Loss and reduction of paired appendages is common in vertebrate evolution. How often does such convergent evolution depend on similar developmental and genetic pathways? For example, many populations of the Threespine Stickleback and Ninespine Stickleback (Gasterosteidae) have independently evolved pelvic reduction, usually based on independent mutations that caused reduced Pitx1 expression. Reduced Pitx1 expression has also been implicated in pelvic reduction in manatees. Thus, hind limb reduction stemming from reduced Pitx1 expression has arisen independently in groups that diverged tens to hundreds of millions of years ago, suggesting a potential for repeated use of Pitx1 across vertebrates. Notably, hindlimb reduction based on reduction of Pitx1 expression produces left-larger directional asymmetry in the vestiges. We used this phenotypic signature as a genetic proxy, testing for hindlimb directional asymmetry in six genera of squamate reptiles that independently evolved hindlimb reduction and for which genetic and developmental tools are not yet developed: Agamodon anguliceps, Bachia intermedia, Chalcides sepsoides, Indotyphlops braminus, Ophisaurus attenuatuas and O. ventralis, and Teius teyou. Significant asymmetry occurred in one taxon, Chalcides sepsoides, whose left-side pelvis and femur vestiges were 18% and 64% larger than right-side vestiges, respectively, suggesting modification of Pitx1 expression in that species. However, there was either right-larger asymmetry or no directional asymmetry in the other five taxa, suggesting multiple developmental genetic pathways to hindlimb reduction in squamates and vertebrates more generally.

Micro Computed Tomography (μ-CT) scanning. Each specimen was wrapped in plastic and scanned for 4 minutes using a Perkins-Elmer Quantum GX2 micro-CT Imaging System. To prevent beam hardening, an image artifact, we used either an Al 1.0 mm or an Al 0.5 mm + Cu 0.06 mm filter, as needed. Because individual size varied within and among taxa, we adjusted voltage, current, and voxel size for specimen size (Table 1). We used the smallest voxel size that still contained the pelvis and femurs (if present).

Image processing. We used the open-source program InVesalius v.3.1 to reconstruct 3D images from raw DICOM files. We applied a reconstruction threshold to reduce image noise. Thresholds varied by voxel size and tissue density and were adjusted by eye to create surfaces with reduced artifact, maximal bone quality, and high definition. Because these criteria are subjective, no single setting optimizes 3D surface reconstruction for all specimens. For consistency, the thresholds were determined by the same individual (S.S.) across all samples.

Landmarks. Homologous landmarks were placed on each 3D surface using open-source software MeshLab 2016. All landmarks were digitized by S.S. for consistency. Each image was rotated during landmark placement to minimize parallax, and the images were rotated after landmark placement to ensure visually that they had been placed correctly. We marked the anterior point of the pubis and the posterior point of the ilium on right and left sides to measure the pelvis (Fig. 1). In specimens with a less well-developed pubis, the anterior- medial point of the epipubis was used as the anterior landmark. In species without femurs, the pubis, ilium, and ischium bones of the pelvis were usually indistinguishable; we therefore landmarked the anterior and posterior points of the pelvic vestige.

For species with femurs, we placed landmarks on the proximal point of the femur head (at the hip) and the most distal point of the femur (at the knee; Fig. 1). CT resolution was not fine enough to distinguish distal limb bones in B. intermedia due to extensive reduction (Fig. 1B). Therefore, proximal landmarks were placed on femur head and distal landmarks were placed on the most distal skeletal point of the limb, which might include non-femoral elements. However, in mice, altered Pitx1 expression resulted in reduction in the tibia, fibula, and metatarsals, as well as the pelvis and femur (Lanctot et al. 1999; Szeto et al. 1999; Marcil et al. 2003). Thus, including elements distal to the femur should yield a suitable measure of hind limb length in B. intermedia, especially since our metric is asymmetry. We exported landmarks from MeshLab as picked_points.pp files.

R statistical software is all that is required.