Pbx1/Pbx2 govern axial skeletal development by controlling Polycomb and Hox in mesoderm and Pax1/Pax9 in sclerotome

Abstract

The post-cranial axial skeleton consists of a metameric series of vertebral bodies and intervertebral discs, as well as adjoining ribs and sternum. Patterning of individual vertebrae and distinct regions of the vertebral column is accomplished by Polycomb and Hox proteins in the paraxial mesoderm, while their subsequent morphogenesis depends partially on Pax1/Pax9 in the sclerotome. In this study, we uncover that Pbx1/Pbx2 are co-expressed during successive stages of vertebral and rib development. Next, by exploiting a Pbx1/Pbx2 loss-of-function mouse, we show that decreasing Pbx2 dosage in the absence of Pbx1 affects axial development more severely than single loss of Pbx1. Pbx1/Pbx2 mutants exhibit a homogeneous vertebral column, with loss of vertebral identity, rudimentary ribs, and rostral hindlimb shifts. Of note, these axial defects do not arise from perturbed notochord function, as cellular proliferation, apoptosis, and expression of regulators of notochord signaling are normal in Pbx1/Pbx2 mutants. While the observed defects are consistent with loss of Pbx activity as a Hox-cofactor in the mesoderm, we additionally establish that axial skeletal patterning and hindlimb positioning are governed by Pbx1/Pbx2 through their genetic control of Polycomb and Hox expression and spatial distribution in the mesoderm, as well as of Pax1/Pax9 in the sclerotome.

Graphical abstract

Introduction

The vertebrate post-cranial axial skeleton consists of a metameric series of elements comprising the vertebral bodies, intervertebral discs, and proximal ribs (primaxial component), as well as the sternum and sternal ribs (abaxial component)(Christ and Wilting, 1992, Liem et al., 2001; Nowicki and Burke, 2000). The primaxial pattern, consisting of distinct cervical, thoracic, lumbar, sacral, and caudal vertebrae as well as proximal ribs, is initially laid down during somite formation and differentiation, whereas the abaxial pattern derives from the lateral plate mesoderm (LPM). During somite formation blocks of paraxial mesodermal (PM) cells on both sides of the neural tube arise from undifferentiated pre-somitic mesoderm, while in somite differentiation a ventral compartment of this mesoderm forms the mesenchymal sclerotome (Christ et al., 2004). At the onset of vertebral formation, ventro-medial sclerotomal cells migrate medially, condense into a uniform density, and form a segmented pattern that gives rise to the distinct vertebral bodies and the annuli fibrosi of the intervertebral discs. Likewise, lateral sclerotomal cells migrate, but do so dorsally to form the vertebral pedicles and the laminae of the neural arches as well as the proximal portions of ribs (Christ and Wilting, 1992).

Both somite formation and differentiation are dependent upon molecular signals from the notochord (Cleaver and Krieg, 2001; Liem et al., 2000; Pourquie et al., 1993; Stemple, 2005; Watterson et al., 1954). Indeed, the availability of mouse notochord mutants, such as undulated (un; Balling et al., 1988; Chalepakis et al., 1991; Dietrich and Gruss, 1995; Wallin et al., 1994) and Danforth short tail (Sd; Koseki et al., 1993), which display specific malformations in vertebral column development, has led to the firm conclusion that notochord signaling to the somites is required for proper formation of the cartilaginous axial skeleton. In vertebrates, the notochord is eventually replaced by the vertebral column, and its remnants form the nuclei pulposi of the intervertebral discs (Hunter et al., 2004; Smits and Lefebvre, 2003).

Numerous genes have been shown to play roles in post-cranial axial skeletal development and, in this regard, Hox transcription factors were among the first discovered (Kmita and Duboule, 2003; Krumlauf, 1994; Wellik, 2007). While it has been demonstrated in zebrafish that four Hox genes (hoxb1, hoxb5, hoxc6, and hoxc8) are expressed within the developing notochord where they may function in patterning (Prince et al., 1998), Hox expression and function in the mammalian notochord has not been extensively analyzed. Most research on Hox indicates that their patterning roles take place in the developing PM and LPM during axial differentiation where they establish rostral-caudal expression boundaries that lead to the regionalization of the vertebral metameric pattern and to the establishment of the distinctive characteristics of vertebrae and ribs (reviewed in Favier and Dolle, 1997; and Wellik, 2007; Nowicki and Burke, 2000; Wellik and Capecchi, 2003). For example, compound mutants for specific Hox paralogs, such as Hox5, Hox6, and Hox9, exhibit drastic transformations in the thoracic region, including rib and sternal defects (McIntyre et al., 2007). Importantly, these global aspects of Hox-based axial skeletal patterning are not apparent in mice that are mutant for only single paralogous alleles, due to the high redundancy among the different paralogous family members (Wellik, 2007).

Hox-based patterning of the vertebral column partially depends on the action of Polycomb group genes, whose function is to repress Hox gene transcription in somites rostral to Hox expression domains and to aid in the maintenance of the boundary conditions that help specify the distinctively patterned vertebrae (Kim et al., 2006). Recently, mutations for the mouse Polycomb genes, Bmi and Eed, have demonstrated that in both their individual and compound absence, Hox gene expression is shifted rostrally and vertebral identities are homeotically transformed and shifted across several regions of the vertebral column (Faust et al., 1995; 1998; Kim et al., 2006; Lessard et al., 1999; van der Lugt et al., 1994).

Later in mouse gestation, members of the Pax family of transcription factors (Robson et al., 2006) are also functionally required to regulate axial skeletal development. Among these, Pax1 is active in the sclerotome (Deutsch et al., 1988), and the severity of the Pax1 mutant phenotype in mice with different Pax1 mutations indicates that its actions may be strongest in the lumbar region and the tail (Koseki et al., 1993). Indeed, Pax1 mouse mutants, such as un, exhibit highly dysmorphic vertebral columns with lack of (or split) vertebral bodies, absence of intervertebral disks, and missing or severely malformed proximal ribs (Wallin et al., 1994). Interestingly, the severity of the Pax1 vertebral phenotype is exacerbated when Pax9 is concurrently lost (Peters et al., 1999) due, at least in part, to altered control of Bapx1 expression in the sclerotome (Rodrigo et al., 2003). The finding that both Pax and Hox mutants have defects in similar skeletal structures suggests that they likely synergistically pattern structures of the axial skeleton (Aubin et al., 2002).

Recently, critical roles of Pbx genes (Pbx1–3), which encode the TALE class of homeodomain-containing transcription factors (Burglin, 1997; 1998; Mukherjee and Bürglin, 2007), have been established in skeletal development (Capellini et al., 2006; Selleri et al., 2001). During the past decade, Pbx TALE homeodomain proteins have been mainly regarded as Hox cofactors, i.e. ancillary factors that increase Hox DNA-binding specificity and selectivity (reviewed by Mann and Affolter, 1998; Mann and Chan, 1996; Moens and Selleri, 2006). It has further been demonstrated that Pbx proteins, when forming complexes with Hox proteins (and with other partner proteins of the Meis family), lead to the transcriptional regulation of Hox genes themselves as well as other target genes (Capellini et al., 2006; Ferretti et al., 2000; Jacobs et al., 1999; Maconochie et al., 1997; Popperl et al., 1995). However, more recent findings also suggest that Pbx proteins can function more broadly in a Hox-independent manner (Berkes et al., 2004; Knoepfler et al., 1999), as well as act as hierarchical regulators of 5’ Hox genes (Capellini et al., 2006). Previously, it was demonstrated that Pbx1 is required for proper patterning of both the axial and limb skeleton (Selleri et al., 2001). Conversely, it was found that absence of either Pbx2 or Pbx3 does not determine skeletal phenotypes (Rhee et al., 2004; Selleri et al., 2004). However, by exploiting a Pbx1/Pbx2 loss-of-function mouse model, it was established that decreasing Pbx2 dosage in the absence of Pbx1 affects limb development more severely than the loss of Pbx1 alone, with the appearance of novel distal limb abnormalities (Capellini et al., 2006). These data indicate that Pbx have overlapping functions during skeletal development.

In this study, it is first uncovered that Pbx1/Pbx2, as well as Meis1/Meis2, are highly expressed in the notochord and in the sclerotome and vertebral anlagen, unlike Pbx3 and Meis3, which exhibit negligible expression in all of these tissues. Next, by exploiting a Pbx1/Pbx2 loss-of-function mouse model, it is shown that despite the lack of axial skeletal abnormalities in Pbx2^−/− embryos (Selleri et al., 2004) decreasing Pbx2 dosage in the absence of Pbx1 affects axial skeletal development more severely than the loss of Pbx1 alone (Selleri et al., 2001). In fact, compound Pbx1/Pbx2 mutant embryos exhibit a homogeneous vertebral column morphology characterized by the marked flattening of all vertebral bodies, the thinning of the laminae, the loss of identity of the transverse processes in all regions of the vertebral column, as well as the presence of rudimentary ribs. Additionally, compound Pbx1/Pbx2 mutant embryos display shifts in hindlimb positioning, which are concurrent with the alterations in vertebral identity and morphogenesis. Of note, all of these defects do not appear to arise from perturbed notochord function, as morphology, cellular proliferation, and apoptosis, as well as expression of critical regulators of notochord signaling are normal in Pbx1/Pbx2 mutants. While the observed defects are consistent with loss of Pbx activity as a Hox-cofactor in the mesoderm, we additionally establish that axial skeletal patterning and hindlimb positioning are governed by Pbx1/Pbx2 through their genetic control of Polycomb and Hox expression and spatial distribution in the mesoderm, as well as of Pax1/Pax9 in the sclerotome.

Materials and Methods

Mice

Intercrosses of Pbx1^+/−(Selleri et al., 2001) and Pbx2^+/− (Selleri et al., 2004) were carried out in order to obtain Pbx1^+/−Pbx2^+/− double heterozygotes. On a C57BL/6 background the number of double heterozygotes obtained was well below the expected Mendelian ratio. To increase their number, C57BL/6 double heterozygous males were crossed to an outbred strain, Black Swiss [NIH-BL(S)]. Next, double heterozygous C57BL/6 females and mixed double heterozygous C57BL/6-Black Swiss males were intercrossed and their progeny was analyzed for lethality as well as soft and hard tissue morphologies. In subsequent generations on mixed genetic backgrounds, a significant amelioration of the C57BL/6 limb but not axial skeletal phenotype was observed.

Histology and immunohistochemistry

For histological analysis, embryos were fixed in formalin and embedded in paraffin for sectioning using standard procedures. 5 µm sections were stained with Hematoxylin and Eosin, mounted in DPX, and photographed using a Magnafire digital camera (Optronix). Immunohistochemistry was performed on somite-matched mouse embryos at different gestational days, fixed with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) overnight (O/N) and processed for immunohistochemistry (Capellini et al., 2006). Paraffin sections were dewaxed and then treated by a microwave antigen retrieval step. The primary antibodies used to detect Pbx1, Pbx2, Pbx3, Meis, and Prep1 proteins were: a Pbx1b-specific monoclonal antibody (αPbx1b; Jacobs et al., 1999); a Pbx2-specific monoclonal antibody (G-20 sc-890; Santa Cruz); a Pbx3-specific monoclonal antibody (αPbx3a; Rhee et al., 2004); a Pan Meis-specific monoclonal antibody (αMeis; Jacobs et al., 1999) and a Prep1-specific monoclonal antibody (αMeis4; Upstate USA, Inc., Lake Placid, NY). To detect the presence of Phospho Histone H3, an antibody for Rabbit polyclonal Anti-phospho histone H3 (Ser10; Upstate USA) was utilized (Hendzel et al., 1997). To detect neurofilament staining on whole embryos at E11.5 and E12.5, an anti-2H3 neurofilament monoclonal antibody was acquired (as developed by T. Jessell and J. Dodd, and obtained from the Developmental Studies Hybridoma Bank, under the auspices of the NICHD, maintained by The University of Iowa, Department of Biological Sciences, Iowa City, IA 52242) and whole-mount immunohistochemistry was performed according to the protocols described in Capellini et al. (2006).

Skeletal preparations

Differential staining of cartilage and bone in mouse embryos (E13.5) was visualized using Alcian Blue and Alizarin Red (Selleri et al., 2001). Individual vertebral skeletal elements were dissected, isolated, and photographed (as described above) in glycerol in an agarose plate to ensure proper and consistent orientation.

Whole-mount in situ hybridization

Whole-mount in situ hybridizations were performed on somite-matched embryos at different gestational days using digoxygenin-labeled antisense RNA probes as previously described (Di Giacomo et al., 2006; Selleri et al., 2001). Probes for Bmi (Kim et al., 2006), Col2a1 (Cheah et al., 1991), Eed (Kim et al., 2006), Hoxa4 (Toth et al., 1987), Hoxb5 (Krumlauf et al., 1987), Hoxb8 (Charite et al., 1994), Hoxc6 (Sharpe et al., 1988), Hoxd3 (Condie and Capecchi, 1993), Hoxd9 (Dolle et al., 1989), Noggin (Valenzuela et al., 1995), Pax1 (Chalepakis et al., 1991), Pax9 (Neubuser et al., 1995), and Sox9 (Wright et al., 1995) were used.

Section in situ hybridization

Somite-matched embryos from E10.5 to E13 were harvested and fixed O/N at 4°C in PBS containing 4% PFA. Single-stranded sense and antisense riboprobes for in situ hybridization on frozen sections were synthesized for Bapx1 (Tribioli et al., 1997), Brachyury (Wilkinson et al., 1990), Col2a1 (Cheah et al., 1991), Hoxb1 (Frohman et al., 1990), Hoxb2 (Hunt et al., 1991), Hoxc6 (Sharpe et al., 1988), Hoxd9 (Dolle et al., 1989), Meis1–3 (Capdevila et al., 1999), Noggin (Valenzuela et al., 1995), Pax1 (Chalepakis et al., 1991), Pbx1 (Brendolan et al., 2005), Pbx2 (Selleri et al., 2004), Pbx3 (Di Giacomo et al., 2006), Shh (Echelard et al., 1993) and Sox9 (Wright et al., 1995).

Analysis of cellular proliferation

BrdU incorporation and cell density counting

Pregnant mice were injected intravenously with 50 µg BrdU/gram of body weight 3 hours before sacrifice. Embryos were fixed and embedded in paraffin as described above to obtain transverse sections. BrdU was detected by immunohistochemistry as described (Nowakowski et al., 1989; Selleri et al., 2001) and sections were counterstained lightly with hematoxylin and eosin. Monoclonal anti-BrdU was purchased from Neo Markers. All BrdU-positive (dark brown) and negative (blue) nuclei were counted over various sections. Five different sections (per axial level) containing sclerotome and vertebral anlage were analyzed for each wildtype (WT) and Pbx1^−/−;Pbx2^+/− littermate embryo; this comparison was performed on three embryos for each genotype per gestational day (E9.5, E11.5, E12.5). Sections were digitized using an Eclipse E600 microscope (Nikon). For each section, outlines of the notochord and sclerotome areas were traced manually, corresponding areas were computed, and positive and negative cells counted digitally (Stereo Investigator Software Package, Version 5.05.4, Micro-brightfield). Approximately 400–600 cells (dependent on the genotype and gestational day analyzed) in total were counted per each section. To compare cell densities between WT and mutant embryos, total cell numbers (both positive and negatively stained nuclei) were counted within a standard area encompassing the sclerotome and then normalized using this area measurement. Next, they were compared using Student’s t-tests as described elsewhere (Capellini et al., 2006).

Anti-phospho H3 labeling

Embryos were treated and sectioned according to the protocol described above for section in situ hybridization. Slides were rinsed in PBS and incubated first with PBS and goat serum, and later with rabbit-polyclonal anti-phospho histone H3 (pHH3; Upstate Biotechnology)(1:1000) O/N. The next day, slides were incubated with an anti-Rabbit biotinylated IgG secondary antibody (Jackson ImmunoResearch) diluted 1:500 in PBS at RT for 30 min, then treated with Streptavidin conjugated peroxidase ABC kit (Vector Laboratories, Cat# SA-5004), and finally signal was detected using diaminobenzidine (DAB) substrate.

Caspase3 apoptosis assays

Caspase3 antibody staining was used to distinguish apoptotic cells from non-apoptotic cells in compound Pbx1/Pbx2 mutant versus littermate WT embryos. Initially, paraffin embedded embryos were sectioned to 10 µm using a Leica microtome and treated for immunohistochemistry as described above. After blocking, the Rabbit anti-Caspase3 (R&D Systems, Cat# AF-835) primary antibody (Biogenesis) was diluted 1:2000 in PBS/0.1% TritonX-100 and applied at 4°C O/N. The sections were next incubated with biotinylated Goat anti-Rabbit IgG secondary antibody (Vector Laboratories, Cat# BA-1000) diluted 1:500 in PBS at RT for 30 min, and the signal was detected using HRP-Streptavidin (Vector Laboratories, Cat# SA-5004) as described above. Three embryos per genotype were examined at each gestational time point including the key controls, Pbx1^−/−; Pbx2^+/+ and Pbx1^+/−; Pbx2^−/−.

Results

TALE homeodomain proteins are present in the somitic mesoderm, the notochord, and the surrounding mesenchyme during mid-gestation

Previous studies have uncovered the presence of Pbx1 and Pbx2 in the somitic mesoderm and the LPM of developing embryos (Capellini et al., 2006; Schnabel et al., 2001). In situ hybridization, as performed here, revealed that PBC- and MEIS-encoding genes were variably expressed in the notochord and condensing pre-vertebral mesenchyme from E10.5–E12.5 (Suppl. Fig. 1). Specifically, Pbx1 was strongly expressed in the developing notochord and detectable in the surrounding mesenchyme (Suppl. Fig. 1A–C) and Pbx2 was present, albeit at lower levels, in the notochord (Suppl. Fig. 1D–F). In addition, Pbx2 was diffusedly expressed in the adjacent mesenchyme, in agreement with previous results indicating widespread expression of Pbx2 throughout the embryo (Selleri et al., 2004). Conversely, Pbx3 expression levels in the notochord and surrounding mesenchyme were either extremely low or undetectable (Suppl. Fig. 1G–I). Despite the varied expression of these Pbx genes in the notochord and mesenchyme, all three were highly expressed in the ventral columns of the spinal cord (Suppl. Fig. 1). Meis genes were also expressed in the spinal cord and at higher levels in the ventral horns (Suppl. Fig. 2). However, unlike Pbx1, both Meis1/Meis2 were expressed at relatively lower levels in the notochord and adjacent somitic mesoderm and mesenchyme (Suppl. Fig. 2A–F), while Meis3 was absent from these tissues (Suppl. Fig. 2G–I). Although mid-thoracic level transverse sections are shown depicting Pbx and Meis expression patterns, similar results were obtained at all axial levels (data not shown). The above results for Pbx and Meis gene expression were corroborated with immunohistochemistry using Pbx- and Meis-specific antibodies from E10.5 to E12.5 (see Fig. 1A; Pbx1 localization at E10.5 is not shown).

Figure 1. Localization of Pbx and Meis/Prep proteins during notochord and vertebral development.

Colocalization of Pbx and Meis/Prep at each gestational day is shown in the 10× view (top panels in A and left panels in B) and the 40× red insets (bottom panel in A and right panels in B). (A) Pbx and Meis proteins are present in the notochord and vertebral anlagen throughout mid-gestation as demonstrated by immunohistochemistry on mid-thoracic transverse sections at E10.5 and E12.5. While at these gestational days Pbx1, Pbx2, Meis and Prep1 colocalize (black arrowheads) in the notochord and also in the adjacent mesenchyme, Pbx3 is absent or present at very low levels in these tissues (empty black arrowhead). (B) At E13.5, Pbx1/Pbx2 colocalize (black arrowheads, red inset) in the notochord as well as in the early vertebral cartilage, while Pbx3 levels remain very low or absent in the latter tissues (empty black arrowhead), as compared to its high levels in the spinal cord (black arrow). P, perichondrium; SC, spinal cord; V, vertebra. Dorsal (top).

To determine if Pbx protein localization persisted in the notochord and condensing mesenchyme of the vertebral bodies, additional immunohistochemistry at E13.5 (i.e., the gestational day of lethality of compound Pbx1/Pbx2 mutants reported below) was performed. Indeed, Pbx1 and Pbx2 proteins persisted in both the notochord and condensing mesenchyme of the vertebral bodies and were present in the forming perichondria (Fig. 1B). Also, consistent with Pbx3 expression patterns and protein localization at earlier gestational days (Suppl. Fig. 1G–I and Fig.1A), Pbx3 protein, while strongly present in the ventral horns of the spinal cord, was undetectable in the notochord, the condensing mesenchyme of the vertebral bodies, and the perichondria (Fig. 1B).

Loss of Pbx1/Pbx2 causes drastic post-cranial axial skeletal defects

To investigate whether Pbx1 and Pbx2 have roles in the developing primaxial and abaxial skeleton, compound Pbx1/Pbx2 mutants were generated (see methods). Skeletal and/or soft tissue abnormalities were evident only in Pbx1^−/−;Pbx2^+/+, Pbx1^−/−;Pbx2^+/−, and Pbx1^−/−;Pbx2^−/− embryos. Since Pbx1^−/−;Pbx2^−/− embryos die at E9.5/10 due to multiple organogenesis defects (Capellini et al., 2006) our analyses focused on Pbx1^−/−;Pbx2^+/− (hereafter, Pbx1/2 mutant) embryos, which survive until E13.5, a gestational time-point sufficient to investigate the early development of the axial skeleton.

Pbx1/2 mutant embryos exhibited drastic exacerbations of Pbx1^−/− axial skeletal defects (Selleri et al., 2001) at E13.5 (Fig. 2). Unlike WT (Fig. 2A–C, J,M), single Pbx1^−/−;Pbx2^+/+ embryos displayed several vertebral and rib defects, such as loss of identities of cervical vertebrae number 1 (hereafter, C1) and C2 (Fig. 2D–F), proximal-distal shortening of ribs that lacked their most ventral domains, and fusion of ribs at the midshaft (Fig 2K,N). Interestingly, despite the lack of vertebral and rib defects in Pbx1^+/+;Pbx2^−/− and Pbx1^+/−;Pbx2^−/− embryos, Pbx1/2 mutant embryos displayed more severe defects in those axial domains affected by the loss of Pbx1 alone (Fig. 2G–I, L,O) and significant malformations at additional axial levels (Fig. 2L,T,U). Specifically, in the cervical region, unlike WT (Fig. 2B–C), Pbx1/2 mutant C1 and C2 were more rudimentary in appearance versus Pbx1^−/−;Pbx2^+/+ C1 and C2 (Fig. 2E–F) and lost all features typical of these vertebral elements (Fig. 2H–I). Additionally, in the Pbx1/2 mutant the remaining cervical vertebrae exhibited thinned transverse processes, which gave the appearance of a loss of identity. Therefore, based on morphological comparisons, the transition from the cervical to thoracic vertebrae was poorly defined at this developmental stage (Fig. 2H–I). In the thoracic region, unlike in WT (Fig. 2J,M) and Pbx1^−/−;Pbx2^+/+ (Fig. 2K,N) embryos, all vertebral centra were more flattened and medial-laterally attenuated in Pbx1/2 mutant embryos (Fig. 2L,O). The rib defects observed in Pbx1^−/−;Pbx2^+/+ embryos (Fig. 2K,N) were more severe in Pbx1/2 mutant embryos, as individual elements appeared as extremely shortened, rudimentary cartilage condensations (Fig. 2L,O). In addition, the rib cartilages for rib #1 and #13 were barely detectable. When coupled with the general rudimentary morphology of the vertebrae, as in the cervical region, the transition from thoracic to lumbar segments was also difficult to discern. In the lumbar-sacral and caudal regions, transverse processes were markedly attenuated or flattened in Pbx1/2 mutant embryos (Fig. 2T–U), unlike in WT (Fig. 2P–Q) and Pbx1^−/−;Pbx2^+/+ (Fig. 2R–S) embryos and the overall posterior regionalization of the vertebral column was difficult to demarcate.

Figure 2. Vertebral and rib skeletal malformations in Pbx1^−/−;Pbx2^+/+ and Pbx1^−/−;Pbx2^+/− embryos at E13.5.

(A,D,G) When compared to WT (A), Alcian Blue staining of E13.5 littermates reveals post-cranial axial defects in Pbx1^−/−;Pbx2^+/+ (D) and Pbx1^−/−;Pbx2^+/− (G) embryos, as demonstrated in the different colored insets depicting higher magnification views of each vertebral segment. In the cervical region (red insets), when compared to WT (B, C; black and red arrowheads), Pbx1^−/−;Pbx2^+/+ (E, F) and Pbx1^−/−;Pbx2^+/− embryos (H, I) both display thinning/reduction in the transverse processes of cervical #1 (C1) and C2 (empty black arrowheads), which is more severe in Pbx1^−/−;Pbx2^+/− embryos (H; double empty black arrowheads). Additionally, the latter display marked reductions in the transverse processes of lower cervical vertebrae (H, empty red arrowheads). In the thoracic region (J–L), when compared to WT (J; red arrowheads) and Pbx1^−/−;Pbx2^+/+ embryos (K), the vertebral centra of the Pbx1^−/−;Pbx2^+/− mutants (L) are rostral-caudally flattened and medial-laterally thinned (empty red arrowheads). Also, in the thoracic region (M–O), as compared to WT (M, red arrowheads), Pbx1^−/−;Pbx2^+/+ embryos (N) display distally shortened and markedly dysmorphic ribs (empty red arrowheads), which are often fused, while Pbx1^−/−;Pbx2^+/− embryos (O) exhibit more severe rib defects with drastically shortened and/or rudimentary elements (double empty red arrowheads). In the lumbar (L1) - sacral (S1) region (P, R, T), the transverse processes of Pbx1^−/−;Pbx2^+/− embryos (T) are more attenuated and flattened (empty red arrowhead) than seen in both WT (P) and single Pbx1 mutant embryos (R) (red arrowheads). Finally, no major defects were observed in the caudal vertebral domain (Q, S, U) of these embryos, although the development of their posterior-most domains may be delayed and/or stunted. Rostral (top).

To more deeply characterize the axial defects present in Pbx1/2 mutant embryos, detailed dissections of isolated vertebral elements were performed (Fig. 3). At E13.5, Pbx1/2 mutant C1 and C2 were fused and their transverse processes were more acutely angled (Fig. 3A, right, all views) compared to WT C1 and C2 vertebrae (Fig. 3A, left, all views). Furthermore, due to this fusion, a distinguishable dens (odontoid process) was absent from the mutant C2 (Fig. 3A, right, ventral view) when compared to WT (Fig. 3A, left, ventral view). Finally, in the Pbx1/2 mutant the transverse processes were markedly thinned and lacked the anlagen of the anterior and posterior tubercles (Fig. 3A, right, lateral view). Several of the alterations present in the most rostral cervical vertebrae of Pbx1/2 mutant embryos were also apparent in more caudal cervical vertebrae, such as C4. For example, an acute angulation (Fig. 3A, right, cranial view) and thinning (data not shown) of the transverse processes were present, unlike in WT (Fig. 3A, left, cranial view). Additionally, unlike WT C4 (Fig. 3A, left, cranial view), the groove for the vertebral artery was absent and the centrum was more rounded (Fig. 3A, right, cranial view), similar to that of thoracic vertebrae (compare C4 Pbx1/2 mutant cranial view to T3 WT cranial view), and potentially suggestive of a homeotic transformation. Lastly, the vertebral centrum of C4 (as well as other, more caudal, cervical vertebrae) was rostral-caudally flattened (Fig. 3A, right, ventral view) compared to WT (Fig. 3A, left, ventral view).

Figure 3. Morphological abnormalities in dissected vertebral elements from Pbx1^−/−;Pbx2^+/− embryos.

Alcian Blue staining of developing vertebrae spanning several axial regions reveals that all vertebral types lose specific features of their axial level identity. (A) Compared to WT (left), in Pbx1^−/−;Pbx2^+/− embryos (right), C1 and C2 are grossly mis-shapened and fused together as evident from the cranial, ventral, and lateral views. Because of this fusion, no distinct dens (odontoid process) is detectable in Pbx1^−/−;Pbx2^+/− embryos (right) compared to WT (left, red arrowhead). As evident in the cranial view, the transverse processes in Pbx1^−/−;Pbx2^+/− C1 and C2 vertebrae (right) project from the body at more acute angles when compared to WT (left). In addition, these mutant transverse processes are thinned (lateral view, empty black arrowheads) and lose specific morphologies that define C1 and C2 identity, compared to WT (lateral view, black arrowheads). The morphology of mutant C4 is also drastically altered compared to WT. In the cranial view, its centrum is more rounded (empty red arrowhead) when compared to the WT (red arrowhead), and the groove for the vertebral artery is absent (empty black arrowhead) compared to WT (black arrowhead). In the ventral view the entire rostral-caudal height of the body is also reduced (short arrow) compared to WT (long arrow). (B) Compared to WT, all of the developing thoracic vertebrae (represented here by T3) are smaller, the transverse processes project from the body at more acute angles when compared to WT, and the ribs are either rudimentary elements or completely absent (dashed red rectangle) compared to WT (red rectangle) seen in the cranial view. Also, there is a marked reduction in the height of the mutant centra (short arrow) compared to WT (long arrow). (C) Compared to WT centra (red arrowhead and long arrow respectively), the mutant centra of all developing lumbar vertebrae (represented here by L3) are more rounded and flattened (cranial view, empty red arrowhead) and thinner in their overall height (ventral view, short arrow). In the lateral view, this height reduction is evident (short arrow) compared to WT (long arrow), as is the loss of anterior and posterior tubercles (empty black arrowheads) emerging from the transverse process, specifically compared to WT (black arrowheads). In cranial views, ventral views, and lateral views, ventral, rostral, and rostral are towards the top, respectively.

Defects were also present in Pbx1/2 mutant thoracic vertebrae, as represented here by T3, which possessed more acutely angled and attenuated transverse processes, a markedly reduced rostral-caudal thickness of the vertebral centrum anlagen, and loss of more distal rib structures (Fig. 3B, right, all views) when compared to WT (Fig. 3B, left, all views). In addition, its centrum was more rounded (Fig. 3B, right, cranial view) than in WT (Fig. 3B, left, cranial view). Cartilage defects were also present in the lumbar region, as represented here by L3 (Fig. 3C, all views). Specifically, its centrum was more rounded (Fig. 3C, right, cranial view) and more markedly rostral-caudally flattened (Fig. 3C, right, ventral and lateral views) than in WT (Fig. 3C, left, all views). Furthermore, in a similar pattern to that observed for more rostral vertebrae, the transverse processes of the Pbx1/2 mutant L3 were extremely thinned and missing superior, middle, and inferior turbercles (Fig. 3C, right, lateral view) compared to WT (Fig. 3C, left, lateral view). In this manner, both the thoracic and lumbar vertebrae more similarly resembled one another than their WT axial counterparts. Overall, despite the persistence of one Pbx2 allele, the morphology of the vertebral column in Pbx1/2 mutant embryos appeared homogeneous.

Pbx1/2 mutant axial defects are not associated with alterations in cellular proliferation and apoptosis in the notochord and surrounding mesoderm and mesenchyme

To examine if detectable changes in cellular behaviors in the notochord and surrounding mesenchyme were responsible for the observed axial skeletal defects, tissue histology, cell density, cell proliferation, and apoptosis assays were performed on Pbx1/Pbx2 compound mutant and WT embryos from E9.5 to E13.5 (Fig. 4A, Suppl. Fig. 3, and data not shown). Hematoxylin and Eosin staining on E13.5 transverse sections revealed that in Pbx1/2 mutant embryos the structure of the notochord and of the adjacent mesenchymal condensation were relatively normal compared to WT (Fig. 4A and Suppl. Fig. 3A). Visual and statistical examination of cell densities between WT and Pbx1/2 mutant embryos from E9.5 to E13.5 revealed that at each gestational day analyzed Pbx1/2 mutant embryos possessed lower densities of cells around the notochord compared to WT controls (e.g., at E11.5, p < 0.01)(Fig. 4A and Suppl. Fig. 3A–C). The lower cell density observed in Pbx1/2 mutant embryos could not be causally related to detectable changes in cell proliferation or apoptosis at these gestational days (see below). Furthermore, this difference could not be attributed to edema. Indeed, while edema was slightly present at E12.5 and more markedly present at E13.5 (as previously reported in Capellini et al., 2006), it was barely detectable from E9.5 to E11.5 (Fig. 4A, Suppl. Fig. 3A–C, and data not shown). To examine if the loss of Pbx1/Pbx2 affected rates of cellular proliferation in the notochord and adjacent mesenchyme, both BrdU (Fig. 4A and Suppl. Fig. 3B) and anti-phospho H3 (data not shown) assays were performed from E9.5 to E12.5. Despite the observed differences in cellular density, at all gestational days (represented here by transverse sections at E11.5 and E12.5), no differences in BrdU- or H3-positive cells were detected in Pbx1/2 mutant versus WT embryos (Fig. 4A and Suppl. Fig. 3B). Likewise, analyses of apoptosis using Caspase3 assays at E9.5–E13.5 revealed no differences between Pbx1/2 mutant and WT embryos (Fig. 4A and Suppl. Fig. 3C). Overall, the observed abnormalities in axial skeletal morphology (Figs. 2–3) and histology (Fig. 4A and Suppl. Fig. 3A) in Pbx1/2 mutant embryos could not be explained by alterations in proliferation or apoptosis in the somitic mesoderm, notochord, and surrounding mesenchyme.

Figure 4. Unperturbed cellular proliferation and programmed cell death in Pbx1^−/−;Pbx2^+/− notochord and vertebral anlage; unperturbed expression of notochord markers in Pbx1^−/−;Pbx2^+/− embryos.

(A) Unless otherwise noted, all images are transverse sections at the mid-thoracic level and are shown in 10× (upper) panels and 40× (lower) insets. Hematoxylin and Eosin staining of E13.5 transverse sections of Pbx1^−/−;Pbx2^+/− and WT embryos (left) does not reveal significant differences in the structure and cellular morphology of the notochord but does reveal that the cellular density of vertebral anlagen is reduced compared to WT embryos (black arrowheads depict the notochord). At E13.5, edema is evident between the spinal cord and the vertebral anlagen of Pbx1^−/−;Pbx2^+/− embryos. Next, at E11.5, cell proliferation as measured by BrdU incorporation is not different in Pbx1^−/−;Pbx2^+/− notochord and vertebral anlage compared to that present in WT (middle, black arrowheads, insets). Finally, at E11.5, levels of cellular apoptosis, as assayed by Caspase3 antibody staining, do not differ between Pbx1^−/−;Pbx2^+/− and WT notochord and vertebral anlage (right, black arrowheads, insets). In contrast, higher levels of cellular apoptosis are equally present in Pbx1^−/−;Pbx2^+/− and WT dorsal root ganglia at these gestational days. (B) In situ hybridization on E11.5 transverse sections reveals that the expression of Brachyury (T) remains unaltered in Pbx1^−/−;Pbx2^+/− notochord (red inset in 20× (left) and 40× (right) panels) as shown at the level of the forelimb (FL). Similarly, at E11.5, Shh expression remains unaltered in the notochord and neural floor plate in Pbx1^−/−;Pbx2^+/− mutants as shown at both forelimb (FL) and hindlimb (HL) axial levels. Drg, dorsal root ganglion; E, edema; FP, Floor Plate; SC, spinal cord; V, vertebra. Dorsal (top).

Pbx1/2 mutant vertebral defects are coupled with shifts in hindlimb position along the embryonic axis

Interestingly, unlike WT and Pbx1^−/−;Pbx2^+/+ embryos, Pbx1/2 mutant embryos also displayed rostral shifts in hindlimb position (Fig. 5A and data not shown). Specifically, at E13.5, the pelvic girdle of Pbx1/2 mutant embryos was positioned laterally approximately 2–4 vertebrae caudal to the last rib-bearing thoracic vertebra. In contrast, in WT littermates, the pelvic girdle resided laterally, but approximately 5–6 vertebrae caudal to the last thoracic vertebra. Due to the early lethality at E13.5 of these mutants, before sacral induction occurs (Kaufman, 1992), it was not possible to ascertain the exact position of the pelvic girdle and hindlimb with respect to the lumbar-sacral boundary. However, in agreement with this rostral hindlimb shift, at earlier gestational days (e.g. E10.5) the position of the hindlimb bud was found to be shifted rostrally by approximately 1–3 somites (Table 1). Conversely, the position of the forelimb bud (and skeleton) of Pbx1/2 mutant embryos was found to be usually unchanged compared to WT, although on rare occasion a slight forelimb shift was also observed (Table 1 and Fig. 6A).

Figure 5. Hindlimb bud position is shifted in Pbx1^−/−;Pbx2^+/− embryos and is associated with alterations in posterior neurogenesis and Hox expression.

(A) Alcian Blue staining of E13.5 Pbx1^−/−;Pbx2^+/− and WT littermates reveals that with respect to the most caudal rib-bearing thoracic vertebra (black arrowheads), the mutant pelvic girdle (and hindlimb) is shifted rostrally by at least two vertebrae (dashed red-lines). (B) Earlier at E12.25, 2H3 antibody staining for neurofilament localization demonstrates that in Pbx1^−/−;Pbx2^+/− embryos hindlimb bud position is also rostrally shifted and, while forelimb bud position remains relatively stable, the distance (described by the white dashed lines) spanning both limb types is markedly reduced compared to WT. Neurofilament staining also demonstrates that despite similar staining in the neuroectoderm (long black arrows in mutant and WT), in the mutant neurogenesis becomes severely compromised more caudally along the embryo, specifically in the lateral and ventral extent of projecting neural fibers (short red arrows) when compared to WT patterns (short black arrows). Furthermore, the neuronal fibers that normally enter the hindlimb field, as seen in WT (black arrowheads), are not present in the mutant (empty black arrowheads) at this day. (C) In situ hybridization at E10.5 reveals that Hoxd9 expression is rostralized and up-regulated in the somites (red arrowheads) and LPM (blue arrow) in Pbx1^−/−;Pbx2^+/− embryos when compared to WT (black arrowheads and arrow, respectively). HL, hindlimb. Rostral (top)

Table 1.

Gene	Genotype1	Gest. Day2	FL range3	HL range4	Signal Start3
Hoxd3	WT	E10.5	7 to 11	24 to 29	3
	Pbx1−/−;Pbx2+/−	E10.5	7 to 11	22 to 27	1 to 2
Hoxb8	WT	E10.75	7 to 11	24 to 29	11
	Pbx1−/−;Pbx2+/−	E10.75	7 to 11	23 to 28	9
Hoxc6	WT	E10.75	7 to 11	24 to 29	11
	Pbx1−/−;Pbx2+/−	E10.75	6/7 to 11	22 to 27	9

¹3 mutant/WT pairs were observed for each day

²All embryos were somite-matched

³Values refer to the rostral-most level in somite position(s) in each embryo

⁴Values refer to the caudal-most level in somite position(s) in each embryo

Figure 6. Alterations in axial vertebral morphological identity and hindlimb bud position in Pbx1^−/−;Pbx2^+/− embryos are associated with rostral shifts in the expression of Hox and down-regulations of Polycomb group genes, Bmi/Eed, along the embryonic axis.

(A) When compared to WT, at E10.5–10.75, the rostral extent of expression of Hoxd3, Hoxb8, and Hoxc6 is shifted rostrally by approximately 1–3 somites in Pbx1^−/−;Pbx2^+/− embryos (compare somite numbers in WT and mutant) specifically with respect to the first somite (denoted by somite #1). Note: the Pbx1^−/−;Pbx2^+/− embryos shown for Hoxb8 and Hoxc6 display a slight anteriorization of the forelimb bud position. (B) When compared to WT (black arrowheads in all panels), in Pbx1^−/−;Pbx2^+/− embryos, the expression of Bmi is severely reduced at E10.5 along the entire posterior embryo (top; red arrowhead and empty black arrowheads) and specifically reduced, if not absent, in the somites adjacent to the hindlimb buds at E10.5 (below; empty black arrowheads) and at E11.5 (right; empty black arrowheads), though it remains expressed in the neuroectoderm (E10.5, below; black arrow). At these same gestational days, when compared to WT (black arrowheads), Eed expression (below and right) is also severely reduced posteriorly in the mutant (empty black arrowheads), and even from the neuroectoderm (E10.5, below; empty black arrow). FL, forelimb; HL, hindlimb; ov, otic vesicle. Rostral (top).

To ascertain if the position of the hindlimb bud was shifted with respect to their developing peripheral nerves, immunohistochemical staining with an anti-2H3 neurofilament antibody was performed at E11.5 (data not shown) and E12.25 (Fig. 5B). First, it was uncovered that the Pbx1/2 mutant hindlimb bud was rostrally shifted and the distance between forelimb and hindlimb buds was reduced compared to WT (Fig. 5B). Second, as detected by this staining, it was evident that neurogenesis was compromised in the more posterior domains of the Pbx1/2 mutant embryo. Indeed, developing peripheral nerves entering the WT hindlimb were not detectable in the Pbx1/2 mutant hindlimb bud (Fig. 5B).

Finally, as compound Hox9 paralogous mutants have been recently shown to possess posterior hindlimb shifts (McIntyre et al., 2007; Wellik, 2007), in situ hybridization with a Hoxd9 probe (as well as other Hox probes, see below) was performed to determine if alterations in its expression pattern were associated with the hindlimb rostralization in Pbx1/2 mutant embryos. Indeed, the anterior boundary of Hoxd9 expression was rostralized by at least three somites and up-regulated in the LPM in mutants, compared to somite-matched WT embryos (Fig. 5C).

Pbx1/2 mutant vertebral defects are associated with rostral shifts in the expression of Hox genes and down-regulation of Polycomb group genes, Bmi/Eed, along the embryonic axis

Given the reported roles of the spatio-temporal Hox activation (Dubrulle et al., 2001) on vertebral patterning (reviewed in Wellik, 2007) and hindlimb bud positioning (Cohn et al., 1997), additional whole-mount in situ hybridizations were performed on Pbx1/Pbx2 mutant and WT embryos using other Hox probes (Hoxa4, Hoxb5, Hoxb8, Hoxc6, and Hoxd3). For all Hox genes analyzed, in contrast to their WT expression, PM expression was shifted rostrally by 1–3 somites in somite-matched Pbx1/2 mutant embryos at E10.5–E10.75 (Fig. 6A displays representative results for Hoxd3, Hoxb8, and Hoxc6). These rostral shifts occurred independently of changes in neuroectodermal Hox expression, which remained relatively unperturbed in mutant embryos, and resembled the pattern described above for Hoxd9 (Fig. 5C). Overall, these findings are concurrent with the observed defects in vertebral morphology as well as the rostral shifts in hindlimb bud positioning (Table 1) in Pbx1/2 mutant embryos.

Recently, it has been uncovered that the loss of the repressive effects of Polycomb group protein function, as reported in Bmi and Eed single and compound mutant mice, leads to rostral shifts in Hox gene expression and alterations in vertebral identity and morphology (Kim et al., 2006). To ascertain if shifts in Hox gene expression in Pbx1/2 mutant embryos were concurrent with perturbations in either Bmi and/or Eed gene expression, in situ hybridizations with specific Polycomb probes were performed. Indeed, the expression of both Polycomb group genes was severely reduced to absent in Pbx1/2 mutant embryos at E10.5 and E11.5 (Fig. 6B). Specifically, at E10.5, despite their widespread expression in most embryonic tissues (Kim et al., 2006), which accounts for low-level background signal, both Bmi/Eed transcript levels were reduced in the somites of Pbx1/2 mutant embryos across the entire rostral-caudal embryonic axis and such reductions were more severe posteriorly (Fig. 6B and data not shown). The perturbation in the Bmi/Eed expression worsened at E11.5, as weaker and more diffused signal was detected along posteriorly located somites in Pbx1/2 mutant embryos (Fig. 6B and data not shown).

Pbx1/2 mutant axial skeletal defects are not associated with perturbed expression of notochord markers

To address potential roles of Hox genes in the notochord during early axial morphogenesis, in situ hybridization of selected Hox genes, such as Hoxb1, Hoxb2, Hoxc6, Hoxd9, as well as the Hox genes mentioned above, was performed on WT embryos from E9.5 to E12.5. Unlike for Hox gene expression in the zebrafish notochord (Prince et al., 1998), very weak to undetectable expression was observed for the above-mentioned Hox genes in the mouse notochord (data not shown). These data support the notion that an alteration of Hox gene expression in the PM, but not in the notochord, of Pbx1/2 mutant embryos underlies their vertebral defects.

Given the demonstrated roles of notochord signaling on somite differentiation and morphogenesis of the axial skeleton (Cleaver and Krieg, 2001; Pourquie et al., 1993; Watterson et al., 1954), in situ hybridization using notochord-specific markers such as Brachyury, Noggin, and Shh was performed on transverse sections of Pbx1/2 mutant and WT embryos at different axial levels (cervico-thoracic, thoracic-lumbar, and lumbar-sacral) from E11.5–E13.5. At all gestational days analyzed and at all axial levels, the expression of these notochord markers remained unperturbed in Pbx1/2 mutant embryos (Fig. 4B; Noggin data not shown). Furthermore, as previously reported (Capellini et al., 2006), Shh expression remained normal in the notochord in Pbx1/2 mutant embryos at E9.5–E10.5.

Pbx1/2 mutant axial skeletal defects are associated with perturbation of Pax1/Pax9 expression in early somites and sclerotome, but not with alterations in mesenchymal and cartilage cell differentiation

Recent research has demonstrated the roles of Pax1/Pax9 in somite formation, sclerotomal morphogenesis and differentiation, and development of the posterior vertebrate axial skeleton (Peters et al., 1999; Rodrigo et al., 2003). To ascertain whether the expression of Pax1/Pax9 was reduced during somite formation, whole mount in situ hybridization was performed on WT and late surviving Pbx1^−/−;Pbx2^−/− mutant embryos spanning gestational days E10–E10.75 (i.e., time-points just prior to their demise). Pbx1^−/−;Pbx2^−/− embryos had in general significantly fewer somites; and indeed in those somites that did form, the expression of Pax1/Pax9 was absent, specifically when compared to WT control and littermate embryos at earlier as well as similar gestational days (Fig. 7A). Interestingly, Pax9 remained expressed in the caudal pre-somitic mesoderm of these mutants despite the absence of both Pax9 and Pax1 in mature somites, indicating that the initiation steps in somite formation may have occurred normally (Fig. 7A). Detailed analyses of these events were hampered in Pbx1/Pbx2 double homozygous mutants by their early lethality and global malformations.

Figure 7. Alterations in vertebral morphogenesis are associated with down-regulations of Pax1 and Pax9 somitic mesodermal and sclerotomal expression in Pbx1^−/−;Pbx2^−/− and Pbx1^−/−;Pbx2^+/− embryos but not with changes in marker gene expression for mesenchymal and cartilage cell differentiation.

(A) In situ hybridization of Pax1 (top) and Pax9 (bottom) on WT and Pbx1^−/−;Pbx2^−/− embryos spanning gestational days E9–E10.75 demonstrates that the expression of both genes is absent from the developing somites of double knockout embryos. Pax1 expression is completely absent from the developing somites of double mutant embryos (2 embryos shown) unlike the WT expression pattern at several different gestational days. The signal present in the cranial region represents background trapping of the probe. Similarly, when compared to WT embryos at different gestational days, Pax9 is absent in the somites of double mutant embryos (2 embryos shown), yet low levels of expression are detected most caudally in their pre-somitic mesoderm. (B) In situ hybridization of Pax1 on WT and Pbx1^−/−;Pbx2^+/− embryos at E11.5 demonstrates that it is down-regulated in the sclerotome of the compound mutant. Specifically, on whole embryos (left), Pax1 expression is markedly reduced posteriorly in the mutant as its expression is reduced along the flank (red arrow) and forelimb (red arrowhead), absent from the proximal hindlimb bud (empty black arrowhead), and spatially restricted to caudal somites (black arrow) compared to the levels seen in WT embryos (black arrows and arrowheads, respectively). Additionally, section in situ hybridization at the axial levels of the forelimb (FL, red inset on right) and hindlimb (HL, blue inset on right) demonstrates that Pax1 is spatially diffused in the sclerotome (Scl) in Pbx1^−/−;Pbx2^+/− embryos. (C) Pax9 expression on whole mount E11.5 mutant embryos is reduced along the entire axis but not to the same extent as seen for Pax1; specifically, it is reduced in the anterior forelimb and hindlimb buds (short red arrows) and is down-regulated in the differentiating somites (long red arrow) of the mutant compared to the WT (black long and short arrows, respectively). (D) Sox9 and Col2a1, markers of mesenchymal condensation and cartilage cell differentiation, respectively, remain relatively unchanged between WT and Pbx1^−/−;Pbx2^+/− embryos as revealed by whole-mount (left) and section (right) in situ hybridization. By whole-mount, Sox9 (blue arrows) and Col2a1 (black arrows) are well-expressed in WT and Pbx1^−/−;Pbx2^+/− vertebral bodies at E12 as shown in lateral and dorsal views, although Sox9 is absent in areas where ribs should form in Pbx1^−/−;Pbx2^+/− embryos (red arrow) compared to WT (black arrow). By section at E13 (right, top), Sox9 is slightly reduced in the thoracic vertebral body mesenchyme of Pbx1^−/−;Pbx2^+/− embryos (gray arrowhead) compared to WT (black arrowhead), while strong and normal signals are detected for this gene in both thoracic and caudal neural arches (black arrows), and caudal body mesenchyme (black arrowheads) and notochords (red arrowheads). Also at E13, Col2a1 expression (right, bottom) remains unaltered in developing thoracic and caudal vertebral structures in Pbx1^−/−;Pbx2^+/− embryos compared to WT embryos, even in those embryos where neural tube defects are evident. Scl, sclerotome. In (A) WT Rostral (top), Pbx1^−/−;Pbx2^−/− (left); In (B–C) Rostral (top); In (D, whole-mounts) Rostral (top); In (D, sections) Dorsal (top).

To explore whether the expression of Pax1/Pax9 was altered in Pbx1/2 mutant embryos at later stages of sclerotomal morphogenesis, both whole-mount and section in situ hybridizations were performed at E11.5. Due to the observed axial shifts in hindlimb bud positioning in Pbx1/2 mutant embryos, section in situ hybridizations were performed on multiple axial levels from both WT and Pbx1/2 mutant embryos and equivalent axial levels were carefully compared. In whole-mount embryos (Fig. 7B–C), both Pax1 and Pax9 expression was down-regulated along the embryonic axis, although the latter was not as reduced as the former (Fig. 7C). Specifically, Pax1 expression was down-regulated in lumbar-sacral domains, as well as along the flank and in both fore- and hindlimbs (albeit more reduced in the anterior hindlimb than forelimb) (Fig. 7B). Interestingly, Pax1 expression was also spatially altered in the sclerotome of Pbx1/2 mutant embryos at all gestational days analyzed, as shown by section (Fig. 7B). The observed spatial perturbation in Pax1 expression became more diffused throughout the sclerotome at the lumbar-sacral (hindlimb) axial level (Fig. 7B, middle). At E11.5, the down-regulation of these genes occurred in the noticeable absence of edema, which only became evident after E12.5 in Pbx1/2 mutant embryos. Furthermore, analysis of additional genes such as Sox9 (Wright et al., 1995) and Col2a1 (Cheah et al., 1991) at E12 and E13 (Fig. 7D, right), revealed relatively normal patterns of expression in Pbx1/2 mutant embryos despite the presence of slight edema in their expression domains (see below). The observations that Pax1/Pax9 are down-regulated and spatially diffused in the sclerotome of Pbx1/2 mutant embryos sheds light on the genetic control of Pax by Pbx in this tissue; specifically since Pbx1/2 mutants, unlike Pbx1^−/−;Pbx2^−/− embryos, possess relatively normal somite numbers and morphology during mid-gestation. Finally, despite the perturbation in Pax1 and Pax9 spatial distribution, expression of Bapx1, a known target of these Pax genes in the axial mesoderm (Rodrigo et al., 2003), remained relatively unperturbed in Pbx1/2 mutant embryos (data not shown).

To ascertain if the observed skeletal defects in Pbx1/2 mutant embryos were associated with disruptions in mesenchyme of the sclerotome and cartilage cell differentiation in all domains of the forming vertebrae, additional whole-mount and section in situ hybridization were performed at E12 to E13 using Sox9 (Wright et al., 1995), Noggin (Valenzuela et al., 1995), and Col2a1 (Cheah et al., 1991), markers of this differentiation process (Fig. 7D and data not shown). By whole-mount in situ hybridization, as shown at E12, both Sox9 and Col2a1 remained similarly expressed in all vertebral domains in both WT and Pbx1/2 mutant embryos (Fig. 7D, left). Expression of Noggin also remained unchanged (data not shown). Noticeable differences were only evident for Sox9 in the pre-rib mesenchyme, as in Pbx1/2 mutant embryos its expression was absent, coincident with the presence of rudimentary ribs in these embryos. In order to further ascertain whether cellular differentiation occurred normally in Pbx1/2 mutant embryos in all vertebral domains (i.e., vertebral centra, laminae, and neural arches), in situ hybridization on multiple axial levels from both WT and Pbx1/2 mutant embryos was performed to avoid poor matching of sections due to the observed axial shifts in hindlimb bud positioning in the latter genotype. At both E12 and E13, only minor differences were observed in the expression of Sox9 in the anterior (thoracic), but not posterior (caudal) mesenchyme that gives rise to the vertebral centra, whereas Sox9 remained normally expressed in the notochord and all condensations that give rise to the lamina and neural arches (Fig. 7D, right). At these same gestational days, Noggin and Col2a1 remained similarly expressed in these domains in comparable sections of WT and Pbx1/2 mutant embryos (Fig. 7D, right and data not shown), even at those axial levels where dysmorphologies of the cartilaginous condensations were evident (see Col2a1 panel). Overall, mesenchymal and cartilage cellular differentiation, as assessed by these markers, remained relatively unaltered in Pbx1/2 mutant embryos.

Discussion

Pbx1/Pbx2 are required in the axial mesoderm for primaxial and abaxial skeletal patterning

Our results establish that, while Pbx1 is primarily required for patterning of the cervical vertebrae and the distal ribs (Selleri et al., 2001), Pbx1/Pbx2 share overlapping functions in the development of both the primaxial and abaxial skeleton. As for primaxial domains, when compared to Pbx1^−/− embryos, Pbx1/2 mutant embryos displayed more severe dysmorphologies in all vertebral compartments, as their vertebrae were hypoplastic and lacked many of the characteristic features of each specific axial region (Figs. 2–3). The vertebral dysmorphologies and the early lethality of Pbx1/2 mutant embryos made it difficult to discern if the demarcations between vertebral regions were properly maintained. However, it was evident that at transitions between vertebral segments strong similarities were present in adjacent vertebrae. For example, in Pbx1/2 mutant embryos there are at least eight vertebrae in the cervical region that display similar transverse and centra morphologies, while only seven are normally present in WT embryos (Fig. 2). In addition, C4 exhibited transverse and centrum morphologies similar to upper thoracic vertebrae, such as T3 (Fig. 3). Likewise, all lumbar-sacral vertebrae were morphologically similar and together with the rostralization of the hindlimb, these phenotypes may indicate a general shift in lumbar-sacral morphology (Figs 2, 3, and 5). Thoracic patterning and development were also severely compromised in Pbx1/2 mutant embryos, as cartilaginous rib rudiments persisted but were heavily truncated distally in abaxial domains (Figs. 2–3). Besides their anatomical location, these rudiments lacked any characteristics that would ascribe them to a specific rib number. In addition, sternal primordia were also absent in the Pbx1/2 mutant embryos (data not shown). When these data are taken together, the entire axial vertebral skeleton of Pbx1/2 mutant embryos may be considered homogeneous, indicating that its global patterning was disrupted in the absence of Pbx1/Pbx2.

As in the limb, except for Pbx1^−/− (Selleri et al., 2001), Pbx1/2 mutant, and Pbx1^−/−;Pbx2^−/− embryos (Capellini et al., 2006), all other compound Pbx1;Pbx2 genotypes lacked axial skeletal defects. Furthermore, only Pbx1/2 mutant embryos displayed significant alterations in gene expression (see above). These findings highlight the primary role of Pbx1, and the critical impact that overlapping Pbx1/Pbx2 spatio-temporal expression patterns have on axial development. Indeed, the early co-localization of Pbx1/Pbx2 in developing and differentiating somites and in the LPM may be critical in axial skeletal patterning (Fig. 1 and Suppl. Fig. 1). Our data indicate that, unlike the limb, where after initial bud formation Pbx1/Pbx2 occupy complementary domains (Capellini et al., 2006), Pbx1/Pbx2 are expressed in overlapping mesodermal (and mesodermal-derived) domains throughout successive stages of vertebral and rib morphogenesis (Fig. 1 and Suppl. Fig. 1). However, in the mesoderm during rostral to caudal patterning of the axial skeleton, Pbx1/Pbx2 function most likely early (i.e., during somite patterning) and reiteratively (see below), while later sclerotomal and cartilage cell differentiation are unaffected in the developing vertebrae of Pbx1/2 mutant embryos (Fig. 7D). Differences were only evident in the mesenchymal condensations of developing ribs, as shown by the loss of Sox9 expression in these domains in Pbx1/2 mutant embryos (Fig. 7D), not excluding a possible role of Pbx at later stages during rib morphogenesis and differentiation. Tissue-specific inactivation of Pbx1 on a Pbx2-deficient background will help clarify Pbx roles in these domains.

Finally, we have revealed that even though Pbx1/Pbx2 are expressed in the mouse notochord, their compound loss does not influence its morphological development or the expression of notochord-specific markers (Fig. 4). These data support our hypothesis that Pbx1/Pbx2 function in the mesoderm during primaxial and abaxial patterning.

Pbx roles in vertebral patterning are partially mediated by their genetic control of Polycomb group gene expression in the PM

We observed a striking down-regulation of the expression of Bmi/Eed in Pbx1/2 mutant embryos (Fig. 6B). The proteins that these genes encode function additively, in separate Polycomb complexes during the “maintenance phase” of Hox expression boundaries, to repress Hox gene transcription in somites rostral to the anterior limit of Hox expression (Kim et al., 2006). Indeed, single Bmi^−/−, and Eed^−/− mutants and several compound Bmi;Eed mutants display rostralizations of Hox gene expression domains, and exhibit homeotic transformations and alterations in vertebral patterning (Faust et al., 1995; 1998; Kim et al., 2006; Lessard et al., 1999; van der Lugt et al., 1994). Importantly, these patterning alterations were found across all axial levels and were dosage-dependent (Kim et al., 2006; Fig. 2). When taken in this context, Pbx may control Hox spatial distribution indirectly, via their hierarchical control of the Polycomb group genes. Accordingly, as shown in Figure 8 (A–B), as a consequence of Pbx loss, Bmi/Eed are down-regulated along the rostral to caudal axis in the Pbx1/2 mutant and their repressive effects relieved, leading to the rostralization of Hox expression boundaries during somite and sclerotomal differentiation.

Figure 8. Proposed model of Pbx1/Pbx2 function during vertebral patterning and morphogenesis.

(A–C) In all panels, black arrows, inhibitory lines, text, bars, and diagrammatic sketches represent wildtype conditions and the proposed genetic roles of Pbx, while all corresponding red signatures reflect events or processes that occur in the Pbx1^−/−;Pbx2^+/− (Pbx1/2 mutant) embryo. (A) Proposed genetic roles of Pbx1/Pbx2 during vertebral patterning. As represented in a transverse section at the early hindlimb bud (HL) axial level at E10.5 (blue inset), Pbx1/Pbx2 are normally expressed in the notochord (Nc) and developing somite (Som). The loss of Pbx from the Nc (red cross), as occurs in the Pbx1/2 mutant, does not alter signaling to the ventro-medial somite, as indicated by the black arrow. Conversely, the loss of Pbx1/Pbx2 from the somite causes perturbed vertebral patterning via their primary genetic control on the expression of Polycomb group genes (i.e., Bmi/Eed) and concomitant secondary control of Hox genes, although this latter relationship may also be indirect via a loss of Polycomb repression on Hox transcription (see (B) and text for more details). (B, left) The proposed genetic relationships between Pbx1/Pbx2, Bmi/Eed, Hox, and vertebral patterning in the paraxial mesoderm (PM) are also summarized in a cranial-caudal depiction of a developing embryonic axis at E10.5. According to the wildtype pattern (black bars overlying each gene name), Pbx1/Pbx2 and Bmi/Eed are normally expressed along the entire axis of the embryo, while Hox genes are expressed in nested and graded domains spanning cranial to caudal somites (only mid-thoracic to sacral somites are shown on the right). In the absence of Pbx1 (empty red bar) and partial absence of Pbx2 (shaded pink bar due to its heterozygous state in the Pbx1/2 mutant), Bmi/Eed expression is reduced (short red arrow) along the axis, most notably posteriorly (graded red to white bars), which may primarily lead to the vertebral patterning defects along the entire axis (red arrow). Concomitantly, loss of repression of Hox gene transcription and rostralization of Hox signal (red bars above black bars, spanning higher somite numbers, shown to the right) may also in part contribute to the etiology of these defects. In this diagram, the solid blue arrow indicates previously published relationships, whereas the solid red arrows depict roles revealed in this paper. (B, right) The changes identified in the PM are coincident with, but not necessarily causative of, the alterations in limb bud positions as occurring in the LPM. Specifically, in the absence of Pbx1/Pbx2 (red line diagram) and their control on Hox, HL are rostrally shifted by 2 or more somites compared to the wildtype pattern (black line diagram). (C) Proposed genetic roles of Pbx during sclerotomal formation and vertebral morphogenesis. As represented in the green inset of a transverse section at the HL bud axial level at E11.5/12, Pbx1/Pbx2 are normally expressed in the developing sclerotome (Scl) where they are upstream genetically of Pax1/Pax9. The loss of Pbx1/Pbx2 from this tissue (red cross) influences vertebral morphogenesis via perturbations of the expression of Pax1/Pax9, critical markers of the sclerotome (red cross). See text for details. D, dermomyotome; FL, forelimb; HL, hindlimb; LPM, lateral plate mesoderm; Nc, notochord; NT, neural tube; PM, paraxial mesoderm; Scl, sclerotome; So, somatopleure; Som, somite. In (A and C), Ventral (bottom) and Lateral (right); in (B), Rostral (top).

While Pbx genetically control Polycomb, the malformations in Pbx1/2 mutant embryos are more extreme than those observed in single and compound Polycomb or Hox mutants. Therefore, Pbx likely control additional factors in axial skeletal development. Importantly, Pbx proteins also act as cofactors for Hox members (Hox1–10), and by doing so they increase the DNA binding specificity and selectivity of Hox at target genes (reviewed by Mann and Affolter, 1998; Mann and Chan, 1996; Moens and Selleri, 2006). Accordingly, the loss of Pbx may result in an overall decrease in Hox functionality at all target genes across every region of the axial skeleton, likely leading to the overall loss of identity of all vertebrae.

As mentioned, we have also found that the rostral-most expression domains of all analyzed Hox genes representing different paralogous groups (Hox3, 4, 5, 6, 8, 9) spanning all four Hox clusters are shifted rostrally, typically by at least 2 somites in Pbx1/2 mutant embryos (Fig. 6A and 8B). Hox roles in primaxial (Burke et al., 1995; Iimura and Pourquie, 2006; 2007; Krumlauf, 1994) and abaxial patterning (McIntyre et al., 2007; Wellik, 2007) have been extensively documented. Knockout mice harboring null mutations for single and multiple paralogous Hox genes demonstrate that different paralogs pattern specific segments of the vertebral column and ribs and that partial functional redundancy exists between paralogous (and between non-paralogous) members (Iimura and Pourquie, 2007; Wellik, 2007). For example, compound mutants for Hox4 group paralogs display additive changes in cervical vertebral morphology dependent on Hox dosage (Horan et al., 1995), while in mice lacking Hox10 function (i.e., in which Hoxa10, Hoxc10, and Hoxd10 have been deleted), distinct lumbar vertebrae do not form (Wellik and Capecchi, 2003). Given that Pbx1/2 mutant embryos exhibit relatively similar vertebrae lacking axial-level characteristics indicative of their WT counterparts, it cannot be ruled out that the associated rostral shifts in axial Hox expression contribute, at least in part, to the etiology of these defects (Fig. 8A–B). Overall, Pbx genes may function reiteratively throughout axial development by genetically controlling Polycomb group and Hox gene expression, as well as by acting as Hox cofactors (Fig. 8A–B).

Pbx roles in the LPM govern hindlimb positioning

Little is known about the mechanisms by which limb bud position is determined, although evidence for Hox roles in limb bud positional specification comes from several single and compound Hox mouse mutants. For example, Rancourt et al. (1995) demonstrated that the axial level of scapula formation was shifted in single Hoxb5^−/− and compound Hoxb5;Hoxb6 mutant mice suggesting that forelimb level may also have been altered. McIntyre et al. (2007), on the other hand, found a posterior shift in the axial skeleton of compound homozygous Hox9 paralogous mutants (i.e., for Hoxa9/b9/c9/d9), which placed hindlimbs more caudally than in WT controls. Furthermore, in the chick embryo, Cohn et al. (1997) demonstrated by experimentally altering the expression boundaries of Hox9 genes (i.e., Hoxb9, Hoxc9, and Hoxd9) that ectopic limbs formed at different axial levels, although in these experiments shifts in the axial level where limbs normally form were not observed.

Interestingly, associated with Pbx1/2 mutant primaxial and abaxial defects, we found that hindlimb position was rostralized by at least two somites at early stages (e.g., E10.5; Table 1) and by two to three vertebrae at later stages (E13.5; Fig. 5). While we previously reported that Pbx genes control Hox expression in the limb, a derivative of the LPM (Capellini et al., 2006), we now report here that this hierarchical relationship also exists for axial mesodermal tissues and the LPM proper. For example, we found that the loss of Pbx1/Pbx2 is associated with a rostralization of Hoxd9 expression in both somites and the LPM (Fig. 5C); and consistent with this potential gain of function, we found that hindlimb location was also rostralized (Fig. 5A–B). This finding is consistent with the posterior shifts in hindlimb bud position described above for Hox9 loss-of-function mutants. In this regard, the basis for determining hindlimb position may reflect the actions of Pbx on Hox in the LPM prior to limb induction (see Fig. 8B). At this time, however, Pbx may act upstream of other parallel pathways driving limb positioning. For example, after careful re-examination of the reported skeletal defects of single and compound mutants for Bmi/Eed (reported in Kim et al., 2006; Fig. 3G), we noticed subtle but noteworthy rostral shifts in hindlimb position. Specifically, despite the posterior transformation of L6 to S1, the location of the sacral attachment of the iliac blades in Bmi^−/−, Eed^−/−, and Bmi^−/−;Eed^−/− mutants occurred more rostrally, at the level of L5, rather than at L6 as in the WT. While there is a discrepancy in the degree of hindlimb rostralization between these sets of mutants, at least two points emerge whose consideration may help clarify these differences, as well as the roles of these genes during axial and limb specification: (1) Kim et al. (2006) reported that the Eed mutant allele (l7Rn5^1989SB/+) used to generate Eed^−/− and Bmi^−/−;Eed^−/− compound mutants was hypomorphic and when present in its homozygous state led to an ameliorated phenotype compared to a true null allele (l7Rn5^3345SB/+); and (2) The Pbx1/2 mutant model used in this paper is heterozygous for Pbx2, which results in residual Pbx2 protein and thus implies that the genetic control of Pbx on these Polycomb genes is only partially eliminated in Pbx1/2 mutant embryos.

Pax1/Pax9 mediate, in part, Pbx1/Pbx2 roles in the sclerotome and somites, as well as in other mesodermal-derived tissues

Early stages in primaxial development involve the differentiation of the somitic mesoderm into sclerotomal mesenchyme and the maturation of sclerotome into vertebral elements. These processes are mediated by at least two members of the Pax family of transcription factors, Pax1 and Pax9 (Peters et al., 1999; Wallin et al., 1994; Wilm et al., 1998). The functional relationships among these genes during vertebral development is complicated by the finding that in Pax1 absence, Pax9 expression is up-regulated and spatially expanded in the sclerotome (Peters et al., 1999), partially rescuing the Pax1 null vertebral phenotype (Wallin et al., 1994; Wilm et al., 1998). However, in Pax1^−/−;Pax9^−/− mutants more severe defects in centrum and intervertebral disc morphologies are observed across lumbar, sacral, and caudal domains of the vertebral column.

In Pbx1/2 mutant embryos we observed a modest down-regulation and marked spatial perturbation (diffusion) of Pax1 expression in the sclerotome at all axial levels (Fig. 7B). This pattern was not duplicated for Pax9, which while still slightly reduced, persisted in all vertebral axial domains (Fig. 7C). In accordance with Peters et al. (1999), the slight down-regulation of Pax9 that we observed may reflect the up-regulation of this gene’s expression in the sclerotome after Pax1 down-regulation. Alternatively, the relatively higher Pax9 expression in Pbx1/2 mutant embryos (compared to Pax1 expression) may result from differences in the regulation of Pax1 versus Pax9 by Pbx. Our findings that both Pax1 and Pax9 are down-regulated in other non-axial domains in Pbx1/2 mutant embryos and absent in the somites of Pbx1^−/−;Pbx2^−/− mutants (see Fig. 7A–C) further support our hypothesis of a genetic control of Pax1 and Pax9 by Pbx family members (see Fig. 8C). These findings are also consistent with previous reports showing that Pbx genes can regulate other Pax family members during embryogenesis, e.g., during thymic development (Manley et al., 2004) and pancreatic islet cell differentiation (Zhang et al., 2006).

Interestingly, from E9.5 to E13.5, we found that the ventro-medial sclerotome in Pbx1/2 mutant embryos exhibited significantly lower cellular density, a difference that was not coupled with changes in cell proliferation or apoptosis (Fig. 4A and Suppl. Fig. 3). Conversely, in Pax1^−/−;Pax9^−/− mutants, similar changes in the density of the sclerotome were reported, but were associated with decreased cell proliferation (E12.5) and increased apoptosis (at E13.5) (Peters et al., 1999). In addition, Rodrigo et al. (2003) discovered that Bapx1 (a sclerotomal marker) may be a direct transcriptional target of Pax1/Pax9, since its expression was lost in compound Pax1;Pax9 mutants. In contrast, we found that Bapx1 expression remained normal in Pbx1/2 mutant embryos. These discrepancies may be attributed to our findings that: (1) in Pbx1/2 mutant embryos, Pax1 and Pax9 are not completely absent but modestly down-regulated and spatially perturbed to different degrees; and (2) Pbx1/Pbx2 appear to govern early processes that may pleiotropically influence later morphogenetic events, including cellular density in the developing somites (Fig. 8A; see below).

The genetic control of Pax1 and Pax9 by Pbx1/Pbx2 may also be influenced by Hox themselves. Activation of Hox target genes may be prevented by the loss of their cooperative interactions with Pbx and/or Pax. With regard to the latter, it has been reported that Pax and Hox can form complexes to regulate common target genes (Benassayag et al., 2003; Gong et al. 2007; Plaza et al., 2001). Specifically, Gong et al. (2007) have recently uncovered that the metanephric expression of Six2 and Gdnf requires a complex involving Hox11, Eya1, and Pax2. Furthermore, both Hox and Pax have been shown to physically and genetically interact and to cooperatively regulate downstream target genes in Drosophila organogenesis (Benassayag et al., 2003; Plaza et al., 2001). Thus, many of the observed vertebral defects, arising from altered somite and sclerotomal formation in Pbx1/2 mutant embryos may result from disruptions in Pax and Hox and their cooperative effects on target gene expression (Fig. 8C).

Summary: Hierarchical, reiterative, and overlapping functions of Pbx genes in multiple developmental processes

We have shown that Pbx1/Pbx2 have overlapping hierarchical genetic roles in the axial mesoderm that gives rise to the skeletal elements of the vertebral column, as well as ribs, sternum, and clavicle. In the context of the appendicular and girdle defects reported here and in Capellini et al. (2006), these results additionally establish that Pbx1/Pbx2 genetically regulate the development of all post-cranial skeletal structures. However, Pbx genetic control is not limited to skeletal development, as in domains where Pbx1/Pbx2 are lost, posterior neurogenesis is severely compromised (Fig. 5B). Indeed, a growing body of literature demonstrates that Pbx family members act as hierarchical regulators of many organogenetic processes including, for example, development of hematopoietic (DiMartino et al., 2001; Sanyal et al., 2007), thymic (Manley et al., 2004), splenic (Brendolan et al., 2005), pancreatic (Kim et al., 2002; Zhang et al., 2006), nephrogenic (Schnabel et al., 2003), and adrenal and urogenital (Lichtenauer et al., 2007; Schnabel et al., 2003) systems. In many of these contexts, Pbx has been shown to act as a direct regulator of target gene expression, a control often requiring cooperative interactions with other TALE family members, such as Meis and Prep, as well as Hox proteins. Given the co-localization of many of these partner proteins with Pbx1–3 (Fig. 1 and Suppl. Figs. 1–2), it is likely that the differential regulation of target genes by Pbx requires the formation of context-specific complexes, a process that at this stage is poorly understood. As Pbx, Meis, and Prep global and tissue-specific knockout mice become available, the mechanisms and molecular networks hierarchically regulated by Pbx in many different developmental processes will be clarified.

Supplementary Material

01

Supplementary Figure 1. Expression of Pbx genes in the developing notochord from E10.5 to E12.5. Expression of individual Pbx genes at each gestational day is shown in a 10× panel (above) and a 40× red inset (below). (A–C) Pbx1 is strongly expressed (black arrowheads, red insets) in the developing notochord throughout mid-gestation. (D–F) At these time points, Pbx2, which is diffusely expressed in the embryonic mesenchyme, is also expressed in the developing notochord (grey arrowheads, red insets), albeit at lower levels when compared to Pbx1. (G–I) Pbx3 expression levels in the notochord (dashed circle) and surrounding mesenchyme are either extremely low or undetectable (empty black arrowheads, red insets) during these gestational days. (A–I) All three Pbx genes are also expressed in the ventral columns of the spinal cord as shown in the 10× view. Representative mid-thoracic transverse levels are shown. SC, Spinal cord. Dorsal (top).

02

Supplementary Figure 2. Expression of Meis genes in the developing notochord from E10.5 to E12.5. Expression of individual Meis genes at each gestational day is shown in a 10× panel (above) and a 40× red inset (below). (A–C) Meis1 and (D–F) Meis2 are expressed (grey arrowheads) in the developing notochord and surrounding mesenchyme at low levels as compared to their expression levels in the spinal cord, and (G–I) Meis3 expression levels in the notochord (dashed circle; empty black arrowheads, red insets) and surrounding mesenchyme are either extremely low or undetectable during these gestational days. Representative mid-thoracic transverse levels are shown. SC, Spinal cord. Dorsal (top).

03

Supplementary Figure 3. Unperturbed cellular proliferation and programmed cell death in Pbx1^−/−;Pbx2^+/− notochord and vertebral anlage. Unless otherwise noted, all images are transverse sections at the mid-thoracic level and are shown in 10× (upper) panels and 40× (lower) insets. (A) Hematoxylin and Eosin staining of E13.5 transverse sections of Pbx1^−/−;Pbx2^+/− and WT embryos does not reveal significant differences in the structure and cellular morphology of the notochord but does reveal that the cellular density of vertebral anlagen is reduced compared to WT in thoracic and caudal domains of the embryo (black arrowheads depict the notochord). Edema is evident between the spinal cord and the vertebral anlagen of Pbx1^−/−;Pbx2^+/− embryos. (B) Cell proliferation as measured by BrdU incorporation is not different in Pbx1^−/−;Pbx2^+/− notochord and vertebral anlage compared to that present in WT at E11.5 and E12.5 (black arrowheads, insets). (C) Levels of cellular apoptosis, as assayed by Caspase3 antibody staining, do not differ between Pbx1^−/−;Pbx2^+/− and WT notochord and vertebral anlage at E11.5 and E12.5 (black arrowheads, insets). In contrast, higher levels of cellular apoptosis are equally present in Pbx1^−/−;Pbx2^+/− and WT dorsal root ganglia at these gestational days. Drg, dorsal root ganglion; E, edema; SC, spinal cord; V, vertebra. Dorsal (top).