Introducing Trait Networks to Elucidate the Fluidity of Organismal Evolution Using Palaeontological Data

Abstract

Explaining the evolution of animals requires ecological, developmental, paleontological, and phylogenetic considerations because organismal traits are affected by complex evolutionary processes. Modeling a plurality of processes, operating at distinct time-scales on potentially interdependent traits, can benefit from approaches that are complementary treatments to phylogenetics. Here, we developed an inclusive network approach, implemented in the command line software ComponentGrapher, and analyzed trait co-occurrence of rhinocerotoid mammals. We identified stable, unstable, and pivotal traits, as well as traits contributing to complexes, that may follow to a common developmental regulation, that point to an early implementation of the postcranial Bauplan among rhinocerotoids. Strikingly, most identified traits are highly dissociable, used repeatedly in distinct combinations and in different taxa, which usually do not form clades. Therefore, the genes encoding these traits are likely recruited into novel gene regulation networks during the course of evolution. Our evo-systemic framework, generalizable to other evolved organizations, supports a pluralistic modeling of organismal evolution, including trees and networks.

Keywords: animal evolution; complexes; network; palaeontology; rhinocerotoids; tinkering.

Fig. 1.

—Principle of the matrix analysis. Our approach exploits existing phylogenetic data matrices featuring taxa as rows and homologous characters as columns. Each original column is replicated in as many new columns as there are character states (e.g., A2, B2), defining a new matrix of taxa by traits, where the presence of each trait is indicated by a “+” and its absence by a “−”. All pairs of columns of this new matrix are then compared with one another, distinguishing four types of distribution of traits across taxa, therefore characterizing four possible types of relationships between all pairs of traits.

Fig. 2.

—Some important network patterns and their biological meaning. The first column displays the relationships between a pair of traits (here character states). The second column represents the corresponding network pattern. The third column introduces the terms specifically used to describe and analyze these patterns. The fourth column highlights some possible biological meanings of these patterns.

Fig. 3.

—Composite phylogenetic tree of selected Rhinocerotidae, resulting from the parsimony analyses of (Antoine 2002; Antoine et al. 2003; Boada-Saña et al. 2008; Antoine et al. 2010), based on 282 cranio-mandibular, dental, and postcranial characters, depicting: (a) eight trait complexes. Each complex is represented by its corresponding motif (each node represents a trait, each green edge represents the type I relationships between two traits) along the phylogeny, based on its taxonomic distribution. Each complex is also identified by a circled number; blue circles representing complexes shared by a common ancestor and all its descendants (putative synapomorphy), yellow circles representing a complex whose distribution does not map simply onto the phylogeny (homoplasy). The top left squared box identifies the distribution of complexes over the main regions of the rhinocerotoid body plan (S, skull; T, teeth; J, jaw; BP, body plan; FL, forelimb; and HL, hind limb). Blue letters highlight complexes of traits from different regions. (b) Phylogeny of Rhinocerotidae showing two exemplary traits with type II relationships. The distribution of trait 44 is nested in that of trait 23 (clade within clade). The distribution of trait 160 is nested in that of trait 217 (nonclade within nonclade). 23: Frontal bone: aspect|‘rugose’; 44: Corpus mandibulae: base|‘very convex’; 160: Lower molars: hypolophid|‘transverse’; 217: Astragalus: orientation trochlea/distal articulation|‘very oblique’.

Fig. 4.

— a) Example of a complex, involved in type II relationships. Nodes 23 and 175 forming the complex are directly connected by a green edge (type I). The distribution of these nodes is nested within the distribution of 33 other nodes (to the left, connected by directed type II edges), and includes the distribution of 12 other nodes (to the left, connected by directed type II edges). b) A selection within these nodes, mapped onto the body plan of rhinos. Node 23: Frontal bone: aspect|‘rugose’; node 175: Pyramidal: distal facet for semilunate|‘L-shaped’; node 47: Ramus|‘inclined backward and upward’; node 191: Femur: trochanter major|‘low’.

Fig. 5.

—Example of nodes involved in type III and IV relationships. Node 12: Nasal bones: rostral end|‘very broad’; Node 38: Symphysis|‘massive’; Node 2: Skull: dorsal profile|‘very concave’; node 43: Corpus mandibulae: base|‘convex’. Orange edges correspond to type IV edges, red edges to type III edges. Node 38 is involved in a triangle of type III edges, and occupies what we defined as a central position in 2 type D triplets.

Fig. 6.

—Mapping of a type D triplet along the phylogeny of rhinocerotids. Each trait is represented by a different color. The distribution of trait 27 overlaps with that of trait 99; the distribution of trait 27 overlaps with that of trait 115; however, the distributions of trait 99 and 115 are disjoint. 27: Zygomatic/frontal widths|‘less than 1.5’; 99: Upper molars: crochet|‘always present’; 115: M1-2: metastyle|‘short’.

Fig. 7.

—Schematic mapping of morphological traits on the rhinocerotoid body plan. Main regions are indicated in boxes. Red squares are relatively unstable traits (i.e., type II in-degree is null); blue squares are relatively stable traits (i.e., type II in-degree is positive); yellow squares indicate traits with significant relative stability (P value <0.05, following a Bonferroni correction for multiple tests, equiprobable and phylogenetic permutation tests). Numbers in squares correspond to NodeID. Black boxed squares correspond to traits that are significantly central in type D triplets (P value <0.05, following a Bonferroni correction for multiple tests, equiprobable and phylogenetic permutation tests). The barplot indicates the relative frequencies of traits in main regions of the rhinocerotoid body plan, observed in all species. Areas in red/blue/yellow are versatile/relatively stable/significantly stable traits, respectively. The main regions are T, teeth; S, skull; J, jaw; BP, body plan; FL, forelimb; and HL, hind limb.