Evolution of vertebrate appendicular structures: Insight from genetic and palaeontological data


  • Amir Ali Abbasi

    Corresponding author
    1. National Center for Bioinformatics, Faculty of Biological Sciences, Quaid-i-Azam University, Islamabad, Pakistan
    • National Center for Bioinformatics, Faculty of Biological Sciences, Quaid-i-Azam University, Islamabad 45320, Pakistan
    Search for more papers by this author

  • The author declares that there is no conflict of interest.


The new body of evidence from fossils and comparative-developmental analysis of subset of appendicular patterning genes has revealed that limb elements seen in tetrapods are assembled in fish fin over evolutionary time. However, despite of deep homology in basic structure and underlying developmental system, there remains a large morphological gap between distal elements of tetrapod limb and distal fin skeleton of tetrapodomorph fish. Understanding the genetic basis of major transformations in distal-limb morphology is the next challenge for evolutionary developmental biologists. Here by integrating data from fossils, comparative-developmental and genetic studies, models are proposed describing the evolution of cis-regulatory elements as a basis for diversification of appendicular architecture. Instead of emphasizing the subset of developmental genes, for instance Hoxd genes, the focus here is on the significance of elucidating cis-regulatory elements for multiple other key molecular players of limb/fin development and genetic/molecular interactions among them, for a better understanding of the developmental and genetic basis of limb evolution. Developmental Dynamics 240:1005–1016, 2011. © 2011 Wiley-Liss, Inc.


Vertebrates limbs are serially iterated structures (fins in fish and limbs in tetrapods), and during evolution there has been a trend towards their morphological and functional diversification both within and between taxa. The vertebrate appendages have been the focus of intense genetic and molecular investigations, because they are not essential for embryonic survival and can be experimentally manipulated in model organisms to define the important cellular and molecular interactions that regulate patterning and skeletal development (Niswander, 2003; Tickle, 2006).

Here, the genetic mechanisms responsible for the origin of novel structures during animal evolution will be discussed. A brief overview of fin/limb history and diversity will be provided. From a developmental, genetic, and palaeontological perspective, the major transitions in appendicular morphology over the course of vertebrate evolution will be discussed. Finally, a hypothesis will be developed that the origin and elaboration of distal limb elements during sarcopterygian history involve the cis-acting element triggered redeployment of an ancient appendage-specific developmental genetic toolkit. Among anciently conserved appendicular specific genes, small differences in the spatiotemporal aspects of expression through subtle changes in existing regulatory regions might have contributed significantly in elaborating the limb architecture over evolutionary time.


The early events in vertebrate development generate modularity in the body plan by dissociating the embryo into developmentally autonomous compartments. This modularity, in turn, allowed the local, developmentally autonomous embryonic structures to accommodate genetic and molecular variations without altering the developmental events within adjoining compartments (Carroll, 2005; Wagner et al., 2007). This regionally restricted accumulation of variations without affecting the general body plan conferred on animals the ability to experience morphological innovations leading to great morphological diversity (Wray, 2007).

Genetically determined morphological diversity could be based on amplification and mutation of coding genetic material or the expansion and evolution of regulatory components that act in cis to control expression of developmental regulators. Within well-studied animal subgroups, such as vertebrates among the chordates, the variability in the coding contents of nuclear genetic material is minimal implicating a crucial role of regulatory elements in phenotypic evolution (Levine and Tjian, 2003; Abbasi, 2008). In fact, with an increasing body of empirical evidence emerging from the fields of evolutionary developmental biology and comparative genomics, it is now broadly accepted that increased morphological complexity and diversity in animals is associated with the evolution of cis-acting DNA elements that regulate the expression of developmental regulators (Carroll, 2008).

It has been proposed that morphological differences are easier to achieve through cis-regulatory mutations than through coding mutations because unlike the coding sequences where insertions, deletions, or substitutions within non-synonymous sites are often intolerable, the cis-acting regulatory modules exhibit plasticity (Carroll, 2005). They often harbor short degenerate binding sites for multiple trans-acting factors (Remenyi et al., 2004; Maston et al., 2006). Combinatorial, simple base pair changes within cis-acting DNA can potentially alter the binding affinities for existing factors, while insertions and deletions can alter the site spacing, and create new or delete existing binding sites (Wray et al., 2003). Thus, contrary to coding sequences, many sequence variations within cis-acting regulatory elements potentially impose tolerable affects on their activity resulting in incremental variations in timing, pattern, and level of the expression of the associated gene, and can work as a fuel for evolution in morphological diversity and complexity in several ways (Carroll, 2001).

Transcriptional evolution, impacting organismal phenotypes, not only involves substitutions, insertions, and deletions in cis-regulatory elements but also includes expansion in the number of cis-regulatory modules and transcriptional factors that bind to them (Carroll, 2001; Levine and Tjian, 2003). For instance, the unicellular yeast genome encodes a total of ∼300 transcription factors, while the genome sequences of Caenorhabditis elegans and Drosophila reveal at least 1,000 transcription factors each (Ruvkun and Hobert, 1998; Aoyagi and Wassarman, 2000). There may be ∼2,000 transcription factors in humans (Vaquerizas et al., 2009). Similarly, morphologically simple animals like Ciona intestinalis have been estimated to contain 10,000–20,000 tissue-specific enhancers (Harafuji et al., 2002). Drosophila contains several enhancers per gene scattered over an average distance of 10 kb. It is difficult to estimate the cis-regulatory contents of the human genome; however, only 5% is estimated to be evolving slower than the neutral rate (suggesting functional constraints) (Waterston et al., 2002) and less than 2% actually encode proteins. The remaining non-coding, slowly evolving sequences (∼90 Mb) of the human genome are predicted to regulate temporal, spatial, and quantitative aspects of gene expression among other roles.

The regulatory evolution in terms of increase in number of transcription factors and expansion in quantity and architectural complexity of cis-acting regulatory elements created new combinations of gene expression. Therefore, elucidating the regulatory underpinning of developmental traits is a prerequisite for understanding the genetic basis of organismal diversity and evolution of morphological characters like vertebrate limbs.


One of the remarkable innovations in vertebrate history was the origin of body appendages to facilitate the feeding and locomotion in Ordovician seas (Fig. 1). This innovation first took place in jawless vertebrates, with some having had ribbon-shaped fins extending laterally along the body wall, others having had paired pectoral fins only, protruding immediately posterior to the head region (White, 1946; Forey and Janvier, 1993; Morrissey et al., 2006) (Fig. 2).

Figure 1.

Evolutionary history of vertebrates in the context of approximate time points of molecular and morphological innovations leading to greater complexity in the limb architecture of modern vertebrates. Diagram illustrating the approximate history of major molecular and morphological innovations leading to diversification of limb elements during the evolutionary history of vertebrates. Genetic and expression analysis with model and non-traditional model vertebrates suggests the presence of GCR/ZRS in the ancestral lineage leading to chondrichthyans and osteichthyans, whereas Gli3 transcript was shown to be present in the fin field of osteichthyans ancestors (Dahn et al., 2007; Davis et al., 2007; Freitas et al., 2007). However, given the well-developed paired fins (at the pectoral level) of osteostracans (closest jawless relative of jawed vertebrates) and placoderms, containing appendicular endoskeleton (including metapterygium) and a/p polarity (Mabee, 2000; Coates, 2003), here it is hypothesized that Gli3, Shh, and phase II Hoxd expression was present in the fin field at least prior to the divergence of the osteostracans from the lineage leading to jawed vertebrates. The presence of two sets of fin pairs is characteristic of jawed vertebrates and distinct identities of pectoral and pelvic fins were proposed to be defined in the last common ancestor of chondrichthyans and osteichthyans by regulatory subfunctionalization of Tbx5 and Tbx4 paralogs (Ruvinsky and Gibson-Brown, 2000; Tanaka et al., 2002). The data from fossils revealed the presence of an elaborate distal endoskeleton (distal to wrist bones) in the pectoral appendage of tetrapodomorph fish (Shubin et al., 2006). Here it is hypothesized that transformation of these distal bones to the distal limb pattern of modern tetrapods must have entailed considerable morphological restructuring and developmental repatterning, such as imposing the pentadactyly constraints through refinement of Shh/Gli3 functions.

Figure 2.

Body appendages serving locomotion and feeding first appeared in jawless vertebrates. The early jawless craniates (Myllokunmingia) in Cambrian seas were using purely undulatory motions. Vertebrate appendages first appeared in the Ordovician/Silurian period either in the form of continuous lateral fin-folds (Jamoytius) or paired fins (osteostracans) extending from the rear of a bony head shield in pre-jawed fishes. The presence of two fin pairs (pectoral and pelvic) is unique to jawed vertebrates.

The two sets of paired appendages are the synapomorphy of gnathostomes (Shubin et al., 1997; Coates and Cohn, 1998) and first made their appearance in the Ordovician/Silurian boundary in the form of pectoral and pelvic fins. Since their origin, the pectoral and pelvic fins are believed to be evolved in parallel (serially homologous) in almost all major gnathostome clades (Shubin et al., 1997) (Figs. 1 and 2). In all jawed vertebrates, the pectoral fin skeletons share the following ancient structures; A series of radials (internal skeleton support) articulates with the girdle, the posterior-most radial further articulates with secondary radials along its anterior edge. This complex posterior radial is known as metapterygium (Fig. 3), which was lost in one major vertebrate group, the teleosts (for example, zebrafish and Fugu) (Mabee, 2000). Soft fin rays, constructed of collagen-like protein, constitute the periphery of both pectoral and pelvic fins. In bony fishes, these soft fin rays incorporate dermal bony rays (lepidotrichia) (Fig. 3). Pelvic fins in contrast to pectoral fins are small and anatomically simpler, and some investigators doubt the serial homology of pelvic and pectoral appendages (Coates and Cohn, 1998).

Figure 3.

Comparative anatomy of the pectoral fin of a chondrichthyan, basal actinopterygian, tetrapodomorph fish, and tetrapod forelimb. Fin-to-limb transition involved addition of some structures (e.g., digital bones) and loss of others (lepidotrichia and non-metapterygial bones). In tetrapodomorph fish, distal bones bearing resemblance to tetrapod autopodial bones (digits) were unknown until the discovery of a well-preserved pectoral appendage of Tiktaalik, the late Devonian sarcopterygian fish. However, there remains a large morphological gap between the distal bones of Tiktaalik/panderichthyes and autopodial skeleton of tetrapods, suggesting that considerable developmental repatterning occurred during the transformation of tetrapodomorph fish-fin to modern tetrapod-limb. According to the classical model, during the course of limb evolution, in lobe-finned fish there has been a turning of complex posterior radial metapterygium (fin/limb skeleton shown in black) toward the anterior of the proximodistal axis. This turning results in the origin of digital arch (Da) in modern tetrapods where anterior digits (e.g., d1) correspond to the most distal region of the limb. However, anterior bending of metapterygium and the origin of digit arch might not be directly correlated with the Phase-II pattern of Hoxd expression, as currently available empirical data suggest the presence of this pattern in the common ancestor of chondrichthyans and osteichthyans (when metapterygium axis was straight). Metapterygium and its associated radials are shown in black and red, respectively. mtr, metapterygial radials; dr, distal radials; mer, mesopterygial radials; lep, lepidotrichia; Da, digit arch; Ra, radius; d1–5, digit1 to digit5; Ul, ulna; Hu, humerus. Drawings adapted from Hinchliffe (2002), Shubin et al. (2006), Davis et al. (2007), Freitas et al. (2007), Boisvert et al. (2008).

For the fin-to-limb transition and the origin of digits in tetrapods, Shubin and Alberch proposed a model to define homologous regions of tetrapod limbs and paired fins of sarcopterygian fish (lobe-finned fishes) (Shubin and Alberch, 1986). According to this model, the proximal parts (humerus/femur, radius/tibia, ulna/fibula) are homologous in tetrapod limbs and sarcopterygian fish fins (Panderichthyes) (Fig. 3). In order to explain the evolutionary origin of distal parts in tetrapod limb, Shubin and Alberch (1986) proposed that during limb development, the distal row of carpal/tarsals arise from digit arch (Fig. 3). The evolution of the digital arch itself resulted from a bending of the metapterygial axis (main stem of branched pattern formed by fin metapterygial and associated radials) of the sarcopterygian fin. This axis is believed to have developed from proximal to distal. According to this model, what was originally distal in the sarcopterygian fin is now anterior in tetrapods. Thus, during fin-to-limb transition, there has been a turning of the proximodistal axis towards the anterior (Fig. 3). The anterior digits, therefore, correspond to the most distal region of the limb.

The concept of the metapterygial axis further suggests a posterior dominance during vertebrate limb development and thus the branching of the cartilaginous elements of the radius and ulna from the more proximal humerus is proposed to be asymmetrical (Coates and Cohn, 1998; Cohn et al., 2002). The anterior side (radial/tibial) does not branch in general, while the posterior (ulnar/fibular) may branch. The majority of the distal elements are hypothesized to arise from the posterior region of the limb. The digital arch model proposes that metacarpals/metatarsals arise by bifurcation from carpals/tarsals, whereas the phalanges arise by segmentation from more proximal elements.


Fin-to-limb transition was one of the most important macroevolutionary changes that took place in the evolutionary history of vertebrates. Molecular studies of the main fin axis during cartilaginous fish, teleost fin, and tetrapod limb development have uncovered striking conservation of the genetic control of pattern formation along the fin/limb axis (Sordino et al., 1995; Ruvinsky and Gibson-Brown, 2000; Tanaka et al., 2002) (Fig. 1).

The fins of teleost fish and mouse limbs have been under intense molecular investigation to draw a map of molecular events that ultimately led to fin-to-limb transition. Among the genes shown to be expressed in both teleost fins and tetrapod limbs are Shh, Ptc1, Bmp4, Fgfs, Pitx1, Dlx, Bmp, Hoxa, Hoxd, Hoxc6, Msx, En1, Sal1, Gli3, dHand, Tbx, Meis, and Wnt,. The similarities in molecular architecture are not only reflected in the simple presence and absence of gene expression, but their spatial distribution in the early fin bud bears a striking resemblance to those found in tetrapod limb development.

The molecular investigations with cartilaginous fish further suggest the acquisition of the expression of Hoxd, Shh, Tbx4, Tbx5, En1, Bmp4, Fgf8, and dHand as early limb patterning regulators deep in vertebrate history at least before the diversification of gnathostomes (Tanaka et al., 2002; Dahn et al., 2007; Freitas et al., 2007).

In the subsequent portions of this article, the genetic basis of the major morphological changes that took place during fin/limb history will be discussed with particular focus on the origin and elaboration of the autopodial skeleton.

Identity of Each Pair of Appendages

In vertebrates, the anterior and posterior appendicular structures appeared at specific levels of the lateral plate mesoderm (LPM), as two pairs of small bud-like structures containing undifferentiated mesenchymal cells covered by a layer of ectoderm (Capdevila and Izpisua Belmonte, 2001). Many genes that participate in initial limb/fin budding and their subsequent growth and patterning have identical expression patterns in forelimbs and hindlimbs and are involved in molecular cascades that are common to both anterior and posterior appendages. However, there are exceptions to this rule. For instance, in both fish and tetrapods the T-box family transcriptional factors Tbx5 and Tbx4, are expressed in the presumptive pectoral and pelvic buds, respectively. Both genes have fundamental roles in limb induction, patterning, and growth and contribute in the specification of limb identity during embryogenesis (Logan, 2003). Evolutionary history of vertebrate Tbx5 and Tbx4 and comparative functional/expression analysis of these genes in fish fin and tetrapod limb have advanced our understanding of how the difference between serially homologous paired appendages (forelimbs and hindlimbs) is established in vertebrates.

Phylogenetic analysis indicated that amphioxus (cephalochordate) has only one T-box gene, AmphiTbx4/5 (Ruvinsky et al. 2000), which by duplication (deep in the vertebrate linage) gave birth to vertebrate Tbx4 and Tbx5. The innovation of the genetic circuit for limb specification occurred by recruiting the expression of these duplicated copies of T-box genes to vertebrate paired appendages and this might have taken place prior to radiation of jawed vertebrates (Fig. 1). The evidence for this early acquisition of limb specification is provided by the fact that the expression domains of Tbx5 and Tbx4 genes in dogfish embryos (cartilaginous fish) are consistent with their expression patterns within the anterior and posterior appendages of higher vertebrates (Tanaka et al. 2002).

The expression analysis of Tbx4 and Tbx5 genes in cartilaginous fish fin led Tanaka et al. (2002) to propose that a single AmphiTbx4/5 like gene in vertebrate ancestory was expressing throughout the rostrocaudal axis of lateral fin folds. The origin of Tbx5 and Tbx4 genes before the chondricthyans-osteichthyans split followed the suppression of their expression in the medial region and establishment of complementary expression domains in the rostral and caudal ends of lateral fin fold. This regulatory subfunctionalization then led to the simultaneous origin of discrete paired fins in gnathostomes ancestory where two independent genes, Tbx5 and Tbx4, specified the anterior and posterior appendages, respectively (Fig. 1).

In contrast to the above scenario, based on the expression pattern of Tbx4/5 in amphioxus, it has been proposed that during the early history of vertebrates this gene was co-opted from its original function in heart development to evolve paired pectoral fins in pre-jawed fishes (e.g., Osteostracans) and then was recruited to a more caudal domain in lateral mesoderm, resulting in the origin of pelvic fins caudally (gnathostome ancestor) (Horton et al., 2008). This model advocates the independent origin of pectoral and pelvic appendages in vertebrate ancestory before the diversification of AmphiTbx4/5. Subsequently, after the divergence of cephalochordates from vertebrates but before the divergence of chondricthyans from osteichthyans, duplication of AmphiTbx4/5 occurred, giving birth to vertebrate Tbx4 and Tbx5 genes. The regulatory subfunctionalization of Tbx5 and Tbx4 paralogs then led to partitioning of their expression domains among pectoral and pelvic appendages, respectively.

Innovation of the Autopod With Digits in Tetrapods

In today's limbed tetrapods, during the early phase of limb development, the (at the onset of limb budding) Hoxd gene expression occurs in a time-dependent manner with anterior-most and middle gene transcripts (Hoxd1–9) deployed early and throughout the limb bud whereas posteriorly restricted expression of the centromeric gene sets (Hoxd10-13) occurs later in time (Nelson et al., 1996; Tarchini and Duboule, 2006) (Fig. 4). Temporal collinearity and progressive restriction of Hoxd gene transcripts towards the posterior of early limb bud is employed by two distinct poorly defined cis-regulatory modules positioned telomeric (Early limb control regulation, ELCR) and centromeric (Posterior restriction, POST) to the Hoxd cluster (Zakany et al., 2004; Tarchini and Duboule, 2006). ELCR induce transcription in a time-dependent manner, i.e., the closer the gene is located to ELCR, the earlier its expression is induced, whereas the POST impose spatial restriction on 5′Hoxd genes (d13–10) transcripts in the posterior region with increasing influence on genes positioned more centromeric of the cluster (Fig. 4). ELCR-POST-induced nested patterns of 5′Hoxd genes to the posterior of limb bud cells (E9–E10) is a prerequisite for the posterior localized expression of Shh transcript (Tarchini et al., 2006). The Shh expression subsequently triggers the second phase of 5′Hoxd genes regulation (E10.5–E12) distally in the presumptive digit-forming region. Inversion of early collinear expression pattern of five centromeric genes (d9–d13) is the hallmark of a second wave of Hoxd gene expression: Hoxd13 is expressed at the highest level and in a domain that extends most anteriorly (Kmita et al., 2002; Tarchini and Duboule, 2006) (Fig. 4). This Shh-mediated anterior-to-posterior asymmetry in the 5′Hoxd genes transcript distribution along the distal edge of the limb is translated into the anterio-posterior (AP) polarity of autopodial bones of tetrapods (Fig. 4). The second phase of regulation is independent of the first wave and is mediated by a distinct cis-acting regulatory module known as “Global Control Region (GCR)” positioned 250 Kb centromeric to the Hoxd genes cluster (Spitz et al., 2003) (Fig. 4). The early pattern of expression persists during later stages of limb bud development and progresses in parallel with a second phase along the proximal, presumptive distal stylopod and zeugopod domain, of the growing limb bud (Tarchini and Duboule, 2006). Thus, it appears that distinct cis-regulatory underpinnings (ELCR-POST and GCR) can trigger the expression of a similar set of developmental regulators (Hoxd genes) in distinct times (early and late) and distinct domains (proximal and distal) to impose differential identities (proximal and distal skeletal elements of mature limb) to different regions of an initially morphologically homogenous developmental entity (early limb bud) (Fig. 4).

Figure 4.

Changes in Hoxd gene expression during the development of tetrapod limb. Hoxd gene expression during early limb bud development is partitioned into two independent phases with different regulatory underpinnings. Left: First wave of Hoxd expression starts with the emergence of limb bud and is mediated by regulatory regions on 3′ side (ELCR) and 5′side (POST) of the cluster. The ELCR element acts (blue arrow) in a time-dependent manner and implements sequential deployment (t1–t5) of genes in the nascent limb bud, i.e., genes near to it activate early (d1–d4) whereas centromeric genes activate later. The POST acts (blue arrowhead) negatively and imposes anterior repression on 5′Hoxd genes (d10–d13). The strength of POST action is distance dependent, i.e., anterior repression is maximum for genes positioned near to this regulatory region. As a result, during the early phase, Hoxd13 expression is confined to a posterior patch of mesenchyme cells whereas Hoxd10 expression is broad but excluded from a small anterior patch of mesenchyme cells. Posteriorly localized 5′Hoxd gene products probably interact with ZRS (limb specific regulatory element of Shh) to trigger the localized expression of Shh along the posterior margin of limb (shown as a light green spot). The Shh subsequently (E10.5 onward) engage 5′Hoxd genes in the second phase of expression (red-shaded region in the limb bud) along the distal margin of the limb bud (presumptive digit-forming region). Bottom: This phase is under the influence of a distinct regulatory region known as GCR positioned 5′ side of the cluster. The GCR revert the 5′Hoxd genes expression with the Hoxd13 expression domain extended most anteriorly. The early phase expression progresses in parallel (E10.5–E12) with the late phase; however, the early phase generated spatial distribution of Hoxd gene transcript is non-overlapping with the late phase pattern and is confined to proximal and intermediate regions (distal stylopod and zeugopod) of the growing limb bud. Right: Distal stylopod and zeugopod elements of limb (blue) derived from early phase Hoxd gene expression whereas autopod skeleton (red) is derived from the second phase of expression. The regulatory underpinnings of early and late phase, and Hoxd expression pattern associated with these two distinct waves is depicted in blue and red, respectively. Dotted lines in the growing limb buds depict the expression boundaries of the respective Hoxd genes. A, anterior; P, posterior.

From the available comparative genetic studies, it appears that molecular patterning of proximal elements (stylopod and zeugopod) is achieved by the similar mechanism in tetrapod limb and teleost fin, but important differences in gene expression patterns occur later, during the development of distal elements. For example one major molecular difference in patterning of the distal region of tetrapod limb and teleost fin is the Shh expression in the posterior region of the limb mesenchyme, controlling AP patterning of the limb all the way to digit development, while in zebrafish Shh expression in the fin bud disappears prior to ray formation (Lopez-Martinez et al., 1995; Sordino et al., 1995). This early loss of Shh expression in the posterior of fin bud is associated with the lack of a second phase of 5′Hoxd genes expression in the fin bud. Therefore, it has been suggested that acquisition of a second phase of expression by the 5′ genes of the Hoxd cluster occurred at the origin of tetrapods and might have been crucial for the bending of the metapterygium axis towards the anterior of the proximodistal axis during limb evolution, which has led to the origin of the autopod where the anterior digits correspond to the most distal region of the limb (Sordino et al., 1995; Shubin et al., 1997). It is worth mentioning here that polarized expression of Shh in vertebrate appendicular structures is governed by a cis-acting regulatory sequence, famously known as ZRS (zone of polarizing activity regulatory sequence) (Lettice et al., 2003; Maas and Fallon, 2005). Intriguingly, the sequence of ZRS and even its relative position with respect to the Shh transcription initiation site is highly conserved in tetrapods, teleosts, and chondrichthyans, whereas spatiotemporal aspects of Shh expression differ considerably among the fish fin (cartilaginous fish/telesots) and tetrapod limbs (Dahn et al., 2007). This might suggest that subtle changes in the anatomy of ZRS (transcriptional factor binding site differences) have diversified specific aspects of its action over the course of vertebrate evolution without aborting its appendicular specificity, which was defined prior to diversification of gnathostomes and has remained under tight constraints afterwards (Fig. 1).

In contrast to comparative molecular and genetic data from representative members of teleost (zebrafish) and terrestrial vertebrate lineages (mouse), the data from fossils have shown that the autopod is not a novelty of tetrapods (Shubin et al., 2006; Boisvert et al., 2008). If this was the case, then the second phase of Hox gene expression might have been acquired by the bony fishes before terrestrial invasion. In agreement with the fossil data, recent analysis has revealed a late phase, inverted collinear expression of 5′Hoxd genes in fish fin including sarcopterygian, basal actinopterygian, and chondrichthyan. However, spatiotemporal and quantitative aspects of Phase II expression may vary slightly among taxa (Davis et al., 2007; Freitas et al., 2007; Johanson et al., 2007) (Fig. 3). The discovery of late phase Hoxd gene expression in non-model finned vertebrates not only illuminates the fact that that the pattern is not a tetrapod novelty but also provides genetic support to the palaeontological work, which suggests that distal apparently modern elements of tetrapod limb were assembled in fishes over evolutionary time (Shubin et al., 2009) (Fig. 1).

Although the data from fossils revealed the presence of an elaborate distal endoskeleton (distal to wrist bones) in the pectoral appendage of sarcopterygian fish (Tiktaalik & Panderichthys), there is no direct morphological relationship among these distal bones and digital elements of primitive tetrapod Acanthostega (Coates, 1996). If the digital elements of modern tetrapods are evolved from distal radial bones of tetrapodomorph fish, then this evolutionary trajectory must have entailed considerable morphological restructuring and developmental repatterning (Figs. 1 and 3). For instance, the Devonian tetrapods were polydactylous (6–8 digits), the elements of carpus/tarsus are patterned oddly, whereas the metacarpals/tarsals and phalanges are morphologically weakly differentiated (Coates and Clack, 1990; Coates, 1996). The transformation from five distal radials of Tiktaalik pectoral fin to modern autopodial elements of tetrapods in late and post-Devonian times, might have involved the origin of additional endoskeleton elements and joints distal to the intermedium and ulnare and proximal to five radials of Tiktaalik, as well as extensive multiplication of radial or digital elements and joints, which should have occurred to evolve an intermediate polydactylous limb condition during the late-Devonian (exemplified by Acanthostega). Subsequently, in post-Devonian times, extensive elaboration of distal limb elements must have occurred, to refine carpus/tarsus and metcarpals/tarsals and digital phalange morphology, to reduce the digit number to five or fewer (pentadactyly constraints), and to give proper identity to digital bones.

The ancient origin of Phase II 5′Hoxd gene expression (predating the origin of autopod) and the existence of a large morphological gap among the distal elements of the Tiktaalik/Panderichthys fin and the autopod of primitive tetrapods suggest that there in no clear-cut relationship between the acquisition of late-phase of Hoxd expression and assembly or evolution of distal elements of modern limbs (Fig. 3). Nevertheless, digital evolution during fin-to-limb transformation might have entailed, firstly, introducing subtle changes in spatiotemporal and quantitative aspects of late phase 5′Hoxd gene expression, over the course of evolution; secondly, the evolution of a new regulatory connection between Hoxd genes (major limb patterning regulators) and the other appendage-specific primitive genes (Table 1 and Fig. 5). The last scenario would suggest that, for better understanding of genetic mechanisms underlying autopod evolution, regulatory changes experienced by other genes acting both upstream and downstream of 5′Hoxd during distal limb patterning should also be taken into account (for instance, see Table 1).

Table 1. Subset of Candidate Genes That Might Have Played a Vital Role in Evolving and Elaborating the Autopodial Elements During Sarcopterygian Historya
Distal limb regulatorsRole in autopod patterningMutant phenotype (limb-autopod)Temporal expression with respect to 5′Hoxd genesLimb specific cis-regulatory elements
  • a

    A critical role of these genes in autopod evolution was deduced through their indispensable functions in distal limb patterning and mutant phenotypes. In the nascent limb bud, these genes act in collaboration (both upstream and downstream genetic interactions) with 5′Hoxd to participle in the precise patterning of autopodial skeleton along the three axes. In-depth understanding of cross-regulatory interactions in limb/fin bud is a prerequisite for understanding the genetic basis of autopod innovation and this demands the genetic and functional exploration of cis-regulatory sites for these and other key players (not listed here) that participate in handplate development/growth. A-P, anterior-posterior; D-V, dorsal-ventral; Pr-D, proximal-distal; AER, apical ectodermal ridge.

ShhA-P patterningLimbs have only single digit (Chiang et al., 2001)UpstreamKnown (Lettice et al., 2003)
Gli3A-P patterningUnpatterned multiple extra digits (te Welscher et al., 2002)UpstreamKnown (Abbasi et al., 2010)
dHandA-P patterningLoss of the autopod (Galli et al., 2010)UpstreamCurrently unknown
Bmp2/Bmp4AER induction/Interdigital apoptosis/D-V patterningDefective digit patterning (Maatouk et al., 2009)DownstreamKnown for Bmp2 (Dathe et al., 2009)
Fgf4/Fgf8Pr-D growthWrist and digit elements greatly reduced (Sun et al., 2002)UpstreamKnown for Fgf4 (Fraidenraich et al., 1998)
Grem1Pr-D growth/AER maintenanceReduced digit number (Khokha et al., 2003)DownstreamKnown (Zuniga et al., 2004)
Wnt7aDorsal and posterior limb developmentLoss of dorsal phenotype and posterior most digits (Adamska et al., 2005)Upstream or parallelCurrently unknown
Alx4A-P patterningAnterior digit deformity (Qu et al., 1997)Downstream or parallelKnown (Kuijper et al., 2005)
Tbx2/Tbx3A-P patterningPosterior digit defects (Suzuki et al., 2004)UpstreamCurrently unknown
En1D-V patterningD-V patterning defects/digit deformity (Adamska et al., 2004)DownstreamCurrently unknown
Figure 5.

Model depicting the evolution of cis-regulatory circuitry as a basis for co-opting the ancient limb/fin-specific genetic toolkit for the origin of autopodial elements in tetrapods. Modern structures in tetrapod limbs such as digits are patterned by gene-sets first deployed in primitive vertebrate appendages (lacking structures comparable to zeugopod/autopod). Schematics showing genes acting both upstream and downstream (see Table 1) of 5′Hoxd during autopod patterning in tetrapods and probably present in the appendage of the common ancestor of Osteichthyes and Chondrichthyes. Black-labeled genes are those whose expression is well established in each of the respective appendages depicted here, whereas red-labeled genes are those that are most likely to be expressed in fish fin, but have yet to be shown. Regulatory cross-talk among these genes is largely unknown for fish fin (left and middle panel), whereas genetic studies in mice have elucidated some aspects of cross-regulatory interactions (right panel). For tetrapod limb bud (right panel), regulatory links depicted in red and green are the ones confirmed functionally (Capellini et al., 2006; Vokes et al., 2008; Galli et al., 2010) whereas black links indicate putative cross-talk through cis-regulatory sequences, speculated here. Solid lines (right panel) indicate direct regulatory interactions where a given gene product (transcriptional factor) mediates the expression by directly binding to the concerned cis-acting site, whereas dashed lines indicate indirect interactions where a given gene product is not the transcriptional factor (Shh, Gre, Bmp, Fgf, Wnt7a), but interacts with cis-acting sites of concerned genes through their downstream effectors. Given the evolutionarily deep conservation of known autopod-patterning genes, it is speculated here that autopod invention in sarcopterygian might entail major shifts in cross-regulatory interactions through changes in their cis-acting sites. These cis-acting element triggered changes in cooperative interactions within the appendage-specific primitive genetic toolkit might then act as fuel for the origin and elaboration of distal limb elements at the root of tetrapod history. Therefore, it is argued here that the elucidation of precise differences in collaborative interactions among autopod regulators in fish-fin and tetrapod-limb bud and correlating these differences with differences in developmental patterning would shed important insight into the genetic underpinning of autopod invention. For fish, the depictions (left and middle panel) do not reflect the spatial or temporal aspects of gene expression, but simply highlight the confirmed or putative presence of given gene products in fins. Asterisk symbol (*) depicts genes with known limb-specific cis-regulatory elements. AER, apical ectodermal ridge; zpa, zone of polarizing activity; pr, proximal; dist, distal; ant, anterior; post, posterior.

Taken together, the origin and elaboration of distal limb elements during sarcopterygian history might have entailed major shifts in cross-regulatory interactions through changes in cis-acting sites of an appendage-specific primitive genetic toolkit (model highlighted in Fig. 5). Therefore, it is argued here that the elucidation of precise differences in collaborative interactions among autopod regulators in fish-fin and tetrapod-limb buds and correlating these differences with differences in developmental patterning would shed important insight into the genetic basis of autopod invention in middle-Devonian times and its subsequent elaboration in late-Devonian and post-Devonian times.

Constraints on Digit Number and Identity

Gli3 takes part in governing the development of all limb elements. It has been implicated in positioning the limb along the main body axis in combination with dHand and Tbx3 (Rallis et al., 2005). Interaction of Gli3 and Plzf is essential for precise patterning of limb skeletal elements (proximal and intermediate) along the proximo-distal (P-D) axis up to the distal margin of the zeugopod (Barna et al., 2005). Additionally, genetic antagonism between Gli3 and Shh within the nascent limb bud regulates the anterio-posterior (A-P) patterning of autopod morphology by constraining the digit number and identity (Litingtung et al., 2002; Towers et al., 2008). Intriguingly, Gli3 functions for the patterning of evolutionary ancient limb elements (stylopod and zeugopod) are independent of its roles in the normal patterning of evolutionarily younger limb skeletal elements (autopod) (Barna et al., 2005). In agreement with these data, Gli3 has distinct cis-regulatory controls for its expression in presumptive stylopod/zeugopod and autopod domains of nascent limb bud (Abbasi et al., 2007, 2010; Paparidis et al., 2007).

Shh signaling acts in two ways to restrict Gli3R to the anterior domain of the developing limb bud, (1) by preventing the formation of the truncated form of Gli3, and (2) by repressing its expression posteriorly. Thus, sonic hedgehog creates a Gli3 gradient, such that the truncated form is seven times more abundant in the anterior third than in the posterior third of the emerging limb bud (Wang et al., 2000). In autopod patterning, this Shh-mediated anterior restriction of Gli3R is crucial for correct digit identity and number as evident from skeletal morphology of Shh−/−, Gli3−/−, and Shh−/−: Gli3−/− null mutants (Fig. 6) (Chiang et al., 2001; Litingtung et al., 2002; te Welscher et al., 2002).

Figure 6.

The default state of the limb is to form many unpatterned digits. Schematic representation of limb elements of wild type and genetically manipulated mice. The knockout data suggest that genetic interactions between Shh and Gli3 define digit number and identity (Chiang et al., 2001; Litingtung et al., 2002; te Welscher et al., 2002). Top: The developing limb bud is shown with dotted ovals representing the presumptive digit-forming regions (5-4-3-2 from posterior to anterior direction) under the influence of Shh, whereas the patterning of digit-1 is under the combined regulation of Gli3 and Alx4 transcriptional factors (Harfe et al., 2004; Panman et al., 2005).

In the Shh−/− mutants, there is a dramatic loss of intermediate and distal limb elements, with the zeugopod containing a single bone (radius/tibia) and a reduced digit 1 (Chiang et al., 2001) (Fig. 6). In the Shh−/− limb bud, Gli3R expression expands posteriorly, resulting in a down-regulation of posterior genes (Gremlin, Fgf4, 5′Hoxd) and initiating apoptosis, which leads to blockage of distal limb development and A-P patterning (Litingtung et al., 2002; te Welscher et al., 2002).

The limbs of Gli3−/− mutant mouse embryos show severe polydactyly with many unpatterned digits (Fig. 6), A-P polarity is lost and no apoptosis occurs in inter-digital regions (te Welscher et al., 2002). The polydactyly of Gli3-deficient limbs is Shh independent, as limbs lacking both Gli3 and Shh display morphology identical to Gli3−/− mutants (Fig. 6). Polydactyly in Gli3−/− and Shh−/−:Gli3−/− double null mutants has been attributed to the anterior extension of Gremlin expression and general inhibition of BMP signaling. This contributes to the expansion of the mesenchyme and formation of additional digits (Aoto et al., 2002).

The abnormalities in limb patterning of the above-mentioned null mutants suggest that the default state of the limb is to form many digits. Gli3 and Shh impose the pentadactyly constraints so that five digits are formed in the mouse limb. Furthermore, Gli3 and Shh also specify the correct digit identity (Litingtung et al., 2002).

A study elucidating the cholesterol modification of Shh concluded that the formation of posterior digits 4 and 5 is governed by autocrine Shh signaling, while half of the digit 3 and the complete digit 2 is under the influence of paracrine Shh signaling (Harfe et al., 2004) (Fig. 6). In the case of autocrine signaling, the differential digit identities depend on the length of time the cells are exposed to a high level of Shh signal. The identities of digits under paracrine signaling are concentration dependent. Digit 1 is not reliant on Shh, rather, a recent investigation suggests that digit 1 is specified by the interaction of the Alx4 transcription factor with Gli3 (Panman et al., 2005) (Fig. 6).

All modern tetrapods have limbs characterized by five or fewer digits (Tabin, 1992). In contrast, recent fossil evidence indicates that Devonian tetrapods had a greater number of digits. For example the Acanthostega forelimb had eight digits (Coates and Clack, 1990). Thus, it appears that early tetrapods were polydactylous and the digit number had been reduced and stabilized at a maximum of five in the subsequent evolution of limb (Fig. 1) perhaps through the recruitment of Gli3 and Shh functions to impose constraints on autopod for digit numbers 5 or fewer (Towers et al., 2008).

The digit identity constraint is that a maximum of five different types of digits might have been imposed earlier than the constraints on digit number itself, as Acanthostega possessed five morphological types of digits even though it had a total number of eight digits (Tabin, 1992).


The elucidation of many crucial genes that regulate growth and patterning of limbs has now made it evident that despite their extreme morphological and functional diversification from fish fin to human limb, the vertebrate appendicular architecture is build upon a fairly similar repertoire of regulatory genes. A growing body of evidence from detailed studies on a subset of genes like Hoxd cluster and Shh suggests that evolution of the cis-regulatory circuitry played a key role in morphological diversification of the vertebrate appendicular skeleton. However, the picture is still incomplete as we know little about the cis-acting regulatory contents of other crucial genes involved in limb patterning and development. The detection and comparative functional analysis of cis-acting regulatory sites of key developmental regulators would thus be an essential contribution towards a composite map of evolutionary events involved in the amazing architectural and functional diversification of limbs between tetrapod and fish lineages and within tetrapods.


I am thankful to Prof. Dr. John F. Fallon and two anonymous reviewers for providing helpful comments and suggestions on this manuscript.