Retinoic Acid Receptor α Function in Vertebrate Limb Skeletogenesis: a Modulator of Chondrogenesis

A RARβ2 RARE double-stranded oligonucleotide probe was prepared by annealing the following oligos: 5′ TCGAGGGTAGGGTTCACCGAAAGTTCAC 3′ and 3′ CCCATCCCAAGTGGCTTTCAAGTGAGCT 5′. The probe was filled in using [³²P]dCTP and Klenow DNA polymerase. All proteins were purified using the pGEX-2T bacterial expression system (Pharmacia LKB Biotechnology, Piscataway, NJ) according to the manufacturer's instructions. Protein–DNA binding reactions were carried out in 20 μl of binding buffer (10% glycerol, 10 mM Hepes, pH 8.0, 50 mM KCl, 2.5 mM MgCl₂, 1 mM ZnCl₂, 1 mM DTT) containing 1 μg poly dI-dC and 0.15 μg of hRARα or hRARα-CB with or without 0.15 μg of mRXRβ on ice for 10 min. After this preincubation, 0.1 ng of labeled probe (67,000 cpm) was added to each tube, and the reactions were allowed to proceed for an additional 20 min at room temperature. Unlabeled probe (50 ng, 500×) was included in some reactions as a specific competitor. The samples were resolved on a prerun 4% nondenaturing polyacrylamide gel at 4°C.

The transgene was constructed in the pGEM9zf(−) derivative, pW1, which contains a BamHI restriction site juxtaposed between the HindIII and SpeI restriction sites and the polyadenylation sequence from SV-40 in the NsiI site (Balkan et al., 1992a). A 3.3-kb fragment of the Hoxb-6 promoter that contains sequences immediately upstream of the transcriptional start site was liberated from pKSIIHoxb-6 (Schughart et al., 1991) by restriction endonuclease digestion with EcoRI and BamHI and directionally subcloned into pW1, thereby generating pW1Hoxb-6. pW1Hoxb6tgRARα1 was made by subcloning the constitutively active RARα1 (Balkan et al., 1992 b) into the BamHI restriction site of pW1Hoxb-6. Hoxb-6tgRARα1 DNA for microinjection was released from the vector backbone using NotI and SfiI, separated on agarose gel, and gel purified using geneclean (BIO 101, La Jolla, CA) according to the manufacturer's instructions. Transgenic mice were made by microinjection of C57Bl6F₁/J- (Jax Laboratories, Bar Harbor, ME) fertilized mouse eggs with DNA at a concentration of 2–3 ng/μl. Mice that carried the transgene were identified by slot-blot hybridization of tail DNA with an Escherichia coli β-galactosidase DNA probe.

Embryos were stained for β-galactosidase activity using a protocol described by Balkan et al. (1992a). The morning of the day of the copulation plug was considered E0.5.

Differential skeletal staining of embryos with alcian blue 8GS and alizarin red S was performed as described by McLeod (1980). Postterm mice were stained using a similar protocol with extended periods of incubation.

Western analysis was performed using standard procedures. Hind limb bud protein extracts were prepared from E11.5 animals. A cut parallel to the body wall was made to release the limb buds, and hind limb bud pairs from individual animals were transferred to 100 μl of treatment buffer (0.125 M Tris-HCl, pH 6.8, 4% SDS, 20% glycerol, 10% 2-mercaptoethanol) and boiled for 5 min with intermittent mixing. Protein extracts (10 μl for J limbs, 20 μl for I limbs) were separated by SDS-PAGE and transferred to nitrocellulose. Blots were incubated with an anti–β-galactosidase antibody (Promega Corp., Madison, WI) for 2 h and washed, and an anti– mouse secondary antibody conjugated to alkaline phosphatase was added for 1 h. Alkaline phosphatase activity was detected using a mixture of nitro blue tetrazolium and 5-bromo-4-chloro-3-indolyl-phosphate (Promega Corp.) according to the manufacturer's instructions.

Hind limbs of E11.5 embryos were collected aseptically and teased apart into small pieces with forceps in Puck's saline G buffer. Cells were dissociated by incubation in the presence of dispase (1 U/ml) (GIBCO BRL, Gaithersburg, MD) with shaking at 37°C in Puck's saline A buffer containing 10% chicken serum. Digestion was terminated by resuspension in SCM (60% Ham's F12, 40% DME supplemented with antibiotics and 10% FBS) followed by gentle trituration to obtain a fine cell suspension. Cells were passed through a No. 20 Nitex (Tetko Co., Elmsford, NY) mesh to remove cell clumps and to obtain a single cell suspension. Cell number was adjusted to 1 × 10⁷ cells/ml and micromass cultures were intiated by carefully placing 10 μl of the cell suspension into the center of a well in a 24-well plate. Cultures were placed in a 37°C incubator containing a humidified atmosphere and 5% CO₂. After attachment (∼1.5 h), 1 ml of SCM was added to each culture, and cultures were returned to the incubator. Media changes were performed at least once a day. Cultures were incubated for a period of 1 to 6 d, at which time they were processed for β-galactosidase activity, and duplicate cultures were stained for the presence of sulfated glycosaminoglycans found in cartilage matrix using alcian blue at acidic pH (Lev and Spicer, 1964). Alcian blue staining was carried out as follows. Cells were washed twice with PBS, fixed for 15 min in Kahle's fixative, washed once with 0.1 N HCl, and stained overnight with 0.1% alcian blue in 0.1 N HCl. After staining, the cultures were washed twice with 70% ethanol and stored in the same. Nodule numbers were determined by counting the number of alcian blue–stained nodules within a fixed area (3 mm²). Statistics were determined using the Student's t-test.

Single-stranded RNA probes with incorporated digoxigenin were transcribed from linear DNA templates according to manufacturer's instructions (Boehringer Mannheim Corp., Indianapolis, IN). Probes were quantified by spotting various dilutions of probe in 10× SSC (1.5 M NaCl, 150 mM sodium citrate) and a control digoxigenin-labeled RNA on a nylon membrane; detection of probe using a sheep anti–digoxigenin conjugated alkaline phosphatase antibody (Boehringer Mannheim Corp.) was as described by the manufacturer.

Probes for mRARα and mRARγ were produced from templates that included part of the 3′ coding and 3′ untranslated region of mRARα and the 3′ DEF region of RARγ. Hybridization probes were synthesized from linearized plasmid DNA templates using in vitro transcription. An ∼600bp riboprobe for mRARα was made using T3 RNA polymerase and a pKSII (Stratagene, La Jolla, CA) plasmid cleaved with EcoRV (cleaves internally) that carried a 2.2-kb mRARα insert. mRARγ riboprobe was synthesized from a pBS (Stratagene) template containing a 1.0-kb mRARγ insert that had been linearized with EcoRI.

Whole-mount in situ hybridizations of micromass cultures were performed as described by Conlan and Rossant (1992) with modifications. Micromass cultures were washed twice with PBS and fixed in 4% paraformaldehyde overnight at 4°C. After fixation, cultures were washed three times with cold PBS and dehydrated through a PBS/methanol series to 100% methanol and stored in 75% methanol at 4°C until use. Before hybridization, cultures were rehydrated in PBT (PBS, 0.05% Tween-20) and treated with proteinase-K for 6 min, washed twice with 2 mg/ml glycine in PBT, and fixed for 20 min in 4% paraformaldehyde. Cultures were washed three times with PBT and twice in hybridization buffer (50% formamide, 0.75 M NaCl, 10 mM Pipes, pH 6.8, 1 mM EDTA, 100 μg/ml tRNA, 0.05% heparin, 0.01% BSA, 1% SDS) and prehybridized for at least 1 h at 55°C. Hybridizations were carried out overnight at 55°C in a programmable dry oven (Bellco Glass Inc., Vineland, NJ) with 1 μg/ml riboprobe. Conditions for washing and antibody incubation were similar to that previously described (Conlon and Rossant, 1992). Color development was carried out overnight in the dark in 10% polyvinyl alcohol (Aldrich, average mol wt 13,000–23,000) (Barth and Ivarie, 1994). Addition of polyvinyl alcohol was found to increase signal to noise ratio.

Transgenic mice were generated that expressed a constitutively active human RARα1 (a chimeric receptor with β-galactosidase fused to the carboxy terminus, denoted tgRARα herein) (Balkan et al., 1992,b) to high levels in the developing limb. In vitro analysis of tgRARα binding to a DNA probe containing the βRARE showed that efficient binding was dependent upon the presence of an RXR heterodimer partner (Fig. 1), consistent with that observed for wild-type RARs (Yu et al., 1991; Leid et al., 1992b). The murine Hoxb-6 (Hox 2.2) promoter fragment, which has been shown previously to direct expression of a LacZ reporter gene to the developing limbs (Schughart et al., 1991), was used to target expression of the fusion receptor in the transgenic animals.

Transgenic mice carrying tgRARα under the control of the murine Hoxb-6 promoter fragment were generated by injecting one-cell embryos with the construct shown in Fig. 2 A. Owing to the presence of RAREs in several Hox promoters (Langston and Gudas, 1992; Popperl and Featherstone, 1993; Marshall et al., 1994), we checked the ability of both hRARα and tgRARα to trans-activate the Hoxb-6 promoter fragment in transient expression assays. Neither hRARα nor tgRARα was able to trans-activate the Hoxb-6 promoter fragment in the absence or presence of 500 nM all-trans RA (data not shown). Screening of transgenic mice was done by slot-blot hybridization of DNA isolated from mouse tails. 12 founders were generated and used to establish eight lines of transgenic mice. Of the 12 founders generated, four exhibited similar congenital limb malformations (A–D, see below), while eight (E–L) appeared phenotypically normal. Original founders as well as male and female progeny from all eight lines have been studied extensively, and a range of phenotypic abnormalities have been observed, depending on the level of transgene expression in the developing limbs.

The expression of tgRARα in the developing limbs of embryos derived from the eight phenotypically normal transgenic founders was analyzed in detail. At E10.5, the transgene was expressed in the mesenchyme of the emerging limb buds and in the ventrolateral mesenchyme between the limb buds. In the fore limb bud, β-galactosidase activity was restricted to the posterior region whereas the entire hind limb bud expressed tgRARα (Fig. 2, B and D). While no differences in the transgene expression pattern were seen between the eight lines, significant differences in tgRARα expression levels were observed. In three of the eight lines (J, K, and L), tgRARα was expressed at high levels in the posterior half of the fore limb and across the entire hind limb (Fig. 2,B). In the other five lines (E–I), however, low levels of transgene expression were seen in both the fore and hind limbs (Fig. 2,D). At E11.5, differences between the two levels of transgene expression persisted, which was further confirmed by Western analysis with an anti–β-galactosidase antibody (data not shown). Specifically, in the hind limbs of embryos derived from high expressing founders, transgene expression was seen across the whole anteroposterior axis and extended distally into the progress zone (Fig. 2,C). In embryos derived from low expressors, tgRARα expression was restricted to anterior and posterior stripes in the mesenchyme of the developing hind limb with no significant expression of the transgene in the progress zone of the fore or hind limb (Fig. 2,E). Tissue sections from E12.5 high expressor hind limbs confirmed that transgene expression was not present in the ectoderm but was present in chondrogenic condensations that will become the tibia and fibula and strongly expressed in the underlying mesenchyme, including the progress zone (data not shown). These results are summarized in Table I. Furthermore, a transgenic line that expressed a potent dominant-negative RAR to a level similar to that of the high transgene-expressing lines described herein exhibited no perceptible phenotype suggesting that the tgRARα is not functioning in a dominant-negative manner (our unpublished observations).

While founders J, K, and L were phenotypically normal, they did pass the transgene in a Mendelian fashion to their progeny, and all transgene positive offspring exhibited limb malformations like those seen in the four affected founders (Fig. 3). Because some founder males with limb malformations had difficulty mating, lines established from male J and female L were relied upon for most of the analysis of the effects of ectopic tgRARα expression on limb development. Interestingly, no limb malformations were observed in the offspring of the five phenotypically normal founders (E–I) with the lower levels of tgRARα expression, suggesting that high level expression of the fusion receptor is required for alterations in normal limb development.

The limb malformations associated with expression of the transgene were analyzed in detail in embryos and animals derived from a representative line (J) (Figs. 4 and 5); similar malformations were seen in four founders and two other independent transgenic lines. Pronounced bilateral defects were always present in the hind limb, whereas defects in the fore limb were variable. In the affected animals, there was a predilection towards the loss of postaxial structures, and, in general, the degree of postaxial deflection in the fore limb appeared to be a good indicator of the severity of the defect.

Malformations in the fore limb (Fig. 4) were primarily restricted to the field of transgene expression that included the ulna, ulna carpal bone, and their respective articulating bones. Although the transgenic animals exhibited a marked variability in the severity of the defect in the fore limbs, all the transgenic animals contained a deformed ulna, which varied from mild with a slight postaxial curvature to severe. The ulna in the severely affected animals was dramatically thickened (in some cases to twice that of the size of the radius), postaxially deflected, and foreshortened (compare Fig. 4, A and E). In comparison, the radius and humerus appeared to be only mildly affected. Furthermore, the radius was always found to be longer than the ulna, and the degree of this disparity appeared to correlate well with the amount of postaxial bending, suggesting that the disproportionate growth of the ulna and the radius in the proximodistal axis most likely contributed to the postaxial deflection. Malformations in the wrist consisted of a reduction in the size of the ulnar carpal and fourth carpal bone with the loss of the fourth carpal–fifth metacarpal articulating surface (Fig. 4, D and E). The pisiform bone also appeared to be sometimes absent or not mineralized. Most of the metacarpals were shortened, and in some of the mildly affected animals, metacarpal 5 was reduced to a mineralized nodule with no corresponding phalanx (Fig. 4,D). Additionally, in some of the more severely affected animals we periodically observed a very broad fourth medial phalanx with a complete duplication of the fourth distal phalanx (Fig. 4 E).

Founders A, B, C, and D, as well as those transgenic animals derived from founder J, all exhibited bilateral hind limb malformations (Fig. 5). In all animals examined, the fibula was malformed; this ranged from complete absence, except for a small distal nodule, to thickened partial fibulas that extended more proximally (Fig. 5, B and D). In addition, the extent of the fibular defect was found to be nearly perfectly symmetrical and to be 100% penetrant, with variability in the degree of severity of the effect on the foot and ankle. No proximal fibula–associated structures were observed using alizarin red or alcian blue staining. In contrast, the tibia was found to be only marginally affected, exhibiting a dorsal curvature and slight shortening and thickening. The calcaneus was found to be reduced in size, displaced laterally to the outside of the fibula, and missing an articulation surface for the fifth metatarsal. In some animals, the navicular and the cuboidum were found to be fused to the second and third tarsals, respectively. Occasionally, the fifth proximal and medial phalanges were missing, with the presence of the distal phalanx (Fig. 5 F). As observed in the fore limb, the defects in the hind limb appear to be primarily restricted to the posterior half of the limb and to involve the distal long bones and proximal foot bones.

To establish that the transgene does not have nonautonomous effects on chondrogenesis in regions where it is not expressed, the axial bones of transgenic and control animals were stained with alizarin red and alcian blue. No significant differences were observed in these bones between control and transgenic animals of similar embryonic age (data not shown).

To understand what changes have occurred during development of the limb to give rise to the aforementioned malformations, a series of E13.5 through E18 fetuses were collected and stained with alizarin red and alcian blue (Fig. 6). Transgenic animals were readily distinguishable from their nontransgenic littermates as early as E11.5. The hind limbs of the transgenic animals were somewhat smaller and contained a disproportionately large outgrowth in the region of the anterior margin, which will give rise to digit 1 (data not shown). Similar outgrowths are also observed in the chick limb bud after placement of an RA-soaked bead in the anterior margin (Summerbell, 1983). At later times in development, digit 1 was found to be duplicated, thereby generating a 1, 1, 2, 3, 4, 5 digit pattern (Fig. 6,B, E15.5, E17, and E17.5). This duplication was seen in ∼40% of the animals, and the extent of the duplication correlated well with the size of the outgrowth. In most instances, this duplicated digit was not maintained, as an analysis of many (n > 200) adult animals revealed no such digit duplication. As can be seen from the embryonic panel, the digit duplication appears to be lost during the course of development with the duplicated digit 1′ becoming fused with digit 1 (Fig. 6,B, E17 and E18). This was evident in the hind limbs of the adult animals, where digit 1 appeared to be composed of several individual and overlapping bones (Fig. 5 F).

Development of the various elements of the hind limb of transgenic animals was significantly delayed in comparison to wild-type animals. This was most apparent at later embryonic stages, where ossification of the proximal phalanges of the hind limbs from transgenic animals was observed to be at least 0.75–1 d behind that of wild-type animals. However, earlier in development the disparity in the size of the limbs was also evident; in this case, the bone primordia were shortened and slightly misshapen. Analysis of hind limb histological sections prepared from E13.5 fetuses revealed that chondrogenesis of the tibia and fibula was delayed in transgenic animals (Fig. 7). The primitive fibula of the wild-type animals contained maturing chondrocytes circumscribed by a well-defined perichondrium. In contrast, the fibula from transgenic animals was not as well organized. The fibula was primarily composed of closely packed chondroprogenitor cells with a lower cytoplasmic/nuclear ratio with a morphology consistent with condensed mesenchymal cells. Chondrogenesis of the digital rays were also found to be delayed in the transgenic animals (data not shown). Hence, overexpression of tgRARα appeared to inhibit or interfere with chondrogenesis in the hind limb by delaying chondroprogenitor differentiation.

To examine more thoroughly the defect in transgenic animals that interferes with chondrogenesis in endochondral bone formation, the ability of the mesenchyme from transgenic limbs to form cartilage nodules was assayed in vitro using micromass culturing methods. Limb mesenchyme cultured under micromass conditions closely follows the progression of limb mesenchyme in vivo, precartilaginous condensations are first evident after 2 d in culture with cartilage nodules appearing ∼1 d later (Ahrens et al., 1977). Cultures were seeded at high density (1 × 10⁵ cells/ 10 μl) in a 24-well plate with dissociated limb mesenchyme pooled from either fore or hind limbs of E11–11.25 transgenic or non-transgenic embryos. These cultures were allowed to grow for up to 3, 4, or 6 d. Cartilage matrix production was assayed for by fixing and staining the cultures with alcian blue at low pH , while both the extent and distribution of transgene-expressing cells were determined by staining fixed cultures with X-gal.

Within 3 d, cultures initiated from wild-type fore or hind limb buds showed numerous well-defined alcian blue– stained cartilage nodules (Fig 8, A and B). No alcian blue nodules were observed in cultures after 2 d; instead, both fore limb and hind limb cultures contained numerous condensations (data not shown). Similar to wild-type cultures, transgenic fore and hind limb micromass cultures after 2 d displayed many condensations (data not shown). Transgene-expressing cells were found to be distributed throughout the cultures and within condensations. In the hind limb cultures, the transgene-expressing cell was the predominant cell type found within the condensations, while in the fore limb there was a mixture of cells, with some condensations being devoid of transgene-expressing cells. The number of transgene-expressing cells from fore limb– derived cultures were much lower than those of the hind limb cultures, which is consistent with the differences observed in transgene expression between the fore and hind limb buds in vivo.

In contrast to wild-type cultures, cultures initiated from the hind limb buds of transgenic animals exhibited few if any alcian blue–stained nodules at 4 d (Fig. 8,C) and only slightly more nodules (<5) at 6 d postinitiation (Fig. 8,G). Examination of these cultures with dark-field microscopy confirmed the absence of any apparent nodules or discrete condensations (data not shown). In contrast, cultures initiated from the fore limbs of transgenic animals displayed numerous well-defined alcian blue–stained cartilage nodules after 4 d (Fig. 8,A) and slightly more after 6 d (Fig. 8,G). However, their appearance was delayed and their number was significantly reduced in comparison to wildtype cells (Fig. 8,G). In addition, the nodules were mostly confined to those regions containing nontransgene-expressing cells (Fig. 8,E). Furthermore, the differences in chondrogenesis observed between the cultures of fore and hind limb buds from transgenic animals correlated well with the observed differences in the severity of fore and hind limb malformations seen in these animals. It should be noted that the transgene continued to be abundantly expressed throughout the culture period in both fore- and hind limb– derived cultures (Fig. 8, E and F), and that there was no significant growth disparity between wild-type and transgenic cultures. Consistent with the above in vivo observations, these results suggest that expression of tgRARα delayed or inhibited chondroblast differentiation subsequent to condensation.

Previous studies have shown that RARα and RARγ are expressed throughout the fore limb mesenchyme between E9.5–E11.5, at which time RARγ expression becomes restricted to precartilage and cartilage (Dollé et al., 1989,b; Ruberte et al., 1990). Whole-mount in situ hybridization methodologies were adapted to the analysis of gene expression in micromass cultures so that the expression of endogenous RARα could be compared to tgRARα. In agreement with previous studies, by day 4 of culture, RARγ was found to be expressed in the core of the nodule, where chondroblasts surrounded by cartilaginous extracellular matrix were observed (Fig. 9, B and D). RARα, on the other hand, was found to be expressed in the peripheral region of the nodule in condensed mesenchymal cells that had not yet differentiated into chondroblasts (Fig. 9, A and C). RARα expression was either absent or expressed at very low levels within the core of the nodule, in chondroblasts (Fig. 9, A and C).

Transgenic animals that ectopically express a constitutively active hRARα1 in the developing murine limb exhibited a spectrum of defects that included polydactyly, syndactyly, ectrodactyly, fibular deficiencies, and tarsal and carpal fusions. The phenotypes observed in these animals are similar to those that have been observed in a number of vertebrate systems in which embryos were administered systemic or localized doses of RA (Shenefelt, 1972; Kochhar and Aydelotte, 1973; Summerbell, 1983; Tickle et al., 1985, 1989). A number of different roles have been proposed for the function of the RARs during limb development (Mendelsohn et al., 1991, 1992), some of which include precartilage and cartilage formation, apoptosis, and patterning of the limb. We have found that RARα is an important regulator of chondrogenesis during limb development.

The RARs are ligand-activated receptors that function as heterodimers. Hence, transgene activity would rely on the presence of a suitable heterodimeric partner. This condition is satisfied during the period of transgene expression before overt cytodifferentiation (E13–13.5) within the limb. RXRα and RXRβ are expressed throughout the limb mesenchyme at these stages and as such could provide a suitable heterodimeric partner for the transgene.

Postnatal transgenic lines that expressed the transgene to high levels during limb development (E9.5–E15) presented with a number of paraxial skeleton deficits, while no apparent phenotype was observed in transgenic lines expressing the transgene to a lower level. Although the expression patterns for the transgene in the two lines were similar, the transgene in the high expressing lines is expressed to a much greater level relative to that of the low expressing lines in the core mesenchyme (along the median axis). This mesoderm will contribute to the future skeletal elements, suggesting that transgene expression in this region may intefere with skeletogenesis and contribute to the observed defects. Consistent with these observations, the limb defects became first evident during embryogenesis within the early stages of skeletogenesis (E13.5) during formation of the cartilaginous template. Furthermore, histological sections prepared from E13.5 limbs from transgenic and a nontransgenic littermate confirmed that development of the limb cartilage was significantly delayed in the transgenic animals. Taken together, these observations suggested that expression of the transgene interfered with chondrogenesis.

Micromass cultures have been previously shown to closely follow those events of chondrogenesis in vivo and have been used to study the role of various molecules in chondrogenesis during endochondral bone formation. Cartilage nodules are first evident as a coalescence of mesoderm cells that form a condensation. These aggregates continue to expand both by cell division and selective recruitment of neighboring cells. After a period of cellular growth, cells within the interior of the condensation begin to differentiate and elaborate cartilage matrix that can be specifically stained with alcian blue. Micromass cultures derived from wild-type and transgenic hind limbs appear similar up to ∼2–2.5 d. Before this time, there is a similar number of well-defined condensations in both cultures. It is at the onset of chondroblast differentiation when differences between the two cultures become readily visible. In the wild-type cultures, chondroblasts begin to appear within the core of the condensations after 3 d in culture as shown by alcian blue staining for cartilage matrix. Whereas, in transgene-expressing cultures, especially those from hind limbs, little or no alcian blue staining is observed after 4 or 6 d in culture, which is indicative of an absence of chondroblast differentiation and elaboration of extracellular cartilage matrix. Hence, transgene expression appears to interfere with the transition from a committed chondroprogenitor cell, found within condensing mesoderm, to a chondroblast located within the core of the nodule.

Analysis of RARα gene expression in micromass cultures after 4 d has shown that RARα mRNA is absent in the core of the nodules but is abundantly expressed in the condensing mesoderm. Therefore, RARα mRNA is downregulated during chondroblast differentiation. In contrast, RARγ appears to be upregulated in the core of the nodules in concert with chondroblast differentiation. In cultures initiated from transgene-expressing limb mesenchyme, RARα is found to be abundantly expressed in condensing mesoderm throughout the culture period and continues to be expressed within the core of condensations well after endogenous RARα expression is downregulated. Together these results suggest that continued expression of RARα during mesoderm-chondroblast differentiation interferes with the appearance of chondroblasts. Hence, downregulation of RARα appears to be a necessary step in chondroblast differentiation (Fig. 10). This model has consequences for both understanding the role of the receptors in teratogenesis as well as in the phenotypes observed in RAR null mutants.

RA has been shown to modify limb development in a number of phylogenetically distinct species (Tabin, 1991; White et al., 1994). Many of the phenotypes associated with overexpression of tgRARα in the developing limb are similar to those observed in the fetuses of dams treated with teratogenic doses of RA. RA teratogenesis causes a spectrum of defects within the limbs that are most commonly reductionist in nature and are thought to be, in part, a consequence of a delay in chondrogenesis (Kwasigroch and Kochhar, 1980; Zimmerman and Tsambos, 1985; Kwasigroch et al., 1986). These malformations include, within the hind limb, fibular agenesis, brachypodism, syndactyly, ectrodactyly, bowing of the tibia, and syntosis of the zeugopods (Shenefelt, 1972; Kochhar, 1973; Kochhar and Aydelotte, 1974). All of these defects are found to varying extents in the hind limbs of the tgRARα transgenic mice described herein.

Several studies have shown that RA inhibits chondrogenesis and that this inhibition is most likely RAR-mediated (Eckhardt and Schmitt, 1994; Von Schroeder et al., 1994). Further studies with receptor-specific agonists have shown that an RARα-specific agonist is significantly more potent in inducing limb malformations than either RARβ- or RARγ-specific agonists (Elmazar et al., 1996). RARγ is expressed during chondrocyte maturation, whereas RARα and RARβ are not; RARα appears to be expressed in most other tissues of the developing limb. Therefore, under conditions of retinoid excess where expression of RARβ (Mendelsohn et al., 1991; Jiang et al., 1994) and possibly RARα (both genes are RA responsive) are elevated within the limb mesenchyme, a delay or inhibition of chondroprogenitor differentiation could ensue, thereby interfering with normal skeletogenesis. No differences in the expression of RARβ were observed between the limbs of tgRARα transgenic mice and wild-type animals (our unpublished observations), and expression was restricted to the anterior and posterior necrotic zones and in the interdigital spaces, suggesting that RARβ is not contributing to the observed phenotype in the transgenic mice.

Application of RA to the anterior margin of the developing chick limb has been shown to induce digit and skeletal element duplications (Tickle et al., 1982). Minor digit duplications are apparent in the transgenic animals described herein. Duplications were observed in the fourth distal phalanx within the forelimb and digit 1 in the hind limb. Both of these duplications occurred at the anterior boundary of transgene expression: digit 4 in the fore limb and digit 1 in the hind limb. Numerous genes have been shown to be important in outgrowth and patterning of the vertebrate limb, and some of these include sonic hedgehog (Riddle et al., 1993), HoxD cluster (Dollé et al., 1989a), and fibroblast growth factors (Niswander et al., 1993, 1994). Analysis of the expression patterns of sonic hedgehog, fibroblast growth factor 4, and Hoxd9 through d12 mRNAs with whole-mount in situ hybridization showed no dramatic changes between transgenic and control animals; staining patterns were found to be consistent with those reported by others (our unpublished observations). These results suggest that the change in patterning observed in the transgenic animals may not involve changes in the expression of one of these gene products, or because the duplications are relatively minor (fourth distal phalanx in the fore limb and periodic partial digit 1 duplication in the hind limb), any changes in expression of these genes might be fairly subtle and therefore difficult to detect.

Recently, mice have been made with null alleles of RARα, RARβ, RARγ, and RXRα (Lohnes et al., 1993; Lufkin et al., 1993; Sucov et al., 1994; Luo et al., 1995). Mice homozygous for either mutant allele do not show any phenotype in the paraxial skeleton, whereas mice homozygous for both, (RARα−/−, RARγ−/−), have severe limb malformations (Lohnes et al., 1994). The limbs of these animals display proper anteroposterior symmetry, suggesting that the primary defect is not at the level of pattern formation (Lohnes et al., 1994). With the exception of the RXRα null mice, treatment of RAR null embryos with teratogenic doses of RA results in limb defects comparable to that seen in heterozygous or wild-type embryos. The observation that RXRα null mutants are resistant to RAinduced limb defects underscores the importance of RXRα as an integral component of the RA signaling cascade in vivo (Sucov et al., 1995). It is interesting to note that in both RARα/RARβ2 and RARα/RARγ null mice, the mice display ectopic cartilages in a number of sites throughout the animal. Furthermore, ectopic cartilages were sometimes present in the interdigital region, and the synovial joints were poorly defined in RARα/RARγ null mice. Given that the presence of RARα and possibly the other RARs may be important in regulating cartilage formation, this suggests that their absence creates permissive conditions and could be expected to give rise to ectopic cartilages, as observed in these animals.

We have provided evidence that the loss of RARα expression appears to be an important regulatory component of mesoderm differentiation into chondroblasts in that chondroprogenitor cells that express RARα failed to differentiate into chondroblasts. Continued expression of RARα in cells destined to become chondroblasts contributes to a different cellular fate, which is currently under investigation. Overall, downregulation of the RARs may be an important determinant in regulating cellular differentiation which has practical implications in understanding their role in normal and abnormal development.

The authors would like to thank Drs. T.J. Montine, M.C. Colbert, A.L. Darrow, and L.E. Kotch for helpful discussions, as well as J. Burchette.

E

embryonic day

RA

retinoic acid

RAR

RA receptor

RXR

retinoid X receptor

tgRARα

transgenic RARα