Mosaic Evolution of Craniofacial Morphologies in Ghost Electric Fishes (Gymnotiformes: Apteronotidae)
Ghost electric knifefishes (Gymnotiformes: Apteronotidae) are a dominant component of the species diversity and biomass of large lowland rivers in Greater Amazonia, including 77 species that exhibit diverse craniofacial morphologies associated with trophic and secondary sexual traits. Here we use open-source computed microtomography (μCT) scans to generate 3D reconstructions for a majority of apteronotid species and almost all valid genera, and geometric morphometric and phylogenetic analyses to explore patterns of skull evolution. As with most vertebrates, principal component 1 (PC1) summarizes variance from brachycephalic to dolichocephalic morphologies, previously described as heterocephaly, and PC2 summarizes variance from recurved (upturned) to decurved (downturned) snout morphologies, described here as heterorhynchy. Phylomorphospace and traitogram analyses found instances of both convergent and divergent evolution along both of the first two PC axes, as well as a preponderance of clades characterized by morphological stasis or phylogenetic conservatism. Certain phenotypic combinations predominate among species and clades, including elongated-downturned snouts and foreshortened-upturned snouts, while other phenotypic combinations are not observed, including elongated-upturned snouts and foreshortened-downturned snouts. These results highlight the power of 3D geometric morphometrics in the study of craniofacial evolution and indicate developmental or functional constraints in the evolution of craniofacial phenotypes in an ecologically dominant clade of riverine Amazonian fishes.
Diversity of head and skull shape in the family Apteronotidae (Teleostei: Gymnotiformes), with an instance of convergence (Porotergus duende [A–B] and Sternarchogiton nattereri [C–D]) and an example of skull shape diversity within a lineage (Orthosternarchus tamandua [E–F] and Sternarchorhamphus muelleri [G–H]). Black scale bar is 5 mm.
Three-dimensional (3D) model of an Apteronotus rostratus skull with the 22-point landmark scheme used for geometric morphometric analyses. (A) Lateral, (B) dorsal, and (C) ventral views of the 3D model. See Table 1 for landmark definitions. Black scale bar is 10 mm.
Principal component analysis with the two major axes of shape variation across 13 genera within Gymnotiformes. PC1 shows heterocephaly, the inverse relationship between the length of the snout and the size of the neurocranium. Extreme morphologies include a relatively short snout with a large braincase and a relatively long snout with a small braincase. PC2 shows heterorhynchy, the relative flexion of the snout along the dorsal-ventral axis. Extreme phenotypes include upturned snouts and downturned snouts. There are morphological similarities across genera, shown by overlap on the morphospace. There is also an inverse relationship between PC1 and PC2 values; the longer the snout the more ventral the snout flexion.
Phylomorphospace based on the phylogeny of Bernt et al. (2019), colored by genus. Several phenotypes evolved multiple times across the Apteronotidae, including a relatively intermediate-length snout with an upturned snout and a relatively short snout with a terminal mouth. Some phenotypes are not seen in species in this family (visualized by dotted polygons), including relatively long, upturned snouts and relatively intermediate, downturned snouts. Species included in this analysis (n = 31) were subsampled from the original morphological dataset to include only those in the apteronotid phylogeny.
An ancestral state estimation completed using the phytools package, showing the number of times certain phenotypes have evolved within a lineage. Extremely brachycephalic phenotypes (low PC1 values) have evolved multiple times from a hypothesized ancestor with an intermediate-length snout. Dolichocephalic phenotypes (high PC1 values) evolved at least twice within this lineage, in the Sternarchorhynchus and Sternarchorhamphinae clades. Species included in this analysis (n = 36) were subsampled from the original morphological dataset to include only those in the apteronotid phylogeny.
A traitogram showing the uncertainty about ancestral character states using transparent probability density in phytools. The darker nodes indicate a higher confidence in estimated phenotype. These results show an extreme uncertainty about the clade's ancestral phenotype and high confidence in more recent ancestral character states. (A) Traitogram for PC1 values (the relative size of the snout and neurocranium); (B) traitogram for PC2 values (the relative flexion of the snout). 1, Sternarchorhynchus montanus; 2, Orthosternarchus tamandua; 3, Adontosternarchus baleanops; 4, Sternarchorhynchus hagedornae; 5, Apteronotus rostratus; 6, Sternarchella duccis.
Contributor Notes
Associate Editor: W. L. Smith.