Expression of Pbx and Other TALE-Encoding Genes During Axis and Limb Development
Research conducted in zebrafish suggested that the functional differences among pbx genes are likely to be due to differences in their expression rather than in their biochemical activities (Pöpperl et al., 2000). Indeed, ectopic expression of any of the zebrafish pbx genes was able to completely rescue the embryonic phenotype of pbx4 mutants. As it turns out, Drosophila exd can accomplish the same rescue, underscoring the high degree of functional conservation across this protein family. Consistent with these findings, the in vitro DNA-binding properties of different Pbx proteins appear very similar: intriguingly Pbx1, Pbx2, and Pbx3 exhibit identical DNA cooperative binding with a subset of Hox proteins in vitro (Chang et al., 1995). These findings predicted that where the expression of Pbx family members tightly overlaps, their functions may be partially redundant. Different Pbx proteins are expressed in temporally and spatially distinct patterns during organogenesis and in adult mice, and in these contexts mouse single mutant phenotypes have been extremely informative. However, overlapping expression patterns of more than one Pbx protein in multiple developing tissues and organs has been observed, including the appendage (Schnabel et al., 2001; Selleri et al., 2001, 2004; Wagner et al., 2001; Kim et al., 2002; Waskiewicz et al., 2002; Rhee et al., 2004). Therefore, analysis of Pbx compound mutants was required to understand Pbx function in these contexts. The limb is a prime example (see below).
Like many transcription factor–encoding genes, Pbx and their TALE-encoding family members have dynamic expression patterns during development reflecting their multi-faceted roles in embryogenesis. Within the developing limb, the dynamic nature of Pbx expression provides the backdrop in which to envisage the multiple genetic and molecular interactions described above. Before discussing their expression patterns prior to and during limb and girdle development, it is important to highlight some general commonalities in Pbx expression in the vertebrate embryo. To begin, mRNA transcripts for Pbx and other TALE-encoding genes, such as Meis and Prep, are not restricted to a single germ layer (i.e., mesoderm, endoderm, or ectoderm), nor to a single histological type (e.g., mesenchyme versus epithelium), but present in all tissue types. Additionally, their expression is not confined to one developmental stage but found dynamically from early embryogenesis (e.g., gastrula) through organogenesis and into adulthood (G. Di Giacomo and L. Selleri, unpublished observations). Pbx and Meis family members often considerably overlap in their expression territories, likely reflecting their known function as Pbx heterodimers as described above. Accordingly, their expression domains can also, either separately or collectively, overlap with those of other homeodomain transcription factors, such as members of the Hox family, reflecting their ability, as discussed above, to form either heterodimeric (e.g., Pbx-Hox or Meis-Hox) or heterotrimeric (Pbx-Meis-Hox) complexes.
With regard to Pbx genes proper, Pbx1–3 exhibit wide expression territories including the limb and girdle, while Pbx4 is expressed solely in the testes (Wagner et al., 2001). However, despite Pbx1–3's broad and often overlapping expression patterns in limb domains, the loss of any one Pbx gene does not substantially alter the expression of the others. This is specifically the case for the limb and girdle, as in these structures, single Pbx gene loss is not compensated for by spatial up-regulations in other Pbx genes (i.e., at least as assessed by in situ hybridization) (Capellini et al., 2006, 2010). The expression of Meis was also examined in single Pbx1 and Pbx2, as well as compound Pbx1−/−; Pbx2+/−, mutant limbs and their expression remained unaltered spatially in the proximal forelimb and hindlimb in all mutant genotypes listed above (Selleri et al., 2001; Capellini et al., 2006).
Specific Pbx and Meis Expression Domains in the Limb and Girdle
Our previous studies established that Pbx genes are expressed throughout girdle and limb development in five general phases (Phases I–V). During Phase I, Pbx1 and Pbx2 show overlapping expression domains as early as E7.5 in the developing somites, in the Lateral Plate Mesoderm (LPM) along the embryonic axis, and through E9.0 in the forelimb bud prominences (Fig. 1). In this phase, they also display overlapping expression in similar domains of the hindlimb, which are typically delayed in their development by approximately half a day (Fig. 1). In contrast, Pbx3 is not expressed in the LPM and is restricted solely to the early forelimb field (Di Giacomo et al., 2006). Less is known regarding Meis expression during this phase, although preliminary data indicate that Meis1 and Meis2 are also expressed in early somites and along the LPM, while Prep has a more restricted domain in the dermomyotome (unpublished results).
Phase II begins when the limb prominences become recognizable limb buds and ends at the first onset of a morphologically identifiable autopod. During this period, Pbx gene expression patterns dynamically change. As limb bud induction and initial outgrowth commence (E8.5–E9.5), Pbx1 transcripts are restricted to LPM and its derivative medial somatopleuric and lateral proximal forelimb mesenchyme, and by E9.25 are lost from the distal mesenchyme (Fig. 1). Henceforth, during this phase from E9.5–11, Pbx1 transcripts remain in flank and proximal limb bud mesenchyme only (Fig.1). Conversely, during this same phase, Pbx2 transcripts become localized distally once the forelimb bud is formed (E9.25). From E9.5 to E11, Pbx2 expression remains distal, and as the bud matures it is confined only along the peripheral rim of the distal mesenchyme (Fig. 1; unpublished data). However, Pbx2 expression still partially overlaps with Pbx1 (and Pbx3 in forelimb only) in the anterior proximal limb field, a domain important for girdle development. During this phase, besides their mesenchymal expression, Pbx1 and Pbx2 also display overlapping expression in the developing AER (our unpublished results). Unlike Pbx1 and Pbx2, Pbx3 expression is present in the flank of early (E9.5) embryos (Fig. 1) but shortly thereafter becomes restricted to only the forelimb bud mesenchyme (until E11.0–11.5). Interestingly, Pbx3 is not present in the AER (our unpublished observations). Regarding Meis, Meis1–3 expression is typically restricted during these stages to the proximal fore- and hindlimb mesenchyme (Capellini et al., 2006; unpublished data), although Prep overlaps in its distal domain with Pbx2 (our unpublished results). Overall, these early expression patterns for Phase I and Phase II suggest a scenario wherein Pbx1 and Pbx2 provide first an overlapping and then a complementary “code” along the early limb and girdle fields and in the entire limb bud (Fig. 1). In contrast, Pbx3 displays unique expression patterns that make this family member an excellent candidate for a highly specific role in early forelimb development (Di Giacomo et al., 2006). Experimental work on compound Pbx1/Pbx3 mutants suggests this may be the case (see below).
Despite their dynamic, and unique expression domains from E9.5 to E11, later in development, Pbx1–3 transcripts again substantially overlap in their spatial extent. During Phase III in the forelimb, expression of all three Pbx (1–3) overlaps in the anterior and posterior mesenchyme proximal to the autopod (E11–11.75, Fig. 1). In the hindlimb, transcripts for Pbx1 and Pbx2 are present at high levels within these domains, unlike Pbx3, which begins being expressed in the hindlimb only after E11. Phase IV occurs during handplate and footplate formation, when all three Pbx are co-expressed in the interdigital mesenchyme, suggesting that the three genes play important roles in later digit specification and morphogenesis (Di Giacomo et al., 2006). Lastly, during Phase V, once the skeletal elements of the limb are formed, Pbx1, Pbx2, and Pbx3 are similarly expressed in proliferative chondrocytes of long bones, while their expression wanes in prehypertrophic and hypertrophic chondrocytes in endochondral ossification (Selleri et al., 2001).
Limb Phenotypes of Pbx Mutant Mice
The multi-phase expression patterns described above suggest that Pbx family members have potentially multiple roles during limb positioning, induction, and outgrowth, as well as limb and girdle patterning, along with functions in digital patterning and endochondral ossification. The use of global mutant mice and the generation of compound mutant embryos, in which multiple Pbx alleles are concomitantly lost, have shed light on Pbx critical functions in these developmental processes.
With regard to the limbs and girdles of single Pbx mutant embryos, only Pbx1 homozygotes exhibit phenotypes (Selleri et al., 2001), while Pbx2 and Pbx3 homozygotes exhibit phenotypically normal appendages (Rhee et al., 2004; Selleri et al., 2004). This finding, along with those obtained from the study of compound Pbx mutants, demonstrate the prime role of Pbx1 in multiple developmental processes (Capellini et al., 2006, 2008, 2010). Pbx1 mutants exhibit forelimbs with a superior-to-inferior reduced scapular blade that is fused at the scapular head to an adjoining humerus, which bears a detached and rudimentary deltoid tuberosity (Fig. 2; Selleri et al., 2001). Their hindlimbs display a pelvis that often lacks an ilium or exhibits one that is detached from the pubis and ischium, and which is fused at the hip or femoral-acetabular joint to a truncated femur. Despite these proximal alterations, the distal fore- and hindlimbs of Pbx1 homozygous embryos appear phenotypically normal (Fig. 2; Selleri et al., 2001). Pbx1 roles in limb development are also evident in zebrafish, in which loss of Pbx1 ortholog pbx4 leads to the reduction and absence of the pectoral fin in the Lazarus mutant (Pöpperl et al., 2000). The roles of other pbx paralogs in zebrafish remain unknown.
Figure 2. Girdle and limb phenotypes of single and compound Pbx mutant embryos at E13.5. FL, Cartilage preparations of forelimbs. Pbx1 homozygous mutants exhibit scapula with superior-to-inferior reduced blade (red arrowhead), and fusions of the scapula to the humeral head with an expanded coracoid process (black arrowhead). Pbx1−/−;Pbx2+/− mutant forelimbs additionally show exceptionally hypoplastic scapular blades (red arrowhead in red inset) with heads that appear duplicated (double black arrowheads in red inset). Distally, they display contorted humeral shafts with deltoid tuberosities that are expanded and projected at an angle to the shaft (black arrowhead). Both the radius and ulna are truncated (red arrows) and there is a significant reduction to loss of digits 1 (open red arrowhead) and 5 (open black arrowhead). Pbx1−/−;Pbx3+/− mutant forelimbs display scapular blade hypoplasia (red arrowhead in inset), fusions, and expansion in the proximal humerus (black arrowhead) and humeral head duplications (double black arrowheads in red inset). However, distal to the humerus, these mutant forelimbs appear grossly normal. HL, Cartilage preparations of hindlimbs. Pbx1 homozygous mutants exhibit reduced pelvic girdles, with an absent ilium (empty black arrowhead) and reduced pubis and ischium (red arrowheads), which are fused to the proximal femur. Distal to the femur, these mutants are grossly normal. Pbx1−/−;Pbx2+/− mutant hindlimbs bear pelves with only one remaining element, likely an ischium (red arrowhead) that is fused to the shortened and malformed femur. They display only one zeugopodial element, likely a malformed tibia, lack the fibula (red arrow), and show one single autopodial ray (open black arrowhead in inset). Pbx1−/−;Pbx3+/− mutants phenocopy single homozygous Pbx1 mutants in all proximal girdle and femur abnormalities (red and open black arrowheads), and like Pbx1 mutants are grossly normal distal to the femur. fe, femur; fi, fibula; FL, forelimb; HL, hindlimb; hu, humerus; pg, pelvic girdle; ra, radius; sc, scapula; ti, tibia; ul, ulna; WT, wildtype. In all panels, proximal is up and anterior is to the right.
Download figure to PowerPoint
To date, analysis of available loss-of-function mouse models for the genes encoding the Pbx TALE dimerization partners Meis and Prep, has not been informative. Indeed, single Meis1 and Prep mutants, along with compound Meis1/Prep mutants, do not reveal limb skeletal phenotypes. All mutants listed above bear normal limbs, while Meis2 mutants have not been obtained yet (Fernandez-Diaz et al., personal communication). Interestingly, work conducted in chick reported that Meis genes are involved in limb development, since over-expression of Meis in distal limb buds leads to proximalization of the limb (Capdevila et al., 1999; Mercader et al., 1999, 2000, 2005). While these data do indicate that Meis homeoproteins are important for proximal limb patterning, their requirements in mammalian limb development are still mostly unknown, as they cannot be inferred solely from over-expression experiments. In sum, to date there is no functional study in the mouse demonstrating that Meis genes are sufficient and/or required for limb bud outgrowth and patterning in mammals.
While the analyses of single Pbx mutants reveal a role for Pbx1 in patterning proximal limb and girdle structures, by generating compound Pbx mutants, we have addressed Pbx roles in the entire appendage and girdle. In this regard, we have begun to investigate issues of functional cooperation and redundancy within domains in which multiple Pbx genes are co-expressed. Below we address Pbx involvement in limb positioning, induction, and outgrowth, as well as limb and girdle patterning, along with their functions during endochondral ossification.
Pbx and Limb Bud Positioning
One of the first events involved in limb formation is the specification of the axial levels where limbs form. It was demonstrated by the work of Harrison (1921) and Slack (1976), for example, using surgical manipulations of chick and salamander embryos, that limb bud position was determined well before bud induction and outgrowth. Yet, despite the seminal work of the early days, the determination of limb bud position along the embryonic rostral-caudal axis still remains enigmatic. A sparse number of mouse mutant models exhibit shifts in the axial position of their limb buds allowing the study of critical factors required for positional determination. Such mouse mutants, as for example single and compound Hox mice, exhibit hindlimbs that are caudally shifted along the embryo (McIntyre et al., 2007). These studies are supported by work in the chick, demonstrating that the expression of Hox9 paralogs along the axial LPM is important in maintaining the boundary conditions involved in limb positioning (Cohn et al., 1997). In these studies, however, little has surfaced regarding the molecular functions of these genes in axial specification. Interestingly, among Polycomb genes, compound loss of Eed and Bmi leads to modest rostral shifts in hindlimb axial position (Kim et al., 2006). As Eed and Bmi typically act as repressors of Hox in anterior axial domains, rostralization of the expression domains of multiple axial Hox genes is observed when these Polycomb genes are experimentally ablated (van der Lugt et al., 1994; Faust et al., 1995, 1998; Lessard et al., 1999; Kim et al., 2006). Thus, Polycomb proteins may provide essential genetic control of limb positioning along the axis, via the modulation of Hox territories.
Importantly, Pbx loss from the embryonic axis (i.e., in Pbx1−/−;Pbx2+/− mutants), from Phase I onwards, leads to reductions in the expression of the Eed and Bmi Polycomb genes and to simultaneous rostralizations in Hox expression territories within the somites and adjacent LPM (Capellini et al., 2008; Fig. 3). These changes are indeed coincident with rostral shifts in the position of the hindlimb bud and limb skeleton in Pbx mutants (Capellini et al., 2008; Fig. 3). For example, unlike single Pbx1−/− and other compound Pbx mutant embryos (including compound triple Pbx1+/−;Pbx2+/−; Pbx3+/− heterozygotes), Pbx1−/−; Pbx2+/− embryos show a rostral shift of the hindlimb buds along the somites and, subsequently, limb skeletons that are shifted rostrally by several lumbar-sacral vertebrae (Fig. 3; and data not shown). These findings suggest that Pbx1 and Pbx2 cooperate in domains wherein the two genes exhibit overlapping expression to position the limb bud. The lack of axial shifts in examined compound Pbx1/Pbx3 mutant embryos is in agreement with our findings that Pbx1 and Pbx2 but not Pbx3 are expressed in the posterior hindlimb embryonic field (Fig. 1). Thus, hindlimb positioning along the embryonic axis is controlled specifically by genetic interactions of Pbx1 and Pbx2, which may accomplish this task via their control of Polycomb gene expression. While this possibility is intriguing, it remains unclear whether Pbx also cooperate with Hox as co-factors in this process (see above). If this were the case, then the observed limb shifts in Pbx mutants could also be the result of a more marked and broad loss of Hox DNA-binding specificity within the LPM, coincidental with alterations in Polycomb gene expression.
Figure 3. Pbx homeoproteins control limb bud and skeletal positioning during limb outgrowth. A: Hierarchical role of Pbx on Polycomb (Bmi and Eed) gene expression along the rostral-caudal LPM and paraxial mesoderm (PM) in the axis. In Pbx1−/−;Pbx2+/− mutants, Bmi and Eed expression is downregulated or perturbed, such that their activity is reduced across the rostral-caudal axis. These changes may be responsible for the shifts in Hox expression along the entire axis, which in turn may account for the vertebral patterning defects also present across the entire axis. Concomitantly, loss of Pbx may dysregulate Hox activity independently of Polycomb (dashed arrow). Importantly, these shifts in Hox expression may be responsible for the rostralization in the position of the hindlimb to more anterior somite levels in Pbx1−/−;Pbx2+/− mutants compared to WT embryos (red line drawing/red somite numbers indicate shifts in the mutant condition, versus black lines/black somite numbers in the WT condition). B: Cartilage preparations showing shifts in limb position leading to the more anterior or rostral location of the pelvic girdle and hindlimb skeleton in Pbx1−/−;Pbx2+/− mutants (right) compared to WT controls (left) (red dashed lines demarcate distance of girdle to the last rib-bearing thoracic vertebrae). Proximal is to the left; rostral is up. Adapted from Capellini et al. (2008).
Download figure to PowerPoint
Pbx and Girdle Development
Prior to and during initial limb bud induction, the LPM along the flank is partitioned into a medial portion, the mesodermal somatopleure, and a lateral portion, which gives rise to the limb proper. The mesodermal somatopleure, at the level of the respective limb fields, gives rise to specific structures of both the pectoral and pelvic girdles (Huang et al., 2000; Malashichev et al., 2008). In the forelimb, cells from specific axial levels of the dermomyotome migrate into this somatopleure, and via cues within this tissue, as well as from the paraxial mesoderm and ectoderm, become specified to form superior to inferior aspects of the scapular blade (Huang et al., 2000). Also, cells of the LPM are specified during the early limb bud stages to form the scapular head and neck, bony structures that directly articulate with the proximal humerus of the limb proper (Huang et al., 2006). In contrast, at the level of the hindlimb, only cells originating from the mesodermal somatopleure have been shown to give rise to the three main elements of each pelvic bone, i.e., a rostral oriented ilium, a ventral facing pubis, and a dorsal positioned ischium (Malashichev et al., 2008). These three elements fuse during development at the location of the hip joint, whereby each pelvic bone articulates with the proximal femur of the limb proper (Pomikal and Streicher, 2010). A number of genetic pathways are known to govern patterning of the bony elements of both pectoral and pelvic girdle during development, though knowledge of the genetic control of girdle development is still rudimentary in comparison to that of limb development (Kuijper et al., 2005; Huang et al., 2006). In the following, we will discuss Pbx contributions to the genetic control of girdle development and how Pbx loss during Phases I–III has marked phenotypic consequences on the development of each girdle. Importantly, like in axial development and limb positioning, Pbx1, Pbx2, and Pbx3 appear to have distinct and differential functions in pectoral versus pelvic development, likely a consequence of differences in their respective expression patterns.
In the pectoral girdle, we have found that Pbx1−/−, Pbx1−/−; Pbx2+/−, and Pbx1−/−; Pbx3+/− embryos exhibit alterations of the scapular blade and head. The latter two genotypes show markedly more severe phenotypes compared to the former (Capellini et al., 2010). While the scapular defects are more severe in Pbx1−/−; Pbx2+/− compared to Pbx1−/−; Pbx3+/− mutants, both bear reduced blades, fusions of the scapular head to the humerus, and apparent duplications of the head/neck complex. We reported that blade patterning genes within the mesodermal somatopleure, but not within the dermomyotome, are severely down-regulated in compound Pbx mutants. Specifically, we identified marked down-regulation of genes that pattern the superior (Alx4, Alx1) (Kuijper et al., 2005), central (Tbx15) (Kuijper et al., 2005; Lausch et al., 2008), and inferior (Gli3) (Johnson, 1967; Kuijper et al., 2005) scapular blade. Also, within the mesodermal somatopleure (and LPM of the proximal limb bud) scapular head and neck markers, such as Hoxc6 (Oliver et al., 1990), were severely reduced and mis-expressed in Pbx mutants, along with genetic pathways controlling mesenchymal condensation and chondrogenesis (Wright et al., 1995; Bi et al., 2001) (i.e., the Emx2 Sox9 pathway; and Col2a expression) (Capellini et al., 2010). Importantly, while a major theme of research has focused on Pbx roles as Hox co-factors, we revealed that their functions in scapular development reside, at least in part, in their ability to form heterodimeric complexes with homeodomain proteins other than Hox. Within the context of the complex morphogenesis of the scapula, we identified critical genetic interactions of Pbx1 with Emx2, a gene responsible for scapular blade patterning, and established that both Pbx1 and Emx2 proteins form a heterodimeric complex capable of directly regulating in vivo and in vitro a conserved Alx1 non-coding element (Capellini et al., 2010). Thus, in the forelimb pectoral field, Pbx homeoproteins cooperates with Emx2 affecting the transcriptional activation of key pathways in scapular blade and head formation.
As for the pelvic girdle, we recently found that Pbx1−/−; Pbx2+/− mutants exhibit only one rudimentary ischial fragment that remains fused to the femoral head (Capellini et al., pages xxx–xxx, this issue). Compared to our findings in the pectoral girdle, we did not observe a strong genetic interaction of Pbx1 with Pbx3 or Emx2, but found that Pbx genes lie upstream of Emx2, controlling its expression in the mesodermal somatopleure. We also established that Pbx homeoproteins hierarchically regulate multiple genes and pathways involved in the formation of each element of the pelvic girdle. Notably, we demonstrated that Pbx genes act upstream of Pitx1, Tbx15, and Emx2, genes that control ilium formation (Lanctot et al., 1999; Pellegrini et al., 2001; Marcil et al., 2003; Lausch et al., 2008); Alx1, Prrx1, and Twist1, genes that affect pubis development (ten Berge et al., 1998; Kuijper et al., 2005; Krawchuk et al., 2010); and Pax1, a marker of ischium formation (Timmons et al., 1994; LeClair et al., 1999). These results suggest that Pbx homeoproteins play critical roles in the early somatopleure, likely underlying the substantial down-regulation of genes involved in pelvic girdle development. Through the use of novel bioinformatic analyses (McLean et al., 2010) we additionally revealed that Pbx genes control girdle and hindlimb morphogenesis by potentially regulating multiple genes (including those listed above) with expression domains within the pelvic field/hindlimb. By this approach, we also found potential Pbx1/Emx2-binding sites in non-coding, conserved regions adjacent to genes whose disruption (by loss of function mutations in mouse, or spontaneous mutation in humans) results in pelvic/hindlimb phenotypes, such Prrx1 and Tbx15 (Capellini et al., pages xxx–xxx, this issue).
Pbx roles in developmental programs as cell fate specification have long been considered primarily as those of Hox cofactors (reviewed in Moens and Selleri, 2006). However, it is of note that no known single or compound Hox mutant mouse model displays scapular blade defects. Only Hox5 mutants exhibit rostral blade shifts, with normal blade morphology (Capellini et al., 2010). Similarly, misexpression or loss of Hox genes result at best in mild alterations of the pelvic girdle, including defects in mice misexpressing Hoxd12 in lateral plate derivatives; modest malformations of pelvic bones and sacrum in Hoxc10 mutants; and lack of uterosacral ligaments in Hoxa11 mutant mice (Capellini et al., pages xxx–xxx, this issue). Therefore, it is unlikely that Pbx homeoproteins affect their roles in girdle formation solely as Hox cofactors, suggesting instead cooperation with other proteins, as Emx2, in scapula development. Thus, we cannot yet invoke a mechanism whereby loss of Pbx proteins, as Hox co-factors, leads to the girdle phenotypes of Pbx compound mutants via altered Hox function.
Roles of Pbx in Limb P-D, A-P, and D-V Axis Formation, Outgrowth, and Skeletal Patterning
When compared to single Pbx mutants, compound mutants exhibit significant alterations in most skeletal elements of each limb type (Capellini et al., 2006, 2010). At E13.5, Pbx1−/−;Pbx2+/− limbs display marked exacerbations of the proximal limb phenotype of single Pbx1−/− mutants as well as novel distal limb malformations (Fig. 2). For example, Pbx1−/−;Pbx2+/− forelimbs have severely dysmorphic humeri that are fused to the girdle at the scapulohumeral joint (Capellini et al., 2010). Distally, these mutants exhibit a truncated and dysmorphic radius and ulna and bear substantial reductions in digits one and five (Capellini et al., 2006). On the other hand, Pbx3 functions at this time are limited to the formation of the humerus and scapula, as Pbx1−/−; Pbx3+/− mutant forelimbs are affected only proximally (Fig. 2; (Capellini et al., 2010). Compound Pbx1/Pbx3 double homozygous mutants die during early limb formation, thus limiting a deeper assessment of Pbx3 roles in this process.
The severity of appendicular phenotypes due to compound Pbx loss is more apparent in the hindlimb, wherein only Pbx1 and Pbx2 are expressed during limb outgrowth (Phase I and II; Fig. 1). Therefore, in the hindlimb dosage reductions have more substantial effects, since loss of Pbx1 and Pbx2 enables a Pbx-null state. Accordingly, Pbx1−/−;Pbx3+/− hindlimbs do not present phenotypic exacerbations versus single Pbx1 null mutants, and exhibit normal limb skeletons distal to the femur (Capellini et al., pages xxx–xxx, this issue). On the other hand, Pbx1−/−;Pbx2+/− hindlimbs at E13.5 exhibit a markedly truncated femur fused to a residual pelvic rudiment (Fig. 2; Capellini et al., 2006). Distally, there is only one zeugopodial element, a moderately malformed but truncated tibia, and a single autopodial ray consisting of one digit, likely digit number one (Fig. 2). This severe digit reduction is similar to that observed in Sonic hedgehog (Shh) null mutants (Chiang et al., 2001). The absence of distal limb elements and the drastic reduction in digit number, along with the severe proximal limb malformations, reveals that both Pbx1 and Pbx2 are required for patterning skeletal elements within all limb compartments. Despite the documented early gestational lethality of Pbx1−/−;Pbx2−/− homozygous mutants at E10.5, these embryos have forelimb buds that are highly dysmorphic and hindlimb buds that are barely detectable (Capellini et al., 2006).
Roles of Pbx in P-D Outgrowth
The process by which the limb prominences undergo lateral, distal outgrowth occurs through the coordinated regulation of the three main P-D, A-P, and D-V developmental axes (see Zeller et al., 2009). Disruptions in any one of these axes typically alters the proper development of the others and often leads to cessation of limb bud outgrowth and/or mis-patterning of soft tissue structures and skeletal elements. One of the earliest events in the establishment of the P-D axis is the induction of a fibroblast growth factor, Fgf10, in the early mesenchyme subjacent to the limb ectoderm of both limb prominences (E8.5–9.5) (Cohn et al., 1997; Ohuchi et al., 1997; Min et al., 1998; Sekine et al., 1999). While being induced by Tbx5 and Tbx4 in the forelimb and hindlimb fields, respectively (Agarwal et al., 2003; Minguillon et al., 2005; Koshiba-Takeuchi et al., 2006; Saito et al., 2006), Fgf10 expression in these domains may also be under the control of collinear Hox gene expression in the early limb bud (Zakany et al., 2007). Once induced, however, Fgf10 expression promotes the expression of Fgf8 within a maturing strip of specialized ectodermal epithelium at the distal tip of the limb bud, the pre-AER and AER (Barrow et al., 2003). Fgf8 expression in the AER then further promotes cellular outgrowth and patterning in the limb via a feedback mechanism on Fgf10 expression in the subjacent mesenchyme (Fig. 4) (Xu et al., 1998; Lizarraga et al., 1999). This Fgf10-Fgf8 loop likely aids the early specification of the skeletal elements of the three limb compartments, and promotes and/or maintains the expression of multiple genes critical for limb development, such as Hox, Wnt, and Shh (Niswander et al., 1994; reviewed in Niswander, 2003; Zeller et al., 2009). For example, Hox expression territories along the P-D axis are also under the influence of Fgf expression in the distal limb (Sun et al., 2002) and are essential for P-D limb outgrowth, as evinced by extreme distal limb truncations in mutants lacking entire HoxA/HoxD cluster genes (Kmita et al., 2005). In these mutants, there is also a coordinated down-regulation and cessation of the Fgf10-Fgf8 feedback loop demonstrating the complex circuitry between these factors.
Figure 4. Model of Pbx gene functions during early limb development. E8.5–E9.0: During early limb field induction, the proximal-to-distal (Pr-D) axis becomes established as Pbx genes control Fgf10-Fgf8 signalling in limb mesoderm (black arrows) and/or possibly in pre-AER ectoderm (light blue arrow; light blue line represents pre-AER). E9.0–E9.5: During Pr-D axis establishment, AER formation and limb outgrowth, Pbx continue to regulate Fgf8-Fgf10 signaling, at least in part due to their upstream control of Hox expression in the early limb bud (black arrows) and possibly via their control of Fgf8 in the AER (blue arrow; blue line represents AER). This upstream regulation leads to Hox expression throughout the limb bud and establishment of the pre-Shh ZPA domain (pink) in the posterior limb. E10–E10.75: Pbx control of Hox expression leads to the direct regulation of Shh expression onset and maintenance (black arrows) in the ZPA (terracotta-colored zone). Shh expression in turn promotes Fgf expression in the AER (blue) and maintains the ZPA-AER signalling network (black arrows). This genetic network leads to the proper control of Shh-dependent distal limb (digital) development and proximal-distal patterning and growth. A, anterior; D, distal; P, posterior; Pr, proximal; ZPA, zone of polarizing activity.
Download figure to PowerPoint
As the proper spatio-temporal regulation of Hox genes and Fgfs is critical to the establishment of the P-D axis in the early limb prominence, we have identified essential roles for Pbx as upstream regulators of Hox gene expression in the early limb bud (Fig. 4). Indeed, in Pbx1−/−;Pbx2+/− mutants, transcripts of both 5′ HoxA and 5′ HoxD genes are substantially reduced and dysregulated in the forelimb bud, while they are absent in the early hindlimb bud specifically prior to and during initial P-D limb outgrowth (during Phase I and II; Fig. 1) (Capellini et al., 2006). Pbx1−/−;Pbx2+/− embryos exhibit truncated and mispatterned proximal-to-distal hindlimbs, defects that partially phenocopy those observed when entire HoxA and HoxD clusters are simultaneously lost from the limb (Kmita et al., 2005). Furthermore, in addition to the observed reductions in Hox expression in the hindlimb, we also identified reductions in Fgf10 and Fgf8 expression in the distal limb mesenchyme and AER, keeping in mind that the residual expression of Fgf10 and Fgf8 in their respective domains likely results from the remaining copy of Pbx2 in this mouse model (Capellini et al., 2006; data not shown). Accordingly, a hierarchical role for Pbx on P-D axis formation and the Fgf10-Fgf8 loop is evinced by the failure of hindlimb formation in compound Pbx1−/−;Pbx2−/− mutants (Capellini et al., 2006). Interestingly, in this mutant genotype Fgf8 expression is completely lost in the hindlimb bud and extremely reduced in the forelimb bud (unpublished data; Fig. 5). It is of note that loss of Fgf8 expression is not observed in other tissues of these mutant embryos, and is maintained, for example, at the midbrain to hindbrain junction (Fig. 5; unpublished data).
Figure 5. Pbx hierarchically regulate early limb P-D and D-V axis establishment. Top left: In situ hybridization at E11 reveals that Fgf8 expression remains mostly unchanged (white arrowhead) at the mid-to-hindbrain junction in Pbx1−/−;Pbx2+/− embryos (30 som.) compared to somite-matched E10 WT (30 som.) and littermate E11 WT (44 som.) embryos (black arrowheads). Littermate WT embryos are included to demonstrate that Pbx1−/−;Pbx2−/− mutants are delayed in somitogenesis. Middle left: Fgf8 expression is barely detectable in Pbx1−/−;Pbx2−/− forelimb buds (red arrowhead) compared to somite-matched WT and littermate WT forelimb buds (black arrowheads). Bottom left: Fgf8 expression is absent in Pbx1−/−;Pbx2−/− hindlimb buds (blue arrowhead) compared to somite-matched WT and littermate WT hindlimb buds (black arrowheads). Top right: In situ hybridization at E10.25 reveals that En1 expression is absent from the anterior and posterior ventral ectoderm in Pbx1−/−;Pbx2+/− forelimbs (blue arrowheads) compared to controls. Middle right: In Pbx1−/−;Pbx2+/− hindlimbs at E10.25, En1 expression is only lost from the posterior ventral ectoderm (blue arrowhead) compared to controls. Bottom right: In hindlimbs at E10.75, Lmx1b expression is significantly up-regulated in the ventral domain in Pbx1−/−;Pbx2+/− mutants (red arrowhead), although such up-regulation may reflect a delay in limb bud outgrowth (E10 limb bud on right). Left, panels show distal limb bud tip with anterior to the top and dorsal to the left; right, panels show ventral limb bud with anterior to the top. A, anterior; FL, forelimb; HL, hindlimb; P, posterior.
Download figure to PowerPoint
Roles of Pbx in Limb A-P Axis Formation
Concurrent with the initiation of the Fgf10–Fgf8 feedback loop, early asymmetry in the A-P axis occurs due to inhibitory interactions in the anterior and posterior limb field mesoderm, specifically between Gli3 and Hand2, respectively (Fig. 4). This mutual antagonism ultimately leads to the restriction of Hand2 expression to the posterior aspect of the early limb prominence (te Welscher et al., 2002). Simultaneously, collinear HoxA and HoxD gene expression patterns overlap at first in a wide A-P distributed pattern along this field. However, as the limb bud undergoes P-D outgrowth, transcripts of HoxA and HoxD genes become increasingly restricted to the distal and posterior limb bud (Zakany et al., 2004). This expression pattern, along with the presence of Hand2 (Charite et al., 2000), leads to the activation of Shh along the posterior limb bud mesenchyme in the ZPA (Fig. 4) (Zakany et al., 2004), which is essential to distal limb and digital patterning (Riddle et al., 1993). Shh signaling in the ZPA mesenchyme promotes a cascade of gene expression leading to the induction of Fgf4 in the AER (Chiang et al., 1996, 2001). Fgf4 further promotes ZPA function, maintains Shh and Hox gene expression along the A-P and P-D axes, and thus facilitates further outgrowth and patterning of the distal limb (Niswander et al., 1993, 1994; Fallon et al., 1994) In this regard, Hox genes serve to coordinate early development of the AER and ZPA via their effects of Fgf10 and Shh, both of which then serve to maintain positive regulatory interactions in the limb bud along both A-P and P-D axes (Fig. 4).
By showing that Pbx genes are required for regulating Hox gene expression in the limb, we have revealed a critical control of A-P axis formation by Pbx. Indeed, we have observed that the loss of Pbx1 and Pbx2 in the hindlimb field (i.e., in Pbx1−/−;Pbx2−/− mutants) leads to the complete failure of Shh induction in the ZPA, unlike in other tissues such as the cloaca and notochord wherein Shh is expressed at normal levels in these mutants (Capellini et al., 2006, 2008) (Fig. 4). Interestingly, in Pbx1−/−;Pbx2+/− hindlimbs, Shh absence is not coupled with marked reduction of Fgf8 expression in the posterior AER at early days, or of Hand2 expression in the mesenchyme, but is correlated with the near complete loss of all HoxA and HoxD transcripts in the pre-ZPA. Furthermore, the down-regulation of Hox and absence of Shh expression leads expectedly to the complete breakdown of the Shh-Fgf4 loop and the cessation of limb bud development. We have revealed that Fgf4 is not expressed in Pbx1−/−;Pbx2+/− hindlimb AER and that all effectors of this feedback loop are substantially disrupted (for review see Zeller et al., 2009). Not surprisingly, the distal hindlimb zeugopodial and autopodial phenotype of compound Pbx1−/−;Pbx2+/− mutants phenocopies that observed in Shh null mutants (Chiang et al., 2001), while the striking proximal-to-distal phenotype across all limb compartments results at least in part from significant disruption of HoxA/HoxD expression (Kmita et al., 2005). In the forelimb, Hox genes are down-regulated and mis-expressed, but not as severely as in the hindlimb, and Shh is only mildly down-regulated in the Pbx1−/−;Pbx2+/− mutant genotype, likely a consequence of Pbx3 expression in the forelimb. Notably, the importance of Hox control by Pbx in the pre-ZPA was further substantiated by our finding that both Hox and Pbx proteins are directly bound in vivo to a Shh long-range limb enhancer (Capellini et al., 2006). This enhancer, when experimentally deleted in mice, recapitulates a global Shh null limb phenotype (Lettice et al., 2002, 2003; Sagai et al., 2005). Thus, Pbx homeoproteins are essential upstream regulators of both P-D and A-P axes formation in the limb, in part via their coordinated control of Hox expression in the bud mesenchyme.
The severe alterations of 5′ HoxA/HoxD gene expression, especially in posterior hindlimb bud domains of Pbx1/2 mutant embryos, occurred in limbs that exhibited relatively normal early AP patterning. Therefore, at least in the limb, Pbx1/Pbx2 act upstream of Hox genes and not solely as Hox-cofactors, in a manner that appears to occur independently from known molecular regulators required for the establishment of early AP asymmetry. Furthermore, this Pbx-mediated Hox gene control occurs early, perhaps in the limb field, because both 3′ and 5′ HoxA/HoxD genes were already altered at the onset of limb bud development. The mechanisms by which Pbx hierarchically control and maintain Hox genes' spatial distribution in early limb mesenchyme remain elusive. Pbx cooperative roles with other transcription factors upstream of Hox genes (Berkes et al., 2004) (reviewed by Moens and Selleri, 2006) may have bearings. Indeed, we cannot exclude the possibility that in the limb Pbx might directly regulate Hox gene colinearity, given their co-expression with Hox genes in the limb bud.
Roles of Pbx in Limb D-V Axis Formation
Unlike for the P-D and A-P axes and their coordinated regulation, fewer molecular regulators of D-V axis formation have been identified. It is also unclear whether Hox genes have roles in D-V axis establishment. Two known genetic regulators of the D-V axis are Engrailed1 (En1), expressed in the ventral ectoderm and essential for ventral fates, and Wnt7a, expressed in the dorsal ectoderm and critical for dorsal fates. Genetic studies using single and compound loss-of-function models for En1 and Wnt7a demonstrated roles for these genes in D-V patterning and AER formation (Loomis et al., 1998). Specifically, En1 has been shown to inhibit Wnt7a signaling ventrally, restricting it to the dorsal ectoderm, whereas Wnt7a spatially limits the expression of the transcription factor Lmx1b to the dorsal mesoderm (Loomis et al., 1996). In turn, Lmx1b has been shown to be critical for dorsal mesenchymal and ectodermal patterning (Riddle et al., 1995; Vogel et al., 1995; Chen et al., 1998). Interestingly, loss of proper D-V axis establishment, particularly via the disruption of Wnt7a in the dorsal ectoderm, does lead to the reduction of Shh expression in the ZPA and failure of A-P axis formation. This finding indicates that some degree of developmental coordination between these axes exist. Likewise, as En1 and Wn7a also genetically interact to direct proper AER development in the distal limb ectoderm (Loomis et al., 1998; Kimmel et al., 2000), they are important factors in P-D axis establishment and limb bud outgrowth (Fig. 4).
Recent results (unpublished) in our laboratory suggest that Pbx genes may also have a role in the formation of the D-V axis. Indeed, we examined the expression of the key D-V regulators discussed above in Pbx1−/−; Pbx2+/− embryos from E10 to E11 (note: we could not examine double Pbx1−/−;Pbx2−/− embryos as they succumb in utero prior to these gestational days). In E10.25 Pbx1−/−; Pbx2+/− forelimbs, we found that En1 expression was reduced in the most anterior and posterior domains of the ventral ectoderm, while in hindlimbs it was specifically down-regulated in the posterior aspect of the ventral ectoderm (Fig. 5). However, despite En1 reduction, we found that the expression of genes such as Fgf8 remained mostly unperturbed in the posterior AER of the early bud (Capellini et al., 2006). The expression of Lmx1b was also examined. Lmx1b is normally expressed throughout the early mesenchyme prior to E9.5–10.5, but later becomes restricted to the dorsal mesenchyme. At E10.75, we found that Lmx1b expression was restricted to the dorsal mesenchyme of Pbx1−/−;Pbx2+/− forelimbs as in wildtype littermates, indicating that the D-V axis was relatively well established in this mutant limb type. Furthermore, no overt skeletal or soft-tissue phenotypes, which could reflect mis-patterning along the D-V axis, were identified in the mutant forelimb. In Pbx1−/−;Pbx2+/− hindlimbs, Lmx1b expression was up-regulated compared to somite-matched littermate controls (Fig. 5). Specifically, this up-regulation was evident not only throughout the dorsal limb mesenchyme but also in the posterior ventral mesenchyme, suggesting a possible localized perturbation of Wnt7a signaling in the overlying ectoderm (Fig. 5). However, we observed only a minor down-regulation of Wnt7a expression in this domain (data not shown). Interestingly, we found that Lmx1b was also expressed in ventral posterior and anterior domains in earlier staged wildtype hindlimb buds. This observation suggested that the ectopic Lmx1b expression in posterior Pbx1−/−;Pbx2+/− hindlimbs potentially results from a developmental delay (Fig. 5). Currently, it remains unclear if control of P-D and A-P axes formation by Pbx factors leads to a concurrent disruption in early D-V polarity, or if the latter D-V disruption is cell autonomous and results from Pbx loss in the ectoderm.
Overall, the loss of multiple Pbx genes from the limb substantially affects the establishment of all three developmental axes and disrupts major signaling centers. It is also highly probable, given the alteration of limb bud positioning and loss of key signaling centers in the early limb bud of Pbx compound mutants, that Pbx homeoproteins exert their roles in early development during limb initiation, prior to the formation of the A-P, P-D, and likely D-V axes (Fig. 4). Experiments by tissue-specific and temporally controllable Pbx deletion in the mouse are ongoing to address all of these open questions.
Pbx Limb Phenotypes and Models of Limb Specification
We have established that the dosage of Pbx1 is paramount, compared to that of Pbx2 and Pbx3, in the developmental control of multiple organs, including the limb and girdle (Capellini et al., 2006, 2008, 2010). We have also observed that Pbx2 is the only Pbx family member that overlaps markedly with Pbx1 in the early limb field (and ectoderm) and is the only Pbx expressed in the distal limb mesoderm after initial bud formation (Fig. 1). Of all compound mutants analyzed, including Pbx1+/−;Pbx2+/−, Pbx1+/−;Pbx2−/−, Pbx1−/−;Pbx3+/−, Pbx1+/−;Pbx3−/−, Pbx2−/−;Pbx3+/−, Pbx2+/−;Pbx3−/−, and Pbx1+/−;Pbx2+/−; Pbx3+/−, only compound Pbx1−/−; Pbx2+/− and Pbx1−/−;Pbx2−/− mutants have drastic P-D limb defects in which each compartment of the forelimb and hindlimb is markedly affected. Multiple known markers of early limb P-D and A-P axis formation are markedly perturbed in these mutants. These experimental findings are in conflict with our observations that single Pbx2−/− and compound Pbx1+/−;Pbx2−/− mutants have phenotypically normal limbs (Selleri et al., 2004; Capellini et al., 2006), and lack changes in limb gene expression. Taken together, these data constitute strong genetic evidence that Pbx1 and Pbx2 control early limb bud formation, in domains wherein they overlap and that Pbx1 plays a prime role in this interaction. We argue that they most likely govern specification and patterning of the limb skeletal elements early in development within the limb field proper.
As several models have been put forth over the last five decades describing how cells become specified to form the individual P-D segments of the limb skeleton (for a detailed review, see Benazet and Zeller, 2009), each of these models (e.g., progress zone model, early specification model, two signal gradient model, and differentiation front model) has specific predictions as to when and where cells of each of the limb compartments are specified, as well as to the roles that specific limb centers (i.e., proximal limb, PZ, AER) have during development (Summerbell et al., 1973; Dudley et al., 2002; Tabin and Wolpert, 2007; Benazet and Zeller, 2009). In this regard, it will be relatively straightforward to test against these temporal requirements using conditionally targeted Pbx alleles, together with specific Cre deleter lines that permit excision of Pbx at specific time points (prior to, or during, early limb field induction versus during limb bud formation) and in select limb spatial compartments (mesoderm versus AER). It is clear from predictions based on preliminary experimental evidence that two key features of Pbx expression (i.e., localization in mesoderm versus AER and differential temporal specificity) must be addressed in depth by genetic dissection. These critical experiments are currently underway. They will pinpoint Pbx functions in the limb and reveal when and where the dominant control of limb axes formation by Pbx homeoproteins occurs during development.
Pbx Roles in Limb Endochondral Ossification
Soon after generating the single Pbx1 loss-of-function mouse model, we reported that the affected embryonic domains of limbs (and ribs) in these mutants displayed markedly diminished chondrocyte proliferation, associated with a notable increase in the number of hypertrophic chondrocytes, accompanied by premature bone ossification (Selleri et al., 2001). Intriguingly, the pattern of expression of genes known to regulate chondrocyte differentiation was not perturbed in Pbx1-deficient cartilage at early days of embryonic skeletogenesis. However, precocious expression of Col1a1, a marker of bone formation, was found in single Pbx1 homozygous mutant embryos. These initial studies established an iterative role for Pbx1 in multiple developmental programs in skeletal formation and revealed a novel function in coordinating the extent and/or timing of proliferation with terminal differentiation during limb endochondral ossification. These later functions of Pbx impact on the rate of endochondral ossification and bone formation (Selleri et al., 2001).
Additionally, it was recently shown that Pbx1 represses osteoblastogenesis by blocking Hoxa10-mediated recruitment of chromatin-remodeling factors in mesenchymal cells. Conversely, targeted depletion of Pbx1 by short hairpin RNA (shRNA) in cultured cells increased expression of osteoblast-related genes. Chromatin-associated Pbx1 and Hoxa10 were present at osteoblast-related gene promoters preceding gene expression, but only Hoxa10 was associated with these promoters during transcription. These interesting results further revealed that Pbx1 is associated with histone deacetylases normally linked with chromatin inactivation. Loss of Pbx1 from osteoblast promoters in differentiated osteoblasts was associated with increased histone acetylation and CBP/p300 recruitment, as well as decreased H3K9 methylation. Based on these studies, conducted in cultured cells, it was proposed that Pbx1 plays a central role in attenuating the ability of Hoxa10 to activate osteoblast-related genes in order to establish temporal regulation of gene expression during osteogenesis (Gordon et al., 2010). It will now be important to corroborate these exciting findings in a system in vivo, possibly in animal models deficient for Pbx proteins.