Regulation of limb bud initiation and limb-type morphology

Authors

  • Veronique Duboc,

    1. Division of Developmental Biology, MRC-National Institute for Medical Research, London, United Kingdom
    Search for more papers by this author
  • Malcolm P. O. Logan

    Corresponding author
    1. Division of Developmental Biology, MRC-National Institute for Medical Research, London, United Kingdom
    • Division of Developmental Biology, MRC-National Institute for Medical Research, The Ridgeway, London, NW7 1AA, UK
    Search for more papers by this author

Abstract

While the paired forelimb and hindlimb buds of vertebrates are initially morphologically homogeneous, as the limb progenitors differentiate, each individual tissue element attains a characteristic limb-type morphology that ultimately defines the constitution of the forelimb or hindlimb. This review focuses on contemporary understanding of the regulation of limb bud initiation and formation of limb-type specific morphologies and how these regulatory mechanisms evolved in vertebrates. We also attempt to clarify the definition of the terms limb-type identity and limb-type morphology that have frequently been used interchangeably. Over the last decade, three genes, Tbx4, Tbx5, and Pitx1, have been extensively studied for their roles in limb initiation and determining limb-type morphologies. The role of Tbx4 and Tbx5 in limb initiation is clearly established. However, their putative role in the generation of limb-type morphologies remains controversial. In contrast, all evidence supports a function for Pitx1 in determination of hindlimb morphologies. Developmental Dynamics 240:1017–1027, 2011. © 2011 Wiley-Liss, Inc.

INTRODUCTION

Tetrapods have two sets of paired appendages that emerge as budding outgrowths from the lateral plate mesoderm (LPM) at fixed positions along the rostro-caudal axis of the body. Initially, the forelimb and hindlimb buds are morphologically homogeneous bulges of cells. As outgrowth of the nascent limb continues, the expanded pools of progenitors differentiate into the inter-connected array of bones, tendons, muscles, vasculature, and skin. Each individual tissue element attains a characteristic morphology that ultimately defines the constitution of a forelimb or hindlimb. While the bones, tendons, and some of the limb vasculature are derived from the LPM, all of the limb muscles are derived from precursors that migrate from the epaxial dermomyotomal compartment of the somites medial to the limb buds (Pearse et al., 2007). The emergence of the array of limb skeletal elements is regulated by signalling molecules expressed in key signalling centres and their targets (Butterfield et al., 2010; Duboc and Logan, 2009; Zeller et al., 2009), which are thought to be acting equivalently in both forelimb and hindlimb buds. Therefore, the development of forelimb and hindlimb-type morphologies is a useful paradigm to study how diversity of form can be generated through modulation of conserved gene regulatory networks.

Several classical experiments performed in the chick demonstrate that the forelimb- or hindlimb-forming potential is specified prior to limb bud formation and that the cells of the limb bud are autonomously determined. Cells transplanted from a wing-forming region give rise to a wing when grafted to an ectopic location and, conversely, comparable leg grafts develop into a leg (Hamburger; 1938; Stephens et al., 1989; Saito et al., 2002, 2006). In addition, recombination experiments in the chick have shown that limb-type morphology is determined by the mesenchyme and is independent of the overlying ectoderm. Transplants of wing bud mesenchyme combined with leg bud ectoderm transplanted to the flank will develop into wing structures, while reciprocal grafts of leg mesenchyme covered with wing ectoderm produce leg elements (Zwilling, 1959; Logan, 2003; Saunders and Gasseling, 1968). Both classical embryology and more recent molecular data indicate that limb muscles are patterned by extrinsic cues (Hasson et al., 2010; Christ et al., 1977; Kardon et al., 2003). Muscle progenitors that migrate into a forelimb bud develop into forelimb muscles in response to environmental signals present in the forelimb and hindlimb muscle progenitors respond to equivalent cues in the hindlimb environment.

Two T-box transcription factors, Tbx5 and Tbx4, and a paired-type homeodomain transcription factor, Pitx1, have expression profiles that are consistent with these genes playing a role in determining formation of forelimb or hindlimb structures. Tbx5 is expressed in the forelimb but not the hindlimb, whereas Pitx1 and Tbx4 are expressed in the hindlimb buds but not the forelimb. In addition, each gene is initially expressed throughout the limb mesenchyme, but not in the limb ectoderm, and these limb-type restricted expression patterns are conserved across vertebrate species (Lamonerie et al., 1996; Szeto et al., 1996; Gibson-Brown et al., 1998; Isaac et al., 1998; Logan et al., 1998; Ohuchi et al., 1998). A more historical review of the factors regulating limb initiation and limb-type morphology has been published previously (Logan, 2003).

FGFS: A COMMON NODE IN FORELIMB AND HINDLIMB FORMATION

One of the most dramatic demonstrations of which factors regulate limb bud initiation came from experiments carried out in the chick that showed that a source of FGF applied to the interlimb LPM (spanning somites 20 to 26, between stage 13 to 17) of the embryo is sufficient to induce formation of a complete ectopic limb (Cohn et al., 1995, 1997). This and subsequent experiments (Ohuchi et al., 1997) not only implicate FGF signalling as critical for limb initiation but also demonstrate that FGF ligands, acting during a relatively brief period, are sufficient to establish the series of events that lead to formation of a mature limb.

The identity of the FGF ligand critical for limb initiation was revealed by analyses of Fgf10 knockout mice (Min et al., 1998; Xu et al., 1998; Sekine et al., 1999). Forelimb and hindlimb buds fail to form in Fgf10 mutants and all limb skeletal elements are absent apart from a rudimentary scapula and pelvis (see Fig. 2Ab–d). Fgf10 expressed by cells of the limb mesenchyme signals via receptors in the apical ectodermal ridge (AER), a specialised ridge of cells in the overlying ectoderm, to induce and maintain Fgf8 expression. In turn, the Fgf8 ligand expressed by cells of the AER signals via receptors in the limb mesenchyme to positively regulate Fgf10 expression in the limb mesenchyme (Fig. 1C). Thus, a positive feedback-loop of FGF signalling common to both forelimb and hindlimb is established that drives limb bud outgrowth (Xu et al., 1998).

Figure 1.

Schematic illustration of the limb initiation process in the mouse. A: From left to right, dorsal view of the right half of a mouse embryo prior to limb initiation (around [E] 8.75). Axial cues confer limb-forming potential to a region of the LPM. A combinatorial Hox code determines the levels at which the limbs will form. B: Mouse embryos at the onset of forelimb budding (15–20 somites) (top) and hindlimb budding (30–35 somites) (bottom). Schematic of the presumptive limb-forming areas of the forelimb (top) and hindlimb (bottom) showing the expression domains of Tbx5 and Pitx1/Tbx4, respectively. C: Initiation of Fgf10 expression in the limb mesenchyme by Tbx factors leads to establishment of the positive feedback-loop between mesenchymal Fgf10 and Fgf8 in the ectoderm/AER. A stable FGF signalling positive feedback-loop marks the end of the initiation phase and the onset of the outgrowth process. D: Chronologies of the major events starting from a uniform flank to the emergence of the limb buds.

A BROAD REGION OF THE LPM HAS LIMB-FORMING POTENTIAL

FGF-soaked bead implantation experiments in the chick also demonstrated that a strip of the LPM, encompassing the forelimb- and hindlimb-forming regions as well as the intervening interlimb LPM, has limb-forming capacity (Cohn et al., 1997) (Fig. 1A). This territory is further subdivided into a rostral, forelimb-forming region and a caudal, hindlimb-forming region. A source of FGF placed in the rostral region will produce an ectopic wing while a bead placed in the caudal region produces an ectopic leg. Strikingly, an ectopic limb that forms at a level that straddles the junction of these two territories will be comprised of wing elements in the anterior and hindlimb elements in the posterior reflecting the origin of the progenitors. Consistent with these observations, the pattern of Tbx4/Pitx1 and Tbx5 expression in the ectopic buds correlates with the morphology of the structures that subsequently develop. Ectopic wings induced close to the wing express Tbx5. Hindlimbs induced near the level of the hindlimb express Tbx4/Pitx1. An ectopic hybrid limb that forms at a level that straddles the junction of these two territories expresses Tbx5 in the anterior and Tbx4/Pitx1 in the posterior (Gibson-Brown et al., 1998; Isaac et al., 1998; Logan et al., 1998; Ohuchi et al., 1998). Therefore, the expression of Tbx5 and Tbx4/Pitx1 is correlative to the ultimate limb-type morphology and, thus, these genes serve as clear markers of forelimb and hindlimb progenitors.

RETINOIC ACID IS REQUIRED FOR LIMB BUD INITIATION

Retinoic acid (RA) appears to have multiple, temporally distinct roles during limb development. It is first required for limb bud initiation and, subsequently, influences anterior-posterior and proximal-distal patterning, although these later roles are subject to controversy (Lewandoski and Mackem, 2009). Here, we focus solely on the role of RA during limb initiation.

The requirement for RA signalling during forelimb initiation is conserved in vertebrates. In chick as well as in zebrafish, application of chemical inhibitors of retinaladehyde dehydrogenases, enzymes essential for RA synthesis, prevents limb bud formation (Stratford et al., 1996; Grandel et al., 2002; Gibert et al., 2006). Mice lacking the retinaladehyde dehydrogenase, Raldh2, do not form forelimb buds and die by embryonic day (E) 10 (Niederreither et al., 1999). Similarly, the zebrafish raldh2 mutants, no fin and necklace, fail to form pectoral fins (the equivalent of the forelimb in fish) (Grandel et al., 2002; Gibert et al., 2006). Zebrafish raldh2 mutants die during early larval stages (Grandel et al., 2002) prior to the stage of pelvic fin budding that occurs after three weeks of development (Grandel and Schulte-Merker, 1998) and, consequently, the role of RA during pelvic fin (the equivalent of the hindlimb in fish) development has not been addressed. Significantly, transplantation of wild-type cells that contribute to the somites in raldh2 mutant hosts is sufficient to rescue pectoral fin formation, suggesting that this axial tissue is an important source of RA signal during normal pectoral fin induction (Gibert et al., 2006). More recent work using the zebrafish raldh2 mutant has proposed that RA is required for fin bud formation at two time-points. The first phase occurs much earlier than originally predicted, during gastrulation, and specifies a population of precursors to a fin-forming fate. Subsequently, RA is required to regulate expansion of these progenitors during initiation of the fin bud (Grandel and Brand, 2010).

Classical experiments performed in chick demonstrate that initiation of wing formation is dependent upon signals from adjacent axial structures (Kieny, 1970; Kieny et al., 1972). Placement of a foil barrier between the somites and LPM at forelimb levels blocks wing bud formation (Murillo-Ferrol, 1965; Sweeney and Watterson, 1969; Stephens and McNulty, 1981). Equivalent experiments placing a foil barrier between LPM and somites at hindlimb levels has not been reported and, therefore, whether axial tissues are the source of an inductive signal required for hindlimb initiation remains an open question that should be addressed in the future.

Together these observations identify RA as a candidate for the axial influence on limb bud initiation. RA may act as a competence factor that confers limb-forming capacity to a broad region of the LPM encompassing both forelimb- and hindlimb-forming regions as well as the inter-limb LPM (Fig. 1A).

ROSTRO-CAUDAL HOX CODE DEFINES THE LEVELS OF FORMATION OF APPENDAGES

Genes of the HoxC cluster exhibit restricted expression patterns along the LPM of the flank and are thought to specify the levels at which the limb bud will appear and, thus, their identity (Cohn et al., 1997; Cohn and Tickle, 1999) (Fig. 1A). For example, Hoxc4-5 is expressed mostly in the mesenchyme of the presumptive forelimb area (Savard et al., 1988; Tabin, 1989; Molven et al., 1990; Burke et al., 1995), Hoxc9, Hoxc10, Hoxc11 are, predominantly, restricted to the mesenchyme of the hindlimb domain (Simon and Tabin, 1993; Peterson et al., 1994), whereas Hoxc6 and Hoxc8 expression correlates with the interlimb flank. Hox9 paralogues are also believed to contribute to the combinatorial code that specifies forelimb and hindlimb territories (Rancourt et al., 1995; Cohn et al., 1997). Hox transcription factors have been suggested to be responsible for the restricted expression patterns of Tbx4 and Tbx5 (Minguillon et al., 2005) (Fig. 1B). Currently, no functional evidence has been presented demonstrating a direct role of Hox genes in the regulation of limb initiation. Mice that lack the entire HoxC cluster do not exhibit obvious defects in limb position (Suemori and Noguchi, 2000). This observation does not, however, rule out the possibility that they participate in this process in collaboration with other functionally redundant Hox genes. Other indirect observations are consistent with such a model. Alteration of Hox gene expression patterns correlates with, and has been implicated in, the evolution of limbless in snakes (Cohn and Tickle, 1999) while a more recent study suggests that different downstream interpretation of the Hox code is important in the loss of paired appendages in this lineage (Woltering et al., 2009).

A ROLE FOR INTEGRIN SIGNALLING IN LIMB INITIATION

The frog Xenopus differs from many other vertebrates in that its limbs develop at the free-swimming tadpole stage after several weeks of development and that its hindlimbs develop prior to the forelimbs. An insertional mutation disrupting nephronectin (npnt) in Xenopus tropicalis results in a frog lacking all forelimb elements while the hindlimb develops normally (Abu-Daya et al., 2010). This phenotype can be reproduced by knock-down of npnt using morpholino-oligo nucleotides. Npnt is a small, secreted integrin ligand and is necessary for metanephros development in the mouse (Muller et al., 1997; Linton et al., 2007). The amphibian kidney (pronephros) is located adjacent to the presumptive forelimb-forming region and although npnt is ubiquitously expressed during frog development at the time forelimb outgrowth is initiated, transcripts are enriched in pronephros. The correlation with pronephros is potentially significant since extirpation experiments in the chick have implicated nephrogenic mesoderm in forelimb initiation (Geduspan and Solursh, 1992), although this model has been recently challenged by studies in chick (Fernandez-Teran et al., 1997) and mouse (Perantoni et al., 2005). In the npnt mutant, Tbx5 expression in the presumptive forelimb area is not initiated, suggesting that nephronectin-integrin signalling acts upstream of the molecular events leading to forelimb formation. This study demonstrates a novel mechanism required for forelimb initiation in X. tropicalis and reveals a difference in the regulatory pathways controlling forelimb and hindlimb initiation in the frog. It remains to be determined if integrin signalling is a conserved mechanism regulating vertebrate limb initiation or if it has been independently co-opted in the amphibian lineage.

TBX GENES ESTABLISH THE FGF SIGNALLING LOOP IN THE NASCENT LIMB BUD

Loss-of-function experiments in mouse and zebrafish have demonstrated that Tbx5 and Tbx4 play critical, conserved roles in forelimb and hindlimb bud initiation, respectively. In mouse embryos mutant or conditionally deleted for Tbx5, the forelimb buds fail to form (Agarwal et al., 2003; Rallis et al., 2003). In the absence of Tbx5, Fgf10 expression is not initiated and as a result the FGF signalling loop between limb mesenchyme and ectoderm is never established. Fgf10 appears to be a direct target of Tbx5 (Ng et al., 2002; Agarwal et al., 2003). In contrast to the Tbx5 conditional mutant, which lack all forelimb skeletal elements, a rudimentary scapula forms in the Fgf10 mutant, suggesting that Tbx5 makes an additional contribution to forelimb development beyond the positive regulation of Fgf10 expression (Rallis et al., 2003). In both the zebrafish tbx5 mutant (heartstrings) and following targeted knockdown using morpholino-oligonucleotides, pectoral fin formation is blocked (Ahn et al., 2002; Garrity et al., 2002; Ng et al., 2002). The absence of the pectoral fin is caused, at least in part, by the failure of migration of tbx5 -positive limb precursors.

The genetic hierarchy leading to bud induction in zebrafish differs slightly from other vertebrates. tbx5 regulates the expression of fgf10 indirectly through the expression of the teleost fish-specific gene, fgf24 (Fischer et al., 2003) and does so via a feed-forward model of transcriptional regulation (Harvey and Logan, 2006). tbx5 is required for the expression of the FGF ligand fgf24. It also regulates expression of zinc finger transcription factors, sall1a and sall4, that are required for expression of the FGF receptor, fgfr2, through which fgf24 signals. tbx5, therefore, has a positive input into both expression of FGF ligand and its receptor to establish the FGF signalling loop that is required for fin outgrowth. A similar feed-forward regulatory relationship between Tbx5, Sall genes, FGF ligands and receptors is thought to occur in amniotes (Harvey and Logan, 2006; Koshiba-Takeuchi et al., 2006).

In gene deletion-gene replacement experiments, Tbx4 can replace the function of Tbx5 in forelimb initiation suggesting that Tbx4 has an equivalent role in hindlimb initiation (Minguillon et al., 2005). In contrast to the absence of forelimb observed in the Tbx5 knock-out, in Tbx4 null mice a hindlimb bud does form although it is drastically reduced in size. The resulting hindlimbs fail to grow due to improper initiation and maintenance of Fgf10 expression (Naiche and Papaioannou, 2003). Therefore, although Tbx4 is necessary for normal Fgf10 expression, it is not exclusively required and other factor(s) have an input into the establishment of FGF signalling in the emerging hindlimb bud (Fig. 1C).

A role for Tbx5 and Tbx4 in the initiation of limb outgrowth is corroborated by the limb defects found in human syndromes associated with mutations in these genes. Mutations in TBX4 and TBX5 are associated with Small Patella syndrome (SPS; OMIM no. 147891) and Holt-Oram syndrome (HOS; OMIM no. 142900), respectively (Basson et al., 1997; Li et al., 1997; Bongers et al., 2004). HOS is a dominant disorder affecting the heart and upper limbs. Haplo-insufficiency of TBX5 causes a range of limb abnormalities, the mildest phenotype being triphalangeal (three-jointed) thumb but that most commonly results in the failure of limb elements to form. SPS, similarly, is a dominant disorder that is characterised by osteodysplasia of the knee, pelvis, and foot.

TBX4 AND TBX5 HAVE TEMPORALLY DISTINCT ROLES DURING LIMB DEVELOPMENT

Two independent studies have investigated the temporal requirement of Tbx4 and Tbx5 during mouse limb bud outgrowth using conditional deletion strategies. Both studies reach identical conclusions: Tbx4 and Tbx5 are required during the initial phase of limb initiation but are dispensable for continued limb outgrowth (Hasson et al., 2007; Naiche and Papaioannou, 2007) (Fig. 2B). During the first phase, Tbx5 and Tbx4 are required to establish Fgf10 expression and the FGF signalling positive feedback-loop. Once established, the FGF signalling loop is self-maintaining and the input from Tbx4 or Tbx5 is no longer required (Fig. 1C). During a brief second phase, when the limb bud can develop autonomously, deletion of Tbx5 in forelimbs or Tbx4 in hindlimbs has no effect on outgrowth of the limb skeleton but rather specifically affects muscle and tendon morphogenesis (Hasson et al., 2010). It is likely that Tbx5 and Tbx4 proteins have distinct targets during these two phases of activity.

Figure 2.

Limb-type identity and limb-type morphology. A, a: Representation of the flank of a mouse embryo after limb bud formation. The axial levels at which the limbs form define their forelimb or hindlimb identity. b, d: 3D renderings of the serially homologous skeletal elements of the forelimb (b) and hindlimb (d), respectively. Scapula/pelvis (grey), humerus/femur (blue), radius and ulna/tibia and fibula (red), carpals and phalanges/tarsals and phalanges (orange); arrow indicates the patella. Images adapted from the interactive 3D mouse limb anatomy atlas http://www.nimr.mrc.ac.uk/3dlimb (Delaurier et al., 2008). c: Colour coding used to describe the serially homologous bones of the limb in b and d. e, f: Immunostaining, using an anti-muscle Myosin antibody, illustrating the forelimb (e) and hindlimb (f) -type muscle patterns in (E) 15.5 Limbs. Fdb, Flexor digitorum brevis; Abq, Abductor quinti; Fds, Flexor digitorum superficialis muscles. g, h: Dermal pad pattern of a (E) 17.5 forelimb (g) and hindlimb (h). B: Schematic representation the chronology of events regulated by the activities of Tbx5, Tbx4, and Pitx1.

In summary, limb bud initiation describes events that lead to establishment of the FGF signalling positive feedback-loop between limb mesenchyme and the AER. In contrast, limb bud outgrowth encompasses events during subsequent expansion of the limb bud after the FGF positive feedback-loop has been established. Forelimb initiation is a Tbx5 -dependent process and although a small hindlimb bud can form in Tbx4 mutants, Tbx4 is required for normal hindlimb initiation. In contrast, limb outgrowth can occur independently of Tbx5 and Tbx4 activities.

DEFINING LIMB-TYPE IDENTITY AND LIMB-TYPE MORPHOLOGY

The terms limb-type identity and limb-type morphology have frequently been used interchangeably in studies describing the establishment of the specific structures of the forelimb and hindlimb. Limb-type identity in tetrapods is primarily defined by the position along the rostro-caudal axis at which the limbs emerge; the pair of limbs positioned at the anterior of the embryo are the forelimbs while the limbs at the posterior are the hindlimbs (Ruvinsky and Gibson-Brown, 2000). This distinction is independent of the morphology of the limbs. There is no generic forelimb or hindlimb-type morphology since there can be tremendous diversity in the equivalent appendages between different species, for example, the wing of a bird and forelimb of a mouse. Many of the characteristic features that define particular limb-type morphologies are, therefore, specific to the species considered.

Forelimb and hindlimb are serially homologous structures reflecting their common evolutionary and developmental histories (Fig. 2A, b–d) (Ruvinsky and Gibson-Brown, 2000). Homologous structures of the forelimb and hindlimb, such as humerus/femur and radius-ulna/tibia-fibula, therefore, share some anatomical similarities but can be distinguished by characteristic limb-type morphologies. Forelimb muscle and tendons also have hindlimb homologous counterparts that differ in size, shape, relative position in the limb, and origin and insertion sites on the limb skeleton (Fig. 2A, e,f). Defining precisely what constitutes an unambiguous characteristic forelimb and hindlimb feature is important when interpreting the apparent disruption of limb-type morphologies.

CONFLICTING DATA ON THE ROLE OF TBX4 AND TBX5 IN ESTABLISHING LIMB-TYPE MORPHOLOGIES.

As discussed previously, two T-box transcription factors Tbx4 and Tbx5 have expression profiles that are consistent with them playing roles in determining limb-type morphologies. Tbx5 is expressed in the forelimb, whereas Tbx4 is expressed in the hindlimb buds and both genes are expressed throughout the limb mesenchyme and not in the limb ectoderm (Gibson-Brown et al., 1998; Isaac et al., 1998; Logan et al., 1998; Ohuchi et al., 1998). Ectopic gene misexpression experiments in the chick appeared to confirm a role for Tbx5 and Tbx4 in determining forelimb and hindlimb morphologies, respectively. Ectopic expression of Tbx5 in the developing chick hindlimb bud has been reported to partially transform the morphology of the leg to a more wing-like type and induce the ectopic expression of forelimb markers (Rodriguez-Esteban et al., 1999; Takeuchi et al., 1999). Conversely, Tbx4 misexpression in the forelimb can apparently partially transform the wing to develop some hindlimb features and the expression of hindlimb markers (Takeuchi et al., 1999). Following the misexpression of Tbx4, the endogenous expression of Tbx5 in the forelimb is maintained arguing that the morphologically transformed forelimbs have retained their forelimb molecular identity.

The observations in the chick are in direct contrast with results of genetic experiments in the mouse using a combination of a Tbx5 conditional mouse and transgenic lines to carry out gene deletion-gene replacement experiments. Following conditional deletion of Tbx5 and simultaneous replacement with transgenically-supplied Tbx4, forelimb outgrowth and morphologies are rescued (Minguillon et al., 2005, 2009). These results demonstrate that Tbx4 is sufficient to compensate for Tbx5 in forelimb initiation and that, in this context, it is not able to determine hindlimb morphologies. They also demonstrate that a morphologically recognizable forelimb can form in the absence of Tbx5 activity and indicate that other factors must determine their formation. Moreover, the Tbx4-rescued forelimbs retain expression of forelimb markers Tbx5, Hoxc4, and Hoxc5, and hindlimb markers are not ectopically expressed. The Tbx4-rescued forelimbs have, therefore, retained their forelimb molecular identity. Taken together, these results demonstrate that in mouse, Tbx5 and Tbx4share common roles during limb initiation but do not determine limb-type morphologies.

LIMB FORMING CAPACITY OF TBX4 AND TBX5 IS EVOLUTIONARILY ANCIENT

Tbx4 and Tbx5 are paralogous genes that arose by duplication of a single, ancestral Tbx4/5 gene. The extant cephalochordate amphioxus possesses a single Tbx4/5 (amphiTbx4/5) and lacks paired appendages, whereas all vertebrates with paired appendages express Tbx5 in the forelimbs and Tbx4 in the hindlimbs (Gibson-Brown et al., 1996). It is potentially significant that the timing of this duplication coincides with the acquisition of paired appendages in vertebrates. In the background of the conditional deletion of Tbx5, an amphiTbx4/5 transgene is able to rescue forelimb formation. This demonstrates that the ancestral gene, present in limbless organisms, has the ability to initiate limb formation if expressed in the appropriate location (Minguillon et al., 2009). In transgenic analysis, constructs containing fragments of the mouse Tbx5 and Tbx4 genomic regions are able to drive gene expression in the forelimb- and hindlimb-forming regions of the LPM while, in contrast, the amphiTbx4/5 genomic locus appears to lack the regulatory modules enabling expression in the LPM. In accordance with broadly accepted evolutionary models, these results suggest that changes at the level of the regulation of Tbx5 and Tbx4expression, rather than the generation of novel protein function, was necessary for the acquisition of paired appendages during vertebrate evolution (Horton et al., 2008; Minguillon et al., 2009).

PITX1 CONTRIBUTES TO SPECIFICATION OF HINDLIMB-TYPE MORPHOLOGIES

One gene, the paired-type homeodomain transcription factor Pitx1, has been unambiguously implicated in limb-type specification. Pitx1 is expressed in the hindlimb-forming region and hindlimb bud, but not in the forelimb region (Lamonerie et al., 1996). Functional studies performed in both chick and mouse support a role for Pitx1 in specification of hindlimb morphology. When misexpressed in the developing wing of a chick embryo, the muscles, skeletal elements, and skin of the wing adopt some characteristic leg features (Logan and Tabin, 1999; Takeuchi et al., 1999). Consistent with these observations, Pitx1 misexpression in the mouse forelimb results in transformation and translocation of specific muscles, tendons, and bones to acquire a hindlimb-like morphology (DeLaurier et al., 2006). Significantly, equivalent transgenic lines that express ectopic Tbx4 in the forelimb or Tbx5 in the hindlimb do not produce any alterations to limb-type morphologies again consistent with neither gene having a role in determining limb-type morphologies (Minguillon et al., 2005, 2009). In Pitx1 null mice, the hindlimb skeleton loses some of its characteristic features (Lanctot et al., 1999; Szeto et al., 1999; Marcil et al., 2003). This is particularly clear in the zeugopodal bones. Normally, the diameter of the fibula is around half that of the tibia while the homologous elements in the forelimb zeugopod (ulna and radius) are roughly equivalent in diameter. In the Pitx1 mutant, the fibula and tibia have equivalent diameters. The knee joint also lacks a patella and size and shape of the calcaneous bone in the ankle are abnormal. The loss of these hindlimb characteristics has been interpreted as indicating that the Pitx1−/− hindlimb has adopted a more forelimb-like morphology. However, the molecular marker of the forelimb Tbx5 is not ectopically expressed and the putatively “transformed” hindlimbs do not develop definitive forelimb morphologies (Marcil et al., 2003). The Pitx1−/− hindlimb phenotype is thus consistent with a loss of hindlimb attributes rather than acquisition of forelimb features and suggests that forelimb is not the default morphology.

The Pitx1−/− mutant hindlimb is not devoid of all hindlimb characteristics. Although Pitx1 clearly plays an important role in determining some hindlimb morphologies, it is unlikely to be acting alone but rather is part of the multiple inputs required for the complete array of hindlimb morphologies to be produced. Ectopic expression of Pitx1 at stages when the forelimb bud has already formed is sufficient to partially transform the morphology of limb structures to those of the hindlimb. This is consistent with Pitx1 acting to establish hindlimb morphologies, from the onset of limb budding until limb progenitors are undergoing terminal differentiation to form distinct tissues (Fig. 2B). In summary, therefore, Pitx1 appears to act as a “modifier” to influence the response of progenitors to signalling/patterning inputs that are common between forelimbs and hindlimbs.

PITX1 REGULATES TBX4 EXPRESSION AND CONTRIBUTES TO NORMAL GROWTH OF THE HINDLIMB

Accurately assigning roles for genes in determination of limb-type morphology based on the characteristics of abnormally developed limb elements that form following gene deletion or gene misexpression can be problematic. A definitive set of limb-type-specific criteria needs to be established that can unambiguously categorise structures and distinguish changes in limb-type morphology from altered morphology arising through abnormal development. A good example of this type of problem is presented by the phenotype of the Pitx1 mutant. Pitx1 is required for normal levels of Tbx4expression in the hindlimb (Lanctot et al., 1999; Szeto et al., 1999) and misexpression of Pitx1 in the forelimb of the chick or the mouse is sufficient to induce Tbx4 ectopic expression (Logan and Tabin, 1999; DeLaurier et al., 2006). Moreover, Tbx4 expression is reduced in Pitx1−/− mutants (Fig. 1B). Therefore, at least some aspects of the Pitx1−/− phenotype can be attributed to hypomorphic levels of Tbx4 transcripts that will lead to the abnormal development of limb structures.

Clarification of the epistatic relationship between Pitx1 and Tbx4 and the effects of hypomorphic Tbx4 expression on hindlimb formation has come from an analysis of the regulatory elements of Tbx4 in mouse. By combining a systematic functional scanning in transgenic mice bearing BAC-modified lacZ-reporters and alignment with the human Tbx4 locus to find evolutionary conserved elements, Menke et al. (2008) identified two hindlimb enhancer sequences that regulate Tbx4 hindlimb expression. A region located upstream of the Tbx4 coding sequence can drive strong expression in the proximal portion of the hindlimb, but not the distal half of the autopod at E12.5. An element, 3′ of the coding sequence, drives strong LacZ expression throughout the hindlimb and genital tubercle. Importantly, both enhancers are active in the hindlimb-forming region before the onset of hindlimb bud outgrowth and thus are likely to be important during hindlimb bud initiation. Three putative Pitx1-binding sites are present in the 5′enhancer and mutation of the most conserved site leads to reduction in enhancer activity. Homologous recombination to excise the 5′ hindlimb enhancer from the mouse Tbx4 locus results in lowered Tbx4 hindlimb expression and animals homozygous for the deletion have smaller hindlimb bones (pelvis, femur, tibia, and patella). In the autopod, anterior digits are more affected than posterior digits and distal elements are more reduced than proximal elements. The reduction in the size of hindlimb bones following loss of Pitx1 input in Tbx4 expression suggests that regulation of Tbx4 by Pitx1 is required for normal initiation of hindlimb outgrowth.

These results demonstrate that a hindlimb bud expressing hypomorphic levels of Tbx4 does form structures with hindlimb characteristics, consistent with the phenotypes observed in human with Small Patella syndrome (SPS). This underlines the importance of distinguishing a disruption in normal growth and formation of structures from a disruption in them attaining their correct limb-type morphology. This can be difficult since normal growth of a structure will ultimately contribute to its morphogenesis.

OLD DOGMA THROWN A NEW BONE

In recent work, Ouimette et al. adopted a strategy to compare the abilities of transgenically provided Tbx5 or Tbx4 to rescue the Pitx1−/− hindlimb defects (Ouimette et al., 2010). In the Pitx1 mutant, the levels of Tbx4 transcripts are reduced and the ilium of the pelvic girdle fails to form normally contributing to a defective rotation of the hindlimb. Delivery of transgenic Tbx4 can rescue formation of the ilium and, consequently, rescues the rotation defect. In the Pitx1−/− knee joint, the fibula makes contact with the femur rather than the tibia in a manner that has some similarities to the articulation of bones at the elbow of the forelimb and a more hindlimb-like pattern of bone contacts are re-established in the Tbx4-rescued knee. In addition, the normal angle between the calcaneus and the footplate is also restored. Aspects of hindlimb-specific muscle position are also apparently rescued by the Tbx4 transgene. In contrast, a Tbx5 transgenic line was unable to rescue the formation of these structural elements to the same extent. These morphological changes in the apparently “rescued” mutant are interpreted as a rescue of hindlimb morphology although they may also be explained by rescue of initiation of normal outgrowth. Interestingly, some hindlimb features absent in the Pitx1 mutant, such as the patella and the small ratio of fibula/tibia, are not rescued by transgenic elevation of Tbx4 levels.

The authors (Ouimette et al., 2010) suggest that these paralogous proteins have evolved distinct functions and that changes in the Tbx4 protein function explain how it determines hindlimb morphologies. In luciferase assays, as previously reported (Agarwal et al., 2003; Rallis et al., 2003), Tbx5 functions as a transcriptional activator. However, Tbx4 is identified to have transcriptional repressor properties. Both activator and repressor activities were mapped to the C-terminal domain. In this work and other studies (Minguillon et al., 2005, 2009), Tbx4 ectopically expressed in the forelimb is not capable of transforming structures to a more hindlimb-like morphology. An explanation is offered that Tbx4 is normally acting in cooperation with a hindlimb-specific transcriptional co-repressor and, therefore, in the forelimb context Tbx4 is an activator and the Tbx4 repressor activity, correlated with attainment of hindlimb morphology, is not functional.

These results have the surprising conclusion that following duplication to generate the Tbx5 and Tbx4 paralogues, the Tbx4 protein evolved a novel functional domain. It also remains unclear how Tbx4 is acting simultaneously in hindlimb progenitors as both transcriptional activator, in its role in hindlimb initiation, and as transcriptional repressor, directing formation of hindlimb morphologies.

MUTATION OF CIS-REGULATORY MODULES CAN GENERATE MORPHOLOGICAL DIVERSITY

Paired appendages have been repeatedly lost in some fish, amphibian, reptile, and mammalian lineages (Shapiro et al., 2006). In stickleback fish, loss of the pelvic girdle occurs in several freshwater populations likely as an adaptive response to predators and/or water chemistry (Protas and Tabin, 2004). Previous studies have mapped stickleback pelvic reduction locus to the Pitx1 gene through genetic linkage analysis and have found that Pitx1 expression is absent in pelvic reduced species (Shapiro et al., 2004). A recent study by Chan et al. (2010) further refines these observations by precisely identifying the cis-regulatory modification responsible. This study efficiently illustrates how morphological changes can arise from alteration in a single regulatory region of a key developmental control gene rather than by modification of the protein sequence (Chan et al., 2010). F1 hybrid stickleback fish were generated by crossing fish with intact pelvic girdles (FRIL) to those with a reduced pelvis (PAXB). The presence of the FRIL pitx1 allele is able to rescue formation of pelvic structures in the hybrid fish. Furthermore, the PAXB allele of pitx1 is expressed at lower levels than the FRIL allele in a pelvic-restricted manner, suggesting that the failure of pelvic development is a consequence of cis-regulatory changes. Using association mapping in a population of sticklebacks in which some individuals had a complete pelvis and others had a reduced pelvis, Chan et al. (2010) found that a 23-kb region upstream of Pitx1 was strongly associated with pelvic reduction. Alignment of the medaka and zebrafish sequences that span this region enabled the authors to isolate conserved regions that could act as potential enhancer elements. Further deletion analysis, by cloning different fragments upstream of a reporter gene, identified a 2.5-kb pelvic-specific enhancer (2.5-kb Pel). Notably, the Pel region is absent from the Pitx1 loci of pelvic-reduced species. This 2.5-kb Pel region upstream of a Pitx1 transgene is able to rescue pelvic girdle development in the pelvic-reduced fishes, confirming that regulatory changes in pitx1 underlies pelvic reduction in the stickleback. This study demonstrates how cis-regulatory changes can be employed to generate diversity in hindlimb size.

DIFFERENCES IN TIMING OF FORELIMB AND HINDLIMB DEVELOPMENT

Apart from the differences in morphology, limb-type specific differences can also be reflected in the timing of the induction of forelimb and hindlimb budding and their subsequent development, a phenomenon known as heterochrony, and this varies in different species (Richardson et al., 2009). For example, in mouse and chick the forelimbs emerge just before the hindlimbs, while in zebrafish the relative gap in time between formation of the pectoral and pelvic fins is much longer. As mentioned earlier, in the frog Xenopus formation of the hindlimbs occurs before the forelimb.

A dramatic example of heterochronic limb formation is found in Marsupials. In contrast with placental mammals, marsupials are born with forelimbs that are developmentally advanced relative to their hindlimbs. Marsupials give birth after a relatively short period of gestation to immature neonates that must crawl to the teat where they attach and continue their development. They are able to do so due to the presence of functional forelimbs. The forelimb bud is first apparent at an earlier stage compared to the formation of other embryonic structures in marsupials (Sears, 2009). Perhaps surprisingly, hindlimb formation also occurs earlier although its subsequent outgrowth is delayed (Sears, 2009). Consistent with these observations, expression of markers of forelimb and hindlimb initiation, Tbx5/Tbx4 and their targets Fgf10 and Fgf8, are detected at earlier developmental stages in the marsupial, the Opossum, Monodelphis domestica when compared to the mouse (Keyte and Smith, 2010). It remains unclear whether the subsequent difference in timing of forelimb and hindlimb development arises through acceleration of the forelimb developmental programme or delay of the hindlimb programme or though a combination of these two processes. Further study will be required to provide a complete mechanistic explanation for heterochrony of limb formation in marsupials.

PERSPECTIVES

The establishment of forelimb and hindlimb morphologies is likely to be determined by multiple factors and at least partially though a modifier mechanism in which the presence (and therefore also by default, absence) of a gene alters the response of limb progenitors to patterning signals that are equivalently expressed in both forelimb and hindlimbs. Divergence of regulatory networks in forelimbs and hindlimbs also likely contributes to the diversification of limb-type specific morphologies. So far the clearest candidate “modifier” contributing to limb-type morphologies is Pitx1 as it is the only gene unambiguously demonstrated to be part of the limb-type determination process being both required for hindlimb morphologies to develop and also able to transform limb-type morphologies following ectopic expression in the forelimb. The Pitx1−/− hindlimb loses aspects of hindlimb morphology but does not acquire forelimb characteristics. Neither the forelimb nor the hindlimb represent a default limb-type morphology since forelimb and hindlimb morphologies arise, in part, as a result of the distinct developmental history of the different regions of the LPM from which the forelimb and hindlimb arise.

The acquisition of forelimb and hindlimb morphologies is, therefore, a multigenic process and likely a culmination of the gene expression profiles in progenitors both at pre-limb bud stages and limb-bud stages. Teasing apart all the contributing factors that determine limb-type morphologies is an enormous task but should provide significant insight into how an embryo generates morphological diversity with a limited genetic tool kit.

Ancillary