Genetic deletion of Cyp26b1 negatively impacts limb skeletogenesis by inhibiting chondrogenesis

Endochondral ossification is a process in which mesenchymal cells differentiate into chondrocytes and produce a cartilage anlagen that serves as a template for subsequent bone formation. Cartilage provides the first structural template in the embryo and plays an important role in patterning the head, trunk, and limbs. During limb development, lateral plate mesoderm-derived cells migrate into the presumptive sites of limb formation. Through poorly defined mechanisms, some of these cells become specified to the chondrocytic lineage. Shortly thereafter, prechondrogenic cells aggregate together and form precartilaginous condensations (Hall and Miyake, 2000). Cell differentiation is initiated within the core of the condensation, resulting in the appearance of chondroblasts and the production of an extracellular matrix abundant in type II collagen (Col2a1) and various proteoglycans, including aggrecan (Acan). As chondrocytes mature, they exit the cell cycle, increase in size and become hypertrophic (reviewed by Kronenberg, 2006; Lefebvre and Smits, 2005).

The size and shape of developing skeletal elements relies on the spatio-temporal control of lineage-specific chondroblast differentiation. Although there is debate on when and how chondrogenic progenitors are specified, their differentiation occurs in a proximal-distal direction, with proximal cells differentiating and elaborating a chondrogenic matrix before more distal cells (Tabin and Wolpert, 2007). Regulation of limb chondrogenesis involves a cascade of factors that act together to specify the positional identity and fate of precursor cells and regulate their progression through the chondrogenic programme. These factors include, but are not limited to, members of the WNT, FGF, BMP and SHH families, and the focus of this study, retinoid signalling (Tabin and Wolpert, 2007; Weston et al., 2003b).

The vitamin A metabolite, retinoic acid (RA), regulates the differentiation of a variety of cell types, and consequently the development, growth and maintenance of a large number of tissues and organs (Mark et al., 2009; Underhill et al., 2001; Zile, 2001). The actions of RA are mediated by members of the nuclear receptor superfamily, RA receptors (RARs), for which three isoforms have been identified, RARα, RARβ and RARγ (reviewed in Germain et al., 2006; Mark et al., 2009). The RARs function in a heterodimeric complex with a retinoid X receptor (RXR) partner to either repress (unliganded RARs) or activate (liganded RARs) gene expression through RA response elements (RAREs) within target genes (Perissi et al., 2010; Weston et al., 2003a). Single receptor knockout mice display relatively minor phenotypes, whereas double knockouts present with a broad spectrum of congenital abnormalities (Lohnes et al., 1993; Lufkin et al., 1993; Luo et al., 1995). In the skeleton, deletion of both Rara and Rarg leads to widespread defects (underossification, loss and gain of bones) in the axial and appendicular skeleton and cranium (Lohnes et al., 1994; Mendelsohn et al., 1994). More recently, analysis of Rara,Rarg and Rarb,Rarg double knockouts has shown that RAR-mediated repression is essential for growth plate function and cartilage matrix homeostasis (Williams et al., 2009).

The transcriptional activity of RARs is modulated by the availability of RA. Evolutionarily conserved mechanisms precisely regulate the synthesis, transport and degradation of RA (Abu-Abed et al., 1998; Duester, 2008; MacLean et al., 2001; Petkovich, 2001; Sonneveld et al., 1998; White et al., 1997). This is accomplished by various binding proteins and two groups of enzymes: the retinaldehyde dehydrogenases (RALDHs: ALDH1A1, ALDH1A2 and ALDH1A3) and CYP26 enzymes (CYP26a1, CYP26b1, CYP26c1), which synthesise and degrade RA, respectively. In mice, ALDH1A2 has a prominent role in embryogenesis, because Aldh1a2-deficient mice die around embryonic age (E) 10 as a result of defects in multiple organs and tissues (Niederreither et al., 1999). Moreover, in the developing limb bud, Aldh1a2 appears to be the predominant RA synthesis enzyme because Aldh1a2-null animals present with a wide spectrum of defects, including severe limb abnormalities, whereas loss of Aldh1a1 or Aldh1a3 has a more limited impact on development (Niederreither and Dolle, 2008). CYP26 enzymes serve to catabolise all-trans RA (at-RA) into more polar inactive metabolites that appear to have negligible RAR transcriptional activity (Pennimpede et al., 2010a). Both Cyp26a1 and Cyp26b1 are expressed in the developing skeleton, and appear in mesenchymal condensations, and subsequently in differentiating chondroblasts and osteoblasts (de Roos et al., 1999; Laue et al., 2008; Maclean et al., 2009; Spoorendonk et al., 2008).

Hyper- and hypovitaminosis A have been reported to adversely affect the skeleton for over 60 years (Fell and Mellanby, 1950; Fell and Mellanby, 1952; Mellanby, 1943; Mellanby, 1947). During that time, numerous reports have shown that vitamin A (retinol) and its metabolites affect skeletal development, growth and homeostasis. Throughout development, too much or too little RA has been associated with skeletal malformations characterised by loss or gain of skeletal elements (Underhill and Weston, 1998; Weston et al., 2003b). Accordingly, Cyp26b1-null mice exhibit pronounced skeletal abnormalities that are characterised by either underossification or loss of endochondral and intramembranous-derived bones (Maclean et al., 2009; Pennimpede et al., 2010b). To understand the cellular and molecular basis underlying these defects, we explored the function of endogenous retinoid signalling in the limb chondrogenic programme using the Cyp26b1-knockout mouse as a model system. We demonstrate that retinoid signalling negatively impacts chondrogenesis in vivo, and that the severity of limb defects in a Cyp26b1^−/− background reflect levels of retinoid signalling. This indicates that the Cyp26b1^−/− limb phenotype is due, at least in part, to defects in expression of the chondroblastic phenotype.

Genetic analysis indicates that Cyp26b1 is the primary RA-metabolising enzyme in the developing limb bud (Abu-Abed et al., 2001; Pennimpede et al., 2010a; Yashiro et al., 2004). To further examine the role of endogenous retinoid signalling in skeletogenesis, both conventional and conditional Cyp26b1-knockout mice were analysed. Cyp26b1^fl/fl mice were crossed with PcxNLSCre⁺ transgenic mice to achieve deletion of Cyp26b1 in the germline and to generate Cyp26b1^−/− animals. These animals exhibited a spectrum of developmental malformations as previously described (Pennimpede et al., 2010b; Yashiro et al., 2004). Cyp26b1^fl/fl mice were also crossed with a line of transgenic mice that express Cre under direction of the Prrx1 promoter, which is expressed as early as E9.5 in the developing limb bud, parts of the head, and sternum (Kimura et al., 2010; Logan et al., 2002). Upon external observation, both Cyp26b1^−/− and Prrx1Cre⁺/Cyp26b1^fl/fl mice exhibited truncated limbs and abnormal digit formation at E18.5, although the limbs of Prrx1Cre⁺/Cyp26b1^fl/fl mice were not as severely truncated and presented with more digits than age-matched Cyp26b1^−/− animals (Fig. 1A). Wild-type littermates in Prrx1Cre⁺/Cyp26b1^fl/fl and Cyp26b1^−/− lines were both morphologically normal and comparable for experimental purposes (supplementary material Fig. S1A).

At E18.5, Cyp26b1^−/− mice exhibit a greatly reduced stylopod and zeugopod, a loss of carpal bones, and oligodactyly, with only two or three digits forming per autopod. Prrx1Cre⁺/Cyp26b1^fl/fl animals present with a slightly longer stylopod and zeugopod than null animals, recognisable wrist elements, and the formation of four digits (Fig. 1B). In both lines, the scapula and ilium appear morphologically normal, and nail-like structures are formed at the digit tips (data not shown). The clavicle is missing in Cyp26b1^−/− but this structure is not deleted in Prrx1Cre⁺/Cyp26b1^fl/fl animals.

The defects described at E18.5 are apparent as early as E15.5, and notably, areas of Alizarin Red staining in wild-type skeletons are not present in either the Prrx1Cre⁺/Cyp26b1^fl/fl or Cyp26b1^−/− animals at E15.5 (Fig. 1C). These areas of staining are observable at E18.5, indicating a delay in the mineralisation of skeletal elements in both Cyp26b1 conditional and null mutants. Furthermore, Prrx1Cre-driven deletion of Cyp26b1 partially rescues both the autopod and to a lesser extent the stylopod defects, whereas the zeugopod is substantially impacted in both transgenic lines.

The limbs from Prrx1Cre⁺/Cyp26b1^fl/fl mice were less affected than Cyp26b1^−/− limbs, and to determine whether differences in RA signalling contributed to these phenotypes, both lines were crossed with an RA reporter transgenic mouse. Limbs from Prrx1Cre⁺/Cyp26b1^fl/fl and Cyp26b1^−/− animals expressing the RARE-lacZ transgene, which harbours a trimerised repeat of RARb2 RARE linked to the hsp68 minimal promoter (Rossant et al., 1991) were stained for X-gal to allow visualisation of regions of activated retinoid signalling. Fore- and hindlimbs exhibited similar patterns of retinoid signalling. At E11.5, very little staining was observed in the limbs of wild-type mice, but was present proximally in Cyp26b1^−/− limbs (supplementary material Fig. S1B). By E12.5, activated RA signalling could be observed in the interdigital region of wild-type limbs; however, in null animals, areas showing activated retinoid signalling extended almost throughout the entire proximal–distal axis of the limb bud (Fig. 2A). In Prrx1Cre⁺/Cyp26b1^fl/fl animals, aberrant retinoid signalling was observed proximally, but this was not as intense, nor did it expand as far distally, as in the null animals. At E13.5 and E14.5, retinoid signalling continued to be observed interdigitally in wild-type limbs, whereas the only area free of activated retinoid signalling in Cyp26b1^−/− limbs was the distal tip, where Cyp26a1 was expressed (Abu-Abed et al., 2002) (Fig. 2A, supplementary material Fig. S1B). Furthermore, at later stages, lacZ staining could be observed in strips flanking the developing digits, indicating not all lacZ⁺ cells undergo cell death (Fig. 2A, data not shown). Again, Prrx1Cre⁺/Cyp26b1^fl/fl limbs exhibited ‘intermediate’ levels (in terms of both distribution pattern and intensity) of activated retinoid signalling between wild-type and Cyp26b1^−/− limbs.

To further investigate changes in RA signalling in the mutant limbs, fore- and hindlimb buds were microdissected into proximal and distal regions for analysis of Cyp26b1 and Rarb expression (Fig. 1B, supplementary material Fig. S1C). Quantitative RT-PCR (qPCR) demonstrated that Cyp26b1 is more highly expressed in the distal region of the developing limb bud at E11.5 and E12.5, as previously reported (Pennimpede et al., 2010b; Yashiro et al., 2004) (Fig. 2B). Furthermore, this analysis revealed that although Cyp26b1 expression was completely absent in Cyp26b1^−/− animals, a small amount of Cyp26b1 expression remained in the distal regions of Prrx1Cre⁺/Cyp26b1^fl/fl limbs. Cyp26b1 was absent in the proximal regions of Prrx1Cre⁺/Cyp26b1^fl/fl limbs. Rarb expression in the proximal and distal regions of fore- and hindlimbs was analysed to determine quantitative increases in retinoid signalling in these tissue populations because Rarb is a direct RAR target gene and is expressed in proximal limb mesenchyme and later in interdigital regions (Mendelsohn et al., 1991; Mollard et al., 2000). Large increases in Rarb expression were observed in both proximal and distal regions of Prrx1Cre⁺/Cyp26b1^fl/fl and Cyp26b1^−/− fore- and hindlimbs, although increases in the Prrx1Cre⁺/Cyp26b1^fl/fl limb were not as large, particularly in the distal regions where small amounts of Cyp26b1 expression were observed.

To assess retinoid status in the limb mesenchyme, high-density primary limb mesenchymal (PLM) cultures were established from E11.5 wild-type and Cyp26b1^−/− limbs. During the preparation of these cultures, the ectoderm is enzymatically separated from the mesenchyme, enabling the analysis of mesenchyme-dependent effects. qPCR analysis revealed that in the absence of Cyp26b1, Cyp26a1 and Crabp2 expression increased, and Aldh1a2 expression decreased, probably to compensate for increased levels of RA (Fig. 2C, supplementary material Fig. S1D). Additionally, the expression of both a direct RAR-target gene, Rarb, and a putative target gene, Fgf18 (Delacroix et al., 2010), increased ~threefold at 3 days, indicating that increased retinoid signalling occurs in culture (Fig. 2C, supplementary material Fig. S1D). Furthermore, the activity of a retinoid-responsive reporter, RARE-LUC, which consists of a trimerised RARE upstream of the firefly luciferase gene (Balkan et al., 1992), increased significantly in Cyp26b1^−/− PLM cultures. Treatment with an ALDH1 inhibitor, diethylaminobenzaldehyde (DEAB), at culture initiation was able to reduce RARE-LUC reporter activity in Cyp26b1^−/− cultures, but not to wild-type levels (Fig. 2D). Altogether, these results indicate increased and sustained endogenous retinoid signalling in Cyp26b1^−/− limb mesenchyme.

It has previously been reported that increased levels of RA negatively affect skeletal development by inhibiting chondrogenesis in vitro (Cash et al., 1997; Weston et al., 2002; Weston et al., 2000); however, the role of endogenous RA signalling in this process has not been fully investigated in vivo. qPCR analysis of microdissected E11.5 and E12.5 limbs from wild-type and Cyp26b1^−/− embryos demonstrated little change in the expression of Sox9, but Acan, a marker associated with chondroblast differentiation, was significantly decreased (Fig. 3).

A Col2-LUC reporter, which is based on the SOX5, SOX6 and SOX9 binding site from the first intron of Col2a1 upstream of the firefly luciferase gene (Lefebvre et al., 1997; Lefebvre et al., 1998) was used to follow chondroblast differentiation. The activity of this reporter has been shown to tightly correlate with chondroblast differentiation and cartilage formation (Hoffman et al., 2006; Muramatsu et al., 2007; Weston et al., 2000). Col2-LUC reporter activity in Cyp26b1^−/− PLM cultures was around seven-times lower than that in the wild type after only 1 day of culture (Fig. 4A). Treatment with DEAB was able to increase Col2-LUC activity in both wild-type and Cyp26b1^−/− PLM cultures; however, even ‘rescued’ Cyp26b1^−/− cultures had significantly lower Col2-LUC reporter activity than untreated wild-type cultures (Fig. 4A). Notably, Col2-LUC reporter activity is tightly correlated with RA reporter activity, because overexpression of Cyp26a1, which degrades RA, led to a 80% decrease in RA reporter activity and a threefold increase in Col2-LUC (Fig. 4B and data not shown). Interestingly, expression of Cyp26a1 in limb mesenchyme from Cyp26b1-null limbs was not sufficient to rescue the chondrogenic defect, even though under these conditions, RARE-LUC activity was decreased (Fig. 4B). By contrast, the addition of an RAR antagonist AGN194310 (4310, 100 nM) 24 hours after culture initiation was sufficient to rescue chondrogenesis in cultures derived from Cyp26b1-null mesenchyme (Fig. 4C). Under these conditions, the addition of 100 nM atRA decreased cartilage nodule formation in both Cyp26b1 replete and deficient cultures (Fig. 4C). These results indicate that the chondrogenic defect in the limb mesenchyme is associated with increased activation of RAR-mediated signalling. However, RAR-mediated signalling does not appear to directly affect SOX9 activity, because overexpression of Sox9 in the presence or absence of Rara and/or ketoconazole (CYP inhibitor, see below) did not impact SOX9-induced Col2-LUC activity (supplementary material Fig. S2).

BMPs have been shown to have an essential role in limb skeletogenesis, both in specifying limb mesenchymal cells to a chondrocytic fate and in enhancing their subsequent differentiation into chondroblasts (Pogue and Lyons, 2006; Wu et al., 2007; Yoon and Lyons, 2004). To determine whether the ‘chondrogenic’ defect of the Cyp26b1^−/− mesenchyme could be rescued by the addition of BMP, Cyp26b1^−/− and wild-type PLM cultures were established and treated with BMP4 or vehicle. As reported previously, BMP4 reduced the activity of a retinoid-responsive reporter gene in wild-type cells and this was also observed, albeit to a lesser extent, in Cyp26b1^−/− mesenchyme (Fig. 4D) (Hoffman et al., 2006). Interestingly, BMP4 exhibited negligible pro-chondrogenic activity in cultures derived from Cyp26b1-null embryos, whereas BMP4 promoted chondrogenesis in wild-type cultures, as determined by following Col2-LUC reporter activity and Alcian Blue staining (Fig. 4D,E). Consistent with these observations, Sox9 was modestly reduced (~30%), whereas both Col2a1 and Acan were substantially reduced in the Cyp26b1-deficient cultures following BMP4 treatment in comparison to Cyp26b1⁺ cultures (Fig. 4F). These results further demonstrate that Cyp26b1^−/− mesenchyme exhibits a defect in chondrogenesis.

Herein, Cyp26b1^−/− mice have been used to systematically examine the role of endogenous RA signalling in chondrogenesis and skeletal development.

The first step in chondrogenesis is the aggregation of pre-chondrogenic cells into precartilaginous condensations. To normalise for potential differences in cell density that would impact chondrogenesis, Cyp26b1⁺ and Cyp26b1^−/− PLM cultures were plated at similar densities. Under these conditions no overt increase in cell death was observed (data not shown). Following culture of limb mesenchymal cells, peanut agglutinin (PNA) staining was used to identify pre-chondrogenic condensations. These analyses revealed very little difference between the ability of wild-type and Cyp26b1^−/− cultures to form prechondrogenic condensations following 1 day of culture (Fig. 5A). In both cultures, diffuse PNA staining was observed throughout the culture from both wild-type and Cyp26b1^−/− limb mesenchyme. By day 5, there is still an appreciable amount of PNA staining throughout wild-type cultures, whereas discrete intensely PNA⁺ regions are observed in the cultures from Cyp26b1^−/− mice. Treatment with DEAB leads to more extensive PNA staining in cultures derived from limbs of both wild-type and Cyp26b1^−/− mice, and this can be more easily observed in the internodular regions of the Cyp26b1-null cultures (Fig. 5A). Typically, with increasing culture time, prechondrogenic cells within the condensations differentiate, and this is associated with reduced PNA staining. This is not observed in the Cyp26b1^−/− cultures in which PNA staining intensity is maintained or possibly increased (Fig. 5A). Consistent with this, qPCR analysis revealed that although there is little change in Sox9 expression between wild-type and Cyp26b1^−/− cultures, the expression of condensation markers Vcan and Tnc is maintained or increased in null cultures (Fig. 5B). These results indicate that the mesenchyme from Cyp26b1^−/− limbs forms discrete prechondrogenic condensations in which chondroblast differentiation (and thus progression from the condensation stage) appears to be impaired.

To further address the impact of a loss of Cyp26b1 on chondrogenesis, and specifically chondroblast differentiation and the associated production of a cartilaginous extracellular matrix, PLM cultures were stained with Alcian Blue. Consistent with the Col2-LUC reporter gene findings in Cyp26b1^−/− cells (Fig. 4A), a marked decrease in the number of Alcian-Blue-stained cartilage nodules is observed in Cyp26b1^−/− cultures. This is obvious as early as day 4 of culture, and is much more pronounced by day 8. Addition of DEAB results in a small increase in cartilage formation in Cyp26b1^−/− cultures (Fig. 6A). qPCR analysis of Cyp26b1^−/− cultures demonstrates that expression of the long form of Sox-5 (L-Sox5) and Sox6 is reduced ~60% in 3 day cultures (Fig. 6B). Similarly, the expression of extracellular matrix molecules such as Col2a1, Acan, Hapln1, Comp and Matn1 were significantly decreased (Fig. 6B). These findings are congruent with the analysis of Sox9 and Acan in vivo (Fig. 3) and are also consistent with the severe cartilaginous defects in the skeletal elements in the Cyp26b1^−/− limbs (Fig. 1). Altogether, these results indicate that chondroblast differentiation is appreciably decreased in the limb mesenchyme of Cyp26b1^−/− mice.

In contrast to the negative role of RAR activation in chondrogenesis, RA signalling has been shown to be required in chondrocyte hypertrophy (Koyama et al., 1999). There was little change in expression of the hypertrophic markers Pthr1, Vegfa and Runx2 in Cyp26b1^−/− cells compared with wild type; however, an increase in Spp1 and a decrease in Col10a1, Mmp13 and Alp1 was observed after 8 days of culture (Fig. 7A). The magnitude of the decrease in the expression of these latter genes (50–80%) was much smaller in comparison to that of the chondroblastic markers (8–27-fold or higher). Notably, the expression of Ihh was substantially downregulated at all time points in Cyp26b1^−/− cultures. Consistent with this, the hedgehog target gene Ptch1 was also decreased. IHH impacts Pthrp expression, and Pthrp was also found to be downregulated at all times. Cyp26b1-null mesenchyme-derived cultures produce fewer cartilage nodules, and as such it might be expected that chondrogenic stages proceeding chondroblast differentiation, such as chondrocyte hypertrophy, might be negatively impacted. To determine whether the nodules that did form underwent chondrocyte hypertrophy, immunofluorescence was performed to examine the distribution of MMP13 expression, a late chondrocyte hypertrophy marker. This analysis revealed that there was little change in distribution of MMP13 (Fig. 6C). Furthermore, alkaline phosphatase staining was used to further assess chondrocyte hypertrophy in the cultures, and although there were fewer nodules in the Cyp26b1-null cultures, the intensity of staining around the nodules was similar to that observed in the Cyp26b1⁺ cultures (Fig. 7C). These findings suggest that the decrease in many of the hypertrophy-associated genes, such as Mmp13, probably reflects a decrease in the number of cartilage nodules, rather than a decrease in hypertrophy.

The impact of manipulation of endogenous retinoid signalling on chondrogenesis and chondrocyte hypertrophy was further explored in PLM cultures treated with ketoconazole, a CYP inhibitor, in which the timing of CYP inhibition could be controlled. Analysis of gene expression in wild-type cells treated with ketoconazole demonstrates similar changes in gene expression to those observed in untreated Cyp26b1^−/− cultures (Figs 5, 6 and 7). Not surprisingly, multiple components within the RAR signalling pathway (Aldh1a2, Crabp2, Cyp26a1, Cyp26b1 and Rarb), are differentially expressed after ketoconazole treatment with all direct targets (Crabp2, Cyp26a1 and Rarb) being upregulated. Markers associated with prechondrogenic mesenchyme were elevated (Col1a1, Tnc), whereas chondroblastic differentiation markers (Acan, Col2a1, Comp, Cspg4, Hapln1, Matn1, Mia1, Sox5, Sox6 and Sox9) were substantially decreased in ketoconazole-treated cultures (Figs 5, 6, 7 and 8). Furthermore, this treatment led to a significant decrease in the expression of Ihh, Pthrp and Ptch1. Decreased expression of Pthrp has been shown to accelerate chondrocyte hypertrophy and consistent with this, ketoconazole treatment induced precocious expression of Col10a1, Mmp13, Spp1 and Vegfa expression in day 1 and/or 3 cultures (Fig. 8). Because ketoconazole inhibits both CYP26A1 and CYP26B1 activity, this might highlight a potential, albeit minor, role for Cyp26a1 in the Cyp26b1^−/− limb. To further address the stage-dependent action of RA on chondrogenesis and chondrocyte hypertrophy, wild-type cultures were treated with ketoconazole at various times after culture initiation. Addition of ketoconazole 1 day after establishment of cultures had a pronounced negative impact on Alcian Blue staining, whereas treatment at later times following chondroblast differentiation (days 3 and 5) had an undetectable effect (supplementary material Fig. S3). These results are consistent with those from analyses of Cyp26b1^−/− mice and suggest that RA signalling impacts the differentiation of chondrogenic progenitors.

Recent studies have shown that at the time of chondrogenic mesenchymal condensation in the limb, SOX9⁺ cells contribute to both the chondrogenic and tendogenic lineages (Soeda et al., 2010). Sox9 expression does not change in the limbs of Cyp26b1^−/− mice or derived cultures; however, chondroblastic markers are significantly downregulated. The continued expression of Sox9 coupled with decreased chondroblast differentiation suggests that increased RA signalling could be acting to redirect the cells to an alternative cell fate and/or maintain the cells as progenitors resulting in the expression of various lineage markers. To evaluate these possibilities, markers associated with the closely linked tendogenic lineage were examined. Analysis of tendon markers in Cyp26b1^−/− PLM culture by qPCR revealed a modest increase (less than twofold) in Scx, Tnmd and Col1a1 expression, and other markers of tendon development in null cultures (Fig. 9A) (Docheva et al., 2005; Schweitzer et al., 2001; Xu et al., 1997). To evaluate tendon cell differentiation, Mohawk (Mkx), a gene expressed in differentiating tendon cells, and important for the upregulation of the different types of Col1 during differentiation, was also examined (Ito et al., 2010; Liu et al., 2010). Mkx expression was decreased in the Cyp26b1-null cultures (Fig. 9A). Consistent with these findings, qPCR analysis of wild-type PLM cultures treated with ketoconazole demonstrated a significant increase in Scx expression after 1 and 3 days of culture (Fig. 8). Furthermore, although cultures treated with ketoconazole exhibited increased expression of Scx, Tnmd expression did not change significantly (Fig. 8). More importantly, consistent with the analysis of the Cyp26b1^−/− mesenchyme, Mkx was decreased. Visualisation of Scx expression, which is expressed in tendon progenitors, in Cyp26b1^−/− animals expressing the ScxGFP⁺ transgene (Pryce et al., 2007; Schweitzer et al., 2001) demonstrated that Scx expression appears to be increased and mostly confined to the proximal region of the forelimb of Cyp26b1^−/− animals when compared with wild-type littermates (Fig. 9B). However, qPCR analysis of Scx expression in fore- and hindlimbs microdissected into proximal and distal regions demonstrated that Scx expression was not significantly changed in wild-type and Cyp26b1-null limb buds (Fig. 9C). Together, these results reflect the chondrogenic process where retinoid signalling promotes the initial stages of the tendogenic programme; however, in this instance, tendon cell differentiation is inhibited.

For many years, it has been believed that RA has an instructive role in limb patterning by acting as a ‘proximalising’ morphogenetic factor, and that the limb malformations in Cyp26b1^−/− mice result from a shortening of the proximo–distal (P–D) axis (Yashiro et al., 2004). However, new evidence suggests that RA is dispensable for early limb patterning, because Aldh1a2,Aldh1a3 double-knockout mice have normal hindlimb development and disruption of Rarg partially rescues the Cyp26b1^−/− limb phenotype despite altered expression of P–D genes (Pennimpede et al., 2010b; Zhao et al., 2009). In this work, we demonstrate that the severity of limb malformations in Cyp26b1^−/− mice is tightly correlated to levels of retinoid signalling, and that a lack of Cyp26b1 negatively impacts chondrogenesis and tendogenesis before the onset of Ihh and Mkx expression, respectively. Similarly to recent reports on X-irradiation-induced phocomelia in chicks, in which the pre-chondrogenic mesenchyme is depleted (Galloway et al., 2009), the impact of RA on chondrogenesis results in an insufficient number of chondrocytes and subsequent defects in the proper formation of skeletal elements. This suggests that the limb phenotype in Cyp26b1^−/− animals is not due to changes in patterning per se, but more likely results from a failure in the subsequent execution of differentiation programmes.

The first step in the chondrogenic programme is the specification of mesenchymal cells to a chondrocytic fate, following which, committed prechondrogenic cells become tightly packed and form precartilaginous condensations (Hall and Miyake, 2000). Cyp26b1^−/− cells express early condensation markers such as Sox9, which is essential for the commitment of mesenchymal cells towards the chondrogenic lineage, and appear to form condensations (Akiyama et al., 2002). However, Cyp26b1^−/− cells exhibit a significant decrease in differentiation and appear to remain at a prechondrogenic stage. Previous studies have demonstrated a requirement for RAR-mediated repression for chondroblast differentiation in culture (Weston et al., 2002; Weston et al., 2000). Consistent with this, markers of chondroblast differentiation are markedly decreased in the limb mesenchyme of Cyp26b1^−/− mice. However, both in vitro and in vivo, cartilages appear and this indicates that some prechondrogenic cells escape retinoid action and differentiate. One possibility is that as Aldh1a2 is not expressed in chondrogenic regions (Hoffman et al., 2006). It is likely that levels of RA decrease in the centre of condensations where RA cannot diffuse to, allowing some cells to differentiate. Fewer, but much larger, cartilage nodules form, indicating that cells remain and expand in the condensation stage for a longer period before their eventual differentiation. Treatment with DEAB is not able to rescue Cyp26b1^−/− cells, demonstrating that in the absence of degradation, sufficient RA remains to interfere with subsequent chondroblast differentiation. Similarly, overexpression of Cyp26a1 failed to restore chondrogenesis in the Cyp26b1-null cultures. Cyp26a1 expression is increased in Cyp26b1-null cultures; however, this also appears to be insufficient to rescue loss of Cyp26b1 expression. Both Cyp26a1 and Cyp26b1 exhibit similar enzymatic activity (Pennimpede et al., 2010a). The inability of upregulated Cyp26a1 (Fig. 2C) or co-transfection of Cyp26a1 (Fig. 4B) to rescue the chondrogenic defect in vitro suggests that the limited number of Cyp26a1-expressing cells are insufficient to overcome the increased level of retinoid signalling in the Cyp26b1^−/− cultures. This might also partly explain why Cyp26a1 expression in vivo is not sufficient to rescue the limb defects resulting from loss of Cyp261. Furthermore, these results suggest that CYP26b1 is the predominant at-RA metabolising enzyme in the developing appendicular skeleton. Importantly, the addition of an RAR antagonist effectively rescued cartilage nodule formation in the Cyp26b1 null cultures, suggesting that the loss of Cyp26b1 affects chondrogenesis through RAR-mediated signalling and that the cells in the null cultures exhibit chondrogenic potential. Altogether, these results indicate that RA affects prechondrogenic cells at the timing of differentiation and blocks their progression to differentiated chondrocytes.

Retinoid signalling has been shown to be required to promote the transition from prehypertrophy to hypertrophy during endochondral ossification in the developing limb (Koyama et al., 1999). Congruent with this, we have demonstrated that Cyp26b1^−/− cells, with the exception of Ptrhp and Ihh, exhibit modest changes in the expression of hypertrophic-associated markers. The small decrease in Col10, Alp1 and Mmp13 expression and the delay in mineralisation probably result from the decrease in chondrogenic differentiation and the formation of fewer, albeit larger, cartilage nodules. Consistent with this, alkaline phosphatase staining around individual nodules is comparable between Cyp26b1⁺ and Cyp26b1-null cultures.

Growth plates sustain skeletal growth through the proliferation of chondrocytes, matrix synthesis and accumulation, and chondrocyte hypertrophy. The growth plate is generally divided into two zones, the upper zone, which contains immature chondrocytes, maintained by PTHrP signalling, and the lower zone, which contains Ihh-expressing prehypertrophic and maturing hypertrophic chondrocytes. PTHrP regulates the expression of Ihh, which controls chondrocyte proliferation and maturation, and Ihh in turn regulates Pthrp expression, creating a feedback system that controls the growth of skeletal elements (Kronenberg, 2006). The hypertrophic portions of the growth plate have been shown to contain higher levels of endogenous retinoid signalling than the upper, immature zones and a requirement for RAR-mediated repression in the upper growth plate zone has been shown to maintain chondrocyte function and the production of cartilage-associated extracellular matrix proteins (Williams et al., 2010; Williams et al., 2009). Previous studies have also suggested that RA-mediated repression of target gene expression in the upper zone is necessary to maintain physiological levels of Ihh expression (Koyama et al., 1999). We demonstrate in this study that an increase in retinoid signalling substantially decreases Ihh expression. Consequently, hypertrophic chondrocytes emerge earlier, with less growth having occurred, and this might partly explain the observed truncation of skeletal elements. The role of RA upstream of Ihh is further exemplified by the striking resemblance between Cyp26b1^−/− and Ihh^−/− limbs (St-Jacques et al., 1999). Our findings indicate that RA negatively impacts chondrogenesis before activation of Ihh signalling; however, RA appears to promote subsequent steps involving chondrocyte hypertrophy.

Multi-potent mesenchymal cells derived from the lateral plate mesoderm have the ability to contribute to lineages such as cartilage, bone, fat, and tendon (Pearse et al., 2007). We investigated whether Cyp26b1^−/− cells have the potential to contribute to lineages other than chondrogenic, and chose to focus on the tenogenic lineage because chondrocytes and tenocytes both arise from SOX9⁺ cells. The two lineages share early molecular mechanisms during mesenchyme condensation and close interactions between cartilage and tendon exist during development (Schweitzer et al., 2010). An upregulation of early tendon markers was observed in Cyp26b1^−/− cells, indicating that the decrease in chondrogenesis reflects an increase in commitment to tenogenesis or the maintenance of mesenchymal progenitors. However, Mkx, a gene associated with late tendogenesis, was downregulated in both Cyp26b1^−/− and ketoconazole-treated cultures, and thus it is unlikely that functional tenocytes are forming.

In the developing limb proper, active RA signalling has been identified in three distinct locations, in the interdigital region (IDR), the zone between the developing zeugopodal elements and later on in the perichondrium. In the IDR, RA has an important role in modifying the chondrogenic potential of IDR cells, with a subset of these cells undergoing apoptosis, whereas others contribute to unidentified lineages (Weston et al., 2003b). RA is thought to have a similar role in the zeugopod, where it suppresses the chondrogenic potential of inter-element mesenchymal cells, enabling the formation of discrete zeugopodal cartilages. Within the developing skeletal elements, RA levels are highest within the perichondrium, which is an important source of undifferentiated skeletogenic cells (Williams et al., 2010), and active RA signalling has also been identified in this tissue (von Schroeder and Heersche, 1998). These observations, coupled with the analysis of the Cyp26b1^−/− mice described herein, indicate that retinoid signalling inhibits expression of the chondroblast phenotype. Furthermore, RAR antagonists promote precocious chondroblast differentiation, and compound RAR mutants present with ectopic cartilages at a number of anatomical sites (Lohnes et al., 1994; Weston et al., 2002; Weston et al., 2000). Marker analysis of Cyp26b1^−/− mice indicates that limb mesenchymal cells retain a prechondrogenic phenotype; however, there is also increased expression of markers of other cell lineages, indicating that RA might be operating to maintain cells at a progenitor stage. This premise is certainly congruent with a recent report showing that retinoid levels are highest in the perichondrium within the developing skeleton (Williams et al., 2010). In this regard, RA is probably maintaining mesenchymal progenitors in part through the upregulation of genes such as Twist1 and Fgf18 (a putative direct RAR target gene), both of which repress chondroblast differentiation (Delacroix et al., 2010; Hinoi et al., 2006), are expressed in the perichondrium and were found to be significantly elevated in Cyp26b1^−/− cells (supplementary material Fig. S1D). At other sites in the developing embryo, RA has also been shown to be important in maintaining progenitor populations, and this has recently been shown to be the case in a subpopulation of precursor cells within the subventricular zone of the developing forebrain (Haskell and LaMantia, 2005). Collectively, these findings highlight an important function for endogenous RA signalling in regulating progenitor cell behaviour in the developing limb.

Generation and genotyping of mice containing Cyp26b1 null and floxed alleles, and ScxGFP⁺ has previously been described (MacLean et al., 2007; Pryce et al., 2007). RARE-lacZ and Prrx1Cre⁺ mice were obtained from Jackson Laboratory (Bar Harbor, ME). The mice used in the studies described were of a mixed genetic background and timed matings were established by housing females with males overnight and checking females daily for vaginal plugs. Noon of the day that a plug appeared was denoted as E0.5, and on the appropriate day, pregnant females were humanely euthanised and embryos collected. Importantly, some germline recombination has been observed to occur in the offspring of female Prrx1Cre⁺-expressing mice, therefore only male Cre⁺ mice were used in these experiments (Logan et al., 2002). Animals were maintained in a facility at the Biomedical Research Centre (UBC) and experimental protocols were conducted in accordance with approved and ethical treatment standards of the University of British Columbia.

Ketoconazole, diethylaminobenzaldehyde (DEAB) and at-RA were obtained from Sigma and dissolved in 95% ethanol. AGN194310 (4310) was kindly provided by Rosh Chandaratna, dissolved in 95% ethanol and used at a final concentration of 100 nM (Johnson et al., 1999). BMP4 was purchased from R&D, resuspended according to the manufacturer's instructions, and added to media at a final concentration of 20 ng/ml and 2 ng/ml, respectively.

Alcian Blue and Alizarin Red skeletal staining of embryos was performed as described (McLeod, 1980) with the slight modification that E18.5 embryos were dissected, and placed in tap water at 4°C for approximately 4–6 hours before being scalded in 70°C water for approximately 30 seconds. This facilitated the skinning and evisceration of embryos.

E11.5-E14.5 embryos were dissected in PBS and fixed in 0.2% glutaraldehyde in PBS for 30 minutes at room temperature. Embryos were covered in staining solution (2 mM MgCl₂, 0.02% NP40, 0.01% deoxycholate, 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, 0.1% X-gal) for 6 hours at 37°C. Embryos were post-fixed in 10% formalin and stored in 70% ethanol.

Primary cells were isolated from murine limb buds as previously described (Hoffman et al., 2006). Briefly, whole limbs were removed from E11.5 mouse embryos and micro-dissected into regions as described in Fig. 2 and supplementary material Fig. S1. Cyp26b1^+/+ and Cyp26b1^+/− limbs were pooled together. Limb pieces were proteolytically digested in a dispase solution, filtered through a cell strainer (40 μM) to obtain a single-cell suspension and resuspended at a density of 2.0×10⁷ cells/ml. 10 μl spots, or ‘micromasses’, were dispensed onto Nunclon delta SI plates, incubated, and allowed to adhere for 1 hour before addition of culture medium (40% DMEM, 60% F12, 10% FBS) with or without addition of factors. Addition of medium was considered time 0 and medium was changed every other day following establishment of cultures.

To assess chondrogenic activity, a SOX-responsive reporter was used as described previously (Weston et al., 2002). A trimerised RARE upstream of the firefly luciferase gene (pGL3-RARE-luciferase) was used to assess RA activity (Hoffman et al., 2006). Primary limb mesenchymal cultures were transiently transfected with Effectene (Qiagen) using a modified protocol as described (Karamboulas et al., 2010). Analysis of reporter gene activity was performed using the Dual Luciferase Reporter Assay System (Promega), according to the manufacturer's instructions. Firefly luciferase was normalised against Renilla luciferase activity to control for differences in transfection efficiency and to generate relative light units (RLU).

For visualisation of condensed mesenchyme, peanut agglutinin (PNA) staining was used. Briefly, primary cultures were fixed in 4% PFA for 30 minutes at 4°C and Rhodamine-labelled PNA (Vector Labs, 10 μg/ml in PBS) was added to cultures and incubated overnight at 4°C. The following morning, cultures were washed and the distribution of PNA-bound cells was visualised with fluorescence microscopy. Alcian Blue staining of cartilage nodules was performed as previously described (Weston et al., 2000). For alkaline phosphatase staining, cultures were fixed in 10% formalin for 15 minutes, and stained for 1 hour in the dark with alkaline phosphatase staining solution (Fast Red Violet and napthol AS-MX phosphate in 0.1 M Tris-HCl, pH 8.3).

MMP13 distribution was visualised using immunofluorescence. Primary cultures were fixed in 4% PFA and antigen retrieval was performed via PK digestion (20 μg/ml PK in TE buffer for 10 minutes at 37°C). Cells were then blocked in 1% BSA with 0.3 M glycine, and incubated with an anti-MMP13 antibody (Abcam, ab39012, 1:500, Cambridge, MA) overnight. The following day, cells were rinsed in PBS, incubated with secondary antibody conjugated to Alexa Fluor 488 (Molecular Probes, A11034, 1:1000) in PBS for 1 hour at 4°C, and MMP13 distribution was visualised and photographed using fluorescence microscopy. Control consisted of incubation with the secondary antibody alone.

RNA was harvested from primary cultures using RNeasy (Qiagen) as per the manufacturer's instructions. RNA was isolated directly from mouse limbs using Trizol (Invitrogen). After isolation, RNA was resuspended in 30 μl of DEPC-treated water and the quality and amount of RNA was assessed using a NanoDrop ND-1000 spectrophotometer.

Total RNA was reverse transcribed to cDNA using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems) as per manufacturer's instructions. Approximately 2 μg or the maximum amount of RNA available was transcribed in a 20 μl reaction volume. To follow the expression of various gene transcripts, quantitative real-time PCR was carried out using the 7500 Fast Real-Time PCR System (Applied Biosystems). Primers and TaqMan minor-groove-binding probes for Aldh1a2, Col1a1, Col2a1, Cyp26a1, Sox9 and Scx, were designed using Primer Designer 2.0 (Applied Biosystems) as previously described (Hoffman et al., 2006; Scott et al., 2010; Weston et al., 2002). The primer and probe concentrations were optimised according to the manufacturer's instructions. Other primer and probe sets were designed using IDT RealTime PCR Assay Design Tool (supplementary material Table S1) or purchased from the TaqMan gene expression collection (Applied Biosystems). Quantification was carried out using ~10 ng of total RNA using the standard curve method and expression of all genes relative to endogenous rRNA was determined using TaqMan Ribosomal Control Reagents (Applied Biosystems).

Images were acquired with a Q-imaging Retiga 1300i camera using Openlab5 Software, through either a dissecting microscope (Stemi SVII, Carl Zeiss Microimaging) or an inverted microscope (Axiovert S100, Carl Zeiss). The Cyp26b1^−/− and ScxGFP⁺ fluorescence images were obtained with a Leica MZFLIII (Richmond Hill, ON), using Q-imaging Retiga 2000R and Q-Capture Pro (Q-Imaging, Surrey, BC) software. Images were edited using Adobe Photoshop CS4 (Adobe Systems Incorporated, San Jose, CA) and all images in one panel were adjusted equally.

Luciferase assays, immunofluorescence, and Alcian Blue, PNA and alkaline phosphatase staining of micromass cultures were performed in triplicate or quadruplicate and repeated using at least three distinct preparations of cells. RT-qPCR of cultured cells was carried out in duplicate and repeated a minimum of three times with independent RNA samples. RT-qPCR using RNA isolated from tissues was performed on single embryos and repeated with similar results for a minimum of n=3. For studies involving lacZ or skeletal staining of embryos, n>3. One representative experiment is shown for all results.

Data was analysed using a Student's t-test with unequal variance or one-way analysis of variance (ANOVA), followed by Tukey post-test for multiple comparisons, using GraphPad Prism, version 5 (GraphPad Software, La Jolla, CA). Significance is represented as follows: *P<0.05, **P<0.01, ^#P<0.001.

We would like to thank Ronen Schweitzer (Shriners Hospital, Portland, OR) for the Scx-EGFP mice and Rosh Chandraratna for providing AGN194310. H.J.D. was supported by a training fellowship from the Skeletal Regenerative Medicine Canadian Institutes of Health(CIHR) Team grant, a graduate fellowship from the Canadian Arthritis Network, and a junior graduate studentship from the Michael Smith Foundation for Health Research. This research was funded by grants to T.M.U. from the CIHR, and T.M.U. holds an Investigator award from The Arthritis Society.