Recent efforts to decipher the enigma of the turtle's shell revealed that distantly related turtle species deploy diverse processes during shell development. Even so, extant species share in common a shoulder blade (scapula) that is encapsulated within the shell. Thus, evolutionary change in the correlated development of the shell and scapula probably underpins the evolution of highly derived shell morphologies. To address this expectation, we conducted one of the most phylogenetically comprehensive surveys of turtle development, focusing on scapula growth and differentiation in embryos, hatchlings and adults of 13 species. We report, to our knowledge, the first description of secondary differentiation owing to skeletal remodelling of the tetrapod scapula in turtles with the most structurally derived shell phenotypes. Remodelling and secondary differentiation late in embryogenesis of box turtles (Emys and Terrapene) yielded a novel skeletal segment (i.e. the suprascapula) of high functional value to their complex shell-closing system. Remarkably, our analyses suggest that, in soft-shelled turtles (Trionychidae) with extremely flattened shells, a similar transformation is linked to truncated scapula growth. Skeletal remodelling, as a form of developmental plasticity, might enable the seemingly constrained turtle body plan to diversify, suggesting the shell is not an evolutionary straitjacket.

Understanding developmental processes is foundational to clarifying the mechanisms by which convergent evolution occurs. Here, we show how a key convergently evolving trait is slowly ‘acquired’ in growing turtles. Many functionally relevant traits emerge late in turtle ontogeny, owing to design constraints imposed by the shell. We investigated this trend by examining derived patterns of shell formation associated with the multiple (at least 8) origins of shell kinesis in small-bodied turtles. Using box turtles as a model, we demonstrate that the flexible hinge joint required for shell kinesis differentiates gradually and via extensive repatterning of shell tissue. Disproportionate changes in shell shape and size substantiate that this transformation is a delayed ontogenetic response (3–5 years post-hatching) to structural alterations that arise in embryogenesis. These findings exemplify that the translation of genotype to phenotype may reach far beyond embryonic life stages. Thus, the temporal scope for developmental origins of adaptive morphological change might be broader than generally understood. We propose that delayed trait differentiation via tissue repatterning might facilitate phenotypic diversification and innovation that otherwise would not arise due to developmental constraints.

A key trend in the 210-million-year-old history of modern turtles was the evolution of shell kinesis, that is, shell movement during neck and limb retraction. Kinesis is hypothesized to enhance predator defense in small terrestrial and semiaquatic turtles and has evolved multiple times since the early Cretaceous. This complex phenotype is nonfunctional and far from fully differentiated following embryogenesis. Instead, kinesis develops slowly in juveniles, providing a unique opportunity to illustrate the postembryonic origins of an adaptive trait. To this end, we examined ventral shell (plastral) kinesis in emydine box turtles and found that hatchling plastron shape differs from that of akinetic-shelled relatives, particularly where the hinge that enables kinesis differentiates. We also demonstrated shape changes relative to plastron size in juveniles, coinciding with a shift in the carapace-plastron structural connection, rearrangement of ectodermal plates, and bone repatterning. Furthermore, because the shell grows larger relative to the head, complete concealment of the head and extremities is only achieved after relative shell proportions increase. Structural alterations that facilitate the box turtle's transformation are probably prepatterned in embryos but require functioninduced changes to differentiate in juveniles. This mode of delayed trait differentiation is essential to phenotypic diversification in turtles and perhaps other tetrapods.

Diversification of the turtle's shell comprises remarkable phenotypic transformations. For instance, two divergent species convergently evolved shell-closing systems with shoulder blade (scapula) segments that enable coordinated movements with the shell. We expected these unusual structures to originate via similar changes in underlying gene networks, as skeletal segment formation is an evolutionarily conserved developmental process. We tested this hypothesis by comparing transcriptomes of scapula tissue across three stages of embryonic development in three emydid turtles from natural populations. We found that alternative strategies for skeletal segmentation were associated with interspecific differences in gene co-expression networks. Notably, mesenchyme homeobox 2 (MEOX2) and HOXA3-5 were central hubs driving the activity of 2,806 genes in a candidate network for scapula segmentation, albeit in only one species. Even so, scapula muscle overgrowth corresponded to the activity of similar myogenic networks in both species. This and other derived developmental processes were not observed in the third species, which displayed the ancestral (unsegmented) scapula condition. Differential gene expression tests against this reference lineage supported histological and network analyses. Our findings illustrate that molecular underpinnings of convergent evolution, including during the diversification of the atypical turtle "body plan," are influenced by variation in underlying developmental processes.

The cover image, by Gerardo A. Cordero et al., is based on the Research Gene network variation and alternative paths to convergent evolution in turtles, DOI: . Photo credit: Gerardo Antonio Cordero.