Skeletal plasticity in response to embryonic muscular activity underlies the development and evolution of the perching digit of birds

Most birds have an opposable digit 1 (hallux) allowing the foot to grasp, which evolved from the non-opposable hallux of early theropod dinosaurs. An important morphological difference with early theropods is the twisting of the long axis of its metatarsal. Here, we show how embryonic musculature and the onset of its activity are required for twisting of metatarsal 1 (Mt1) and retroversion of the hallux. Pharmacologically paralyzed embryos do not fully retrovert the hallux and have a straight Mt1 shaft, phenocopying the morphology of early tetanuran dinosaurs. Molecular markers of cartilage maturation and ossification show that differentiation of Mt1 is significantly delayed compared to Mt2-4. We hypothesize on how delayed maturation may have increased plasticity, facilitating muscular twisting. Our experimental results emphasize the importance of embryonic muscular activity in the evolutionary origin of a crucial adaptation.

morphology. This morphology is arguably an evolutionary intermediate between the straight Mt1 of dinosaurs and the twisted Mt1 of modern birds, and conceivably allowed greater retroversion of Mt1 than Archaeopteryx.
Most discussion on the evolution of the perching digit has centered on the comparison of adult morphology across species, but little is known about development, which can provide crucial information. D1 in the avian embryo is initially not retroverted 9 , and therefore becomes opposable during ontogeny, but no embryological descriptions address the shape of Mt1, and no information is available on the mechanisms of retroversion.
Here, we studied the development of the hallux using Galliformes (Japanese quail and domestic chicken), one of the most basal clades of crown birds that actively use their hallux for perching. We determined when and how the digit changes its orientation in relation to embryonic musculature and cartilage differentiation. Using pharmacological muscular paralysis, we assessed the role of embryonic muscular activity on retroversion. The new embryological and experimental data shows how the development of the opposable hallux has important consequences for discussing its evolution.

Results
Re-orientation of the hallux during development. The cartilaginous precursor of Mt1 originates in the medial side of the foot, almost perpendicular to the limb's main axis (HH30, fig. 2a, black arrowhead). A change in orientation of Mt1 begins at HH32, as the entire hallux becomes tilted, with its distal end slightly displaced towards ventral, and rotates such that its ventral side comes to face the lateral side of the foot, much like a human thumb (see HH32, fig. 2b-d). This orientation is similar to that of most non-avian tetanurans 10,11 . At stage HH36, D1 is re-oriented such that the entire hallux becomes perpendicular to the main limb axis, in an elevated, spur-like orientation ( fig. 2b-d). Also at HH36, the proximal contact of Mt1 to Mt2 moves markedly towards the ventral face of Mt2 ( fig. 2b-d). However, this does not cause retroversion: The ventral side of the hallux continues pointing towards the lateral side of the limb, with Mt1 abducted almost perpendicular to the tarsometatarsus. At HH38 an opposable hallux has been attained: It is no longer spur-like, presenting an opposable orientation (i.e., pointing towards Maturation and ossification of Mt1 is delayed and begins at its proximal end. Embryonic cartilages are initially composed by immature chondrocytes, characterized by the production of collagen type IIA (COLII) 12 . Maturation of metatarsals 2,3 and 4 begins at mid-shaft as cells become flattened, and start to secrete Indian hedgehog protein (IHH) 13,14 . Those cells later exit the cell cycle, begin to grow (hypertrophy), stop producing COLII and synthetize collagen type X (COLX) 15 . Eventually, the hypertrophic cells die by apoptosis, leaving their extracellular matrix for ossification and remodeling. This dynamic of expansion from the center towards the ends generates two fronts of differentiation -the growth plates 16 . This maturation pattern was also the ancestral condition for Mt1, as observed in crocodylians (bird's closest living relatives) and other amniotes 17 . The ossification of Mt2-4 begins around HH34 and their diaphyses are well ossified at HH37. In contrast with the other three avian metatarsals, Mt1 is a small element since the beginning of chondrogenesis ( fig. 2a). The size difference increases at HH32, when the other metatarsi elongate conspicuously and Mt1 keeps its small elliptic shape. Differentiation of Mt1 is greatly delayed, and follows an unusual proximal-to-distal pattern. The presence of COLII as a marker of immature chondrocytes is detected in the entire element as late as HH38 ( fig. 3a). In contrast, metatarsals 2, 3, and 4 exhibit large central regions with no COLII since HH34 ( fig. 3f). IHH is clearly present in HH32 of other metatarsi but only becomes present in Mt1 at HH34, when its production starts in the proximal end ( fig. 3b). Histological sections show that hypertrophic chondrocytes appear at the proximal end of the Mt1 at HH35 (fig. 3c). COLX synthesis begins in the proximal end of Mt1 at HH38 and advances towards distal in the next stages ( fig. 3d). Calcification starts in the proximal end only at HH40 (SI, fig. S1), and the distal half of Mt1 is covered by collar bone at HH42 ( fig. 3e) [18][19][20] . As a result of delayed differentiation from proximal to distal, Mt1 remains completely cartilaginous until exceptionally  late (HH38) with a small diaphysis and just one growth zone in the distal end. Even after the beginning of ossification, the cartilaginous epiphysis continues to occupy about half its total length (HH42).
The torsion of Mt1. In quail and chicken embryos at HH35, the long axis of Mt1 is straight. At HH36 it becomes slightly twisted at its proximal end, and at HH40 torsion is conspicuous along its entire axis Muscle development and activity in the hallux. Three muscles -two flexors and one extensorcontrol the avian hallux. The muscle flexor hallucis longus (FHL) is the only extrinsic muscle of the hallux -its belly is situated outside the foot, in the ventral shank; its tendon goes through the lateral hypotarsus and crosses obliquely to the ventral tarsometatarsus, and its principal insertion is in the ventral base of the ungual phalanx ( fig. 5c). Quail embryos exhibit a well-individualized FHL at HH33, as visualized by whole-mount immunofluorescence against Myosin ( fig. 5a). Its insertion is discernible by Tenascin expression at HH35 (fig. 5b, white arrowhead). The other flexor of the hallux, the muscle flexor hallucis brevis (FHB), is located in the medio-ventral side of the foot; its tendon inserts in the ventral side of Mt1 (or at the base of the proximal phalanx, depending on the species) 21 . The muscle originates from a common mass of muscular fibre for all ventral intrinsic muscles. Its distal part is discernible since HH33 (fig. 5a); its insertion in the ventral Mt1 is visible at HH35 ( fig. 5b, red arrowhead).
The musculus extensor hallucis longus (EHL) is the only extensor of the hallux. It arises from the dorso-medial border of Mt2 and its main insertion is in the base of the ungual phalanx. A secondary insertion in the base of the proximal phalanx is common. In quail embryos, it separates from the dorsal mass of intrinsic muscles at HH35 (fig. 5a), when its insertion is already visible (fig. 5b, white arrow); a secondary insertion is present at HH36 (fig. 5a, red arrow). Our data show that the musculoskeletal system is completely connected to the hallux at HH36, when torsion begins. The onset of digit movements occurs as soon as the muscular system is anatomically functional 22 . Digits are immobile until HH34, when the movements of the foot are restricted to the ankle joint. The first digit movements -synchronous extensions of all digits -appear at HH35, probably due to the action of the musculus extensor digiti longus. The flexion of the digits, including the hallux, starts at HH36 (see SI, Movie 1).
The effect of muscular paralysis on hallux retroversion. The temporal dynamics of muscle development and activity suggests their involvement in the torsion of Mt1. Pharmacological and genetic impairing of embryonic muscular activity have been widely used to show its importance in shaping the vertebrate skeleton [23][24][25][26][27][28] . We tested the hypothesis that muscular activity is influencing Mt1 torsion by paralyzing chick embryos before the twisting of Mt1 (HH34). Paralyzed embryos show normal ventral displacement of Mt1 ( fig. 6b), indicating that the change of the articulation site to Mt2 from medial to ventral is not influenced by muscular activity. Nevertheless, in paralyzed embryos the ventral side of the hallux faces the medial side of the foot and at HH40 its long axis lays parallel to the long axis of Mt2, similar to control embryos before the torsion of the hallux ( fig. 6a,b). Consistent with this, Mt1 of paralyzed embryos fails to twist and has a straight shape (fig. 6c). Importantly, the morphology of Mt1

Discussion
It is well known that the skeleton of vertebrates is plastic in response to altered muscular activity [30][31][32] . Although even the adult skeleton can change as a result of processes such as bone remodeling, morphology is much more affected at juvenile and embryonic stages of active cartilage growth. Several experiments have demonstrated the great early plasticity of the vertebrate skeleton, and a few are known to induce changes that reverse or imitate specific evolutionary innovations. For instance, pharmacological embryonic paralysis leads to agenesis of the tibiofibular crest of birds (an innovation present since early theropod ancestors) 33 , while terrestrialization of juvenile Polypterus (Bichir) results in modifications of the pectoral girdle resembling those of early stem tetrapods 34 .
The development of the avian hallux provides a new clear-cut case of skeletal plasticity underlying a classically discussed evolutionary innovation. The musculoskeletal system controlling the hallux is in place and active at HH36, when torsion begins in the immature Mt1 cartilage. Pharmacological paralysis confirms that embryonic muscular activity is necessary for torsion and opposability of the hallux. Our embryological observations show that torsion of Mt1 is the chief cause of digit reversion: Displacement of the site of articulation of Mt1 to Mt2, from medial to ventral, does not affect the orientation of the hallux. Throughout the twisting process, Mt1 remains an immature, COLII-expressing cartilage, and torsion is over before chondrocytes even begin to express COLX, an indicator of mature cartilage. Differentiation and ossification of Mt1 is also notably delayed with regard to the other metatarsi. This suggests that Mt1 is especially plastic at the stages in which embryonic musculature produces twisting.
Recent work has shown that embryonic muscular activity is required for retroversion of digit 4 in birds with zygodactyl feet 35 . Unlike "normal" feet with a non-retroverted digit 4, zygodactyl feet undergo asymmetric degeneration of the embryonic musculature of digit 4. This difference provides a straightforward explanation (unbalanced forces) for its retroversion. However, in the case of the hallux, the ancestral topology of musculature in tetanuran dinosaurs is inferred to have been similar to the topology found in modern birds 36,37 . If so, some additional factor must have been required for torsion of Mt1. Perhaps, delayed differentiation of Mt1 led to increased plasticity, allowing embryonic musculature to significantly alter its shape. Our data also shows that the differentiation of the hallux in modern birds follows a unique pattern where differentiation progresses from proximal to distal. In forms intermediate between Archaeopteryx and modern birds, the proximal end of MT1 is non-twisted, but the distal end is notably offset. The new context of developmental data suggests that muscle activity in these early Pygostylia could have encountered an already differentiated proximal end, but a still immature, more plastic distal epiphysis, leading to the characteristic J-like shape of their Mt1 (fig. 7). The hypothesis that successive delays in maturation allowed the progressive torsion of Mt1 is consistent with the fact that birds became increasingly paedomorphic since their dinosaur ancestors 38 and that stem Maniraptora fossil embryos -including Enantiornithes -differ from extant birds in that they hatched with almost totally ossified metatarsals, metacarpals, and phalanges [39][40][41][42][43] . In an alternative scenario, embryonic motility may begin earlier or become more intense in modern birds than in their ancestors anticipating and/or amplifying the influence of muscles on Mt1 development. Interestingly, it has been shown that metabolism and rate of bone growth are correlated 44 , and that high metabolism in Aves depends on the increase of musculature to generate heat 45 , making it possible that both increased muscular activity and changes in skeletal development are physiological and evolutionarily related to each other through the increase in avian metabolism. The role of embryonic musculature also provides an explanation for secondary acquisition of a straight, non-twisted Mt1 in penguins and petrels: These are derived among modern birds in that the FHL and EHL muscles are absent (SI, fig. S2).
The hallux has been amply debated in terms of anatomical and adaptive evolution, but no data or experiments were available to include the mechanisms that generate discussed morphologies. Our developmental data provides an important missing piece of its evolutionary history, drawing attention to other potentially involved genetic and environmental factors. In particular, we show that changes in skeletal differentiation rates may secondarily enable influences of the muscular system, yielding a possible link with skeletal heterochrony and plasticity. The subject of skeletal plasticity brings up interesting theoretical questions, that are increasingly discussed in modern evolutionary theory. A common notion is that advantageous phenotypic plasticity (i.e., that allows the development of an adaptive trait) will tend to become canalized or genetically assimilated through natural selection 46 . However, the origin of a retroverted hallux is a case of acquisition of plasticity, rather than loss by genetic assimilation. Further, much like the formation of the fibular crest 33 , torsion of Mt1 in birds is phylogenetically ancient, but continues to develop based on mechanical forces, which have not been replaced/assimilated by some alternative genetic/molecular mechanism. Genetic assimilation of an adaptive trait may not be an inevitable outcome, especially so if development of the trait is based on systemic/epigenetic interactions that emerge reliably, in a recurrent and stable fashion 47,48 . In retrospective, we believe our data supports the need for a conceptual framework that encourages explicit discussion of developmental mechanisms and systemic interactions, further integrating this information with morphological evolution as observed in the fossil record 49 .

Material and Methods
(a) Experimental Animals and Nomenclature. Fertilized Broiler chicken and quail eggs were incubated at 37.5 °C and 70% humidity in an incubator with automatic rotating shelves. All animal Figure 7. The evolutionary transition to a twisted Mt1. In Intermediate forms like the early short-tailed Quiliania, Mt1 has a non-twisted proximal end, but an offset ("bent") distal end (arrow). Because Mt1 in modern birds matures from proximal to distal, we suggest that in (1) non-Pygostylian birds, Mt1 would have matured before the onset of muscular activity; (2) Maturation of Mt1 was delayed in pygostylia, allowing its distal bending by embryonic muscular activity; (3) Ornithuromorpha further delayed maturation of Mt1, allowing full torsion of its long axis by embryonic muscular activity.
Scientific RepoRts | 5:09840 | DOi: 10.1038/srep09840 procedures were in accordance with the Chilean legislation and were approved by Institutional Animal Care and Use Committees. We follow an anatomical terminology based on Baumel and Witmer 50 . The series of chicken and quail development are based on Hamburger and Hamilton (HH) stages 51 .
(b) Skeletal Staining. Developmental series for quail and chicken were prepared for cartilage and bone staining. For cartilage staining embryos were fixed in 100% and stained in a 5:1 ethanol:acetic acid solution with 0.025% 8G Alcian Blue (Sigma-Aldrich) for two days at RT in an orbital shaker. For bone staining, embryos were post-fixed in 10% buffered formalin for 1 hour at 4 °C and stained with 0.04% Alizarin Red (Sigma-Aldrich) in 0.5% KOH for 1 hour at RT. Embryos were cleared in a sequence of 1:3, 1:1 and 3:1 glicerol:water. For fluorescent staining of bones, embryos were fixed in 4% of paraformaldehyde solution, washed and incubated for 2 hours in 10 μl/ml of Calcein (Sigma-Aldrich), and then cleared with Urea 4 M. Paraffin sections were cut at 7 μm thick and stained with Safranin and Hematoxylin employing standard histological protocols. (d) Embryo Paralysis. We treated seventy-two chicken embryos at HH34 in three independent rounds of experiments. Embryos were paralyzed by the application of 100 μl of a solution of 15 mg/ ml of Decamethonium bromide diluted in PBS 53,54 . Control embryos received the same amount of the vehicle, PBS. The drug was delivered with a micropipette through a small hole in the shell above the air camera. Eggs were sealed with adhesive tape and re-incubated until the appropriate stages, verified by morphological criteria.