The trunk replaces the longer mandible as the main feeding organ in elephant evolution--Supplementary codes
Data files
May 28, 2024 version files 266.31 MB
-
code_files.zip
-
README.md
Abstract
The long-trunked elephantids underwent a significant evolutionary stage characterized by an exceptionally elongated mandible. The initial elongation and subsequent regression of the long mandible, along with its co-evolution with the trunk, present an intriguing issue that remains incompletely understood. Through comparative functional and eco-morphological investigations, as well as feeding preference analysis, we reconstructed the feeding behavior of major groups of longirostrine elephantiforms. In the Platybelodon clade, the rapid evolutionary changes observed in the narial region, strongly correlated with mandible and tusk characteristics, suggest a crucial evolutionary transition where feeding function shifted from the mandible to the trunk, allowing proboscideans to expand their niches to more open regions. This functional shift further resulted in elephantids relying solely on their trunks for feeding.
README: Title of Dataset:
The trunk replaces the longer mandible as the main feeding organ in elephant evolutionSupplementary data
Brief summary of dataset contents, contextualized in experimental procedures and results.
Description of the Data and file structure
Codes information: (All codes are written in English)
Code S1 Script file for Bayesian total-evidence dating analysis (.nex file). (Open it in MrBayes Software)
Code S2 Script auxiliary file for most parsimonious analysis, applying for some ordered and irreversible characters (characters 19–24 in the TNT program, please note that these character numbers were automatically assigned by TNT file, and begin from character “0” rather than “1”, therefore, these characters in supplementary Appendix and Data S1 were characters 20–25) (.tnt file). (Open it in TNT Software)
Code S3 Script file for most parsimonious analysis (.tnt file). (Open it in TNT Software)
Code S4 FEA setting file, for distal force tests (this file can be directly submitted in the Abaqus 6.14). Open it in Abaqus 6.14 Software; or open-source software Calculix (http://www.calculix.de/ and FEBio (https://febio.org/
Code S5 FEA setting file, for distal twig cutting tests (this file can be directly submitted in the Abaqus 6.14). Open it in Abaqus 6.14 Software; or open-source software Calculix (http://www.calculix.de/ and FEBio (https://febio.org/
Sharing/access Information
Links to other publicly accessible locations of the data:
Was data derived from another source?
If yes, list source(s): No
Methods
Cladistic analysis
Cladistic analyses were performed to evaluate the phylogenetic hypothesis of trilophodont longirostrine proboscideans. The data matrix contained 37 taxa, including most of the known trilophodont longirostrine taxa at the species level, and Phiomia serridens, an Oligocene basal elephantiform, was selected as outgroup. Additionally, a basal elephantoid, Tetralophodon longirostris was also included to assess which clade the true elephantids originated from. The morphological characters included 5 characters from upper tusks, 9 from mandibular tusks, 37 from cheek teeth, 19 from the cranium, and 10 from the mandible, mainly following Tassy, 1996; Shoshani, 1996; and Wang et al., 2017. For a description of the characters and states, please see Appendix; for the data matrix, please see Figure 1–Source Data 1. Two methods, Bayesian tip-dating (BTD) and maximum parsimony (MP) analyses were performed.
In BTD analysis, the fossil ages were incorporated as tip calibrations (Gavryushkina et al., 2014; Ronquist et al., 2012; Zhang et al., 2016). The Lewis Mkv model (Lewis, 2001), with gamma rate variation across characters (Mkv + G) (Yang, 1994), was initially used; subsequently, the timetree was modelled by the fossilized birth death process (Heath et al., 2014; Stadler, 2010). The process was conducted using the time of the most recent common ancestor (root age) and included hyperparameters of speciation rate, extinction rate, fossil-sampling rate, and extant-sampling probability. The root age was first assigned an offset-exponential, with mean age of 37 Ma and minimum age of 34 Ma, referring to the oldest fossil. The fossil ages were fixed to their first occurrence. The extant-sampling probability was fixed to 1 because no living genera were specified. Apart from the timetree, the other key component was the relaxed clock model, which models the evolutionary rate variation along the branches in the tree. We used the independent gamma rate clock model (Lepage et al., 2007), in which the mean clock rate was initially assigned a lognormal prior (–6, 1) and the variance parameter of the clock rate was exponential (10).
We executed two independent runs and four chains per run (one cold chain and three hot chains) using Markov chain Monte Carlo. Each run was executed for 1 million generations and sampled every 5000 generations. The first 25% of the samples were discarded as burn-in and the rest of the two runs were combined. Good convergence and mixing were determined by effective sample sizes larger than 200 for all parameters and average standard deviations of split frequencies smaller than 0.01 (Geyer, 1992). The BTD analysis was performed in MrBayes 3.2.7 (Ronquist et al., 2012) (Supplementary file 3).
MP reconstruction was performed by TNT1.1 (Goloboff et al., 2008). In MP analysis, all characters were equally weighted. Characters 20–25 pertained to the loph/lophid numbers of cheek teeth (character numbers begin from “1” in our numeration; however, in MP analysis, TNT1.1 automatically numbered characters from “0”; therefore, in the TNT1.1 program, these characters were numbered “19–24”). These were treated as ordered and irreversible, which was performed by setting “step-matrix → of costs” under the menu “Data → Character settings”, assigning the value “9” to the blanks where i > j (i to j in the matrix), and assigning the value “j–i” to the blanks where i < j. The traditional search strategy was performed, and the results were reported based on a 50% majority consensus tree of the most parsimonious trees. Node supports were calculated by symmetric resampling with 0.33 change probability (1000 replicates). The script files are provided in Supplementary file 3. The major consensus tree was then calibrated using the time PaleoPhy function in R package paleotree 3.3.25 (Bapst, 2012), and is shown in Figure 1–figure supplement 1.
FE analysis
We investigated the feeding behaviors of the three longirostrine gomphothere families, i.e., Choerolophodontidae, Amebelodontidae, and “Gomphotheriidae”, using finite element (FE) stimulation. Three species were selected to represent each family, Choerolophodon chioticus, IVPP V23457 (cranium and mandible); Platybelodon grangeri, HMV 0930 (cranium and mandible); and Gomphotherium tassyi, IVPP V22780 (cranium) and IVPP V22781 (mandible), respectively. Note that, in Gomphotherium, we were unable to get access to cranium and mandible that belonged to one individual. We used a handheld Artec Spider 3-dimensional scanner to obtain the surface topology of these specimens. The surface meshes were produced using Artec Studio 14 Professional. These meshes were first repaired with ZBrush 2021; for example, ZBrush 2021 was used to recover the broken edges of mandibular tusks, create the remaining mandibular tusks in the alveolus, create the keratinous cutting plate of the mandible in Choerolophodon, and for retro-deformation of the crushed cranium (i.e., the cranium of IVPP V23457).
The rough surface meshes were edited using Materialise 3-matic Research (V12.0). The surface meshes were smoothed by removing the small knobs and filling the holes for volume mesh generation. Before volume meshing, the models were cut into symmetric halves along the median sagittal plane. To reduce computation, only the right halves were preserved for further analyses. Note that the cranium was only used to define the attachments or insertions of jaw-closing muscles (modeled by many draglines, see below), and was treated as a rigid body in simulation (therefore, the cranial and nasal cavities are not relevant). The mandible contains two parts, the bony structure, including the cheek teeth, and the food acquisition organs (the mandibular tusk in Platybelodon and Gomphotherium, and the keratinous cutting plate in Choerolophodon). These two parts were integrated using the command “Non-manifold Assembly” in Materialise 3-matic Research. Then, the cranium and the mandible were aligned based on their natural position, i.e., the occlusal surface of the upper and lower tooth rows were matched, and the mandibular condyle and glenoid fossa were fitted. The sagittal surfaces of the cranium and mandible were made to coincide with the x-z plane, and the x-positive direction was set along the rostral direction. For comparison between different taxa, the three mandibular models were scaled to the same volume as Choerolophodon (6,563,708.0146 mm3) (Figure 3–figure supplement 1). Finally, volume meshes were generated in the cranium and mandible across the three models, and exported as .inp files that could be loaded into Abaqus CAE (V 6.14), the engineering software for FE analysis. The parameters for volume mesh generation are listed in Supplementary file 1.
The volume meshes, representing the geometric models, were imported into Abaqus CAE (V 6.14) (Figure 3–figure supplement 2–4), and included two parts, cranium and mandible; additionally, a third part was created by Abaqus CAE, a long cylinder (300 mm in length, 50 mm in diameter), to model twigs that were cut by mandibular tusk or keratinous cutting plate. Note that the millimetre (mm)–ton (t)–second (s) unit system was adopted; other units included Newton (N) and million Pascal (MPa). Different materials, including bone, dentine, keratin, and wood were assigned to the corresponding parts. The materials of bone, dentine, and keratin were treated as isotropic linear elastic materials. For the detailed parameters, see Supplementary file 1 (Drake et al., 2016; Huo et al., 2000). However, twigs could not be treated as an isotropic linear elastic material, because the purpose of this simulation was to evaluate the food procuring efficiency of different taxa in different working conditions. Here the twigs were assigned as an orthotropic elastoplastic material, and the parameters (of wet red pine tree) were obtained from a wood handbook (Risbrudt et al., 2010).
The occlusal surfaces were coupled by two arbitrary points on the upper and lower teeth. These two points were connected by a “beam connector”, which constrained all the degrees of freedom (df) between the two points (Figure 3–figure supplement 2–4). In this way, we simulated the occlusal surfaces of the upper and lower teeth. Jaw-closing muscles were simulated by several groups of “axis connectors”. This type of connector does not constrain any df of the two extreme points, and allows exerting force along the connector (Figure 3–figure supplement 2–4). Four jaw-closing muscles were considered, including temporalis, superficial masseter, zygomaticomandibularis, and pterygoideus internus. Ten axis connectors were assigned to the temporalis, and four, three, and three were assigned to the latter three, respectively. These connectors were uniformly arranged along their insertion areas based on their natural anatomy. The areas of the temporal fossa (At) and ascending ramus (AA) were measured in 3-matic Research (V12.0) (Figure 3–figure supplement 1). We estimated the muscle force of temporalis as follows (Tseng et al., 2017):
At (mm2) × 0.3
this force was equally distributed to the 10 axis connectors for temporalis.
Alternatively,
AA (mm2) × 0.3
was considered the gross force for superficial masseter, zygomaticomandibularis, and pterygoideus internus. This force was also equally distributed to the other 10 axis connectors.
Note that the At and AA in the Platybelodon and Gomphotherium models were not true. These models were scaled to the same volume as that of Choerolophodon. In the simulation, we uniformly assigned muscle forces to make it easy to compare models (Supplementary file 1).
The cranium was treated as a rigid body and was fixed. Another boundary condition was assigned to a node of the mandibular condyle, which only allowed the y-direction rotation and constrained any other dfs (simulating the rotation of the mandibular articulation). The dfs of the x- and z- rotations of the mid-symphysis were also constrained by considering the connection to the other half.
Two tests were carried out on the composite models: the distal forces test (dft) and twig-cutting test (tct). The dft includes two steps: 1, applying the muscle force; and 2, exerting a distal 5,000 N force that gradually changes from horizontally to vertically, by which we assessed the optimum direction of the external force for the mandible of each taxon. In this test, the “twig” was not included in the model.
In the tct, the middle point of the “twig” was set in close contact with the distal edge of the mandibular tusk and keratinous cutting plate, and was placed horizontally (Figure 3–figure supplement 2), 45° obliquely (Figure 3–figure supplement 3), and vertically (Figure 3–figure supplement 4). One extremity of the “twig” was fixed, and contact properties were assigned (hard contact normally and 0.3 frictional coefficient tangentially). The tct also includes two steps: 1, applying the muscle force as in dft; and 2, displacing the cranium and mandible 10 mm towards the “twig” to stimulate cutting action of proboscideans, by which we determined the cutting efficiency of directions for each taxon. In the results, the sum of the equivalent plastic strain (EPS) from total twig elements was calculated and reported for each model. The plastic strain represents the irreversible deformation of an element, and the sum of EPS from all twig elements can reflect the cutting effects in each model. The videos for von Mises stress contour colour maps were also generated in the dft (Videos 1–3) and tct (Videos 4, 6, 8, 10, 12, 14, 16, 18) modelling, and for the EPS contour colour maps of twigs in tct (Videos 5, 7, 9, 11, 13, 15, 17, 19) modelling.