High-density 3-D coordinate data of avian and non-avian dinosaur endocasts

How do large and unique brains evolve? Historically, comparative neuroanatomical studies have attributed the evolutionary genesis of highly encephalized brains to deviations along, as well as from, conserved scaling relationships among brain regions. However, the relative contributions of these concerted (integrated) and mosaic (modular) processes as drivers of brain evolution remain unclear, especially in non-mammalian groups. While proportional brain sizes have been the predominant metric used to characterize brain morphology to date, we perform a high-density geometric morphometric analysis on the encephalized brains of crown birds (Neornithes or Aves) compared to their stem taxa—the non-avialan coelurosaurian dinosaurs. When analyzed together with developmental neuroanatomical data of model archosaurs (Gallus, Alligator), crown birds exhibit a distinct allometric relationship that dictates their brain evolution and development. Furthermore, analyses by neuroanatomical regions reveal that the acquisition of this derived shape-to-size scaling relationship occurred in a mosaic pattern, where the ‘avian’-grade optic lobe and cerebellum evolved first among non-avialan dinosaurs, followed by major changes to the evolutionary and developmental dynamics of cerebrum shape after the origin of Avialae. Notably, the brain of crown birds is a more integrated structure than non-avialan archosaurs, implying that diversification of brain morphologies within Neornithes proceeded in a more coordinated manner, perhaps due to spatial constraints and abbreviated growth period. Collectively, these patterns demonstrate a plurality in evolutionary processes that generate encephalized brains in archosaurs and across vertebrates.

Methods
The landmark data were collected using the program Landmark Editor v3.6 to virtually place three-dimensional (3-D) coordinate points on polygon mesh files of endocasts reconstructed from micro-computed tomography (CT) data. The landmark scheme includes a combination of discrete landmarks, curve semi-landmarks, and surface semi-landmarks that characterizes the shape of endocasts and their regions (cerebrum, optic lobe, cerebellum, medulla). Initially, coordinate data were collected from both sides of the endocasts (e.g., left and right cerebrum) which were subjected to generalized Procrustes alignment minimizing total bending energy with sliding semi-landmarks while projecting the semi-landmarks on mesh surfaces. After alignment, the median and left landmarks were extracted (i.e., right landmarks removed) from the aligned dataset to remove redundant shape information and to avoid artifacts associated with aligning one-sided data of bilaterally symmetric structures.