Early amphibians evolved distinct vertebrae for habitat invasions

Abstract

Living tetrapods owe their existence to a critical moment 360-340 million years ago when their ancestors walked on land. Vertebrae are central to locomotion, yet systematic testing of correlations between vertebral form and terrestriality and subsequent reinvasions of aquatic habitats is lacking, obscuring our understanding of movement capabilities in early tetrapods. Here, we quantified vertebral shape across a diverse group of Paleozoic amphibians (Temnospondyli) encompassing different habitats and nearly the full range of early tetrapod vertebral shapes. We demonstrate that temnospondyls were likely ancestrally terrestrial and had several early reinvasions of aquatic habitats. We find a greater diversity in temnospondyl vertebrae than previously known. We also overturn long-held hypotheses centered on weight-bearing, showing that neural arch features, including muscle attachment, were plastic across the water-land divide and do not provide a clear signal of habitat preferences. In contrast, intercentra traits were critical, with temnospondyls repeatedly converging on distinct forms in terrestrial and aquatic taxa, with little overlap between. Through our geometric morphometric study, we have been able to document associations between vertebral shape and environmental preferences in Paleozoic tetrapods and to reveal morphological constraints imposed by vertebrae to locomotion, independent of ancestry.

Fig 1. Temnospondyl vertebral types.

Neural arches in white, pleurocentra in dark gray, intercentra in light gray, spinal canal in packed gray circles, notocanal in striped bars. Left column; rhachitomous vertebra from Eryops megacephalus (FMNH 117) (A) and stereospondylous vertebra from Mastodonosaurus giganteus (AMNH 2994) (B), vertebrae in caudal view. Right column: variations on rhachitomous vertebrae seen in Temnospondyli; (a) Rhachitomous; (b) Reverse Rhachitomous; (c) Stereospondylous; (d) Plagiosaurid. White arrows point to prezygapophyses, black arrows point to postzygapophyses.

Fig 2. Morphospace for temnospondyl neural arches.

Principle component analysis shows vertebral type distribution. A: “consensus” data set; B: “most recent” data set. Dashed circles represent changes between A and B. Convex hulls are grouped according to a priori habitat. Grey shapes are theoretical neural arches representing at a particular point in PC1 and PC2. Shape of the point represents the vertebral type.

Fig 3. Morphospace for temnospondyl intercentra.

Principle component analysis displaying vertebral type distribution. Convex hulls are grouped according to a priori habitat. A: “consensus” data set; B: “most recent” data set. Dashed circles represent changes between A and B. Grey shapes are theoretical intercentra representing at a particular point in PC1 and PC2. Shape of the point represents the vertebral type.

Fig 4. Phylogeny and habitat of temnospondyls.

Tip colors are pooled prior probabilities for clades as gathered from literature and paleobiology database (PBDB; Supplementary Information). Node colors display posterior probabilities calculated from the best-fit (OU; terrestrial-semiaquatic-aquatic) mode in Ancthresh. Thick lines represent stratigraphic range. Reconstructions borrowed with permission from Nobu Tamura.

Fig 5. Traces of neural arch and intercentra shape convergence characterized by environment.

Top row: Reconstructions borrowed with permission from Nobu Tamura. From left to right: Cacops aspidephorus Paracyclotosaurus davidi, Archegosaurus decheni. Middle row: a) Cacops aspidephorus (ancestrally terrestrial); b) Paracyclotosaurus davidi (ancestrally semiaquatic); c) Metoposaurus diagnosticus (ancestrally aquatic). Bottom row: d) Lydekkerina huxleyi (secondarily terrestrial); e) Mastodonsaurus giganteus (variation on semiaquatic form, note no secondarily semiaquatic taxa were reported from this study); f) Archegosaurus decheni (secondarily aquatic).