Fig 1.
Convergent external morphology of the three clades of Malagasy skinks of the sand-swimming ecomorph.
Photographs of (A) Grandidierina rubrocaudata* from Sakabera, (B) G. lineata* from Faux Cap, (C) G. fierinensis*, dark phenotype from Tombohina, and (D) pale phenotype from Anakao, (E) G. petiti* from Sakabera, (F) Voeltzkowia yamagishii** from Ankarafantsika, (G) V. mobydick** from the Bongolava plateau, (H) V. mira from Antsanitia, and (I) P. minimus from the baie de Sakalava. Arrows indicate limbs. All except G represent living specimens. Photographs from A. Miralles, F. Eckhard, A. Rakotoarison and J. Köhler. * Species included in the genus Voeltzkowia until the present study; ** Species included in the genus Sirenoscincus until the present study (see taxonomic accounts in Discussion).
Fig 2.
Phylogenetic tree of Malagasy scincines.
Tree reconstructed using Bayesian Inference, based on concatenated DNA sequences of three mitochondrial (12S, 16S rRNA, ND1) and four nuclear (BDNF, C-mos, Rag2 and phosducin) loci, with posterior probabilities followed by bootstrap support values >50% from a Maximum Parsimony analysis. The three convergent clades including species with regressed eyes of the sand-swimming ecomorph are highlighted by coloured rectangles (note that Pygomeles braconnieri and P. petteri might also be considered as large-sized sand-swimmers). Species are shown with their previous genus-level classification (in quotation marks) while the larger font on the right side shows the revised classification. Within G. fierinensis, pale and dark phenotypes are indicated by yellow and brown circles, respectively.
Fig 3.
Pectoral girdle and forelimbs skeletal anatomy.
Computed tomographic reconstruction of the pectoral girdle (in purple) and forelimbs (in yellow) of seven different species of blind sand-swimming skinks and one species without limb regression (Amphiglossus ornaticeps), based on microtomographic scans, in ventral, ventrolateral and lateral view (not to scale).
Fig 4.
Pelvic girdle and forelimbs skeletal anatomy.
Computed tomographic reconstruction of the pelvic girdle (in purple) and hindlimbs (in yellow) of seven different species of blind sand-swimming skinks and one species without limb regression (Amphiglossus ornaticeps), based on microtomographic scans, in ventral, ventrolateral and lateral view (not to scale).
Fig 5.
Evolution of limb regression over time.
(A) Evolutionary regression of forelimbs and hindlimbs within the Malagasy scincine lineage. Character evolution is reconstructed using three parsimony optimizations implemented in PAUP 4.0b10 with an ordered and irreversible algorithm (Camin-Sokal parsimony). Character evolution is represented on the left side for the forelimbs (red gradient), and on the right side for the hindlimbs (blue gradient). Numbers represent the number of character changes involved in each reconstruction. (B) Time tree of Malagasy scincines, using a Bayesian relaxed-clock phylogenetic approach, with two normal priors used for time calibration: the divergence between Madascincus and Amphiglossus 47 Mya, as retrieved from a previous more comprehensive analysis of Malagasy vertebrates [8], and a mitochondrial substitution rate of 0.0053/site/MYr (see methods for details).Yellow bars at nodes indicate 95% credibility intervals for divergence events. Temporal scale is shown in millions of years. Blue and red bars are maximal temporal windows for limb regression for the four limbless lineages presented in the tree, based on the ancestral state reconstruction with ordered and irreversible characters, and delimited by the last node fitting with a fully pentadactyl inferred ancestor and the first node fitting with an inferred ancestor with no external limbs. Intervals are shown for forelimbs with red bars, and for hindlimbs with blue bars. Where applicable, the maximal temporal frame for initiation of the limb regression process (delimited by the last node fitting with a fully pentadactyl ancestor and the first node fitting with an ancestor with less than five fingers) is represented by light red / blue colors.
Fig 6.
Morphological versus molecular differentiation between the dark and pale phenotype of G. fierinensis.
(A) Number of ventral and dorsal scales along the body. (B) Haplotype network reconstruction including all the species of Grandidierina. Note the elevated amount of shared haplotypes among the two G. fierinensis phenotypes. (C) Mitochondrial divergence within the G. fierinensis complex (subtree extracted from the result of the Bayesian analysis carried out on the concatenated mtDNA sequences, cf. S2 Appendix). (D) Genetic structure inferred from four nuclear loci across 12 individuals of G. fierinensis, with K = 2; assignment probability of each specimen (N = 12) is around 50% for either cluster, independent of its phenotype.