The vertebral column is critical to a vertebrate species' flexibility and skeletal support, making vertebrae a clear target for selection. Anurans (frogs and toads) have a unique, truncated vertebral column that appears constrained to provide axial rigidity for efficient jumping. However, no study has examined how presacral vertebrae shape varies among anuran species at the macroevolutionary scale nor how intrinsic (developmental and phylogenetic) and extrinsic (ecological) factors may have influenced vertebrae shape evolution. We used microCT scans and phylogenetic comparative methods to examine the vertebrae of hundreds of anuran species that vary in body size as well as adult and larval ecology. We found variation in shape and evolutionary rates among anuran vertebrae, dispelling any notion that trunk vertebrae evolve uniformly. We discovered the highest evolutionary rates in the cervical vertebrae and in the more caudal trunk vertebrae. We found little evidence for selection pressures related to adult or larval ecology affecting vertebrae evolution, but we did find body size was highly associated with vertebrae shape and microhabitat (mainly burrowing) affected those allometric relationships. Our results provide an interesting comparison to vertebrae evolution in other clades and a jumping-off point for studies of anuran vertebrae evolution and development.

Data collection and microCT scanning of museum specimens

Of the 209 specimens used in this study (each belonging to a unique species), 203 specimens were microCT scanned for this study, and an additional six specimen scans were downloaded from Morphosource (https://www.morphosource.org/). Our sampling includes at least one species from 47 of the extant 54 anuran families. All specimens used in this study are museum specimens from the National Museum of Natural History (123), Museum of Vertebrate Zoology (80), California Academy of Sciences (2), Centre for Ecological Sciences (1), Museum of Comparative Zoology (1), University of Kansas Biodiversity Institute (1), and University of Florida (1) herpetology collections. We microCT scanned the 203 specimens loaned from the National Museum of Natural History and the Museum of Vertebrate Zoology using a Phoenix vjtomejx M (GE Measurement & Control Solutions, Boston, MA, USA), at the University of Florida’s Nanoscale Research Facility. We performed all scans with a 180 kV X-ray tube containing a diamond-tungsten target, with the voltage, current, and detector capture time adjusted for each scan to maximize absorption range. We reconstructed all scans on GE’s datosjx software version 2.3 and segmented all skeletons using VG StudioMax (Volume Graphics, Heidelberg, Germany). All scans are available for download on Morphosource (https://www.morphosource.org/, project number P967) and information on all specimens used in this study (including those downloaded from Morphosource) can be found in Supplemental Table 1.

We measured snout-vent length (SVL) of all loaned specimens to the nearest tenth millimeter using a digital caliper (31- 415-3, Swiss Precision Instruments, Inc., Garden Grove, CA) and measured SVL of all skeleton models downloaded from Morphosource in Meshlab (Cignoni et al. 2008). We classified larval habitat as direct developer, lotic tadpole, or lentic tadpole, for 176 species from 138 primary literature sources (see Supplemental Table 1) and well as secondary references and databases, AmphibiaWeb.org (AmphibiaWeb 2021), IUCN (IUCN 2021), and the AmphiBio database (Oliveira et al. 2017). We classified adult microhabitat for 207 species using data previously collated (Buttimer et al. 2020; Stepanova and Womack 2020) from primary (Andreone and Luiselli 2003; McCranie and Castañeda 2005; Brito et al. 2012; Matojo 2015) and secondary references (IUCN 2021; Moen et al. 2016; Moen and Wiens 2017; AmphibiaWeb 2021). We used seven of the eight microhabitat categories defined by Moen and Wiens (Moen and Wiens 2017): (1) aquatic—usually in water, (2) arboreal—typically on aboveground vegetation, (3) burrowing—nonbreeding season spent underground or in burrows they dug, (4) semi-aquatic—partially aquatic and partially terrestrial, (5) semi-arboreal—partially arboreal and partially terrestrial, (6) semi burrowing—partially burrowing and partially terrestrial, (7) terrestrial—found on the ground, under rocks, or in leaf litter, and (8) torrential—found in high-gradient, fast-flowing streams. We did not include any semi-burrowing species in this study. All associated data and references are in Supplemental Table 1.

Landmarking Vertebrae

After obtaining skeleton meshes of all scans, seven landmarks were placed on each presacral vertebrae of each specimen using the digit.fixed function in the R package geomorph v4.0.0 (Adams et al. 2021; Baken et al. 2021). All downstream analyses were likewise performed in R v4.1.0 (R Core Team 2021). The seven landmarks corresponded to homologous and repeatable points that defined the outer borders of the vertebrae. Six landmarks were in positions on the dorsal face of the vertebral column. Two of these were placed on either the right or left caudal tip. Another two landmarks were placed on the transverse process tips (one on the left and one on the right). When transverse processes widened towards their terminus rather than coming to a point, landmarks were placed at the center of the tip. Two additional landmarks were placed along each vertebra’s central axis: one at the cervical edge between the transverse processes and another at the caudal edge between the caudal tips. The seventh landmark was also placed on the central axis at the cervical edge but on the ventral face of the column rather than the dorsal face. We assessed landmark placement reliability by having at least two separate researchers landmark 80 of our 209 specimens. We performed a Generalized Procrustes analysis (GPA) on each vertebra that had multiple landmarkers, performed a principal components analysis (PCA) on the aligned landmarks for each vertebra, and compared within-specimen variation in the PCA results. To quantify variation among landmarkers, we added the standard errors of the first four PCA axes for each specimen and compared the sum of these standard errors among specimens, looking for outlier specimens with high standard errors. We also visually inspected the proximity of landmarked vertebrae in the first four principal components to verify inconsistencies among landmarkers. We found the landmarks in this study were reliably placed across all anuran families and provided novel information, since we could find no other study that analyzed vertebrae in three dimensions at the macroevolutionary scale.

R; geomorph package