Derived mode of retinotectal development in E. coqui
Direct developing frogs are known from 12 different anuran families . Although phylogenetic relationships among anurans continue to be debated [44–47], the distribution of direct development on any of the proposed anuran phylogenetic trees indicates that it has evolved many times independently from an ancestral biphasic pattern of development.
The development of the retina and tectum in D. pictus and P. pustulosus as documented here has many similarities with retinotectal development in other biphasically developing frogs including Rana [48–51]), Limnodynastes [52, 53], and Xenopus, which is particularly well studied [40, 54–68]. Since this pattern of retinotectal development is shared between neobatrachian frogs (Rana, Limnodynastes, Physalaemus), including the leptodactylid frog Physalaemus and more basal "archaeobatrachian" lineages such as discoglossids (Discoglossus) and pipids (Xenopus), it probably represents the primitive condition for extant anurans. Thus, the differences in retinotectal development observed in Eleutherodactylus are derived within this direct developing clade of leptodactylid frogs.
In the various biphasically developing species, the different retinal layers develop during embryonic development, when the retina is still small and has reached only a fraction (about one third) of its cross-sectional area at the completion of metamorphosis. The first retinotectal connections are established during early embryonic development, shortly after the major axon tracts have developed in the brain (Tables 1, 2). The expression of NeuroD (only described for Xenopus), which plays a role in regulating neuronal differentiation and the formation of particular neuronal subtypes such as photoreceptors and amacrine cells in the retina [33, 34, 37, 69–75] becomes restricted to the ciliary margin, the outer part of the inner nuclear layer, and the outer nuclear layer as soon as the different retinal layers develop. Proliferation becomes largely restricted to the ciliary margin at the end of embryonic development, from which the retina then grows during larval stages by addition of cells to all cellular layers.
When the first retinofugal fibers reach the contralateral tectum at late embryonic stages, the latter is still very small and consists predominantly of proliferating cells. Retinal fibers cover the tectum in a rostrocaudal direction during larval stages, as the optic tectum continues to grow by proliferation from the ventricular layer, which is most pronounced caudomedially. At early larval stages, the first tectal layers (layer 7–9 and layer 5) differentiate and retinal fibers cover approximately half of the lateral part of the tectum leaving its midline free of fibers. At this stage the tectum has reached only a fraction (about one third) of its size at the completion of metamorphosis. At the end of metamorphic climax, the entire surface of the tectum is covered by retinofugal fibers. In addition to the main, contralateral projection to thalamus, pretectum, and tectum, an ipsilateral projection to thalamus and pretectum, which subserves important functions for binocular vision in postmetamorphic frogs, begins to develop at late larval stages in Xenopus [66, 76]. In D. pictus, a sparse ipsilateral projection develops already at early larval stages.
Retinotectal development in E. coqui has been modified in a number of respects from this ancestral anuran pattern. While the spatiotemporal order of differentiation of retinal and tectal layers and the formation of the retinotectal projection in E. coqui have been conserved, patterns of growth and proliferation of retina and tectum have been modified. Both retina and optic tectum grow rapidly during initial formation of the retinotectal projection in embryonic stages so that at a stage when formation of retinal layers and tectal layers 5–9 are completed, they have already reached around 90 % of their size at hatching. While retinal growth in biphasically developing frogs is mostly due to addition of cells from the proliferative ciliary margin, the retina of E. coqui grows by proliferation of cells throughout the outer part of the inner nuclear layer of the retina until it has almost reached its size at hatching . Thus, although NeuroD expression in the ciliary margin, the outer part of the inner nuclear layer, and the outer nuclear layer resembles the condition in Xenopus, many NeuroD expressing cells in the inner nuclear layer of E. coqui retinae are probably proliferating progenitor cells in contrast to Xenopus. NeuroD expression in some retinal ganglion cells and absence of NeuroD expression in the tectum of E. coqui also differ from X. laevis, suggesting that some of the functions of NeuroD have changed during evolution of E. coqui. A sparse ipsilateral retinothalamic projection develops already at embryonic stage TS 9, prior to completion of retinal layer formation and, thus, significantly earlier than in biphasically developing frogs. This may represent an adaptive heterochronic shift related to precocious adoption of postmetamorphic head shape and eye position in E. coqui, allowing onset of binocular vision already at hatching stages when the optic tectum is still in a relatively immature, larval-like state (note dissociation between suites XI/XII indicating cranial remodeling and suite II indicating retinotectal differentiation in Fig. 1).
Dissociation of growth and differentiation during E. coqui retinotectal development
The modified pattern of retinotectal development in E. coqui indicates that the regulation of growth and differentiation have been dissociated during evolution of Eleutherodactylus as discussed in  (compare suites III/IV with suite II in Fig. 1). Growth has been greatly accelerated by increased cell proliferation, while the timing of early retinotectal differentiation events has been conserved and remains temporally coordinated with early cranial and spinal development (compare suites IX, V, and II in Fig. 1).
In biphasically developing anurans such as Xenopus, thyroid hormones have been implicated in promoting increased proliferation in the retina during mid-larval stages after the thyroid gland develops and thyroid hormone levels rise [77, 78]. However, the precocious growth of retina and tectum in E. coqui is unlikely to be due to the precocious action of thyroid hormones, because acceleration of retinal and tectal growth is observed from early embryonic stages on, long before the thyroid gland matures and the thyroid axis becomes functional around stage TS 10 in E. coqui [30, 31]. Although the mechanisms underlying the increase in proliferation in E. coqui remain at present obscure, the conserved schedule of retinotectal differentiation implies that increased proliferation is not simply due to a general delay in neural differentiation. Changes in the probability of cell cycle exit or the length of cell cycles are better compatible with the pattern observed.
As a consequence of the rapid growth of retina and tectum during the development of the retinotectal projection, retinofugal fibers in E. coqui encounter a much larger target territory as they enter and distribute over the tectum than in biphasically developing frogs. This raises the question, whether the formation of a topographic map of retinal fibers on the tectal surface, as known from other vertebrates, may be compromised in E. coqui. However, the formation of retinotopic maps, which is now known to involve gradients of Eph receptors and their ephrin ligands in retina and tectum, is very plastic and can expand or contract in response to drastic experimental reductions of retinal or tectal size, respectively [79, 80]. In many fishes and amphibians plastic mechanisms of map formation are important to maintain an ordered retinotopic map throughout development, because their tectum grows in a rostrocaudal direction, while new retinal neurons are added in a radial direction [57, 65, 81]. This plasticity probably permitted the drastic acceleration of retinotectal growth in Eleutherodactylus without compromising the ability to form a well ordered retinotopic map.
Increased proliferation and accelerated growth in E. coqui is not confined to the retina and tectum, but is also evident in other parts of the CNS such as the spinal cord  and the brain stem (Schlosser, unpublished observation) (compare suites III, IV, and VIII in Fig. 1). This suggests that altered growth patterns in E. coqui may be due to systemic regulatory changes affecting proliferation of neural progenitors throughout the CNS, although independent regulatory changes in different parts of the CNS cannot be ruled out.
Mosaic evolution of CNS development in E. coqui
Despite altered growth patterns, the schedules of retinal and tectal differentiation and of the formation of retinotectal projections are conserved and remain coordinated with each other and with the formation of early axon tracts in the brain in E. coqui (compare suites II and I in Fig. 1) probaby reflecting their interdependent development for example due to coordinated axon outgrowth. For example, fibers in the tract of the postoptic commissure (TPOC), which is the earliest tract to develop in the embryonic brain, later fasciculate with fibers from multiple other brain tracts as well as with retinofugal fibers. This has led to suggestions that fibers of the TPOC may serve as pioneer axons, on which other brain tracts and retinofugal fibers depend for proper axonal pathfinding [38–41, 67, 82, 83]. However, other studies have shown that retinofugal fibers can properly navigate towards the tectum independent of the TPOC [84, 85] suggesting that both TPOC and other fiber tracts including retinofugal fibers may instead follow common pathway cues and, thus, may form in a temporally coordinated fashion once their pathways become established.
Although schedules of retinotectal differentiation and the formation of early brain tracts remain tightly coordinated with each other in E. coqui, they have become dissociated from differentiation in other parts of the central and peripheral nervous system. For example, the lateral motor columns (LMCs) in the spinal cord develop precociously in E. coqui paralleling precocious development of the limbs, which they innervate [22–25, 27, 28]. As a consequence, in E. coqui LMCs develop simultaneous with the retinotectal system, while in biphasically developing frogs they develop in larval stages after differentiation of most retinal and tectal layers is completed (compare suites VII and II in Fig. 1). Similarly, adult-like cranial skeletal structures and muscles together with their motor innervation precociously differentiate in E. coqui in stages [9, 13, 16, 18, 19, 21], when the retinotectal system is still in an immature, larva-like condition (compare suites XI/XII and II in Fig. 1).
Taken together, this indicates that during evolution of Eleutherodactylus, the schedule of differentiation in CNS areas remained coordinated with the schedule of differentiation of those regions in the CNS or in the periphery, with which they are directly connected (retina-tectum-brain tracts, LMC-limbs, cranial motor neurons-cranial muscles), whereas profound temporal dissociations have taken place between structures that are not or only indirectly connected.