Evolutionary biomechanics: hard tissues and soft evidence?

Abstract

Biomechanical modelling is a powerful tool for quantifying the evolution of functional performance in extinct animals to understand key anatomical innovations and selective pressures driving major evolutionary radiations. However, the fossil record is composed predominantly of hard parts, forcing palaeontologists to reconstruct soft tissue properties in such models. Rarely are these reconstruction approaches validated on extant animals, despite soft tissue properties being highly determinant of functional performance. The extent to which soft tissue reconstructions and biomechanical models accurately predict quantitative or even qualitative patterns in macroevolutionary studies is therefore unknown. Here, we modelled the masticatory system in extant rodents to objectively test the ability of current muscle reconstruction methods to correctly identify quantitative and qualitative differences between macroevolutionary morphotypes. Baseline models generated using measured soft tissue properties yielded differences in muscle proportions, bite force, and bone stress expected between extant sciuromorph, myomorph, and hystricomorph rodents. However, predictions from models generated using reconstruction methods typically used in fossil studies varied widely from high levels of quantitative accuracy to a failure to correctly capture even relative differences between macroevolutionary morphotypes. Our novel experiment emphasizes that correctly reconstructing even qualitative differences between taxa in a macroevolutionary radiation is challenging using current methods. Future studies of fossil taxa should incorporate systematic assessments of reconstruction error into their hypothesis testing and, moreover, seek to expand primary datasets on muscle properties in extant taxa to better inform soft tissue reconstructions in macroevolutionary studies.

Keywords: biomechanics; finite element analysis; macroevolution; multi-body dynamics; rodent mastication.

Figure 1.

Quantitative soft tissue reconstruction and biomechanical modelling of rodent masticatory morphotypes. (a) Muscle volumes are reconstructed using three-dimensional sculpture techniques, as commonly applied in fossils, with values combined with different estimates of fibre length to provide input values for biomechanical models. Incisor bite forces were predicted across 27 ‘fossil’ model iterations of (b) MDA models for comparison to values predicted using real (measured) muscle data. (c) Predicted muscle forces from all model iterations were used to load FE models to compare stresses predicted in fossil models to those from models with real (measured) muscle properties. (Online version in colour.)

Figure 2.

Error magnitudes in the sculptured muscle volume reconstructions by investigators 1, 2, and 3 for the (a) squirrel, (b) guinea pig, and (c) rat. Abbreviations: SM, superficial masseter; Temp, temporalis; AZM, anterior zygomatico-mandibularis; PZM, posterior zygomatico-mandibularis; MP, medial pterygoid; LP, lateral pterygoid; DM/ADM, deep masseter/anterior deep masseter; PDM, posterior deep masseter; Infraorbital zygomatico-mandibularis.

Figure 3.

Error magnitudes in reconstructed (a,b) muscle fibre lengths and (c–h) PCSAs in the three species.

Figure 4.

Comparison of (a) absolute bite forces and (b) percentage error magnitudes in bite forces across the ‘extant’ and ‘fossil’ MDA models. (a) ‘Extant’ model iterations predict the highest incisor bite forces in the squirrel, followed by the rat and then guinea pig. This qualitative pattern across the morphotypes is recovered in all model iterations by investigator 2, by iteration C investigator 3, but in none of the iterations by investigator 1. (b) Quantitative error varied considerably, with most iterations tending to underestimate bite force.

Figure 5.

Stress magnitudes and distributions (represented by von Mises stress) in the FE models across the 30 model iterations. Stress magnitudes along the length of skull in the extant models are compared to those of (a) investigator 1, (b) investigator 2, and (c) investigator 3 and demonstrate significant quantitative and some qualitative error. Some reconstructions, such as (b,e) iteration C those by investigator 2, show a close quantitative match to (d) the extant models, while some reconstructions, such as (f) iteration A by investigator 1 contain both quantitative and qualitative error in relative stress magnitudes and distribution across the morphotypes. (Online version in colour.)