Digit evolution in gymnophthalmid lizards

Abstract

Background The tetrapod limb is a highly diverse structure, and reduction of the limbs accounts for much of the phenotypes observed within species. Squamate reptiles represent one of the many lineages in which the limbs have been greatly modified from the pentadactyl generalized pattern; within the group, limb-reduced morphologies can vary from minor reductions in size of elements to complete limblessness, with several intermediate forms in between. Even though limb reduction is widespread, it is not clear what are the evolutionary and developmental mechanisms involved in the formation of reduced limb morphologies.

Methods In this study, we present an overview of limb morphology within the microteiid lizard group Gymnophthalmidae, focusing on digit number.

Results We show that there are two major groups of limb-reduced gymnophthalmids. The first group is formed by lizard-like - and frequently pentadactyl - species, in which minor reductions (such as the loss of 1-2 phalanges mainly in digits I and V) are the rule; these morphologies generally correspond to those seen in other squamates. The second group is formed by species showing more drastic losses, which can include the absence of an externally distinct limb in adults. We also show the expression patterns of Sonic Hedgehog (Shh) in the greatly reduced fore and hindlimb of a serpentiform gymnophthalmid.

Conclusions Our discussion focus on identifying shared patterns of limb reduction among tetrapods, and explaining these patterns and the morphological variation within the gymnophthalmids based on the current knowledge of the molecular signaling pathways that coordinate limb development.

Introduction

The tetrapod limb has been the subject of extensive investigation in evolutionary and developmental biology for over a century. Since the 19^th century, the evolutionary origin and diversification of the limbs have been greatly debated, and several hypothesis and mechanisms have been proposed to explain how such diversity arose during evolution.

Even though the identification of evolutionary patterns of limb evolution are crucial, the fundamental question of how does such diversity arises is better answered by studying the mechanisms involved in the formation of the limbs. This is not a trivial investigation, because it involves getting access to embryonic material, and being able to experimentally manipulate such embryos, only possible with a more in depth knowledge of the genetics of the organism.

The bulk of knowledge on limb development in tetrapods relies, up to the moment, on the two most extensively studied organisms, the lab mouse and chicken. The signaling networks controlling limb development seem to be fairly conserved among those species, although important differences exist. Great reviews have been published summarizing the current knowledge of how limb development is organized (Niswander, 2003; Zeller et al., 2009; Rabinowitz and Vokes, 2012); for a more detailed picture of such mechanisms, the reader should refer to that literature and references within. However, variations of those mechanisms among natural populations should be expected, and have, in fact, been demonstrated for some species (Thewissen et al., 2006; Cooper et al., 2007; Hockman et al. 2008; Doroba and Sears, 2010; Sears et al., 2011; Cooper et al. 2014; Lopez-Rios et al., 2014).

A group that is underrepresented in developmental studies is the Squamata. Understanding the evolution and diversification of the reptile limb has been the focus of studies in several research fields, including evolutionary biology, paleontology, ecology, anatomy, physiology, functional morphology and, to a lesser degree, developmental biology (Gans, 1975; Lande, 1978; Withers, 1981; Greer, 1991; Caldwell, 2002; Shapiro, 2002; Shapiro et al., 2003; Whiting et al., 2003; Crumly and Sanchez-Villagra, 2004; Kearney and Stuart, 2004; Kohlsdorf and Wagner, 2006; Wiens et al., 2006, Brandley et al., 2008; Kohlsdorf et al., 2008; Russel and Bauer, 2008; Skinner et al., 2008; Bergmann and Irschick, 2009; Jerez and Tarazona, 2009; Young et al., 2009; Leal et al., 2010; Hugi et al., 2012; etc).

A recurrent phenotype among squamate reptiles is limb reduction. From minor losses of phalanges, to the complete loss of the limb, there is a wide spectrum of intermediate morphologies that can occur even among species that belong to the same genus (Choquenot and Greer, 1989; Skinner et al., 2008). Changes like those are extremely frequent, having occurred multiple times independently in almost all major squamate groups (Greer, 1991; Wiens et al. 2006; Skinner et al., 2008). The scincids and anguids top the rank of limb-reduced lineages, but reduction is also seen in the pygopodids, gekkonids, cordylids, gerrho-saurids, dibamids, gymnophthalmids, amphisbaenians, and, obviously, in the snakes, which have attained the greatest degree of limb reduction. Previous studies have successfully identified patterns of limb reduction among squamates and provided evolutionary, functional, anatomical, and biogeographical/environmental hypothesis for the evolution of reduced forms (Greer, 1991; Benesch and Withers, 2002; Wiens et al., 2006; Brandley et al., 2008; Camacho et al., 2014).

Even though an incredible diversity of limbs is observed among squamate reptiles (lizards, snakes and amphisbaenians), little is known about the mechanisms involved in limb development within the group. Most of the studies concern morphological analyses of limb development, which are important for laying the anatomical foundations for further functional investigations (for example, Howes and Swinnerton, 1901; Mathur and Goel, 1976; Rieppel, 1994; Shapiro, 2002; Fabrezi et al., 2007; Leal et al., 2010; Roscito and Rodrigues, 2012a,b), but still few studies (Raynaud, 1990; Raynaud et al., 1998; Cohn and Tickle, 1999; Shapiro et al., 2003; Young et al., 2009) attempted to undercover the molecular mechanisms behind limb development in natural populations of squamate species.

Among the Gymnophthalmidae, a South American group of microteiid lizards, limb-reduced species are widespread, occurring in desert-like environments, highlands, and also forested areas. Despite the diversity of forms, we still lack a clear picture of the diversity of limb morphologies within the group, how are such morphologies distributed across the lineages, and how could they have evolved. Thus, in this paper we present an extensive survey of the limbs of several gymnophthalmid species with the aim of identifying patterns of limb reduction within the group. We also present a preliminary investigation of the molecular signaling involved in the formation of a reduced limb, and analyse the different types of reduction based on the current knowledge of the mechanisms controlling limb development in vertebrates.

METHODS

Material examined:

Our sample is comprised of 34 Gymnophthalmid genera and 70 species, out of a total of 46 genera and 244 species currently recognized for the group (reptile-database.org; august/2014). We included in our sampling only those species for which we could get information of the phalangeal formula, either by observing cleared and stained material, or by information from the literature.

Cleared and double stained specimens are from our personal collection and from the collection of Museu de Zoologia da Universidade de Sao Paulo; the material examined and respective collection numbers is listed in the supplementary information.

The notation used for representing the phalangeal formula is the following:

Phylogenies of the Gymnophthalmidae

Up to the present, there are three phylogenies for the Gymnophthalmidae (Pellegrino et al., 2001; Castoe et al., 2004; Pyron et al., 2013). The same species are consistently grouped together in all three phylogenies (Figure 1). However, each author assigns each group to differently inclusive taxonomic rankings, which generates incongruences over the status of particular gymnophthalmid groups (see discussion in Rodrigues et al., 2007b, 2009). In this paper we do not wish to solve those incongruences, therefore our naming of the major groups and subgroups recognized within the Gymnophthalmid reflects our personal ideas regarding the evolution of these species:

Figure 1.

Relationships within the Gymnopthalmidae, adapted from the phylogenetic hypothesis of Pellegrino et al. (2001), Castoe et al. (2004), and Pyron et al. (2013). Lineages with limb-reduced species are marked in grey.

Alopoglossinae (Pellegrino et al., 2001; Castoe et al., 2004): Alopoglossus, Ptychoglossus

Rachisaurinae (Pellegrino et al., 2001; Castoe et al., 2004): Rachisaurus

Cercosaurinae

Bachiini (Castoe et al., 2004): Bachia

Cercosaurini (Castoe et al., 2004): Cercosaura, Echinosaura, Neusticurus, Placosoma, Pholidobolus, Potamites, Proctoporus.

Ecpleopodinae (Castoe et al., 2004; Pyron et al., 2013): Anotosaura, Arthrosaura, Colobosauroides, Dryadosaura, Leposoma.

Gymnophthalminae

Heterodactylini (Rodrigues et al., 2009): Caparaonia, Colobodactylus, Heterodactylus

Iphisiini (Rodrigues et al., 2009): Acratosaura, Al-exandresaurus, Colobosaura, Iphisa, Stenolepis

Gymnophthalmini (Pellegrino et al., 2001): Tretioscincus, Gymnophthalmus, Micrablepharus, Procellosaurinus, Psilophthalmus, Vanzosaura, Nothobachia, Scriptosaura, Calyptommatus

Embryonic material and whole mount in-situ

We obtained a small developmental series of five Calyptommatus sinebrachiatus embryos during a field trip to Bahia State, Brazil, in 2005, that were fixed in RNAlater after 6, 10, 12, 14, and 16 days after laying. The whole-mount in-situ hybridization for detection of the gene Sonic hedgehog followed a modified protocol used for chicken embryos, starting with two 5-minute washes in 100% methanol, with a 1-hour in between incubation in 6% hydrogen peroxide solution (in 100% methanol). Following rehydration of the embryos in PBT, the embryos were subjected to a proteinase K digestion (20 minute reaction at room temperature) followed by re-fixation in 4% PFA for 20 minutes. Embryos then went through to a series of ‘pre-probe’ incubations in hybridization buffer and finally incubated overnight at 65oC with a chicken-specific probe for Shh (1:10 dilution). After a series of washes with MABT, and a blocking step, embryos were incubated overnight with anti-DIG (1:2000), and then the signal was revealed with NBT/BCIP.

RESULTS

Digit arrangement in gymnophthalmid lizards

The phalangeal formula of both fore and hindlimb, as well as specific comments on the morphology of metacarpal/ metatarsal and phalangeal elements, are presented in Table 1.

Table 1.

Summary of the phalangeal formulas for Gymnophthalmidae species. Entries in grey correspond to those species not directly analyzed. dc/dt, distal carpal/distal tarsal; mc/mt, metacarpal/metatarsal.

A typical squamate pentadactyl condition (Romer, 1956) is that of 5 digits in both fore and hindlimb, with a specific number of phalanges in each digit (Figure 2):

Figure 2.

Forelimb (A) and hindlimb (B) of Cercosaura schreibersi, showing the generalized phalangeal arrangement. Digit identity is indicated as dl-dV. Anterior to the top. Scale bar = 1mm.

Most gymnophthalmid species have limbs following the generalized pentadactyl arrangement, which is the case for all cercosaurinis and alopoglossinis, and a few other species from the Ecpleopodinae. However, several species that show a full pentadactyl arrangement exhibit a reduction in size of one or both phalanges of digit I of the forelimb; this is seen in at least two gymnophthalmid groups.

The majority of species from the Iphisiini (Rodrigues et al., 2009) show a slight reduction in the size of the ungueal phalanx from dI, with Alexandresaurus, Colobosaura (Figure 3D), and Stenolepis (Figure 3E) bearing no claw in this digit; in Iphisa, both phalanges seem reduced. Reduction is not clear in Acratosaura, more material is needed in order to confirm whether its limbs share the morphology of its relatives.

Figure 3.

Forelimb morphologies in gymnophthalmid species with minor limb reductions: Anotosaura vanzolinia (A); Colobodactylus dalcyanus (B); Colobodactylus taunayi MTR 00746 (C); Colobodactylus taunayi MZUSP 94254 (D); Colobosaura modesta (E); Stenole-pis ridley (F); Heterodactylus lundii (G); Heterodactylus imbricatus (H). Digit I to the top. Scale bars: A = 0.5 mm; B-H = 1.0 mm.

Tretioscincus is the other genus in which a clear reduction in size of phalanges in dI can be observed; the other gym-nophthalmini species have lost either one (Gymnophthalmus) or both phalanges from d1 of the forelimb (Micra-blepharus, Psilophthalmus, Procellosaurinus, Vanzosaura) (Figure 4 A-D).

Figure 4.

Limb morphologies in gymnophthalmid species with minor limb reductions: Gymnophthalmus underwoodi forelimb (A); Micrablepharus atticolus (B); Psilophthalmus paeminosus (C); Tretioscincus agilis (D); Rachisaurus brachylepis (E, F). Digit I to the top. Scale bars: A, B, D, F = 1.0 mm; C, E = 0.5 mm

Complete loss of dI of the forelimb, as seen in the beforementioned gymnophthalmini species, only occurs in Rachisaurus brachylepis (Figure 4 E,F). Colobodactylus’ dI of the forelimb is clearly reduced, but we disagree with Kizirian and McDiarmid (1998) and Grizante (2009), who report only the presence of mcI, since we have observed one reduced phalanx following the metacarpal (mc) in both C. dalcyanus and C. taunayi. However, it is possible that there is intrapopulational variation for this character; the presence of a vestigial ossification distal to mc1 in C. taunay (MTR 746) and the apparent absence of such element in the second individual analysed here (MZUSP 94254) argues in favor of this variation (Figure 3B,C).

Interestingly, all those species in which both phalanges from dI of the forelimb are lost, the corresponding metacarpal (mcI) is always present. Also worth of notice is that in all these cases previously mentioned, the minor reductions, or loss of dI, are not accompanied by reductions/losses in the hindlimbs, which maintain the pentadactyl arrangement. The exception is Rachisaurus brachylepis, which has lost one phalanx from dI, and all phalanges from dV of the hindlimb; this condition is not seen in any of the other Gymnophthalmidae analysed.

Another “type” of reduction is seen in some species of the Ecpleopodinae: while most are pentadactylous, Anotosaura and Dryadosaura exhibit a loss of the last phalanx of dlV of the forelimb. In addition, in Anotosaura collaris the last phalanx of dV of the hindlimb is absent (Figure 3A; its sister species, A. vanzolinia, shows the full pentadactyl arrangement). The single individual of Colobosauroides cearensis analysed shows a left-right asymmetry in the number of phalanges in dIV of the forelimb.

Limb reduction reaches extreme cases in two gymnoph-thalmid lineages: the gymnophthalmini, with Nothobachia and Calyptommatus (and possibly Scriptosaura; Rodrigues and dos Santos, 2008), and the Bachiini (Figure 5).

Figure 5.

Limb morphologies in gymnophthalmid species with the greatest degrees of limb reduction: Bachia panoplia fore (A) and hindlimb (B); Bachia bresslaui hindlimb (C); Bachia flavescens hindlimb (D); Nothobachia ablephara fore (E) and hindlimb (F); Scriptosaura catimbau fore (G) and hindlimb (H); Calyptommatus leiolepis forelimb (I); Calyptommatus sinebrachiatus hindlimb (J). Digit I to the top. Scale bars: A-C, F, G, I-J = 0.5 mm; D, E, H = 0.25 mm.

Within the Gymnophthamini (Figure 5E-J), Nothobachia still retains both fore and hindlimbs, although they are very reduced (forelimb styliform and hindlimb with two digits) and presumably not functional for fast walking (Renous et al., 1998). Scriptosaura and Calyptommatus show even further reductions: both lack the forelimb, although retaining a vestigial internal humerus close to the pectoral girdle (and possibly, a vestige of the second limb segment in Scriptosaura; Roscito and Rodrigues, 2013); both species have a styliform hindlimb.

Bachia is a very interesting genus with respect to the diversity of limb morphologies. Limb reduction within the group is more pronounced in the hindlimbs than in the forelimbs, the opposite of all the other gymnophthalmid species in which some kind of reduction occurs. More interestingly, the digits of Bachia species show an apparent loss of anterior-posterior patterning (Figure 5A,B), as previously noted (Kohlsdorf and Wagner, 2006). Digit identity can be determined in species with more developed limbs, but in cases of extreme reductions, no recognizable digit structure is left (Figure 5C).

Bachia species are quite rare in the field, and also in museum collections, therefore, we could only get access to three cleared and stained specimens; the majority of information for these species was compiled from the literature. Because these are rare animals, the taxonomic studies on Bachia species usually don’t contain osteological information, and descriptions of the limbs are mainly based on external morphology. For this reason, we relied solely on the osteological information published by Dixon (1973), Thomas (1965), Kizirian and McDiarmid (1998), and Kohlsdorf and Wagner (2006).

Phylogenetic analysis

The types of limb reduction described above cluster in specific groups (Figure 6 and Figure 7): in the lineage that holds Rachisaurus, heterodactylinis, iphisiinis, and gymnophthalminis, reduction affects digit I of the forelimb, while the hindlimb is frequently pentadactylous (except for Rachisaurus brachylepis). Bachia species exhibit a wide range of morphologies, but none of them following the typical pentadactyl arrangement. Last, reduction in the ecpleopodinis affects digit IV of the forelimb and also digit V of the hindlimb.

Figure 6.

Evolutionary relationships within the Cercosaurini, Heterodactylini, Iphiisini, and Ecpleopodinae subgroups, with a representation of the different limb morphologies observed in each species/groups of species. Limb skeletal elements present in the limbs are colored dark grey, and those that are absent (in reference to the generalized pentadactyl condition) are colored light grey. Species names colored in black indicate those species that have fully pentadactyl limbs; names colored in blue indicate those species that also have pentadacatyl limbs but show reduction in size of one or more phalanges; species colored in red have lost one or more phalanges in relation to the pentadactyl condition. FL/HL, forelimb/hindlimb; ra, radius; ul, ulna; c, carpus; ti, tibia; fi, fibula; t, tarsus.

Figure 7.

Evolutionary relationships within the Gymnophthalmini and Bachiini subgroups, with a representation of the different limb morphologies observed in each species. Limb skeletal elements present in the limbs are colored dark grey, those that are absent (in reference to the generalized pentadactyl condition) are colored light grey, and those with uncertain identity are in red. Species names colored in blue indicate those species that have pentadacatyl limbs but show reduction in size of one or more phalanges; species colored in red have lost one or more phalanges in relation to the pentadactyl condition. FL/HL, forelimb/hindlimb; ra, radius; ul, ulna; c, carpus; ti, tibia; fi, fibula; t, tarsus.

Mapping the distribution of limb morphologies onto the gymnophthalmid subgroups shows shared morphologies among groups of species and allows pinpointing shared or independent origins for similar limb arrangements.

In general, the species that show some kind of limb reduction are grouped together. For example, limb reduction in the Ecpleopodinae is restricted to the genera Anotosaura and Dryadosaura, which are nested within the group, while all sister species show no kind of reduction (Figure 6). The same is true for the Heterodactylini, with the reduced forms grouping together (Figure 6). Cercosaurini and Iphisiini species don’t exhibit variation in limb morphology; the former is comprised of only pentadactyl species, while species from the later group also share the typical pentadactyl arrangement, but show reduction in size of the last phalanx of digit I of the forelimb (Figure 6).

Gymnophthalmini and Bachiini, on the other hand, do not show a clear clustering of species with shared phenotypes (Figure 7). One of the two main branches recognized among the Gymnophthalmini (Figure 7) is comprised exclusively of the lacertiform Tretioscincus, Micrablepharus, Vanzosaura, and Procellosaurinus; while the other holds the lacertiform Gymnophthalmus and Psilophthalmus, and the snake-like Nothobachia and Calyptommatus (Gymnophthalmus is placed as a separate lineage by Castoe et al., 2004). Loss of digit I of the forelimb occurs in Psilophthalmus, Micrablepharus, Vanzosaura, and Procellosaurinus, yet Psilophthalmus does not group directly with the later three species. The snake-like gymnophthalminis Nothobachia and Calyptommatus, which represent the extremes of this subgroup regarding limb morphology, are always grouped together (the topology of Castoe et al., 2004, however, places Psilophthalmus in between them). The Bachiini (Figure 7) show an even more puzzling situation, in which the diversity of limb configurations follows no clear pattern among the species for which limb morphology is known in detail. The limiting amount of data available regarding both the anatomy and evolutionary relationships of Bachia species contribute to this confusing scenario.

Analysis of gene expression in Calyptommatus sinebrachiatus

We analysed the expression of Shh in the fore and hindlimb buds of Calyptommatus sinebrachiatus embryos staged from 6 to 16 days of development (Figure 8). When the egg is laid, both fore and hindlimb are present (Roscito and Rodrigues, 2012b); our selection of developmental stages was intended to sample the early development of both limbs. The forelimb bud develops up to 9-10 days, and then its growth stops and the bud starts to degenerate. The hindlimb bud, which does not degenerate, is already paddle-shaped at 16 days after laying, with femur, tibia, and fibula detected by cartilage staining (Roscito and Rodrigues, 2012a).

Figure 8.

Expression of the gene Sonic Hedgehog (Shh) in embryos of the gymnophthalmini Calyptommatus sinebrachiatus with 6 (A-B), 10 (C-D), 12 (E-G), 14 (H-J), 16 (K) days of development. Images A, C, E, and H show the entire embryo: arrows in A point to the fore and hindlimb, arrow in C points to the forelimb, and arrow in E points to the hindlimb; the images in detail on the upper right corner represent a transversal cut at mid-trunk showing positive staining in the nothocord. Forelimbs are represented in F and I; hindlimbs in D, G, J, and K. Scale bars: A-B, C, E, H = 0.5mm; D, F, G, I, J, K= 0.25mm.

Sonic hedgehog mRNA was consistently observed in the notochord of all embryos (Figure 8A, C, E, H; images in detail). However, it was not detected in the forelimb at any of the stages analysed. In contrast, Shh it was detected in the posterior mesenchyme of the hindlimb bud of the 16-day old embryo (Figure 8K), the oldest stage analysed.

DISCUSSION

In addition to the knowledge of the molecular pathways involved in ‘normal’ limb development, there is also a vast amount of information on the influence of disruptions of such pathways to the resulting limb morphology. These interventions, and the phenotypes resulting from it, provide starting points for investigations of limb development in other species that show different limb morphologies.

Patterns of limb reduction in the Gymnophthalmidae

Limb reduction has evolved repeatedly in squamates, with almost every major group showing the loss of one or more bones in the limbs. The situation is no different among the Gymnophthalmidae, with all but three groups exhibiting loss of one or more phalanges (Figure 1).

Minor reductions, involving the loss of one or two phalanges, account for the majority of cases of limb reduction among the gymnophthalmids. Digit I of the forelimb is the most affected: out of the 10 genera analysed that show these minor losses, 8 have lost phalanges in dI. When a single phalanx is lost, the remaining phalanx is very reduced in size, while the size of the metacarpal seems comparable to that of the other metacarpals. In contrast, when both phalanges are absent, mcI is reduced to a vestigial element. Phalanx loss is also seen in dIV of the forelimb, but with a lower frequency when compared to losses in dI. This scenario contrasts the observations of Greer (1991), that has shown that dIV is more likely to be affected when losses occur in a single digit from the forelimb (26 times), followed by dI (6 times), and last, by dV (only 1 case).

Losses of phalanges in either dI or dIV of the forelimb seem to be specific to distinct lineages: loss in dI is seen in the Rachisaurinae, Heterodactylini, Iphisiini and Gymnophthalmini, while loss of dIV occurs only in some species of the Ecpleopodinae. However, Rodrigues et al. (2013) have recently described a new Leposoma species, L. sinepollex that, as the name suggests, lacks dI of the forelimb. Furthermore, the closely-related L. nanodactylus is reported to have a shorter dI of the forelimb (Rodrigues 1997; Rodrigues et al., 2013). These observations break down the notion of clade-specific types of reduction, since Leposoma are nested within the Ecpleopodinae, a group characterized by reductions affecting dIV of the forelimb. Furthermore, it shows that the scincoides lineage of the Leposoma genus, to which L. nanodactylus and L. sinepollex are allocated, represents one more independent instance of digit reduction among the gymnophthalmids (Rodrigues et al., 2013). A second event of loss of dI of the forelimb among the Ecpleopodinae is seen in the monotypic Amapasaurus tetradactylus (Cunha, 1970; pers. obs.). This species seems to be more related to the parietale group of Leposoma (Rodrigues and Ávila-Pires, 2005), which holds the two Leposoma species analysed here (L. percarinatum and L. osvaldoi), none of which showing any kind of limb reduction. If this relationship is confirmed, then Amapasaurus represents yet another independent event of limb reduction within the group.

Interestingly, the hindlimbs of all of those gymnophthalmid species that show minor losses of phalanges in the forelimb do not show any kind of reduction, and still maintain the ancestral phalangeal formula. The exceptions are Anotosaura collaris, which has lost a single phalanx from dV, and Rachisaurus brachylepis, which has lost the entire dV but the corresponding metatarsal. Digit V of the hindlimb is the digit that accounts for the most cases of losses of phalanges (Greer, 1991; Shapiro et al., 2007), although losses of all phalanges from dV usually co-occur with complete loss of dI (Shapiro et al., 2007), which is not the case for Rachisaurus.

The effect of intrapopulational variation on the number of phalanges lost in both fore and hindlimb, as previously observed in Bachia and Hemiergis (Dixon, 1973; Choquenot and Greer, 1989), cannot be estimated based on our sampling, most of which consisting of a single individual per species. The observation of two specimens of Colobo-dactylus taunayi with a different configuration of dI of the forelimb - depending on the interpretation of the vestigial ossification seen in specimen MZSUP 94254 (Figure 3D), specimens either differ in the number of phalanges, or in the morphology of the single phalanx - argues in favor of the need increased sampling in order to understand the morphology of reduced limbs and its dominant pattern within a species. On the other hand, we have analysed several individuals of Vanzosaura rubricauda, Psilophthalmus paeminosus, and Procellosaurinus tetradactylus and none of those showed any variation in number of phalanges. Interestingly, the three later species have a vestigial mcI, while that of Colobodactylus has a normal size. The extent to which this variability in the number of phalanges represents a clade-specific flexibility, some kind of developmental constraint related to how reduced is the digit, or a simple factor of change in our sampling, will unfortunately remain unknown until increased sampling and more detailed anatomical studies are done.

Major limb reductions are seen in two other gymnoph-thalmid groups: the Gymnophthalmini, with Nothobachia, Calyptommatus, and Scriptosaura (Rodrigues and dos Santos, 2008), and the genus Bachia. In such severe cases of limb reduction, especially regarding losses of multiple digits and of other elements to an almost limblessness state, the remaining skeletal elements (if any) are vestigial and frequently show some kind of loss of anatomical information. Examples of limb elements that don’t share anatomical similarity to its homologous elements from pentadactyl relatives are: the vestigial humerus of Calyptommatus and Scriptosaura; the extremely reduced elements seen laterally to the digits in the forelimb of Nothobachia and in the hindlimb of Calyptommatus, which cannot be identified only based on the adult morphology (Figure 5A, J); and the vestigial elements in Bachia bresslaui hindlimb (Figure 5C, D).

On the other hand, the skeletal elements present in “intermediate forms” usually show some degree of anatomical individuality and may still maintain topological relations to neighboring elements. This is the case for the hindlimb of Nothobachia ablephara, in which the two remaining digits can be identified as dill and dIV based on their relations to the tarsal elements (Roscito and Rodrigues, 2013). Remarkably, digits and phalanges in Bachia seem to be deprived of anatomical information, in the sense that all digits - even in species with high number of digits, look similar to each other.

Two major patterns of limb reduction can be recognized among those Gymnophthalmid lizards that show some degree of limb reduction, whether it is a loss of a single phalanx or the loss of essentially the entire limb. In all species but Bachia, reduction of the forelimb is always more advanced than reduction in the hindlimb; Bachia shows an opposite trend. The first case, in which the forelimb is more reduced than the hindlimb, is the most frequent pattern in limb-reduced squamates, while the second case is only seen in Bachia, in one scincid and one teiid, and in the amphisbaenian genus Bipes (Brandley et al., 2008). As every ‘rule’ has its exception, two recently described Bachia species (B. micromela and B. psamophila; Rodrigues et al., 2007) have the forelimbs more reduced than the hindlimbs. Although no osteological data was presented for either species, the representations clearly show that the hindlimb is more developed than the forelimb (both in length and in number of digits, in the case of B. psamophila).

In addition, reduction in the fore and hindlimbs of Bachia seem to follow distinct morphological patterns: reduction in the forelimb has, apparently, an anterior predominance (dI usually reduced or absent), while in the hindlimb, dV is the most affected. Reduction/loss of dI of the forelimb is common among limb-reduced forms (75% of the cases; Shapiro et al., 2007), and there is not a correlation between reduction/loss of this digit with reduction/loss of any of the other digits, meaning di is relatively independent. Although reduction in dV of the hindlimb is also common (90% of the cases; Shapiro et al., 2007), it is highly correlated with corresponding reduction in dI of the hindlimb (73%). Digit V is lost in the hindlimbs of B. panoplia, B. scolecoides, and B. heteropa, but dI is complete (2 phalanges) in all three species, contrary to the co-loss of dI and dV reported in the above-mentioned study. This predominant reduction of the post-axial side of the limb, with maintenance of the pre-axial side, resembles the pattern seen in archosaurs fore and hindlimbs (Shapiro et al., 2007) and in anurans (Shubin and Alberch, 1986). In addition, a few Bachia species such as B. panoplia, B. scolecoides, B. intermedia, B. bicolor, and B. dorbignyi, show morphologically similar digits with uniform phalangeal numbers, similar to what is seen among the turtles and in some archosaurs (Shapiro et al., 2007). These opposing modes of digit reduction between fore and hindlimbs of Bachia suggest that different developmental mechanisms may be coming into play. Unfortunately, the lack of a detailed anatomical analysis of the genus is a drawback to further hypothesis on limb development.

Developmental biology of reduced limbs

Morphological perspective

The development of the tetrapod limb is a complex process that requires precise special and temporal coordination of many signaling molecules that lead to the differentiation of an initially homogeneous population of cells into the different tissues that form the limbs.

One of those tissues, the limb skeleton, arises from pre-cartilaginous primordial that are formed in a proximaldistal direction. In amniotes, the order of appearance of limb skeletal elements seems to be remarkably conserved, despite the major differences in limb morphology across species. Comparative analyses have shown that the pre-cartilaginous elements form following a primary axis of development that runs through the humerus/femur, ulna/ fibula, distal carpal/tarsal IV, and digit IV; the remaining digits usually form following the sequence III > II/V > I (Shubin and Alberch, 1986). This generality holds true for the squamates analysed so far (Howes and Swinnerton, 1901; Mathur and Goel, 1976; Rieppel, 1994; Shapiro, 2002; etc).

Another generality derived from the analysis of limb-reduced species is that limb elements are lost in the reverse order of their development; thus, the last elements to form are usually lost first. In the case of digits, loss follows the order I > V/II > III/IV (Morse, 1872; Lande, 1978; Greer, 1987; 1991; Shapiro et al., 2007). This scenario could imply that losses of elements that are not essential for the development of other elements could take place more easily than loss of those elements that form the primary axis of the limb. The apparent constraint against the loss of the central digits (III and IV) could also be related to the role that these digits play in hand/foot stability during locomotion (Greer, 1991), but we still lack a detailed and comparative knowledge of how hand/feet width and length could influence on locomotion performance.

Taking into account the prevalence of the primary axis, and the relative common sequence of digit development, one could preliminarily infer that the reduced digits of Calyptommatus, Scriptosaura, and Nothobachia are the most central digits: those in Nothobachia hindlimb would be identified as dIII and dIV, and the single digit in Nothobachia forelimb and in Calyptommatus and Scriptosaura hindlimbs would be identified as dIV. However, if identified as such, then the number of phalanges in these digits does not match to that observed in the corresponding digit if a generalized pentadatyl limb.

The first question that stems from this observation is if these apparently incomplete digits would be the product of an early termination of limb bud development (see Shapiro, 2002, for further discussion). If so, we should expect to find a similar digit/phalangeal configuration in some developmental stage of a pentadactyl limb bud. However, this is not the case for neither species, arguing against a truncation mechanism to explain the resulting limb morphology.

In contrast, there are cases in which the adult phalangeal arrangement corresponds to the arrangement seen in a particular developmental stage, thus reflecting that the reduction process was likely the result of a truncation of embryonic development. This is the case for Hemiergis initialis (Shapiro, 2002), and several other skinks, agamids, cordylids, gekkonids, and others, that show minor losses of one or two terminal phalanges of digits IV and V (Greer, 1991; Russel and Bauer, 2008); among the gymnophthalmids, Anotosaura shows a phalangeal configuration resembling that of a late developmental stage of Calotes versicolor (Mathur and Goel, 1976). Interestingly, all these cases involve minor losses of phalanges (1-3 phalanges lost).

Digit configurations seen in some Bachia species, such as B. panoplia, B. scolecoides, and B. bicolor also do not seem to correspond to truncations of a pentadactyl embryonic developmental program. On the other hand, the forelimbs of B. bresslaui, B. trisanale, and B. heteropa resemble stages of development of other squamates (Shapiro et al., 2007), and could possibly have originated from truncations of development.

Molecular signaling perspective

The understanding of the signaling network that controls limb growth and patterning during embryonic development sheds light on the possible mechanisms by which a limb becomes reduced. Lab-induced mutations, or alterations of individual signaling pathways, result in a variety of limb phenotypes, some of which may resemble those seen in natural populations. Furthermore, the improvement and easier access to genetic and molecular tools led to an increasing interest in the investigation of a wider diversity of species other than the commonly studied chicken and mouse. This increasing knowledge helps to direct investigations on species never so far studied.

Multiple signals interact to control both the growth and patterning of the developing limb. Changes in such system can affect growth, but not the overall pattern (scaling), or can affect the pattern itself (and, hence, localized growth) and result in a different configuration of the skeletal elements.

The forelimb of Dryadosaura looks like the result of a scaling process of limb autopod, since the phalanges are much shorter in comparison to those of the hindlimb, and of the forelimbs of similar-sized gymnophthalmids. However, Dryadosaura also has one phalanx less from the generalized pentadactyl pattern - a condition seen in many other Gymnophthalmids. These frequent asymmetric losses of phalanges reflect some kind of deviation from the common developmental pattern other than a simple scaling mechanism, which should, in principle, affect all digits equally.

Digits form from single cartilaginous condensations in the autopod, which are separated from each other by interdigital spaces. These condensations elongate and segment into phalanges, and the distal-most phalanx differentiates into the ungueal phalanx, associated with epidermal structures. Growth of the digit cartilaginous condensation is controlled by FGF8 signaling from the apical ectodermal ridge (AER), which, in conjunction with WNT signaling from the ectoderm, maintain the distal-most cells in an undifferentiated, proliferative state (Stricker and Mundlos, 2011). BMPs from the mesenchyme are important positive regulators of chondrogenesis. Downregulation in either FGF8 or BMP signaling results in shorter cartilaginous condensations and, frequently, in brachydactyly phenotypes (reduced size or loss of phalanges; Sanz-Esquerro and Tickle, 2003; Stricker and Mundlos, 2011). Conversely, over expression of FGF8 results in increased growth of the digit condensations, and in the formation of additional phalanges in the chicken limb (Sanz-Esquerro and Tickle, 2003).

In this scenario, minor spatial and temporal modulations of FGF8 or BMPs expression could account for the formation of a reduced number of phalanges, or the reduction in size of these elements in the gymnophthalmids analyzed here and also in other amniotes with reduced limbs. Furthermore, the predominance of loss/reduction in the lateral-most digits (dI and dV), a common trend among squamate reptiles (Greer, 1991; Shapiro et al., 2007), could be easily explained by a possible antero-posterior shortening of the extension of the AER, which would put the lateral-most digits under the influence of smaller amounts of FGF signaling. In fact, shortening of the AER is likely one of the mechanisms that contribute to the reduction of dII and dV in the embryonic limbs of some mammals (Cooper et al., 2014; Lopez-Rios et al., 2014).

In contrast, the more infrequent cases of loss of the terminal phalanx from digit IV (seen here only in Anoto-saura species) could be explained by an early termination of AER-derived signaling, which would lead to the truncation of the normal developmental sequence. This hypothesis is supported by experimental manipulations in chicken limbs, in which early interruption of FGF signaling result in the loss of terminal phalanges and in the premature induction of the digit tip developmental program (Sanz-Esquerro and Tickle, 2003) - in fact, the modulation between lower FGF signaling and the switching on of the “tip developmental program”, which involves Wnt signaling, explains the presence of the ungueal phalanx in those digits with reduced number of phalanges. Furthermore, the ungueal phalanx from dIV is the last one to form in development (Mathur and Goel, 1976; Shapiro et al., 2007), hence, would be the first one lost following an early termination of FGF signaling.

Signaling from the AER promotes mesenchymal cell proliferation and limb outgrowth; inhibition of FGF8 and FGF4 at different time developmental stages leads to progressive loss of elements along the proximo-distal axis, and knockdown of these FGFs results in failure of limb formation (Lewandoski et al., 2000; Sun et al., 2002; Dudley et al., 2002). Downregulation of AER signaling results in reduced cell proliferation, resulting in skeletal defects due to the reduced number of cell progenitors. Raynaud (1990) showed that inhibition of DNA synthesis results in a reduction of the number of digits formed in the limbs of the pentadactylous lizard Lacerta viridis, and that the earlier the treatment, the more digits we missing. More interestingly, the digits most frequently affected (lost) were dI and dV and those most frequently retained were dIII and dIV, which corresponds to the naturally occurring pattern described previously (Greer, 1991; Shapiro et al., 2007).

A premature downregulation of FGFs from the AER and the consequent truncation of the developmental pathway could also explain the morphology of the forelimb of Bachia heteropa, which has a phalangeal arrangement (1.2.3.3.2) similar to that seen in a developmental stage of Hermiergis initialis (Shapiro, 2002). However, the arrangement of B. heteropa hindlimb (2.2.3.3.0) is not paralleled by any developmental stage of any squamate studied so far, although it is similar to a developmental stage of the newt Ambystoma (Alberch and Gale, 1985).

The uniform phalangeal number seen in most Bachia species (1-2 phalanges in at least 3 digits) could be the result of truncations of the developmental program coupled with additional mechanisms directing the loss of lateralmost digits - as for example the already-mentioned antero-posterior shortening of the AER and of the signaling derived from it.

The homogeneity of both size and morphology of the digits of those Bachia species led Kohlsdorf and Wagner (2006) to suggest a correspondence of these arrangements with the phenotype resulting from mouse knockout mutants for Gli3. This mesenchymal transcription factor, together with Sonic hedgehog from the ZPA (zone of polarizing activity), are key players in the antero-posterior polarization of the developing limb bud (Wang et al., 2000; Litingtung et al., 2002; te Welscher et al., 2002); their antagonistic interactions determine a signaling gradient along the AP axis of the limb bud over which digit number and identity are laid down (Litingtung et al., 2002). Limbs that develop in the absence of Gli3 are polydactylous, and the resulting digits have 2 phalanges each and seem identical (“appear more serially homologous than in the wild type”; Litingtung et al., 2002). Although we cannot be sure about the developmental mechanisms responsible for the limb phenotypes seen in Bachia, it sounds unlikely that some sort of downregulation of Gli3 could account for these arrangements. Reduction in the amount of Gli3 would imply an anterior expansion of Shh expression, which would result, most likely, in some degree of polydactyly.

Loss or downregulation of Shh, on the other hand, could explain the limb morphologies of Nothobachia and Calyptommatus. Progressively inactivation of Shh expression from mouse limb buds leads to a corresponding progressive loss of digits (Zhu et al., 2008), with more digits being lost the earlier Shh is inactivated. The same effect was observed in scincid lizards from the genus Hemiergis with different limb morphologies: reduction in the number of digits is to a reduction in the duration of Shh expression; the more reduced the limb, the less time Shh is present in the limb buds (Shapiro et al., 2003). Shh maintains a positive feedback loop with growth-promoting FGFs from the AER via Gremlin signaling (Laufer et al., 1994; Niswander et al., 1994; Zuniga et al., 1999; Lewandoski et al., 2000; Sun et al., 2002); thus, decrease in Shh expression is always correlated with a decrease in cell proliferation, resulting in a smaller limb bud (Chiang et al., 2001; Shapiro et al., 2003) and, consequently, in less progenitor cells.

Determining the identity of digits present in partially or greatly reduced limbs can be quite difficult. Even though one might rely on the primary axis generalization, reduced limbs/digits can show a high degree of anatomical divergence from the ancestral condition, often resulting in controversial homology assignments; one of the greatest examples is the over 150-year old debate over the identity of the digits in the chicken wing (reviewed in Young et al., 2011). Therefore, integrating morphological and molecular evidences is essential for the elaboration of an evolutionary scenario to accommodate the divergent observations (for example, Young et al., 2009).

As discussed above, from the morphological point of view, multiple embryological evidences (Shubin and Alberch, 1986; Müller and Alberch, 1990; Chiang et al., 2001; Noro et al., 2009; Young et al., 2009; Leal et al., 2010; Shapiro, 2002, and many others) show that: i) the condensation corresponding to dIV is the first one to form in the limb buds; ii) digits most often form sequentially in a posterioranterior order (IV>III>II/V>I; although Zhu et al., 2008 evidences claim for more detailed observations of digit development); and iii) that digits are lost in the reverse order of their development, which makes dIV the last one to disappear (Morse, 1872; Sewertzoff, 1931; Chiang et al., 2001; Shapiro et al., 2002).

From the molecular biology point of view, there are still standing questions over the exact roles of Shh signaling in the patterning of the AP axis and in determining digit number and identity (Tabin and McMahon, 2008; Harfe, 2011). Nevertheless, it is clear that, in the absence of Shh signaling, the single digit that forms is the Shh-independent dI (Chiang et al., 2001; Ros et al., 2003), whereas progressive loss of Shh leads to dIV being lost last (Zhu et al., 2008). The detection of Shh in the hindlimb of Calyptommatus is an indication that the single-digit phenotype is different from the mutant single-digit phenotype. In addition, Shh-null mice also show severe defects in the zeugopod (Chiang et al., 2001), not seen in Calyptommatus hindlimb. By combining molecular and morphological inferences, we could suggest that dIV is the remaining digit in the vestigial hindlimb of Calyptommatus (as previously interpreted in Roscito and Rodrigues, 2013), and that the absence of other digits may be explained by a failure in maintenance of Shh expression, as observed in Hemiergis lizards.

On the other hand, positional information may not always be a good proxy for determining digit identity and homology, given that patterning of digit identity can be uncoupled from the anatomical positioning of the cartilage condensation. The digits in the chicken wing and in both fore and hindlimb of the lizard Chalcides chalcides are examples where this uncoupling occurs: while anatomical analysis supports the identification of the digits as dI, dII, and dIII (in both the chicken and C. chalcides), gene expression analyses during limb development show that the digits are patterned as dII, dIII, and dIV (Wagner and Gauthier, 1999; Young et al., 2009). The Frame Shift Hypothesis (Wagner and Gauthier, 1999) accommodated these conflicting evidences by proposing a homeotic transformation of character identity.

The phalangeal number in the single digit of Calyptommatus hindlimb is reminiscent of dI from a pentadactyl condition, which might reflect a frame shift-like mechanism taking place in the patterning of this digit. Digit i develops in the anterior-most area of the hand/foot autopod where only Hoxd13, out of the distal Hoxd genes, is expressed (Montavon et al., 2008); thus, the absence of other distal Hoxd genes from the presumptive dI region is a reliable indication of di fate - in fact, Hoxd11 and Hoxd12 expression have been used to identify the homeotic frame shifts of the chicken and the lizard C. chalcides, as previously discussed. The investigation of the expression of the distal Hoxd genes in the hindlimb of Calyptommatus would be decisive in determining if a homeotic shift would explain the digit configuration.

The extent to which disruptions in Shh signaling may be associated with changes in digit identity is still unknown. On one side, a change in downstream Shh signaling in bovine embryos results in a medial-distal shift of Hoxd13 expression, but the resulting digits do not have dI identity (Lopez-Rios et al., 2014). On the other hand, a shift in digit identity was observed in experimentally induced inhibitions of Shh in the chicken wing: digits I and II developed, but these digits formed in embryonic positions of dIII and dIV (Vargas and Wagner, 2009). Therefore, even though Shh was observed in Calyptommatus hindlimb, we cannot exclude the possibility of a downstream effect on its signaling cascade that could potentially induce a frame shift event.

Whether similars mechanisms take place in the singledigit fore and hindlimb of Nothobachia and Scriptosaura, respectively, remains an open question.

The even more reduced forelimb of Calyptommatus, represented in by a vestigial humerus located within the body wall, forms normally in early development but regresses at later stages (Roscito and Rodrigues, 2012b). Sonic hedgehog was not detected in this limb bud at any stage, showing that: i) Shh is not necessary for the emergence of the limb bud (consistent with previous results; Chiang et al., 2001; Ros et al., 2003); and ii) the absence of this important patterning signal may be one of the factors involved in degeneration of the forelimb bud.

Arrested development of limb buds in limb-reduced lizards has been documented previously for the lizards Anguis fragilis, Ophisaurus apodus, and Scelotes brevipes (Rahmani, 1974; Raynaud, 1963; Raynaud et al., 1975); the python snake has vestigial limb buds that do not degenerate (Cohn and Tickle, 1999; Boughner et al., 2007). Cetaceans hindlimb buds also form during embryonic development but degenerate (Sedmera et al., 1997; Thewissen et al., 2006). Remarkably, Shh is not expressed in the hindlimbs of both the dolphin and snake (Thewissen et al., 2006; Cohn and Tickle, 1999). Furthermore, the AER of the dolphin and of the lizard Anguis fragilis are transient structures, which indicates that the failure in maintenance of AER-derived signaling could account for the degeneration of the limb bud. In contrast, the python does not have a morphologically distinct AER nor expresses the genes normally associated with this signaling center (FGF, Dlx, Msx; Cohn and Tickle, 1999), but the limb bud does not degenerate. Shh is also absent from the forelimb buds of Calyptommatus, but further investigation is needed in order to determine the causative factors behind limb bud degeneration.

Conclusions

Considering the great diversity of limbs, it seems equally likely that there exists a corresponding diversity of developmental mechanisms. We are still distant from a comprehensive understanding of the anatomy and evolutionary relationships among species, and we are much farther from understanding the mechanisms by which different limbs are formed, especially considering the dynamic properties of developmental systems. Minor limb reductions can be the result of diverse changes in the developmental program, such as differential regulation of specific signals or of its receptors. However, in greatly reduced limbs, the major signaling pathways (FGF and Shh signaling) seem to be consistently affected, although these analyses do not go further into the dissection of the pathways and its regulations due to the experimental challenges of working with non-model organisms. This apparent convergence regarding the involvement of these major patterning signals as the potential drivers of limb loss can be nothing more than the result of the relaxation of the selective pressures over these key signaling pathways after long evolutionary times.

Acknowledgements

Our work was supported by Conselho Nacional de Pesquisa (CNPq) and Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP). JGR is truly grateful to Cheryll Tickle for helping in establishing the whole mount in situ. We are also greateful for the reviewer’s constructive comments.

Frequent abbreviations

d=

digit

mc=

metacarpal

Shh=

sonic hedgehog

AER=

apical ectodermal ridge

Fgf=

fibroblast growth factor

Bmp=

bone morphogenetic protein

Wnt=

wingless homologue gene