diverse ant genus Pheidole did not find significant relationships between the mandible and head shape with feeding and nesting habits 31 .A phylogenetically broader assessment suggested that predatory ants tend to have longer and straighter mandibles, with predatory and omnivorous species differing in general morphological traits, although this effect was not statistically significant 30 .Also, the mandible length of more than 400 ant species showed no correlation with trophic position 67 .Despite such reports, studies investigating the effects of mandible morphology on biting behavior in ants are still limited.Although there is evidence that mandible shape has an impact on the responses of these structures to bite-loading demands, these studies have often relied on small sets of species 26,33,38 . In ths context, our study stands out by comparing a higher number of species, and indicate that the main feeding habits, such as omnivory versus predation, are not good predictors of stress patterns in ant worker mandibles when considering only the variation in mandible shape and thickness.
Among the species studied, those with stouter mandibles, such as Cephalotes pusillus and majors of Carebara sp.01 and Pheidole aper, are among the ones exhibiting relatively lower stress levels in pressure biting, as predicted.Interestingly, however, these stouter mandibles also showed relatively lower stresses in strike biting, as confirmed when considering normalized and non-normalized stress levels.Those species had mandibles with a broader blade and a subtle constriction near the mandibular base, a more convex curvature on the mandibular external face, and either short teeth or the absence of such structures.The lack of teeth means that the stresses generated in the masticatory margin do not concentrate heavily on smaller regions, as when teeth are present, but spread more evenly along the masticatory margin.A prominent mandibular curvature hampers the stress dissipation towards the delicate mandible blade, generating stress concentrations on the more robust masticatory margin and mandible base.Pheidole major workers are task specialists, often recruited for food processing (e.g.crushing seeds), resource retrieval (e.g.chopping and carrying pieces of large dead arthropods), and colony defense [68][69][70] , behaviors that require strong bite forces and generate higher mechanical demands than other activities.Pheidole aper is not a granivorous species 71 , but there seems to be no strong morphological differences between granivorous and non-granivorous species 31 .A better biting performance of Pheidole major worker mandibles was previously highlighted 72 , mainly in pressure biting 33 , but not in comparison with other ant genera.Majors of Carebara spp.perform similar specialized tasks as those described for Pheidole 73,74 , and their mandibular mechanical responses to bite-loading demands were correspondingly similar.Finally, species of Cephalotes spp.are known by their thicker exoskeleton, culminating in the morphological specialization of major heads to protect nest entrances 75,76 , but the mechanical demands of bite in worker mandibles were never tested before.
Two closely related species, Ectatomma edentatum and Holcoponera striatula, have similar mandibular morphology but show contrasting stress patterns in all biting scenarios tested, with Holcoponera striatula exhibiting relatively higher stress levels than Ectatomma edentatum.Regarding their positioning in the PCAs, both mandibles are more separated along the PC1 in simulations with the masticatory margin, suggesting that this mandibular region could represent an important source of mechanical distinction between the species.In fact, the masticatory margin in Ectatomma edentatum is wider (~ 0.9 mm) than in Holcoponera striatula (~ 0.3 mm), resulting in the bite load being applied along a broader area and hence dissipating over a larger area.However, a deeper investigation is needed to unveil the possible morphological aspects that explain the observed disparity in mechanical performance.
The trap-jaw ants, including Acantognathus brevicornis, Odontomachus chelifer, and Strumigenys denticulata, along with the major worker of Eciton burchellii, possess specialized mandibles, which in general are long, slender, with well-developed teeth, although the major worker of Eciton burchellii differed from the trap-jaw ants in having a hook-shaped mandible with a single tooth.Despite sharing these mandibular characteristics, those four mandibles displayed distinct stress patterns.Strumigenys denticulata and the major worker of Eciton burchellii showed proportionally higher stress levels throughout the mandibles than Acantognathus brevicornis and Odontomachus chelifer.However, when considering the non-normalized stress intervals, all these mandibles were highlighted on the PCAs by showing a proportionally larger volume filled with higher stress intervals, particularly in strike biting.These results do not support the initial hypothesis that trap-jaw ants would show relatively lower stress levels in strike biting.This partial contrast between the results of color maps and stress intervals provides intriguing suggestions about how trap-jaws deal with biting mechanical demands.In general, the long and slender mandibular blade withstands much lower stresses, reducing the risk of failure, but different from what was suggested for the snap-jaw ant Mystrium camillae, its general morphology and cross-section geometry does not indicate that such mandibles are prone to deform and withstand stress 26 , but this requires further investigation.Evidence from Odontomachus monticola highlights the importance of mandibular hollowness in withstanding the stresses generated by their powerful strike bites 38 .The natural hollowness of this species appears to optimize the trade-off between power generation and impact resistance, compared to mandibles with higher indices of hollowness or completely solid mandibles 38 .Accordingly, for all species tested in our study, their natural hollowness and cuticle thickness was kept for the simulations.
Although this study has shown some effects of mandible shape variation on stress distribution, it is important to note that the approach here presented does not fully capture the complexity of the mandible cuticle's material properties.Insects have a gradient of material properties along the layers of their cuticle 77 , which varies across different body regions.This material gradient proved to be functionally significant in some studies [78][79][80] and is particularly relevant to ant mandibles, which can accumulate heavy metals along their masticatory margin, leading to increased cuticular stiffness in those regions [81][82][83][84][85] .However, information about the variation in material properties of ant mandibles is still limited, with only a few studies restricted to a few leaf-cutting ants 81,86 .Moreover, incorporating material property variation in FEA increases the simulation complexity and can lead to confounding effects on structure shape and material properties 56 .Therefore, it is essential to consider the influence of material property variation under specific hypotheses and with appropriate data on the species in question.Further efforts are needed to measure the heterogeneity of cuticle material properties in ant mandibles to provide a better picture of their mechanical behavior.
Biomechanical approaches such as FEA are interesting tools for testing hypotheses about the role of mandible morphology in bite mechanics.Our results suggest that more robust mandibles exhibited lower stress levels under both pressure and strike biting, whereas mandibles of trap-jaw ants did not show signs of adaptation to strike biting in relation to the remaining species.Also, the main feeding habits of the ant species tested do not anticipate the general stress patterns observed in their mandibles under bite loading, being that the interspecific morphological disparity seems more relevant to biting performance among the ants tested.Ant mandibles are essential for food capture and processing but also play a crucial role in other colony tasks performed by workers 10,54 .Our effort builds on previous studies that found weak associations between mandible morphology and feeding ecology in ants 30,31 .Therefore, mandible morphology could reflect the necessity to perform multiple tasks 8,54 or to excel in other activities, such as excavation 32 , hypotheses that demand further studies to be properly tested.
The evolution of ant mandibles has been the subject of intense study in recent years 18,28,29,41 , and although the shovel-shaped mandible is considered the Bauplan of extant ant mandibles 29 , many ant lineages exhibit significant deviations from this general morphology (see also 62 ).Our sampling set covers a significant sample of this variation and relevant functional differences were observed.Further studies may explore the effects of variation in cuticle material properties, particularly considering the interspecific variation in heavy metal accumulation 85 .It would also be meaningful to investigate more widely the relationship between mandible morphology and bite force in ants 87,88 .

Methodology Species selection
To simulate biting behavior, we selected one worker specimen of 25 ant species with a wide range of mandible morphologies, which were representative of two primary feeding habits: predatory (14 species) and omnivorous (10), along with a leaf-cutter ant (Table 1).For a few polymorphic ant species, we also considered a second worker type (Table 1).To establish the general feeding habits of these species, we collected information from online repositories (e.g. 89) and literature sources 90,91 .In cases where we did not find the species' main feeding habit or were able to identify the species only at the genus level (morphospecies), we considered the main feeding habit of the genus.We classify as predators the ant species known to actively hunt other animals (mainly arthropods) because of the mechanical demands of subduing a prey, which is assumed to differ from simply feeding on dead animals.For the predatory category, we included species with specialized mandible morphology and mechanics, the trap-jaw ants.Due to their specialized mandibles, these species were classified here as specialized predators, although not necessarily feeding exclusively on live prey.Omnivorous species do not actively predate other organisms as their primary feeding habit, usually relying on many different food sources without subduing living prey.We also added a leaf-cutting ant (Acromyrmex aspersus), which exhibits a unique biting behavior for cutting leaves.We aimed to broadly categorize species based on their primary differences in feeding mechanical demand, not to describe in detail their natural diets, since many species are known to opportunistically feed on a wide range of trophic sources 71 .In cases where different species displayed similar mandible morphology, we relied on the quality of the µCT scans (see below) to decide which species would enter our final dataset.Species identifications were confirmed with assistance from taxonomists (see Acknowledgments).All scanned specimens are deposited in the Formicidae collection of the "Coleção Entomológica Mítia Heusi Silveira", Universidade Federal de Santa Catarina, Brazil (Supplement S1).

Scanning and reconstruction of volumetric models
Ant specimens were scanned using synchrotron radiation X-ray tomography at two German facilities, Imaging Beamline P05 (IBL) [92][93][94] operated by the Helmholtz-Zentrum-Geesthacht at the storage ring PETRA III (Deutsches Elektronen Synchrotron-DESY, Hamburg, Germany)-and KIT Light Source (Eggenstein-Leopoldshafen).As a result of the tomographic reconstruction the scans were assembled as Tiff image series for subsequent segmentation, with scanning parameters varying depending on the ant species (Table 1).Tomographic reconstruction for P05 data has been done using a custom reconstruction pipeline 95 using Matlab (Math-Works) and the Astra Toolbox [96][97][98] .Volumetric CT data of ant worker mandibles were then pre-segmented using the software Amira 5.4 (Visage Imaging GmbH, Berlin, Germany).A set of slices was manually segmented using the Magic Wand tool, at intervals ranging from 5 to 15 slices, depending on the complexity of the mandible and the quality of the scans.Automatic interpolation between the pre-segmented slices was done using the online Biomedisa 99 .To ensure that the reconstructions accurately represented the morphology of ant mandibles, the outputs from Biomedisa were imported back into Amira 5.4, where we corrected inaccuracies and reduced the complexity of the reconstructed morphology, taking into account the presence of hairs, sharp edges, and holes, among other factors.Cuticle thickness can vary substantially along the ant mandible, and the degree of mandibular hollowness can also differ between species, both being potentially relevant to its bite mechanics 38 .As such, the natural hollowness and cuticle thickness of each mandible were represented in our mandible digital representations.

Finite element mesh generation
The mandible surface models were imported into Blender 2.93 (https:// www.blend er.org/) to position them in a common orientation.We then used FUSION 360 (AUTODESK) to decrease mesh density.By conducting mesh convergence simulations, we determined a minimum mesh density for each mandible that approximates the 3D mandible morphology and provides a lower computational demand to solve the finite element equations 100 .To emulate actual simulation parameters, we generated a simplified version of the final simulation setup, sampling three to six elements from specific mandibular regions to check for convergence (less than 5% of error) on von Mises stress values 101 .We exported meshes from FUSION 360 as .stlfiles and imported them into FEBio 102 , where we conducted convergence tests and the final FEA simulations.The number of elements for the definitive meshes is available in Table 1.Volumetric representations of each worker mandible are available as supplementary material (File S2).
Table 1.List of species considered for finite element analysis in this study.Also shown are the worker type, energy and voxel size of the scans, surface area of the meshes (considered for load normalization), number of elements of each mesh, the applied load and the feeding habit of the species.In this study, we aimed to investigate the effect of mandible morphology on bite-loading demands by simulating four bite conditions that reflect various phases of biting and the employment of different parts of the masticatory margin.Specifically, we simulated mandible use in strike and pressure biting using the entire masticatory margin and employing only the apical tooth.Strike biting emulates the usage of mandibles to hit an object and represents the first phase of bite.To simulate strike biting, we applied a load on the masticatory margin or the apical tooth of the mandible, restricting the nodal displacement on the mandibular articulations with the head in all directions (Fig. 3).Similarly, for the second bite phase, which we refer to as pressure biting and which involves crushing or gripping an object, we applied a load on the mandibular region where the mandibular apodeme inserts, following the direction of forces generated by 0md1 contraction.We also restricted the mandible articulations with the head and the masticatory margin or apical tooth to zero nodal displacements in all directions.To ensure consistency, we applied the same material properties to each simulation.The Young's modulus was defined as 2.75 GPa based on measurements from the mandibles of Atta laevigata available in the literature 86 , and the Poisson ratio was set at 0.3, as commonly considered for ant cuticle 26,33,38,54 .To account for differences in mandible size, we applied a 1 N load to the largest mandible and adjusted the load for the remaining mandibles based on surface area differences 103 (Table 1).Restricting parameter variation to mandible morphology alone, such as mandible shape and cuticle thickness, allows us to compare results between simulations 56 .This approach is commonly used in comparative studies 8,26,33,[48][49][50] .

Intervals method
The Intervals Method is an approach developed to quantify the area (2D) or volume (3D) of structures covered by a specific range of stress 104 .This method involves using information about element stress and volume to define intervals of stress values and calculate the proportion of the structure volume filled with each stress interval.An upper threshold of stress value must be defined, which is a reference for the determination of the remaining stress intervals.It is possible to define any number of stress intervals from this threshold value and to compute the volume occupied by those intervals on each model.The amount of the structure volume filled with each interval represents new variables that can be used in further statistical tests or ordination analyses, such as principal component analyses (PCA) 104 .This approach allows for a direct comparison of stress distribution between the mandibles in a more quantitative way than the color maps, where normalized stress values are being compared.
In comparing the proportion of mandibular volume filled by each stress interval, we can observe which structures are submitted to higher or lower non-normalized stresses, and how relevant those stresses are for the structure as a whole, which is informative when a size-corrected load is considered for the simulations, as was done here.
We applied the method independently to each biting scenario.We extracted data on von Mises stress and volume from the elements of each simulation from FEBio 102 .Then we removed elements that corresponded to the 2% higher stress values in each simulation, as these values often represent artificially high-stress values 104,105 .Even after the removal of such potentially artificial high-stressed elements, a few mandibles showed substantially higher stress values than the remaining ones, achieving values up to one order of magnitude higher (Fig. S1).To account for this variation in the stress scale and avoid eventual biases toward the highly stressed mandibles (Fig. S1), we log-transformed stress values prior to generating the stress intervals.We defined the upper threshold values for each biting scenario to include the 15% highest stress values above the threshold.Defining a higher threshold value could lead to uninformative intervals, as many models would have no elements within the highest interval due to the variation in the range of stress generated among models (Fig. S1).To define the ideal number of stress intervals, we generated datasets with different numbers of intervals (5, 10, 15, 25, 50) and performed PCAs.We considered PC1 and PC2 scores of each dataset in linear regressions with the scores of equivalent PCs of the next interval (e.g., PC1 5_intervals ~PC1 10_intervals ), and we retrieved the coefficient of determination (R 2 ) to analyze the convergence of PC scores.The stop of increase in R 2 defined the final number of intervals 104 .For all biting scenarios, convergence occurred with 15 intervals, so we used this number for the PCAs (File S3).To summarize, we provide our FEA results in two ways: color maps and the intervals method 104 .Color maps were graphical outputs generated by FEA software displaying stress distribution along the mandibles.To compare results between species and biting scenario stress values were normalized using a reference model.The resulting color variation on the maps represents the proportion of stress based on the maximum normalized stress value for each simulation (Fig. 1).In contrast, the intervals method involved quantifying the volume of the mandible that falls within specific stress intervals for each of the four biting scenarios 104 .These proportions served as input variables for the PCAs used to address differences in stress distribution among ant mandibles across the 15 stress intervals (Fig. 2).

Figure 1 .
Figure 1.Colour maps of FEA from all biting scenarios simulated.Species are grouped according to their main feeding habit.Values of von Mises stress were normalized based on a reference model to allow direct comparison between species and biting scenarios.Strike with the entire masticatory margin (a); Pressure with the entire masticatory margin; (c) Strike with the apical tooth; (d) Pressure with the apical tooth.

Figure 2 .
Figure 2. PCA based on the proportion of mandibular volume filled with each of the 15 stress intervals for all bite scenarios simulated: strike (a) and pressure (b) with the entire masticatory margin, strike (c) and pressure (d) with the apical tooth only.Colored circles depict the species main feeding habit, whereas colored arrows depict stress intervals, from the lowest stress value (blue-1) toward the highest (red-15).

Figure 3 .
Figure 3. Diagrams depicting the boundary conditions for each biting scenario.Highlighted are the conditions for strike and pressure biting with the entire masticatory margin.