Pbx1/Pbx2 requirement for distal limb patterning is mediated by the hierarchical control of Hox gene spatial distribution and Shhexpression

Vertebrate limb development occurs along three cardinal axes: proximodistal(PD), anteroposterior (AP) and dorsoventral (DV) (reviewed by Niswander, 2003; Tickle, 2003). These axes are established through molecular networks, beginning when the limbs bud out of the flank and continuing with the organization of signaling centers, such as the zone of polarizing activity (ZPA) in the posterior limb mesenchyme(reviewed by Pearse and Tabin,1998). Continued outgrowth and patterning of the distal limb requires a molecular feedback loop between the ZPA and a strip of specialized epithelium at the apex of the limb, known as the apical ectodermal ridge(AER). Crucial to this interaction is the expression of sonic hedgehog(Shh) in the ZPA (Riddle et al.,1993) and of fibroblast growth factors (Fgfs) in the AER(Sun et al., 2002). Shh signaling establishes the limb AP axis(Chiang et al., 2001) and leads to the formation of distal elements of the posterior zeugopod (i.e. ulna and fibula) and autopod (Harfe et al.,2004).

As of yet, the molecular triggers that activate Shh expression are mostly unknown, although a few candidate genes have been proposed. For example, Hand2 (previously known as dHand) has been shown to act upstream of Shh as Hand2-deficient (hereafter, Hand2^–/–) embryos lack Shh expression(Charite et al., 2000). Hox genes (reviewed by Krumlauf,1994; Capecchi,1997; Deschamps and van Nes,2005) have been more recently proposed as upstream regulators of Shh (Zakany et al.,2004) and of AP and PD axis formation(Kmita et al., 2005). For example, prior to Shh onset, early Hox colinear expression in limb mesenchyme leads to their transcript accumulation in the ZPA(Zakany et al., 2004; Tarchini and Duboule, 2006),while their ectopic expression leads to Shh transcription(Charite et al., 1994; Stratford et al., 1997; Knezevic et al., 1997). Finally, in vertebrates, functional ablation of multiple 5′ Hox leads to distal limb truncations partially mediated by Shh absence(Kmita et al., 2005).

Although early Hox colinearity appears essential for Shh onset,their dynamic expression throughout limb development aids in patterning elements along the PD axis. Indeed, patterning alterations occur in limb skeletal elements of specific developmental modules in mice where multiple paralogous Hox genes have been genetically ablated, suggesting a functional redundancy within paralogous groups(Condie and Capecchi, 1994; Davis et al., 1995; Davis and Capecchi, 1996; Fromental-Ramain et al.,1996a; Fromental-Ramain et al., 1996b; Wellik and Capecchi, 2003). These findings highlight Hox role as global regulators of patterning throughout limb development.

The present understanding of Hox function is that they act partially through the aid of co-factors, such as Pbx TALE homeoproteins(Burglin, 1997; Burglin, 1998), that increase Hox DNA-binding specificity and selectivity (reviewed by Mann and Chan, 1996; Mann and Affolter, 1998; Moens and Selleri, 2006). Pbx proteins, when forming complexes with Hox, are further known to transcriptionally regulate Hox genes themselves(Popperl et al., 1995; Maconochie et al., 1997; Jacobs et al., 1999; Ferretti et al., 2000). However, recent findings strongly suggest that Pbx can also function more broadly in Hox-independent manners(Knoepfler et al., 1999; Berkes et al., 2004) (reviewed by Mann and Morata, 2000; Moens and Selleri, 2006).

Recently, we elucidated roles of Pbx genes in skeletal development and found that although Pbx1 is required for proximal limb patterning(Selleri et al., 2001), Pbx2 or Pbx3 loss does not determine skeletal or limb phenotypes (Selleri et al.,2004; Rhee et al.,2004). Specifically, in Pbx1^–/–embryos, skeletal structures of girdles (i.e. scapula and pelvis) and proximal limb stylopod (i.e. humerus and femur) that normally express nuclear Pbx1 at early developmental stages, are malformed, while their distal elements and joints appear normal. These findings parallel the role of Exd (the Drosophila Pbx ortholog) (Peifer and Weischaus,1990; Rauskolb et al., 1993)in governing proximal domains of the fly appendage, where its expression is restricted (Mercader et al.,1999).

In this study, we first uncover that Pbx1 and Pbx2 are co-expressed in the early vertebrate limb field and that, later, Pbx2is expressed throughout the limb mesenchyme, while Pbx1 is expressed only proximally. Next, by exploiting a Pbx1/Pbx2 loss-of-function mouse model, we determine that, despite the lack of skeletal and/or limb abnormalities in Pbx2^–/– embryos(Selleri et al., 2004),decreasing Pbx2 dose in the absence of Pbx1 does affect limb development more severely than the loss of Pbx1 alone(Selleri et al., 2001). We demonstrate that compound Pbx1/Pbx2 embryos, in addition to their proximal limb defects, exhibit novel and severe distal limb abnormalities; Pbx1^–/–;Pbx2^+/– embryos display loss of distal hindlimb elements, whereas Pbx1^–/–;Pbx2^–/– embryos lack hindlimbs altogether. We establish that in vertebrates distal limb patterning is genetically regulated by Pbx1/Pbx2, at least in part, through their hierarchical control of Hox spatial distribution and Shh expression.

Intercrosses of Pbx1^+/–(Selleri et al., 2001) and Pbx2^+/– (Selleri et al., 2004) were conducted to obtain Pbx1^+/–;Pbx2^+/– mice. On a C57BL/6 background, the number of Pbx1^+/–;Pbx2^+/– obtained was below the expected Mendelian ratio. To increase their number, we crossed C57BL/6 Pbx1^+/–;Pbx2^+/– males to an outbred strain, Black Swiss [NIH-BL(S)], and observed a slight amelioration of the limb phenotype that remained fully penetrant. We then intercrossed these C57BL/6 females and mixed C57BL/6-Black Swiss males and analyzed the progeny for lethality and soft/hard tissue morphologies.

Differential staining of cartilage and bone in whole mouse embryos (E12.5 and E13.5) was visualized using Alcian Blue and Alizarin Red(Selleri et al., 2001).

Whole-mount in situ hybridization was performed on somite-matched embryos at different gestational days using digoxigenin or fluorescein-labeled antisense RNA probes as described (Di Giacomo et al., 2006). These analyses were performed on all compound genotypes, including key controls(Pbx1^+/–;Pbx2^–/– and Pbx1^–/–;Pbx2^+/+), but only shown for those genotypes that displayed alterations in gene expression.

Mouse embryos were dissected at E10.5, fixed with 4% paraformaldehyde in phosphate-buffered saline (PBS) and processed for immunohistochemistry to detect CD44 localization (Sherman et al.,1998). Embryos were permeabilized in 0.1%TritonX-100/PBS, blocked in 3%BSA/0.1%TritonX-100/PBS and then incubated in anti-CD44 (antibody IM7 rat monoclonal, Pharmingen) overnight. The embryos were washed in 0.1%TritonX-100/PBS and incubated with anti-rat Alexa 568 (Molecular Probes-Invitrogen) overnight. After rinsing, embryos were visualized with a fluorescence microscope.

Formaldehyde cross-linking and ChIP (Aparicio et al., 1999) from E10.5 mouse limbs were performed according to described protocols(Orlando et al., 1997), with the following modifications: to disaggregate tissues, samples were forced through 18G needles, cells were fixed for 15 minutes with 1% formaldehyde at room temperature and reactions were quenched with 0.125 M glycine in PBS for 5 minutes. Crosslinked samples were sonicated for 15×25 seconds to obtain average fragment lengths of 500-1000 bp. Immunoprecipitation was performed with 10 μl of protein G-Agarose (KPL) and blocked twice with 1 μg/ml salmon sperm DNA (Sigma) and 1 μg/ml bovine serum albumin, for 2 hours,then overnight. Chromatin was precleared by adding 20 μl of protein G-Agarose for two hours, and incubated with 5 μg of the respective specific antisera, or with 5 μg of anti-Flag (F3165, Sigma) antibody as a control. Incubations were performed overnight at 4°C. These primers were used for PCR amplifications (42 cycles): mouse ShhE1,5′-CTTTGATTTGAAGTCCTGGC-3′; mouse ShhE2,5′-ACTGAGGGGAAAAGTCATC-3′; mouse ShhC1,5′-TCAAGAGAGATCAACAAAAG-3′; mouse ShhC2,5′-TTGGACTCAAGTCCAGAC-3′.

P19 mouse embryonal carcinoma cells(McBurney and Rogers, 1982)were cultured in media (MEMA, GIBCO, Life Technologies) supplemented with 10%fetal calf serum, 2 mM L-glutamine, 100 U/ml penicillin and 100 μg/ml streptomycin. Transfections were performed by CaPO₄ precipitation. In a typical experiment, reporter plasmid (5 mg), expression construct(2.5-5.0 μg) and CMV-βgal (Clontech) (0.1 μg) as an internal control, were used per 6 cm dish. Forty-eight hours after transfection, cells were washed, lysed and assayed for luciferase and β-galactosidase expression (Zappavigna et al.,1994).

Pbx1^+/–;Pbx2^+/– mice were intercrossed to generate progeny with different compound genotypes. Axial and limb skeletal abnormalities were evident only in Pbx1^–/–;Pbx2^+/+, Pbx1^–/–;Pbx2^+/– and Pbx1^–/–;Pbx2^–/–embryos. Pbx1^–/–;Pbx2^–/– embryos died at E9.5/10.5 during the period crucial for limb bud initiation, owing to multiple organogenesis defects, and therefore lacked limbs. Thus, our analyses focused on limb development in Pbx1^–/–;Pbx2^+/– (hereafter, Pbx1/2 mutant) embryos (Fig. 1), which survived until E13.5/14.5. Pbx1/2 mutant embryos exhibited drastic exacerbations of Pbx1^–/– axial and proximal limb defects(Selleri et al., 2001) and novel distal phenotypes (Fig. 1A,B). Foreand hindlimb development in Pbx1/2 mutant embryos were differently affected. For example, although E9.5 Pbx1/2mutant forelimbs already appeared malformed, E10.5 hindlimbs still exhibited relatively normal morphology (Fig. 1A), despite the typical half-day delay in development. However,by E13.5, gross morphology of both limbs was severely affected; specifically,forelimbs displayed shortened zeugopodia and hypoplastic autopodia, whereas hindlimbs appeared rudimentary and lacked true autopodia(Fig. 1B).

Analysis of skeletal development in Pbx1/2 mutant embryos demonstrated that in both limbs, proximal and distal elements were malformed and/or lacking (Fig. 2). Specifically, by E12.5, expression of Sox9, a marker for mesenchymal condensations (Wright et al.,1995), revealed that Pbx1/2 mutant(Fig. 2B) forelimbs displayed only three rudimentary digital condensations, while hindlimbs(Fig. 2D) lacked digital mesenchyme altogether. Furthermore, in hindlimbs, condensations for fibulae were missing. Alcian Blue staining of early cartilaginous elements at E12.5 more clearly demonstrated proximal foreand hindlimb defects (see Fig. 2E-H). By E13.5, Pbx1/2 mutant forelimbs were dysmorphic, but possessed most of their elements, whereas their hindlimbs exhibited striking abnormalities, including loss of distal elements (Fig. 2I-N). Pbx1/2 mutant forelimbs exhibited mis-shapened scapulae that were fused to skeletal elements, probably representing duplicated proximal humeral heads. In addition, humeri had shafts that were thickened and contorted, with cartilaginous anlagen mimicking hypertrophic deltoid tuberosities, more severely than in Pbx1^–/–;Pbx2^+/+(Fig. 2I-K). Although both forelimb zeugopod elements were present, albeit malformed, digits one and five were missing or rudimentary (Fig. 2K). In Pbx1/2 mutant embryos, pelvic girdles were missing most elements except isolated anlagen reminiscent of ischia(Fig. 2L-N). These residual pelvic elements were fused to truncated thickened femurs more severely than in Pbx1^–/–;Pbx2^+/+(Fig. 2L-N). The only remaining hindlimb zeugopodial element appeared as a dysmorphic tibia, whereas its fibula was absent, as were most autopodial elements (except for single rudimentary digital rays) (Fig. 2N).

To determine how specific spatiotemporal differences in Pbx limb expression underlie such abnormalities, we performed in situ hybridization on wild-type embryos with Pbx1and Pbx2-specific probes. From E8.5-9.0, we detected both Pbx1 and Pbx2 in lateral plate-intermediate mesoderm (Fig. 3A,E,I) and early limb fields (Fig. 3B,F,J). From E9.25-E11, Pbx1 was then present in the flank and forelimb proximal mesenchyme(Fig. 3C,K and Fig. 4A,C), while at E9.25 in the hindlimb field it still colocalized with Pbx2(Fig. 3D,H,L). By contrast, Pbx2 remained throughout forelimb (E9.25-11; Fig. 3G,K and Fig. 4B,C) and hindlimb mesenchyme (E9.5-11; Fig. 4B,C), but was particularly intense distally. From E11 to E12, Pbx1 and Pbx2 expression then changed dynamically and became restricted to the anterior and posterior mesenchyme proximal to the autopod(Fig. 4A-C). We also analyzed Pbx1 expression in Pbx2^–/– limbs, and Pbx2 expression in Pbx1^–/– and Pbx1/2 mutant limbs. Pbx1 displayed a slightly broader expression domain but remained proximal (data not shown), while Pbx2 (data not shown) expression remained unaltered in these experiments, thus indicating that Pbx1 and Pbx2 do not markedly control each other's expression.

Starting at early gestational days prior to Shh onset, a mutual genetic antagonism has been shown to exist between Gli3 in the anterior limb mesenchyme and Hand2 in the posterior mesenchyme. In addition, Hand2 has been shown to restrict Alx4 anteriorly(te Welscher et al., 2002a; te Welscher et al., 2002b). These genetic interactions subdivide the early limb into anterior and posterior domains. At E10.5, we found normal Alx4 and Gli3expression in Pbx1/2 mutant hindlimbs(Fig. 5A), indicating that the anterior and posterior domains of the hindlimb mesenchyme were initially preserved. Similarly, Hand2 expression resembled that of Alx4 and Gli3, although it was slightly reduced proximally. Finally, although marked morphological alterations were observed in Pbx1/2 mutant forelimbs prior to E10.5(Fig. 1A), no gross perturbations of Alx4, Hand2 and Gli3 expression were detected (data not shown).

Given the similarities between the distal hindlimb phenotypes of Pbx1/2 mutant and Shh^–/– embryos(Chiang et al., 2001; Kraus et al., 2001), we examined Shh expression in Pbx1/Pbx2 limbs. Strikingly, Shh was never detected in Pbx1/2 mutant hindlimbs from E9.5 to E13.5, as represented at E10.5 and 11.5(Fig. 5B). This absence could not be attributed to either a global Shh downregulation or to a delay in posterior embryonic development, as both Pbx1/2 mutant cloaca and notochord developed normally and displayed normal Shh expression (see Fig. S1A in the supplementary material). Interestingly, we observed low levels of Ptch1 and Gli1 (readouts of hedgehog signaling) (reviewed by Niswander, 2003) at E10.75(Fig. 5C), suggesting that a minimal or transient wave of Shh activity probably occurred. In contrast to Pbx1/2 mutant hindlimbs, in forelimbs Shhexpression was present, albeit at reduced levels, and Ptch1 and Gli1were expressed (see Fig. S1B in the supplementary material), indicating that Shh absence was hindlimb specific.

We next found a relatively preserved CD44 distribution(Sherman et al., 1998) in Pbx1/2 mutant hindlimb AER, despite slight anterior reductions within smaller limbs (data not shown). This finding indicates that Pbx1/2mutant hindlimb AER development proceeded normally, at least until E10.5, and thus was unlikely to be responsible for Shh absence. This result was corroborated by our findings that AER-specific markers, such as Dlx6(Robledo et al., 2002),remained unperturbed in Pbx1/2 mutant posterior hindlimbs (data not shown). Overall, these data demonstrate that the posterior AER was preserved before Shh onset. Conversely at E11.5, after Shh initiation, Fgf8 expression was patchy in Pbx1/2 mutant hindlimbs(Fig. 5B) and resembled that reported for Shh^–/– embryos(Chiang et al., 2001; Kraus et al., 2001).

Intriguingly, normal proximal limb identity and outgrowth were observed (at E11) in Pbx1/2 mutant hindlimbs, as Meis2, a proximal mesenchymal marker (Capdevila et al.,1999), remained unchanged (Fig. 5C). Last, analyses of gremlin(Scherz et al., 2004), Hand2 and Gli3 (te Welscher et al., 2002a; te Welscher et al., 2002b) at later days(Fig. 5D) further confirmed Shh absence in Pbx1/2 mutant hindlimbs, with the latter two expression patterns resembling those observed in Shh^–/–(Chiang et al., 2001; Kraus et al., 2001) and Hoxa/Hoxd-deficient (Kmita et al., 2005) embryos. Overall, these findings establish that Pbx1/Pbx2 are required in the limb mesenchyme for Shhexpression.

Shh absence is intriguing, as Hand2, which has been shown to act upstream of Shh (Charite et al., 2000), remained unaltered in Pbx1/2 mutant limbs(Fig. 5A). However, we found that Pbx expression remained mostly unperturbed in Hand2^–/– forelimbs, although Pbx1 was slightly broadened posteriorly (see Fig. S2A in the supplementary material). Accurate hindlimb analyses were hampered by the poor growth and posterior developmental delay in Hand2^–/– embryos, as well as their lethality at E10.5 (Charite et al., 2000). Given that Hand2, Shh and Hox appear to form a regulatory loop in limb development, wherein Hand2 lies upand downstream of Shh (Zakany et al.,2004), we also examined if Pbx1/Pbx2 and Shhform a similar regulatory loop. Analyses in Shh^–/– embryos revealed relatively unperturbed Pbx1/Pbx2 expression (see Fig. S2B in the supplementary material),despite their abnormal limb morphology and marked apoptosis in distal domains(Chiang et al., 2001).

Recent findings have indicated that Hox expression in limb mesenchyme is crucial for Shh activation(Zakany et al., 2004). Given the lack of Shh in Pbx1/2 mutant hindlimbs and its marked downregulation in forelimbs, we examined Hoxa/Hoxdexpression in compound mutant limbs. Intriguingly, all analyzed Hox genes were severely reduced or absent in future Shh-expressing domains of early Pbx1/2 mutant hindlimbs at E10.25-10.5, despite their relatively normal morphology (Fig. 6A; Hoxd10 data not shown). By contrast, despite their abnormal expression, most Hox genes remained expressed in Pbx1/2 mutant forelimb ZPA (data not shown). In addition, in Pbx1/2 mutant forelimbs, multiple Hox genes were either up-regulated (e.g. Hoxa9and Hoxd9), severely downregulated, or absent (e.g. Hoxa13and Hoxd13), suggesting a multifaceted, hierarchical control of Hox by Pbx. Finally, Hox expression was unaltered in limbs of all other compound mutant genotypes (data not shown).

Given the suggested regulatory loop, wherein Shh regulates digit patterning via its control of autopod-specific 5′ Hox gene reverse colinearity, Hox gene expression was also examined at E11.5. Interestingly,5′ Hoxd gene expression was reduced in Pbx1/2 mutant hindlimb anterior domains, while 5′ Hoxa gene expression remained normal(Fig. 6B; Hoxa10,Hoxd10 and Hoxd12, data not shown). This altered Hoxd gene expression pattern is similar to that reported in Shh-deficient limbs(Chiang et al., 2001). Additionally, Pbx1/2 mutant forelimbs displayed only slight anterior alterations of 5′ Hoxd gene, but not 5′ Hoxa gene, expression(Fig. 6C; Hoxa10,Hoxd10 and Hoxd12, data not shown), coincident with the observed reduction in Shh expression. Therefore, in the context of reduced Pbx1/Pbx2 dose, the mechanisms underlying the spatial regulation of Hoxa versus Hoxd expression differed during autopod development.

To ascertain whether Shh expression was directly regulated by Hox and/or Pbx proteins in developing limbs, we tested for their in vivo binding to an evolutionarily conserved Shh regulatory region (ShhE),which was shown to drive the correct spatiotemporal expression of Shhwithin developing mouse limbs (Lettice et al., 2003). We performed chromatin immunoprecipitation (ChIP)analyses (Orlando et al.,1997) on E10.5 mouse limb cells, using specific antisera against Hoxd10, Hoxd13 and Pbx (Fig. 7A). The enrichment for genomic DNA fragments containing the ShhE sequence was verified by PCR amplification. As shown(Fig. 7A), significant enrichment was observed for anti-Hoxd10 and anti-Hoxd13 antisera in forelimb and hindlimb immunoprecipitated chromatin, indicating that these Hoxd proteins are indeed bound to the ShhE in vivo. No enrichment was observed using a non-specific control antibody (Fig. 7A). Using an antiserum against all Pbx isoforms, Pbx protein was also found to bind in vivo to the ShhE in limb cells(Fig. 7A). The specificity of the Hoxd10, Hoxd13 and Pbx binding to the ShhE was further verified by the lack of enrichment on a genomic region located upstream to this enhancer (Fig. 7A). Finally, as an additional control, no binding was detected for Hoxd13 to the ShhE at E13.5 when Shh expression had ceased in limbs(Fig. 7B).

To determine which Pbx family member colocalized with Shh in limbs we performed in situ hybridization and immunohistochemistry. Notably, Pbx2 was the only Pbx that colocalized with Shh in the ZPA, as it was present throughout the limb(Fig. 4B). By contrast,two-color in situ hybridization revealed that Pbx1 did not colocalize with Shh, even in the absence of Pbx2(Fig. 7C). These results were corroborated by immunohistochemistry, whereby Pbx1 and Pbx2 proteins were detected on adjacent limb sections after Shh expression had been visualized by in situ hybridization (data not shown). Finally, we recently uncovered that Pbx3 is not expressed in hindlimbs(Di Giacomo et al., 2006) and reported findings demonstrated that Pbx4 is present only in testes(Wagner et al., 2001). In conclusion, Pbx2 is the only Pbx family member that colocalizes with Shh, and thus is responsible for the observed binding to the ShhE.

The functional significance of the Hoxd10, Hoxd13 and Pbx2 binding to the ShhE was next tested in transient transfection assays using P19 embryonal carcinoma cells. A luciferase reporter construct (pT81ShhE)was generated, which contained a 745 bp fragment representing the complete ShhE (Lettice et al.,2003). Co-transfection of the ShhE reporter with Hoxd9,Hoxd10 and Hoxd13 expression constructs resulted in significant transcriptional activation above the reporter's basal activity in this system(Fig. 7D). Conversely, Pbx2 co-transfection had virtually no effect on the reporter's activity(Fig. 7D) and its co-transfection with Hoxd9, Hoxd10 or Hoxd13 did not significantly alter their activation of the reporter (Fig. 7D). These findings are consistent with the non-essential function of Pbx2 throughout embryonic development(Selleri et al., 2004). Taken together, our results show that Hoxd proteins can bind in vivo to the limb ShhE and can activate transcription through this regulatory element in cultured cells, suggesting that they directly control Shhexpression.

Our results establish that, although Pbx1 is required for proximal limb patterning (Selleri et al.,2001), Pbx1/Pbx2 share overlapping functions in proximal and distal limb development. In this respect, the fundamental roles of Pbx1/Pbx2 are underscored by our finding that Pbx1^–/–;Pbx2^–/– mutants do not form limbs. Nevertheless, by decreasing Pbx2 dose to one allele on a Pbx1-null background we were able to generate a mouse model where limb development could be studied in depth. Accordingly, Pbx1/2 mutant embryos died earlier in utero than Pbx1^–/– embryos and had limbs with more drastic skeletal defects. Furthermore, compared with those present in Pbx1^–/– embryos, axial skeletal defects were also worsened in Pbx1/2 mutant embryos (data not shown). Overall,these results establish that Pbx1/Pbx2 genetically interacts in axial and appendicular skeletal development.

We observed that the distal Pbx1/2 mutant forelimb and hindlimb morphologies varied. In the forelimb, the ulna and most digits were present,albeit malformed, whereas in the hindlimb, the fibula and most digits were lost. Although limb-specific expression levels of Pbx1/Pbx2could explain this variability, we never detected gross differences in their expression by in situ hybridization or immunohistochemistry. Instead, the less severe distal Pbx1/2 mutant forelimb phenotype is more consistent with our finding that Pbx3 is expressed only in forelimbs up to E11.5(Di Giacomo et al., 2006). Indeed, in Pbx1^–/–;Pbx2^–/–mutants that survive to E11, rudimentary forelimbs, but not hindlimbs, develop(Fig. 1A), coincident with their Pbx3 expression.

Of importance, except for Pbx1^–/–(Selleri et al., 2001), Pbx1^–/–;Pbx2^–/– and Pbx1/2 mutant embryos, all other compound genotypes lacked limb defects and alterations in gene expression. These findings highlight the primary role of Pbx1 and the crucial impact that Pbx1/Pbx2 spatiotemporal expression patterns have on limb development. Indeed, the early colocalization of Pbx1 and Pbx2 in lateral plate and limb field mesoderm(Fig. 3I,J,L), appears crucial for distal limb development. Concomitantly, the later proximal versus distal restrictions of Pbx1 and Pbx2, respectively, suggest a scenario where both genes provide a dynamic spatiotemporal `code' along the entire PD limb axis.

The Pbx1/2 mutant distal limb phenotype markedly resembles those observed in embryos lacking Shh(Chiang et al., 2001; Krause et al., 2001) and Hoxa/Hoxd(Kmita et al., 2005). These similarities are particularly evident in Pbx1/2 mutant hindlimbs,which initially exhibit relatively normal morphology and AP patterning (until E10.25-10.5), but subsequently lack Shh expression. Shhabsence is further supported by Gli3 posterior expansion and the severe reduction of 5′ Hoxd expression, as also documented in Shh^–/– embryos(Chiang et al., 2001; Kraus et al., 2001). Recently,studies on the interaction of 5′ Hoxd and Gli3 proteins(Chen et al., 2004) suggest that their ratio across the limb regulates digit pattern and identity. Therefore, digital loss in Pbx1/2 mutant hindlimbs appears mediated by Shh absence (Fig. 7E; right panel). This exact scenario does not hold for Pbx1/2 mutant forelimbs, which already appear malformed early(E9.5-E10.5), but bear residual Shh expression throughout development. Thus, it is not surprising that forelimbs display less severe digit malformations and/or losses. In fact, digit five absence in Pbx1/2 mutant forelimbs conforms to the expectation that the most posterior digits form under the influence of maximum and sustained Shh signaling (Ahn and Joyner,2004; Harfe et al.,2004). Interestingly, digit one absence in Pbx1/2 mutant forelimbs probably occurs via the disruption of Shh-independent pathways, as early forelimbs display morphological abnormalities prior to Shh onset.

Our findings additionally establish that Pbx1/Pbx2hierarchically control Shh expression and that this process may be independent of regulation of Hand2 (i.e. Pbx and Hand2 may control Shh through parallel pathways)(Charite et al., 2000). We cannot, however, rule out that Hand2 controls Pbx function at the post-transcriptional level and that, therefore, they may act on the same regulatory pathway. In light of the known regulatory interactions among Hand2 and Shh in limb development, wherein Hand2lies both up- and downstream of Shh (see Zakany et al., 2004), we demonstrate that Pbx1/Pbx2 are not downstream of Shh and thus that Pbx/Shh do not crossregulate each other in limb development.

As it had been reported that Shh expression is triggered by Hox genes (Charite et al., 1994; Stratford et al., 1997; Knezevic et al., 1997; Zakany et al., 2004), we analyzed Hox gene expression prior to Shh onset in Pbx1/2mutant limbs. In Pbx1/2 mutant hindlimbs, expression was severely perturbed for all 3′ and 5′ Hoxa/Hoxd genes analyzed, especially in posterior domains, while in forelimbs, the alterations varied depending on the Hoxa/Hoxd paralog. These alterations occurred in mutant limbs that exhibited relatively normal early AP patterning. Therefore, Pbx1/Pbx2 act upstream of Hox genes in a manner that appears to occur independently from known molecular regulators required for the establishment of early AP asymmetry. Furthermore, this Hox gene control occurs early, perhaps in the limb field, because both 3′ and 5′ Hoxa/Hoxd genes are already altered at the onset of limb development.

The mechanisms by which Pbx hierarchically control and maintain Hox genes spatial distribution in early limb mesenchyme remain elusive. Pbx requirements in auto- and/or cross-regulatory interactions(Popperl et al., 1995; Popperl et al., 2000; Jacobs et al., 1999; Ferretti et al., 2000; Waskiewicz et al., 2002), or their cooperative roles with other transcription factors upstream of Hox genes(Berkes et al., 2004) (reviewed by Sagerstrom, 2004; Moens and Selleri, 2006) may have bearings. Indeed, with respect to the former, we cannot exclude the possibility that Pbx might directly regulate early Hox gene colinearity, given their co-expression with Hox genes in the early limb.

The absence of Shh expression in Pbx1/2 mutant hindlimb ZPA is thus in keeping with the severe spatial perturbations of multiple Hox genes in posterior mutant hindlimbs (Fig. 7E; left panel). Conversely, in forelimbs, Shh regulation by Hox requires Pbx1/Pbx2 only in part, as the reduction in Shh expression coincided with a less severe alteration in posterior Hox gene localization. Therefore, Hox gene localization to forelimb ZPA appears to require additional factors, with one potential candidate being Pbx3, the expression of which only in forelimbs partially overlaps with Pbx1/Pbx2 in the mesenchyme(Di Giacomo et al., 2006).

We also found that several 5′ HoxD proteins bind in vivo to the Shh limb enhancer (Lettice et al., 2003) at a gestational stage when Shh is expressed,but not at a later stage when Shh expression has waned. Furthermore,we found that several HoxD proteins can activate transcription through this regulatory element in cell culture. These results provide evidence in favor of the proposed direct involvement of HoxD proteins in the control of Shh expression (Zakany et al.,2004; Kmita et al.,2005) and that such control may be exerted by Hox in an overlapping or synergistic manner (Fig. 7E; top schematic).

In agreement with Pbx2 non-essential function throughout embryonic development (Selleri et al.,2004), we found that despite its colocalization with Shhand its binding in vivo to the ShhE, Pbx2 does not activate transcription or cooperate with Hox proteins in transcriptional activation through this regulatory element in our cellular readout. Therefore, Pbx2 in vivo binding to the ShhE is functionally unessential within the Pbx-Hox-Shh regulatory network(Fig. 7E; top schematic). It may, however, represent a `mark' for the possible recruitment of other yet uncharacterized transcription factors, as reported in the case of the myogenin promoter (Berkes et al., 2004)(reviewed by Sagerstrom,2004). Thus, our data indicate that the primary role of Pbx2 is exerted in concert with Pbx1 in the early limb field via their genetic control of Hox gene spatial expression(Fig. 7E; top schematic).

Our findings that multiple Hox proteins are recruited to the limb ShhE are consistent with observations that single deletions of Hox genes do not produce dramatic phenotypes(Capecchi, 1997) that resemble those caused by the loss of Shh. Indeed, our Pbx1/Pbx2mutant mouse model reveals that multiple Hox gene transcripts must synergistically act in vivo in the ZPA for proper and sustained Shhactivation, as corroborated by previous reports on mice with Hox gene cluster deletions (Zakany et al.,2004; Kmita et al.,2005). The findings that misexpression of single Hox genes(Hoxb8 or Hoxd12) in anterior limbs caused ectopic Shh activation (Charite et al.,1994; Stratford et al.,1997; Knezevic et al.,1997), are not in contrast to our model, given the relaxed DNA-binding specificity of Hox proteins and the high expression levels obtained in misexpression experimental settings. Overall, we establish that Pbx1/Pbx2-dependent Hox gene spatial perturbation in the posterior limb has an equally drastic effect on Shh onset as does the disruption of early temporal Hox gene colinearity and the complete ablation of Hoxa/Hoxd gene function (Kmita et al.,2005).

As Hoxa/Hoxd mutants bear limb truncations(Kmita et al., 2005) and Hox genes are known to synergistically pattern all limb elements(Davis et al., 1995; Davis and Capecchi, 1996; Fromental-Ramain et al.,1996a; Fromental-Ramain et al., 1996b; Wellik and Capecchi, 2003), they must govern limb outgrowth. Similarly, as Pbx1/Pbx2 genetic interactions probably occur in early limb field mesenchyme, where they colocalize, they appear to have crucial bearings on overall limb outgrowth, especially in view of their control of Hox gene spatial distribution. Accordingly, stylopod, zeugopod and autopod elements are grossly mispatterned even at the onset of mesenchymal condensation, suggesting that early Hox-dependent patterning alterations underlie the Pbx1/2mutant defects. Furthermore, Pbx1/Pbx2 homozygous loss results in the absence of hindlimb buds, suggesting that a complete disruption of the Hox gene code occurs in early limb development. In light of the early specification model for limb patterning(Dudley et al., 2002), we suggest that interactions of Pbx genes early in outgrowth may drive specification and patterning of distal limb elements. Limb mesenchyme-specific Pbx inactivation will test this scenario.

Despite similarities to Hoxa/Hoxd and Shh mutant phenotypes, the Pbx1/2 mutant limb defects are indeed more severe. Therefore, our results do not completely negate other concomitant scenarios involving disruptions of different networks that govern limb patterning. For example, as suggested (Kmita et al.,2005), Hox genes may sustain distal limb outgrowth by ensuring proper AER function. Similarly, Pbx1/Pbx2, via control of Hox genes,may allow for the proper AER function required for Shh maintenance. Three factors, however, argue in favor of a primary role for Pbx in Shh control that is independent of AER function: (1) the relatively normal Pbx1/2 mutant hindlimb AER morphology and observed gene expression prior to Shh onset; (2) the finding that ectopic Hox gene expression triggers Shh transcription(Charite et al., 1994; Stratford et al., 1997; Knezevic et al., 1997; Zakany et al., 2004); and (3)the strong resemblance of the Pbx1/2 mutant hindlimb phenotype to those reported in Shh (Chiang et al., 2001; Kraus et al.,2001), as well as Hoxa/Hoxd absence(Kmita et al., 2005). However, Pbx roles in limb development may resemble those in spleen genesis,where Pbx1 hierarchically controls all known pathways required for spleen development (Brendolan et al.,2005).

Evolutionarily, our findings indicate that the mechanisms of PD limb patterning are poorly conserved from insects to vertebrates. Unlike in vertebrates (Capecchi, 1997),the fly leg PD axis develops independently of Hox genes(Mann and Morata, 2000). In addition, in flies, Exd and Hth (the homolog of vertebrate Meis) are restricted to proximal limbs(Rauskolb et al., 1995; Capdevila et al., 1999; Mercader et al., 1999) and Exd is required for patterning of proximal identities(Gonzales-Crespo et al., 1998). By contrast, we demonstrate that in vertebrates distal limb patterning is also Pbx1/Pbx2 dependent. Part of this vertebrate innovation may lie in Pbx1/Pbx2 interaction in early limb field mesenchyme and possibly also in later Pbx2localization to distal limb domains. In addition, in flies, mutations in Exd result in homeotic transformations without altering Hox gene expression, thereby indicating that Exd can act as an ancillary co-factor to Hox proteins (Peifer and Weischaus, 1990; Rauskolb et al., 1993; Rauskolb et al., 1995). Conversely, loss of Pbx1/Pbx2 causes drastic alterations in Hox gene expression, pointing to their primary role as upstream regulators of Hox genes in vertebrate limbs. This vertebrate novelty indeed has crucial bearings on Shh regulation, as well as on the origins of morphological complexity of distal appendages in tetrapods. Finally, we propose that Pbx1/Pbx2 proteins began to exert their functions differentially in hindlimbs versus forelimbs,because in vertebrates only hindlimb patterning is strictly dependent upon Pbx1/Pbx2-Hox-Shh regulation.

We thank Drs Anderson, Behringer, Boulet, Capecchi, de Crombrugghe,Duboule, Harland, Izpisua-Belmonte, Joyner, Lefebvre, Manley, McMahon,Meijlink, Milenkovic, Niswander, Popperl, Scott and Tabin for probes; Dr Mackem for anti-Hoxd13 antisera; Dr Beachy for Shh^–/– mice; Dr Hadjantonakis for E8.5 embryos; Drs Barna, Mann, Martin, Niswander and Tabin for discussions and protocols; Dr R. Hogg and T. Peburn for editing; and Dr L. Lacy for constant support. T.D.C. is a graduate student in the New York Consortium in Evolutionary Primatology and a recipient of the Carole and Morton Olshen Fellowship. The work was supported by grants from the Association for the Study of Malformations and the Italian Association for Cancer Research to V.Z. and by grants from the National Institutes of Health (HD43997) and from the March of Dimes and Birth Defects Foundation (6-FY03-071) to L.S. L.S. is an Irma T. Hirschl Scholar and a recipient of an Alice Bohmfalk Charitable Trust grant.