δ-EF1 is a negative regulator of Ihh in the developing growth plate

Endochondral ossification is a multistep process in which prechondrogenic mesenchymal cells condensate and form distinct cartilaginous elements. Subsequently, this cartilaginous template forms a growth plate necessary for longitudinal growth. During development, the growth plate stratifies into zones of resting, proliferating, prehypertrophic, and hypertrophic chondrocytes, forming a highly interactive structure where each zone is influenced by events in the flanking zones and the perichondrium (Kronenberg, 2003).

Initiation of chondrogenesis begins with induction of the transcription factor Sox9 in limb mesenchyme and subsequent condensation of mesenchymal cells (Bi et al., 1999). Next, the condensed cells differentiate into chondrocytes that, among others, express transcription factors Sox5 and -6 (Smits et al., 2004) as well as matrix protein collagen type II. As the chondroblasts mature, they flatten and align in organized columnar structures along the longitudinal axis of the cartilage template, forming a transient pool of prehypertrophic cells. Finally, these cells exit the cell cycle, enlarge in size, and undergo hypertrophy. At this stage, cells in the perichondrium differentiate into osteoblasts that deposit an extracellular matrix, which will mineralize to form the bone collar. Hypertrophic chondrocytes express collagen type X and produce VEGF that stimulates blood vessel invasion in the calcified cartilage (Provot and Schipani, 2005). Invading osteoclasts will resorb the cartilage matrix, and osteoblasts will deposit bone matrix on the cartilage remnants (Kronenberg, 2003; Provot and Schipani, 2005).

Several factors regulate the process of endochondral bone formation, and one of the more prominent ones is a member of the conserved hedgehog (Hh) family of secreted proteins, Indian hedgehog (IHH; Bitgood and McMahon, 1995; Vortkamp et al., 1996; Hammerschmidt et al., 1997). Hh proteins transduce their signal by binding to a 12-pass transmembrane receptor protein Patched (Ptc). In the absence of Hh, Ptc represses the activation of another multipass transmembrane protein, smoothened (smo; Chen and Struhl, 1996; Marigo et al., 1996; Stone et al., 1996). Binding of Hh ligands to Ptc relieves the repression of smo, activating the Hh signal transduction pathway that is mediated by Gli transcription factors (Gli1, Gli2, and Gli3). Although Gli1 functions as an activator, Gli2 and -3 can act either as transcriptional activators or repressors depending on the cellular context (Matise and Joyner, 1999; Mullor et al., 2002; Cohen, 2003; Nieuwenhuis and Hui, 2005).

Ihh is expressed in prehypertrophic and hypertrophic chondrocytes controlling several important differentiation steps in the development of the endochondral skeleton. Ihh regulates chondrocyte differentiation by activating the expression of parathyroid hormone–related protein (PTHrP; Pthlh) in the periarticular cells. PTHrP is a paracrine signaling molecule that diffuses toward the prehypertrophic zone and signals through its receptor, PTHrP receptor (PPR), to inhibit hypertrophic differentiation, ensuring a supply of proliferating chondrocytes essential for the formation of hypertrophic cells and thus longitudinal growth (Vortkamp et al., 1996; Lanske et al., 1999; Karp et al., 2000). Ihh also regulates chondrocyte proliferation independently of PTHrP. Analysis of Ihh-null mice expressing a constitutively active mutant of PPR showed that the proliferation defect in Ihh-null mice was independent of PTHrP signaling (St-Jacques et al., 1999; Karp et al., 2000; Long et al., 2001). Furthermore, Ihh also regulates ossification of the perichondrium, where it is required for the initial specification of osteoblast progenitor cells but not for the formation of mature osteoblasts (Rodda and McMahon, 2006). Besides its function in osteo- and chondrogenic differentiation, Ihh is also involved in the endothelial cell fate determination and blood vessel formation because Ihh^−/− blood vessels do invade the mutant hypertrophic cartilage but cannot persist (Colnot et al., 2005).

Even though Ihh is one of the key regulators of endochondral ossification, surprisingly little is known about the mechanisms by which the expression of Ihh is regulated. It has been reported that one of the key transcriptional regulators of endochondral bone formation, Runx2, is activating the Ihh promoter in vitro (Yoshida et al., 2004). Ihh expression could also be a target of Wnt signaling because a direct interaction between the Ihh promoter and the β-catenin–Lef1 complex in vivo has been demonstrated (Später et al., 2006).

A precise regulation of Ihh is important for endochondral bone formation, as shown by various skeletal phenotypes induced by modulation of Ihh gene expression. For example, Ihh-null mice exhibit short-limb dwarfism with reduced chondrocyte proliferation and extensive hypertrophy (St-Jacques et al., 1999). Ihh overexpression, however, leads to delayed chondrocyte differentiation, delayed ossification of the long bones, and expansion of the growth plate with increased proliferation (Vortkamp et al., 1996; Kobayashi et al., 2005). Additionally, misexpression of Hh proteins in the cartilage leads to joint fusions (Tavella et al., 2004).

Interestingly, the described phenotypes of Ihh overexpression are very similar to skeletal abnormalities observed in δ-EF1–null mice (Takagi et al., 1998). Targeted inactivation of the δ-EF1 gene leads to craniofacial abnormalities, sternum, rib, and limb defects such as shortening and broadening of long bones, and joint and carpal/tarsal bone fusions (Takagi et al., 1998).

δ-EF1 is a two-handed zinc finger transcription factor (Funahashi et al., 1993). The gene has been described independently as Zfxh1a in mice, ZEB/AREB6/Nil-2-a in humans, BZP in hamsters, and Zfhep in rats (Williams et al., 1991; Watanabe et al., 1993; Franklin et al., 1994; Genetta et al., 1994; Sekido et al., 1996; Cabanillas and Darling, 1996). δ-EF1 binds DNA as a monomer to bipartite CACCT(G) sequences separated by 8–60 nucleotides and located near (±2 kb) the promoter start site. δ-EF1 has been proposed to be a transcriptional repressor–binding E2 box containing sequences and negatively regulating muscle differentiation in vitro (Postigo and Dean, 1997), integrin expression in myoblasts (Jethanandani and Kramer, 2005), and vascular smooth muscle cell differentiation (Nishimura et al., 2006). Additionally, the repressor function of δ-EF1 in vitro has been shown for Col2a1 promoter activity in chondrocytes and for Pro-α1-collagen gene expression in osteoblasts (Murray et al., 2000; Terraz et al., 2001). δ-EF1 is expressed in the lens, the central nervous system, skeletal precursors, the notochord, the myotome, and the limb bud mesenchyme (Takagi et al., 1998; Tylzanowski et al., 2003). Postnatally, δ-EF1 protein expression was detected in the growth plate and articular and meniscal cartilage of mature mice (Davies et al., 2002).

In the present study, we identified δ-EF1 as a candidate regulator gene of Ihh expression in the growth plate of mice. We show that δ-EF1 binds to the Ihh intron 1 transcription regulatory element in vitro and in vivo and that absence of δ-EF1 leads to temporal deregulation of Ihh in developing limbs. Finally we demonstrate that haploinsufficiency for δ-EF1 associated with enhanced Ihh expression leads to increased bone mass in adult mice.

Detailed description of the developing bones of δ-EF1^−/− mice has not been published to date. To better understand the observed skeletal phenotype, we performed histological analysis of limbs from δ-EF1–null mice at different stages of development.

At 12.5 days postcoitum (dpc), no significant differences were detected between mutant and wild-type (WT) limbs (unpublished data). The first signs of calcification in the WT humerus were visible at 13.5 dpc with the formation of a bone collar, whereas in δ-EF1^−/− mice, this process was delayed (unpublished data). At 14.5 dpc, limbs of δ-EF1^−/− mice were shorter than limbs of WT mice (Fig. 1, A and B). Histological analysis revealed that the length of the hypertrophic zone was reduced in δ-EF1^−/− mice as compared with WT mice (Fig. 1, C and D, compare red and black bars; and Fig. 1, I, J, and K, which shows the quantification of the different cartilage zones), whereas the zones of resting and proliferating chondrocytes were enlarged (Fig. 1, E, F, and K), but there was no apparent difference in columnar organization (Fig. 1, G and H). At 16.5 dpc, the delay in hypertrophic differentiation became more pronounced (Fig. 1, A′–K′) and, in addition, we observed defects in columnar cell organization with fewer stacks of cells (Fig. 1, G′ and H′, arrowheads). This phenotype became more pronounced at 18.5 dpc, where a population of enlarged, oval chondrocytes was present instead of stacks of flat-shaped chondrocytes (Fig. 1, G″ and H″, arrowheads).

Chondrocyte hypertrophy was further analyzed by in situ hybridization with probes for Col2a1 and Col10a1 mRNA (markers for nonhypertrophic and hypertrophic chondrocytes, respectively). The Col2a1 mRNA expression domain was considerably longer in δ-EF1–null mice as compared with WT littermates (Fig. 2, A and B, compare red and black lines). Conversely, the Col10a1 mRNA expression domain was smaller in δ-EF1^−/− mice (Fig. 2, C and D, compare red and black lines), which suggests a delay in the onset of hypertrophic differentiation in the δ-EF1–null embryos. The expression domain of Col1a1 mRNA (an ossification marker) was reduced (Fig. 2, E and F, compare red and black lines), and mineralization was delayed as visualized by von Kossa staining (Fig. 2, G and H). The osteoblast precursors were, however, present in the perichondrium, as indicated by alkaline phosphatase staining (unpublished data).

The defects in chondrocyte maturation prompted us to examine the expression of genes associated with Ihh signal transduction. We analyzed the mRNA's expression of Ihh and its receptor and target Ptc, as well as PTHrP and PPR. Ihh mRNA expression was maintained in the prehypertrophic and hypertrophic zones but had a broader expression domain as compared with WT littermates (Fig. 2, I and J, compare red and black arrows). Ptc mRNA was detected in the perichondrium adjacent to the expression domain of Ihh and in the zone of proliferating chondrocytes. Interestingly, in δ-EF1^−/− mice, the domain of Ptc mRNA expression was extended toward the epiphyses (Fig. 2, K and L). The expression of PPR was broadened in δ-EF1^−/− limbs as compared with WTs (Fig. 2, M and N). PTHrP expression was detected in round chondrocytes at the ends of long bones of the WT limbs (Hilton et al., 2007), but there was a marked increase in PTHrP expression around the epiphyses of the long bones of δ-EF1^−/− mice (Fig. 2, O and P, arrows). The changes in the expression pattern of the abovementioned molecular markers were the same in all long bones examined (unpublished data). In summary, δ-EF1 deficiency resulted in a delay of chondrocyte hypertrophic differentiation.

Because δ-EF1–null mice have apparent defects in the growth plates of the long bones, we decided to reexamine δ-EF1 transcripts during bone development. Previous studies showed δ-EF1 mRNA expressed in the perichondrium of long bones at 14.5 dpc (Takagi et al., 1998; Tylzanowski et al., 2003). Our analysis has revealed additional expression domains of δ-EF1 mRNA in the growth plate on sections of 14.5 dpc (Fig. 3) and 16.5 dpc (not depicted) mouse embryos using two independent in situ hybridization methods: colorimetric (Fig. 3, A and B) and fluorescent (Fig. 3, C and D). δ-EF1 transcripts were detected in the entire growth plate (Fig. 3, A and C, red arrows) as well as in some muscles (Fig. 3 A, black arrow; and Fig. 3 C, white arrow; Funahashi et al., 1993). We did not detect expression of δ-EF1 mRNA in limbs of δ-EF1^−/− mice (Fig. 3, B and D). The expression of δ-EF1 mRNA became weaker toward the zone of hypertrophic chondrocytes (Fig. 3 A). These findings indicate that δ-EF1 mRNA is located at the appropriate place and time during bone development in mice to regulate growth plate differentiation.

δ-EF1^−/− mice display an enlarged zone of resting and proliferating chondrocytes. Because this could have been caused by enhanced proliferation, we analyzed the expression of Ki67, an acknowledged molecular marker of that process. There was a significant increase in the number of cells positive for Ki67 in the proliferative zone (both in noncolumnar and columnar cells) of δ-EF1^−/− mice as compared with WT littermates (Fig. 4, A–B′) at 16.5 dpc. The number of proliferating cells in the humerus of the mutant was ∼50% higher than in WT mice (Fig. 4 C). Thus, in the absence of δ-EF1, the balance between proliferation and differentiation of chondrocytes was changed with a higher rate of chondrocyte proliferation.

Having demonstrated that δ-EF1–null mice show an increased cell proliferation in the growth plate, we decided to analyze the expression of Ihh during chondrogenesis in detail. In WT mice at 15.5 dpc, IHH protein was detected in prehypertrophic and hypertrophic chondrocytes, and no IHH-positive cells were detected in the domains of resting and proliferating chondrocytes (Fig. 5, A and A′; Koziel et al., 2004). In δ-EF1^−/− mice, however, a significant amount of IHH-positive cells was detected in the proliferating region of δ-EF1^−/− mice, and IHH-positive cells were detected in the entire growth plate (Fig. 5, B and B′). Similar results were obtained at 14.5 and 16.5 dpc (unpublished data). To analyze the dynamics of IHH and δ-EF1 protein expression during limb development, we performed Western blot analysis (Fig. 5 C). δ-EF1 was detected from 12.5 dpc, with the expression peak at 14.5 dpc. From 16.5 dpc, δ-EF1 expression could no longer be detected. IHH protein expression was detected from 12.5 dpc onwards, and the expression increased over time. IHH expression in δ-EF1^−/− limbs was higher as compared with the corresponding stages of WT limbs (Fig. 5 C′, * and ^), confirming the immunohistochemistry data. Quantification of the Western blot bands showed a twofold increase in Ihh expression in δ-EF1^−/− limbs at 13.5 dpc as compared with WT, and a 1.5-fold increase at 18.5 dpc (Fig. 5 C′).

Because there was a limited expansion of the Ihh mRNA expression domain (as shown in Fig. 2), we measured Ihh mRNA expression by quantitative PCR. Ihh mRNA expression in the developing limbs was increased in δ-EF1^−/− mice at all stages from 12.5 dpc onwards, with a peak at 14.5 dpc (Fig. 5 D). These results show that even though the expression domain of Ihh mRNA was only slightly expanded, transcription levels of Ihh were increased. Importantly, a similar increase in expression of Ihh target gene Ptc was detected in δ-EF1^−/− limbs (Fig. 5 E), which supports the notion that the increase in Ihh transcription may have biological consequences in vivo.

Ihh has both PTHrP-dependent and independent effects on growth plate development. To determine if the defects seen in the growth plates of δ-EF1–null mice were PTHrP mediated, we analyzed the expression of PTHrP mRNA. Similar to Ihh and Ptc, the mRNA expression of PTHrP was increased in δ-EF1^−/− limbs at all stages (Fig. 5 F) as compared with WT controls.

In conclusion, deficiency of δ-EF1 leads to higher levels of Ihh mRNA expression and increased IHH protein levels.

Because the δ-EF1–null phenotype in mice was reminiscent of phenotypes associated with perturbed IHH signaling, we investigated if δ-EF1 could bind directly to regulatory regions of the Ihh gene. Indeed, we found four putative δ-EF1 binding sites in a region of 160 bp at the 5′ end of intron 1 of the mouse Ihh gene (Fig. 6 A).

One of the indications of a potential functional significance of DNA sequences located in noncoding regions of DNA is their evolutionary conservation across species. Therefore, we aligned the genomic sequence of the mouse Ihh to rat, human, dog, cow, chimp, and chicken Ihh sequences using VISTA software (Dubchak et al., 2000; Frazer et al., 2004). We found that the 160-bp region in the Ihh gene containing putative δ-EF1 binding sites was highly conserved across species (Fig. 6 B, black box).

Next, we investigated if δ-EF1 could bind to these sequences in vitro carrying out electrophoretic mobility shift assay (EMSA) using amino-terminally tagged myc–δ-EF1 and a 65-bp fragment of the mouse Ihh intron 1 containing CACCT/CACCTG sites as a probe (see Fig. 6 C and Materials and methods). We could demonstrate that myc–δ-EF1 could bind to the intron 1 fragment of Ihh (Fig. 6 D). Next, we confirmed the specificity of myc–δ-EF1 binding by carrying out a supershift of the δ-EF1 band with a monoclonal anti-myc antibody (Fig. 6 D, lanes 2 and 3). To verify that the binding was dependent on the presence of the intact bipartite CACCT…CACCTG motif, we performed competition assays. As can be seen on Fig. 6 D (lane 4), the WT Ihh fragment competed for δ-EF1 binding with 20-fold the molar excess of the WT, unlabeled, probe. The CATCT…CATCTG mutant (CACCT(G) binding sites are underlined, point mutations in mutated constructs are shown in bold; Fig. 6 D, lane 5) was unable to compete for binding to δ-EF1.

Collectively, the results in Fig. 6 C show that δ-EF1 can bind to the Ihh intron 1 fragment in vitro and that this binding is dependent on the presence of an intact bipartite CACCT/CACCTG binding site.

Although EMSA provides the information about the ability of a transcription factor to bind to its target sequence, it does not indicate if such a binding takes place in vivo and is therefore biologically relevant. To address this issue, we have performed chromatin immunoprecipitation (ChIP) experiments. We have designed PCR primers encompassing the 160-bp region containing δ-EF1 putative binding sites in the Ihh intron 1 and performed ChIP analysis on 16.5-dpc mouse embryos. After protein cross-linking to DNA, we performed immunoprecipitation using δ-EF1–specific antibodies (ZEB; Eger et al., 2005). As can be seen in Fig. 6 E, we were able to amplify the Ihh intron 1 fragment in the DNA templates obtained after immunoprecipitation with anti-δ-EF1 (ZEB) antibodies but not from the IgG controls. Importantly, we did not detect any PCR product in the lysate obtained from δ-EF1–null mice. As seen in the lower panels of Fig. 6 E, we did not detect the presence of unrelated genomic sequences (noggin and Sip1), which ruled out the possibility that the chromatin containing Ihh intron 1 was pulled down nonspecifically with δ-EF1 containing chromatin. These results strongly support the notion that intact δ-EF1 binds in vivo to cognate target sequences present in intron 1 of Ihh.

We demonstrated that binding of δ-EF1 to Ihh intron 1 takes place both in vitro and in vivo; however, whether this binding would repress or activate gene transcription remained unresolved. Therefore, we performed gain-of-function experiments by misexpressing myc–δ-EF1 cDNA in developing chick limbs. Chick limbs of Hamburger-Hamilton (HH) stage 18–20 were electroporated with myc–δ-EF1 in a pCS3 expression construct and incubated until HH stage 36–37, when a skeletal staining was performed. Limbs electroporated with δ-EF1 construct were reduced in size as compared with the contralateral control limbs or limbs electroporated with empty expression vector (Fig. 7, A and A′; and not depicted). The length of the long bones was reduced by a mean of 25% in the zeugopod and by a mean of 40% in the stylopod, as compared with control limbs electroporated with pCS3 alone at the same developmental stage. Additionally, in some limbs, a small reduction in length of the autopod was detected (unpublished data).

To evaluate if the δ-EF1 cDNA was successfully expressed in the chick limb, we performed a Western blot analysis on protein samples obtained from electroporated chick limbs using anti-myc primary antibody. Western blot analysis of the cell lysate of transfected limbs, as well as DF1 cells transfected with the same expression construct, revealed a band of 140 kD. Importantly, we could detect δ-EF1 in the cell lysates from limbs that did show the phenotype but could not detect it in the limbs electroporated with δ-EF1 expression construct but not having the above described phenotype (Fig. 7 D), which supports the notion that the observed phenotype was directly correlated with the successful expression of δ-EF1.

Histologically, the limbs electroporated with δ-EF1 cDNA showed a significant reduction of the zone of hypertrophic chondrocytes, and bone formation was delayed (Fig. 7, A, A′, B, and B′). Furthermore, the distance between the epiphysis and the zone of hypertrophic chondrocytes, as indicated by Col10a1 mRNA expression, is enlarged in limbs electroporated with δ-EF1 cDNA (Fig. 7, C and C′). These observations suggested a delayed onset of hypertrophic and osteogenic differentiation and therefore a possible defect in Ihh signaling induced by misexpression of δ-EF1. To investigate this hypothesis, we performed quantitative PCR to measure Ihh expression levels. Limbs electroporated with δ-EF1 cDNA showed a significant reduction in Ihh expression (Fig. 7 E). To analyze the specificity of this repression, we tested the expression of Col2a1, which is negatively regulated by δ-EF1 in vitro (Murray et al., 2000). Col2a1 mRNA expression was unchanged in δ-EF1 electroporated limbs as compared with control limbs (Fig. 7 E). Therefore, in spite of δ-EF1 overexpression in the entire limb bud of developing chicks, the in vivo repressor effect of δ-EF1 was specific, affecting Ihh expression in the growth plate but not another in vitro identified target gene Col2a1. These results were confirmed using dual in situ hybridization for Col2a1 and Ihh. Limbs electroporated with δ-EF1 cDNA showed a significant reduction in Ihh mRNA expression (Fig. 7 F, red), whereas expression of Col2a1 mRNA was unchanged (Fig. 7 F, green). The overexpression of δ-EF1 cDNA in developing chick limbs provides strong evidence for a role of δ-EF1 as negative regulator of Ihh in the growth plate. To further confirm that δ-EF1 negatively regulates expression of the Ihh gene, we tested whether δ-EF1 interacts in the transcriptional control of Ihh expression. We therefore cotransfected reporter plasmids containing the Ihh intron 1 regulatory region with a δ-EF1 expression plasmid. Overexpression of δ-EF1 in both C17 mouse limb bud cells and DF1 chicken fibroblasts consistently inhibited the activity of the Ihh intron 1 reporter construct (Fig.7, G and H). δ-EF1 could not, however, inhibit the activity of the construct with mutated CACCT(G) binding sites, which indicates that the intact CACCT(G) δ-EF1 binding sites are required for this activity (Fig. 7, G and H).

We have demonstrated that the skeletal phenotype of δ-EF1^−/− embryos was caused by a disturbed Ihh signaling. Several studies implicate Ihh in postnatal bone homeostasis (Maeda et al., 2007; Mak et al., 2008; Ohba et al., 2008). Therefore, we investigated if deficiency of δ-EF1 would lead to a postnatal bone phenotype. Because homozygous mutants die at birth, we analyzed 20-wk-old δ-EF1 heterozygous (HZ) mice. The radiographical analysis revealed normal skeletal patterning in δ-EF1^+/− mice (unpublished data). Next, we performed whole-body dual energy x-ray absorptiometry (DEXA) measurements of muscle, bone, and fat content. This analysis revealed a significant increase in bone mineral density and bone mineral content as compared with WT mice (Table I), and no difference in muscle and fat mass. Peripheral quantitative computed tomography (pQCT) analysis detected a significant increase in the bone trabecular content and trabecular density in δ-EF1^+/− mice (Table II). To analyze if these observations could be associated with a change in Ihh expression, we performed quantitative PCR on dissected growth plates of 8-wk-old WT and δ-EF1 HZ mice. Ihh mRNA expression was significantly up-regulated in growth plates of δ-EF1 heterozygotes as compared with WTs (Fig. 8). Collectively, haploinsufficiency for δ-EF1 causes up-regulation of Ihh in adult growth plates, a higher bone mineral density, and an increase in trabecular bone content.

δ-EF1–null mice exhibit dwarfism and impaired joint formation (Takagi et al., 1998), but no detailed histological analysis of the developing long bones was available. Therefore, after the breeding of mice into a CD1 background, we have reanalyzed the reported phenotype with specific focus on the histology of the growth plate. Despite the delay in expression of Col10a1, a marker gene for hypertrophic differentiation, we still observed hypertrophic chondrocytes in the growth plate of δ-EF1^−/− mice, which indicates that the hypertrophic differentiation was delayed rather than impaired. Additionally, there was an increase in the size of the domain of proliferating chondrocytes and their proliferation rate. Both these defects are consistent with the growth plate phenotypes observed in Ihh-null mice (St-Jacques et al., 1999). Conversely, a misexpression of Ihh under control of Collagen type 2 promoter induced Ptc mRNA and delayed Col10a1 mRNA expression (Long et al., 2004). These observations are very similar to our findings in δ-EF1–null mice, which suggests a link between δ-EF1 and Ihh signaling.

Ihh directly induces proliferation of prehypertrophic chondrocytes (Long et al., 2001). Targeted inactivation of δ-EF1 leads to significant increase in the proliferation of chondrocytes. Conversely, overexpression of δ-EF1 in the developing chick limb leads to reduced length of the zone of proliferating chondrocytes. This effect would be expected if the expression levels of Ihh would be changed, indicating that the growth plate defects observed in δ-EF1^−/− mice are consistent with the phenotypes observed after deregulation of Ihh expression. Next to the direct effect of Ihh on chondrocyte proliferation, Ihh also has PTHrP-dependent effects on chondrocyte differentiation. Ihh induces the expression of PTHrP in the periarticular perichondrium, which will block hypertrophic differentiation (Vortkamp et al., 1996; Karp et al., 2000). The increase in Ihh gene expression in δ-EF1–null mice leads to increased PTHrP expression in the periarticular perichondrium. This augmented PTHrP expression is likely to cause the observed delay in hypertrophic differentiation in δ-EF1–null mice. In support of our hypothesis, it was recently shown that PTHrP signaling activity is negatively regulated by Wnt/β-catenin signaling to initiate chondrocyte hypertrophy (Guo et al., 2009). In this case, the up-regulation of PTHrP in δ-EF1–null mice could be partially counteracted by Wnt/β-catenin signaling, and this could explain a rather mild phenotype of δ-EF1–null mice. Furthermore, final maturation of chondrocytes in δ-EF1–null mice appears to be less affected, as seen by Col10a1 mRNA expression. Our data suggest that terminal chondrocyte differentiation is a separate event that is still intact in δ-EF1–null mice. This is further supported by the observation that Wnt/β-catenin signaling controls final maturation of chondrocytes independently of PTHrP (Guo et al., 2009).

Next, we investigated if the observed phenotypes correlated with the changes in Ihh gene expression. Both Ihh mRNA expression and IHH protein levels were increased in δ-EF1^−/− mice. These data support our hypothesis that δ-EF1 acts as a negative regulator of Ihh transcription in the developing growth plate.

IHH is a long-range signaling molecule and its movement in the growth plate is dependent on heparan sulfate proteoglycans. In mice deficient for glycosyltransferase Ext1, heparan sulfate synthesis is impaired, resulting in an extended range of Ihh signaling, as indicated by changes in the expression levels and/or domains of PTHrP and Ptc mRNAs (Koziel et al., 2004). Ext1-null mice have shorter limbs, with joint fusions and defects in hypertrophic differentiation similar to those detected in δ-EF1^−/− mice. We therefore hypothesize that the excess of IHH protein cannot be sequestered by heparan sulfate, the negative regulator of IHH diffusion. This in turn leads to the extended range of IHH signaling in δ-EF1^−/− mice and, therefore, phenotypes similar to Ext1 deficiency, including the changes in PTHrP and Ptc mRNA expression (Koziel et al., 2004).

In several genes, intron 1 has been associated with the presence of regulatory elements (Reid et al., 1990; Smith et al., 1996; Antoine and Kiefer, 1998). The Ihh gene contains four putative CACCT(G) δ-EF1 binding sites in a very strongly conserved region of intron 1. Our studies demonstrate that δ-EF1 binds to those sequences in vitro and in vivo. The ZFHX1 family of transcription factors is known to be involved in the repression of gene transcription (Funahashi et al., 1993; Kamachi and Kondoh, 1993; Sekido et al., 1994, 1997; Remacle et al., 1999; Comijn et al., 2001; Papin et al., 2002). Binding of δ-EF1 to its target sites in vitro has been demonstrated for several genes (Murray et al., 2000; Terraz et al., 2001; Eger et al., 2005; Jethanandani and Kramer, 2005; Nishimura et al., 2006), and in all these cases, δ-EF1 displayed repressor activity with two exceptions. In one case, δ-EF1 induced ovalbumin expression in estrogen signaling cascades in vitro (Dillner and Sanders, 2004). In the other case, δ-EF1 transactivated promoters of smooth muscle cell differentiation marker genes in vivo (Nishimura et al., 2006).

Here, we demonstrate that δ-EF1 binds to its target sequences in the intron 1 regulatory element of Ihh in vivo, implying that it could be involved in transcriptional regulation of Ihh. This is a novel and important finding because little is known about the transcriptional regulation of Ihh expression. So far, two genes have been reported to directly interact with the Ihh promoter in vivo: Runx2 and Lef1 (Yoshida et al., 2004; Später et al., 2006). Both of them act as transcriptional activators of Ihh, and the identification of a potential in vivo repressor of Ihh adds to our understanding of the complex regulatory circuits governing Ihh expression.

To analyze the effect of δ-EF1 on the transcription of Ihh, we have performed luciferase reporter analyses with the Ihh regulatory element constructs and a δ-EF1 expression plasmid. We have shown in two different cell lines that δ-EF1 has a negative regulatory effect on Ihh transcription. The binding sites for δ-EF1 in the first intron are required for this regulation because a point mutation in the CACCT(G) binding sites prevents this inhibitory effect of δ-EF1 on the Ihh transcription. These in vitro experiments clearly indicate that δ-EF1 is a negative regulator of Ihh in the growth plate; however, this does not necessarily reflect the in vivo situation. Therefore, in addition to these in vitro experiments, we performed in vivo overexpression of δ-EF1 to analyze the regulatory function of δ-EF1 specifically in the growth plate. The developing chick embryo is a versatile and robust model system to study, among other things, limb development and osteochondrogenic differentiation (Tickle, 2004). Therefore, we have misexpressed δ-EF1 cDNA in developing chick limbs to investigate the effect on Ihh mRNA expression. Indeed, we achieved a reduction in Ihh expression in developing chick limbs, which supports the notion that δ-EF1 can repress the expression of Ihh in vivo. Moreover, we did not detect changes in expression of Col2a1, whereas δ-EF1 has been shown to repress Col2a1 transcription in vitro (Murray et al., 2000). Although overexpression experiments have their limitations, the target gene–specific response in this case is supportive of our hypothesis that δ-EF1 is repressing Ihh transcription in vivo. In chick limbs electroporated with δ-EF1 cDNA, we observed a reduction in length of the long bones consistent with the Ihh knockout (KO) phenotype (St-Jacques et al., 1999). Ihh KO mice display a dynamic bone phenotype. At 16.5 dpc, they have a delay in calcification, but at 18.5 dpc, this process is enhanced as compared with the WT littermates. The stage at which we analyzed the chick embryos corresponds to the 16.5-dpc stage in the Ihh mouse mutants, and the observed delay in calcification is consistent with the mouse phenotype. Additionally, it was demonstrated that infection of developing chick limb bud with Nkx3.2 results in down-regulation of Ihh expression, and the bone phenotype is very similar to the one induced by δ-EF1 cDNA (Provot et al., 2006). Namely, there is restriction of the Col10a1 expression domain and a delay in hypertrophy. Therefore it is likely that there are aspects of growth plate biology in the chick that are somewhat different than in the mouse. The reason why the induced phenotype is not as pronounced as in the Ihh-null mice is most likely caused by the delivery method of δ-EF1 cDNA. Chick limb electroporation delivers DNA to target tissues with variable efficiency, and although our system has been optimized for high efficiency using GFP expressing vectors, it is unlikely that we are able to target all the cells in the developing limb. Consequently, some cells will escape δ-EF1 repression, causing a milder limb phenotype. Another way of misexpressing genes in developing chick is by using RCAS virus-based system. Although the system is very efficient and effective, it has a major limitation in the size of the open reading frame that can be misexpressed, namely ∼2.3 kb. Because the δ-EF1 open reading frame is larger than 3 kb, this system cannot be used in our work. Currently, there are no alternative expression systems that could be used in the chick.

In this paper, we show that haploinsufficiency of δ-EF1 in mice leads to an increase in the adult bone mass and trabecular bone in a gender-independent fashion. Moreover, we detected up-regulation of Ihh expression in the growth plates of 8-wk-old δ-EF1 HZ mice, which indicates that the increase in trabecular bone could be the result of Ihh derepression. Our hypothesis that δ-EF1 is a repressor of Ihh during postnatal bone growth is supported by several findings. It has been shown that defects in Hh signaling cause a range of postnatal mouse phenotypes as well as human disorders affecting skeletogenesis (Maeda et al., 2007; Chen et al., 2008; Lodder et al., 2008). Moreover, it has been recently reported that Ptc haploinsufficiency increased adult bone mass in mice (Ohba et al., 2008). Here, the authors have correlated the decrease in Ptc expression to ectopic activation of Smo and, therefore, activation of Ihh signaling. Additionally, it has also been shown that Ihh produced in postnatal chondrocytes is essential to maintain the growth plate and specifically the trabecular bone (Maeda et al., 2007). Our observation that a depletion of δ-EF1 increases trabecular bone is also in line with the studies that both δ-EF1 and Ihh are expressed in the growth plate of the late hypertrophic zone and osteoblasts of 8-wk-old mice (Davies et al., 2002), and in early hypertrophic chondrocytes and osteoblasts in mouse, rat, and human growth plates (Iwasaki et al., 1997; Murakami and Noda, 2000; Kindblom et al., 2002; Maeda et al., 2007).

These results suggest that δ-EF1 is involved in the regulation of multiple steps of skeletal development and postnatal homeostasis. We therefore propose that δ-EF1 acts as a repressor of Ihh transcription involved in the determination of the boundaries of Ihh expression and expression levels.

In summary, we demonstrate that δ-EF1 is a novel in vivo regulator of Ihh expression in the growth plate. This finding adds to our understanding of growth plate biology and provides new insights into bone development and postnatal bone homeostasis.

δ-EF1 HZ mice were kindly provided by Y. Higashi (Institute for Developmental Research, Aichi Human Services Center, Osaka, Japan) and moved into a CD1 background. Mouse colonies were maintained according to the National Animal Welfare Guidelines. δ-EF1^+/− mice in CD1 background (backcrossed >10 times) were mated to generate homozygous δ-EF1^−/− embryos. The pregnant mice were sacrificed by cervical dislocation. All subsequent mouse work, including skeletal stainings, was performed as described previously (Hogan et al., 2004). For analyses on sections, embryonic tissues were fixed in 4% paraformaldehyde overnight at room temperature and embedded in paraffin before sectioning at 5 µm. Safranin O and von Kossa stainings were performed according to standard protocols. Mouse genotyping was performed by PCR on DNA isolated from yolk sacs (Hogan et al., 2004). The PCR primers used to identify δ-EF1 WT allele were δ-EF1 S (5′-AGCACTATTCTCCGCTACTCCAC-3′) and δ-EF1 AS (5′-ACCGCACCTGGTTTACGACACTC-3′), generating a fragment of 193 bp. Primers to identify δ-EF1 null allele were δ-EF1 S as described before and δ-EF1 AS (5′-AACCGTGCATCTGCCAGTTTGAG-3′), generating a fragment of 537 bp.

Immunofluorescence was performed using standard protocols using the following primary antibodies: mouse IgG (Biognost), ZEB for detection of δ-EF1 (E-20), IHH (C-15; both from Santa Cruz Biotechnology, Inc.), and TEC-3 for detection of Ki67 (Dako). For analysis of cell proliferation, the ratio of positive cells to the total cell number in a microscopic field was counted (n = 20, with n equal to number of sections; five different embryos). The microscopic field was determined using histological landmarks on a photograph. Subsequent analysis was performed on the photographs. Statistical analysis was performed using an unpaired student's t test.

In situ hybridization was performed on paraffin sections using standard procedures. The mRNA probes (see the Results section) were used as described previously (Koziel et al., 2004; Smits et al., 2004; Später et al., 2006). The δ-EF1 probe was designed using primers CD183-AS (5′-CATATGCTAAATCCGCTTCAG-3′) and CD184-S (5′-GGTCTTGCTTAAGGCCAAAGG-3′; sequence from nucleotide position 3,751–4,160 from GenBank/EMBL/DDBJ under accession no. D76432).

Probes were provided by Christine Hartmann (The Research Institute of Molecular Pathology, Vienna, Austria), Patrick Smits (Harvard University, Boston, MA), Christa Maas (University of Leuven, Leuven, Belgium), and Laetitia Buelens (University of Essen, Essen, Germany).

Fertilized eggs (Poel-Houben) were incubated at 38.5°C in a humidified atmosphere until they reached the desired developmental stage. Chicks were staged using HH staging criteria (Hamburger and Hamilton, 1992). 1 µg/µl DNA + 0.1% fast green was injected multiple times into the hind limb buds of HH stage 18–20 chick embryos. For the expression studies, δ-EF1 cDNA was cloned into a pCS3 vector (Rupp et al., 1994). The ECM 830 ElectroSquarePorator (BTX) was used to generate electric pulses. Two gold-plated electrodes (ϕ = 5 mm) were placed dorsally and ventrally of the limb bud. Electroporation was performed at the following conditions: voltage, 12 V; length of pulse, 50 ms; and frequency, four times.

Limb buds of 12.5–18.5 dpc WT and δ-EF1^−/− embryos were removed and homogenized for RNA extraction. 1 µg of each RNA sample was transcribed to cDNA. PCR was performed on a Rotor-gene 6000 detection system (Westburg), and analysis was performed using the accompanied software. For Western blot analysis, samples were homogenized in 60 µl of NuPage sample buffer (Invitrogen). Protein concentration was measured using a BCA Protein Assay Reagent kit (Thermo Fisher Scientific), and 20 µg of protein was loaded per lane. Each sample was mixed with 5 µl 0.5M DTT and denatured for 5 min at 95°C. For detection of injected δ-EF1 in chick limbs, the samples were loaded on a 8% Tris-glycine (Invitrogen) gel, and SDS-PAGE was performed in a Tris-glycine running buffer for 1 h and 30 min at 125 V. 200 µg/ml of anti-myc mouse monoclonal antibody (Sanvertech) diluted 1:500 in TBS with Tween 20 was used to detect injected δ-EF1 in pCS3. For detection of Ihh and δ-EF1 in mouse limbs: samples were loaded on a 4–12% Bis-Tris gel (Invitrogen), and electrophoresis was performed in MES running buffer for 1 h and 30 min at 120 V. Primary anti-IHH and anti-ZEB antibodies were diluted 1:200. Densitometry analysis of the Western blot in Fig. 5 was performed using ImageJ software.

DNA fragments of 65 bp containing the putative intron 1 regulatory elements of Ihh were obtained using the following DNA oligonucleotides in a PCR reaction using mouse genomic DNA as a template (CACCT/CACCTG sites underlined): oligo Ihh EMSA: mIhh EMSA S, 5′-CATCGAAGCATCGCCACCTGC-3′; mIhh EMSA AS, 5′-CAGGCAGTCGAGAAAAAAGGTGC-3′. To generate mutated sites, the following primers were used in PCR reactions as described before (mutated base in bold): oligo Ihh EMSA mutant: mIhh EMSA S mutant, 5′-CATCGAAGCATCGCCATCTGC-3′; mIhh EMSA AS mutant, 5′-CAGGCAGTCGAGAAAAAAGATGC-3′. Nuclear extracts were prepared from confluent monolayers of DF1 cells transfected with GFP or δ-EF1 in pCS3 using a standard published method (Tylzanowski et al., 2001). 1 µl of nuclear extract was incubated for 40 min at room temperature with γ[³²P]ATP end-labeled DNA Ihh oligo. Competition analysis was performed with 4- or 20-fold molar excess of unlabeled Ihh or Ihh mutant DNA fragment. For supershift assays, 1 µl of anti-myc mouse monoclonal antibody (200 µg/ml) was added to the nuclear extract. DNA–protein complexes were separated on a 6% nondenaturing polyacrylamide gel in 1× Tris-acetate-EDTA, pH 8.5, at 18 mA. The gel was dried and exposed to film (MS; Kodak) for autoradiography.

WT and mutated 65-bp PCR fragments, as described in the EMSA paragraph, were cloned into the SmaI site of pGL3tk vector (Promega) immediately upstream of a thymidine kinase minimal promoter (modified as described in Verschueren et al., 1999). The cell lines used were C17 cells, an immortalized cell line from a 13.5-dpc limb bud (provided by Vicky Rosen, Harvard University School of Medicine, Boston, MA; Rosen et al., 1993), and DF1 cells, a chicken fibroblast cell line (Cheryl Tickle, University of Bath, Bath, England, UK; Himly et al., 1998). Cells were cotransfected with δ-EF1 expression constructs in pCDNA3 (Remacle et al., 1999) using Fugene HD (Roche) method according to manufacturer's instructions. The amount of transfected DNA was adjusted to a total amount of 600 ng of DNA with the empty expression vector pCDNA3. Luciferase assay was performed 24 h after transfection as described previously (Tylzanowski et al., 2001). As an internal control for transfection efficiency, cells were cotransfected with RSV-LacZ constructs, and β-galactosidase values were determined using a luminescent β-galactosidase assay (Gal-screen; Tropix).

Mouse 16.5-dpc limbs were dissected and frozen at −80°C. Between 40 and 60 mg of tissue was ground in a mortar with the base submerged in liquid nitrogen. Cell lysate was fixed for 10 min in 1% formaldehyde in PBS. Chromatin shearing and ChIP assays were performed using the ChIP-IT enzymatic kit according to manufacturer's instructions (Active Motif). Primary antibodies were mouse IgG (Biognost) for negative control and anti-ZEB, as described in the Immunofluorescence and in situ hybridization section. PCR was performed using primers recognizing a 211-bp fragment of the Ihh regulatory region: mIhh ChIP S, 5′-ATCCCGACATCATCTTCAAGGACG-3′; mIhh ChIP AS, 5′-CCATCTTGGGCAGAAAGACACCC-3′.

The primers used to identify the Noggin WT allele were Noggin-S (5′-GCATGGAGCGCTGCCCCAGC-3′) and Noggin-AS (5′-GAGCAGCGAGCGCAGCAGCG-3′), generating a fragment of 200 bp. Primers to identify WT allele were SIP1WT-S (5′-ATCAGCAGCCTCCTATTTAAACAGAGTGTC-3′) and SIP1KO-S (5′-GAACTAGTTGAATTGGTAGAATCAATGGGG-3′), generating a fragment of 150 bp. The PCR conditions were annealing at 60°C for 10 s, an extension at 72°C for 20 s, and denaturation at 98°C for 3 s.

Total body (excluding the head) and lumbar spine bone density and lean and fat body mass were determined by dual x-ray absorptiometry using a Piximus densitometer (Lunar). Trabecular and cortical bone mineral content were assessed by pQCT using an XCT Research M system (Norland Medical Systems). Slices of 0.2 mm in thickness were scanned using a voxel size of 0.07 mm. Three scans were taken 2.4–2.6 mm from the distal end of the femur or 1.4–1.6 mm from the proximal end of the tibia to determine trabecular bone parameters. An additional scan was taken 4 mm from the distal end of the femur or 7 mm from the proximal end of the tibia to determine cortical bone parameters. Statistical analysis was done using a Student's two-sided t test.

Photography of sections was performed using a SPOT2 digital camera (Diagnostic Instruments, Inc.) placed on microscope (DMR; Leica) with the following objectives: a 2.5× (0.07 NA) N-Plan objective and 5× (0.15 NA), 10× (0.30), and 20× (0.50) HC PL Fluotar objectives (all from Leica). Skeletal stainings were photographed using a SPOT Insight camera (Diagnostic Instruments, Inc.) on a stereo microscope (Discovery V8; Carl Zeiss, Inc.) with a 1.0× free working distance 81-mm Plan-S objective or a 0.63× FWD 115-mm Achromat-S objective. Images were taken in TIFF format with SPOT 4.0.8 software (Diagnostic Instruments, Inc.). Small image modifications (brightness, contrast, and color balance) were performed using Photoshop Elements 2.0 (Adobe) on the entire image.

We thank the group of LSD & Joints, Jenny Peeters, and Roel Vandepoel for help with the mouse work; and Scott Roberts, Jeroen Eyckmans, and Rik Lories for helpful suggestions. Thanks to Laetitia Buelens and Eve Seuntjens for help with in situ experiments. We thank all those who kindly provided probes: Christine Hartmann (The Research Institute of Molecular Pathology, Vienna, Austria), Patrick Smits (Harvard University, Boston, MA), Christa Maas (University of Leuven, Leuven, Belgium), and Laetitia Buelens (University of Essen, Essen, Germany). C17 cells were a kind gift from Vicky Rosen (Harvard University School of Medicine, Boston, MA) and DF1 cells were a generous gift from Cheryl Tickle (University of Bath, Bath, England, UK).

This work was funded by FWO G.214.07 and a PhD grant from the Institute for the Promotion of Innovation Through Science and Technology in Flanders (Instituut voor de Aanmoediging van Innovatie door Wetenschap en Technologie in Vlaanderen) to Ellen Bellon.