Differential contributions of AF-1 and AF-2 activities to the developmental functions of RXRα

Retinoids (vitamin A derivatives) are crucial for many aspects of vertebrate physiology and homeostasis (reviewed in Sporn et al., 1994). They also play essential roles in morphogenesis and organogenesis, as inferred from the large spectrum of developmental abnormalities displayed by vitamin A-deficient (VAD) fetuses (reviewed in Kastner et al., 1995; Dickman et al., 1997; White et al., 1998). Two families of nuclear receptors for retinoids, the RARs (α, β and γ isotypes and their isoforms; activated by all forms of retinoic acid, RA) and the RXRs (α, β and γ isotypes and their isoforms; activated by 9-cis RA only) transduce the retinoid signal. RARs and RXRs act in transfected cells in vitro as ligand-dependent transcriptional transregulators through binding as RAR/RXR heterodimers to cis-acting RA response elements present in cognate reporter genes (reviewed in Leid et al., 1992; Chambon, 1996; Mangelsdorf et al., 1995; Mangelsdorf and Evans, 1995). In addition, RXRs can act as heterodimerization partners with nuclear receptors other than RARs, such as thyroid hormone receptors, vitamin D3 receptor, peroxisome proliferator activated receptors and several orphan receptors (reviews: Mangelsdorf and Evans, 1995; Chambon, 1996; Perlmann and Evans, 1997). Importantly, RAR-specific and RXR-specific ligands synergize to activate target genes or to elicit biological responses in cell systems (Taneja et al., 1996; Taneja et al., 1997; Roy et al., 1995; Chen et al., 1996; Horn et al., 1996; Nagy et al., 1995; Vivat et al., 1997), as well as in whole embryos (Minucci et al., 1996; Minucci et al., 1997; Lu et al., 1997), indicating that both partners of the RAR/RXR heterodimer can be transcriptionally active. However, the liganded RXR is not active unless its RAR partner is itself liganded (Roy et al., 1995; Apfel et al., 1995; Chambon, 1996; Taneja et al., 1996; Taneja et al., 1997; Chen et al., 1996; Vivat et al., 1997; Botling et al., 1997).

To investigate the involvement of RARs and RXRs in the transduction of retinoid signals under physiological conditions in vivo, all RAR and RXR genes, and most of their isoforms have been knocked out in the mouse (for reviews see Kastner et al., 1995; Mark et al., 1999). Altogether, these studies have led to several important conclusions concerning the role of these receptors during morphogenesis and organogenesis: (1) RARs and RXRs mediate the developmental functions of RA; (2) functional redundancies exist among the various RARs and among the various RXRs; (3) RAR/RXRα heterodimers are most likely the main functional units that transduce the retinoid signal during development; (4) transcriptional activation of both partners in RAR/RXRα heterodimers are often required to activate target genes and to mediate under physiological conditions the retinoid effects in morphogenesis and organogenesis.

The N-terminal A/B domain of RXRs contains an autonomous ligand-independent transcriptional activation function called AF-1, whereas the C-terminal, ligand-binding domain E, contains a ligand-dependent transcriptional activation function, AF-2 (reviewed in Leid et al., 1992; Chambon, 1996; Moras and Gronemeyer., 1998). For each RXR, at least two isoforms exist, which differ in their N-terminal region (Leid et al., 1992; Fleischhauer et al., 1992; Liu and Linney, 1993; Nagata et al., 1994). As to RXRα, the major isoform RXRα1 is widely expressed in embryos and adults, whereas RXRα2 and α3 are restricted to the adult testis (Brocard et al., 1996). Furthermore, RXRα can be phosphorylated at several serine and threonine residues in its A/B domain (Adam-Stitah et al., 1999). The AF-2 activity crucially depends upon a conserved amphipathic α helix (the AF-2 AD core; Bourguet et al., 1995; Chambon, 1996; Wurtz et al., 1996; and references therein), whose deletion in the mouse has revealed its requirement for a number of RA-dependent developmental events (Mascrez et al., 1998). However, little is known about the mechanisms through which AF-1 activates transcription or about the relevance of the A/B domain in the global activity of the receptor under physiological conditions in vivo. Depending on the promoter context, the AF-1 of a given RXR modulates the AF-2 activity in cultured cells (Nagpal et al., 1992; Nagpal et al., 1993; Dowhan and Muscat, 1996; Chambon, 1996; Taneja et al., 1997). Thus, the transcriptional activity of a given RXR isoform may ultimately be determined, not only by its AF-2 activity, but also by its isoform-specific A/B domain.

In order to determine the importance of RXRα AF-1 domain A/B during development in vivo, we have engineered a mouse mutant line that expresses a truncated RXRα protein (RXRαΔA/B). The phenotypic analysis of mice carrying this mutation indicate that RXRα AF-1 domain A/B is important for transducing RA signals in vivo. In addition, we have assessed the relative contribution of RXRα AF-1 and AF-2 activities in embryonic development and placentation.

The RXRα targeting vector (pB48) was designed to delete exon 2 (with the exception of its splicing acceptor site), intron 2 and 102 bp of exon 3. On the one hand, a 4.6 kb SpeI-BamHI DNA fragment (from a P19 teratocarcinoma cell genomic library; Clifford et al., 1996), containing a BglII restriction site introduced at the beginning of exon 2 through site-directed mutagenesis, was subcloned into pEMBL19+ (Dente et al., 1983), leading to plasmid pB41. On the other hand, a 3.0 kb BamHI-XbaI DNA fragment (from a 129/sv mouse genomic library; Kastner et al., 1994), containing a BglII site introduced in exon 3 by site-directed mutagenesis, was cloned into pB41, reconstituting then the 7.6 kb SpeI-XbaI region of the RXRα locus (plasmid pB42). The 4.0 kb-long DNA located between the two BglII sites of pB42 was deleted and replaced by an oligonucleotide preserving the RXRα open reading frame, and containing a NheI restriction site (plasmid pB45). The 3.6 kb SpeI-XbaI fragment of pB45 was then cloned into a pBluescript plasmid (Stratagene) containing a 5.7 kb XbaI-HindIII genomic fragment (from a 129/sv mouse genomic library), leading to pB46. Finally, a loxP site-flanked TK-NEO fusion cassette (Metzger et al., 1995) was introduced at the XbaI site located in intron 3 (leading to construct pB48).

Embryonic stem (ES) cells were transfected with NotI-linearized pB48, as previously described (Lufkin et al., 1991). Out of 87 G418-resistant clones, two targeted clones (HS23 and BL19) were obtained. Homologous recombination was confirmed by Southern blot analysis of ScaI-, NheI- and SpeI-digested genomic DNA hybridized with probes A, B and Neo (see Fig. 1 and data not shown). In order to delete the ‘floxed’ TK-NEO cassette, HS23 ES cells were transiently transfected with pSG5-Cre (Gu et al., 1993). Excision was confirmed for clone HS23.26Cre by Southern-blot analysis (Fig. 1 and data not shown). To generate the first RXRαaf1^o mouse line, HS23.26Cre cells were injected into C57BL/6 blastocysts. To generate a second RXRαaf1^o mouse line, BL19 cells were used. In this line, the selectable marker was successfully excised by crossing heterozygotes with transgenic mice expressing Cre early during embryogenesis (CMV-Cre; Dupé et al., 1997). The phenotypes illustrated here are from line HS23.26Cre. However, identical observations were made using line BL19.

Nuclear extracts were prepared from whole E12.5 embryos (Andrews and Faller, 1991). Proteins (15 μg) were separated on 10% gel by SDS-PAGE and transferred onto nitrocellulose membranes. RXRα and RXRαΔA/B proteins were detected using the 4RX3A2 anti-RXRα monoclonal antibody (1/500 dilution; Rochette-Egly et al., 1994) and revealed by chemiluminescence according to the manufacturer’s instructions (Amersham).

Embryos and fetuses were fixed in Bouin’s fluid, embedded in paraffin, serially sectioned and stained with Groat’s Hematoxylin and Mallory’s Trichrome (Mark et al., 1993). Skeletons of E18.5 fetuses were prepared as described (Lufkin et al., 1992). Nile Blue Sulfate staining and in situ hybridization for detection of stromelysin-3 transcripts were described previously (Dupé et al., 1999).

Xαaf1^o mutant mice were in a 129/Sv/C57BL/6 (25/75%) genetic background at the time of testing. All animals were housed in cohorts of three to five mice per cage in 12 hour light/dark cycle, with freely available food and water. Behavioral testing was conducted between 14:00 and 18:00. The open field test was performed as described (Krezel et al., 1998). For the elevated plus maze, mice were placed in the central part of the maze and the percentage of time spent in the open arm, and the number of times animal stretched its head to look down were recorded for 5 minutes. To measure fine locomotor skills, the rotarod, inclined plane and cord tests were employed (Krezel et al., 1998; Wolffgramm et al., 1990; Fehlings, 1995; Perry et al., 1995). The nociception was analyzed by evaluating response to acute thermal stimuli in the tail-flick and hot-plate tests, in which the latency to remove the tail from 50°C water and the time spent on the 54° plate before licking the hindpaws were measured, respectively. Finally in taste preference, mice were presented water and 15% sucrose in different parts of the test cage. The number and the time of visits were scored for 5 minutes over 4 consecutive days, indicating preference for sucrose. For each behavioral paradigm data were analyzed by analysis of variants, ANOVA and Tukey-Kramer post hoc tests.

A targeting vector was designed to alter the RXRα locus in such a way that it encodes a mutant protein lacking amino acids 11 to 132 of RXRα1. This region contains the autonomous AF-1 activation function, several phosphorylation sites (Adam-Stitah et al., 1999), as well as the initiator codon of RXRα2 and α3 isoforms (Brocard et al., 1996). Replacement in ES cells led to a targeted allele in which most of exon 2 (with the exception of the first 3 bp), intron 2 and 102 pb of exon 3 were deleted, and in which a ‘floxed’ TK-NEO cassette was inserted into intron 3. Cre-mediated excision of the selection marker led to the mutant allele (hereafter designated as RXRαaf1), containing a single intronic loxP site (see Fig. 1b,c). Chimeric males derived from two independent ES cell clones (HS23.26Cre and BL19) transmitted the mutation through their germline, and the identity of the mutation was confirmed by sequencing RT-PCR products amplified from RNA of homozygous mutants (data not shown). In nuclear extracts prepared from whole E12.5 embryos, the truncated RXRαΔA/B protein was readily detected as a single species with an increased electrophoretic mobility, when compared with that of RXRα (Fig. 1d). Its overall level of expression in homozygous mutants was comparable with that of RXRα in wild type (WT) embryos (Fig. 1d, compare lanes 1 and 3). Moreover, the RXRαΔA/B and RXRα proteins were present in similar amounts in heterozygous embryos (Fig. 1d, lane 2). Therefore, the present mutation does not drastically alter the steady state level of the RXRα truncated protein. Consequently, its effects are likely to reflect only the lack of RXRα A/B domain.

We describe below the effects of deleting the RXRα A/B domain in mice as well as in mutants additionally null for RXRβ, RXRγ, RARα, RARβ or RARγ. To simplify nomenclature, homozygote mutant mice lacking the RXRα A/B domain are designated as Xαaf1^o, and heterozygotes as Xαaf1/+; RXRα, RXRβ, RXRγ, RARα, RARβ and RARγ homozygote null mutants are designated as Xα, Xβ, Xγ, Aα, Aβ and Aγ, respectively; for example, RXRαaf1^o/RARα^−/− mutants are referred to as Xαaf1^o/Aα mutants.

Deletion of RXRα AF-1 domain A/B does not lead to lethal developmental defects, as viable and fertile Xαaf1^o mutants were obtained at the expected Mendelian ratio (out of 1103 littermates born from heterozygote crosses, 27% were wild type (n=299), 50% were heterozygote (n=553) and 23% were homozygote (n=251)). At embryonic day 18.5 (E18.5), the weight of Xαaf1^o mutants (1.19 g on average; n=20) was similar to that of wild type (1.22 g on average; n=18). In contrast, between 1 and 2 weeks of birth, Xαaf1^o mutants were on average 20% lighter than their wild-type littermates (Fig. 2). Several of the cachectic Xαaf1^o mutants (weight ratio < 0.5) died before weaning, whereas the others lived as long as their wild-type littermates (at least 1 year), but exhibited a weight deficit of about 10% during at least 6 months after birth (Fig. 3). This growth retardation was harmonious as (1) the ratios of adult tissue weights (liver, kidney, heart, lungs, visceral fat) to the total body weight and (2) ratios of the length of skeletal elements (femur and skull) to the total body-length did not reveal significant differences between Xαaf1^o mutants and wild-type animals. We previously showed that RXRα^+/- mice are growth deficient (Kastner et al., 1994). The present data further indicate that the RXRα A/B domain is important for postnatal growth.

Soft tissue syndactyly of the hindlimbs was seen unilaterally (n=33) or bilaterally (n=83) in 70% of the Xαaf1^o adults. It affected mostly the base of the interdigital regions between digits 2-3 and 3-4 (data not shown). Hindlimb interdigital webbing was observed at a low frequency in RXRα^+/− mice, but was absent in Xαaf2^o mutants (Kastner et al., 1997a; Mascrez et al., 1998). These data indicate that RXRα AF-1 domain A/B is involved in the involution of the interdigital mesenchyme (see below).

A persistent hyperplastic primary vitreous body (PHPV) was the only ocular defect observed in Xαaf1^o mutants (two out of 18). In contrast, eye morphogenesis is severely altered in Xα and Xαaf2^o fetuses, which display (in addition to a completely penetrant PHPV) corneal, retinal, lens and eyelid abnormalities (Kastner et al., 1994; Kastner et al., 1997a; Mascrez et al., 1998; see Table 2). Xα and Xαaf2^o fetuses also exhibit cardiac abnormalities, namely a myocardial hypoplasia and an agenesis of the conotruncal septum (Kastner et al., 1994; Kastner et al., 1997a; Sucov et al., 1994). No heart defects were detected in E18.5 Xαaf1^o mutants analyzed by histology (n=3). However at E9.5, electron microscopic analysis of Xαaf1^o embryos (n=5) revealed a single case of precocious differentiation of the cardiomyocytes as in Xα and Xαaf2^o mutants (Fig. 4; Kastner et al., 1994; Kastner et al., 1997b; Mascrez et al., 1998). Xαaf1^o mutants exhibited skeletal abnormalities similar to those found in RARs and Xαaf2^o mutants (Lohnes et al., 1993; Ghyselinck et al., 1997; Mascrez et al., 1998): (1) homeotic transformations and malformations of cervical vertebrae; (2) bilateral agenesis of the metoptic pilar (the posterior border of the optic nerve foramen); and (3) a misshapen cricoid cartilage (Table 1). Altogether, these results indicate that the RXRα AF-1 domain A/B is involved in the involution of the primary vitreous body, the differentiation of cardiomyocytes and the morphogenesis of some skeletal elements; but at first sight this domain also appears dispensable for the majority of the developmental events normally mediated by RXRα.

Retinoid receptors regulate brain functions (Krezel et al., 1998; Chiang et al., 1998). As RXRα is almost ubiquitously expressed in the central nervous system (Krezel et al., 1999), behavioral tests were performed on Xαaf1^o mice. Sensory faculties, motor skills, locomotion activity, stereotypic behaviours, as well as stress responses, were not affected, as Xαaf1^o mice performed as well as wild-type littermates in all these tests (see Materials and Methods for details). Thus, no obvious non-redundant function can be readily ascribed, in the central nervous system, to the RXRα AF-1 domain A/B.

As Xαaf1^o mutants displayed neither the prenatal lethality nor all the congenital defects previously observed in Xα and Xαaf2^o mutants, the RXRα AF-1 domain A/B might be largely dispensable during development. Alternatively, RXRβ and/or RXRγ may functionnally compensate for the absence of the RXRα AF-1 domain A/B. To investigate this possibility, RXRαaf1o mutation was introduced into RXRβ and RXRβ/RXRγ-null genetic backgrounds (Xβ/Xγ mutants develop normally; Krezel et al., 1996). Steady-state levels of RXRαΔA/B in Xαaf1o/Xβ and Xαaf1o/Xβ/Xγ E12.5 embryos were comparable with those of RXRα in Xβ and Xβ/Xγ embryos, respectively (data not shown).

Of the 362 mice genotyped after birth from Xαaf1/⁺/Xβ^+/− intercrosses, Xαaf1o/Xβ mutants were obtained at a Mendelian ratio (23 expected, 19 obtained). However, of the 261 mice born from Xαaf1/⁺/Xβ^+/−/Xγ intercrosses, only 10 Xαaf1o/Xβ/Xγ mutants were obtained (16 expected). But at E14.5, Xαaf1o/Xβ/Xγ fetuses were collected at a Mendelian ratio. This indicates that the absence of all RXR AF-1 domains A/B is not lethal at a developmental stage when organogenesis is almost completed, but results in impaired viability. Xαaf1o/Xβ and Xαaf1o/Xβ/Xγ mutant females were fertile, whereas males were sterile, owing to the loss of RXRβ (Kastner et al., 1996). In adults, the weight of Xαaf1o/Xβ mutants was about 10% less than that of Xαaf1o mutants (data not shown).

A similar bilateral and completely penetrant hindlimb interdigital webbing was observed in Xαaf1o/Xβ and in Xαaf1o/Xβ/Xγ mutants. It was more severe than in Xαaf1o mutants, as it often affected the full length of interdigital spaces (Fig. 5c and data not shown). E14.5 Xαaf1o/Xβ hindlimbs showed a marked decrease in the number of dying cells within interdigital spaces, as assessed by staining with Nile Blue Sulfate (Fig. 5a). Moreover, the expression of stromelysin 3, an RA-regulated matrix metalloproteinase involved in tissue remodeling processes that accompany apoptosis in developing limbs (Lefebvre et al., 1995; Dupé et al., 1999; Ludwig et al., 2000), was markedly decreased in the interdigital spaces of E14.5 Xαaf1o/Xβ hindlimbs (Fig. 5b). These results indicate that the soft tissue syndactyly seen in a majority of Xαaf1o adults, as well as in all Xαaf1o/Xβ, Xαaf1o/Xβ/Xγ adults, is caused by the persistence of the fetal interdigital mesenchyme.

All E18.5 Xαaf1o/Xβ (n=4) and E14.5 Xαaf1o/Xβ/Xγ (n=9) mutants analyzed histologically displayed a bilateral PHPV (compare Fig. 6a,b; Table 2). In addition, one Xαaf1o/Xβ/Xγ mutant displayed a ventricular myocardial hypoplasia, and another one an agenesis of the conotruncal septum comparable with that of Xα mutants (Table 3). These data confirm the requirement of A/B domain of RXR in the involution of the interdigital and primary vitreous body mesenchymes, and indicate that it is also involved in cardiac morphogenesis in some individuals.

The putative start codon of RXRα2 and α3, the two RXRα isoforms specifically expressed in the adult testis (Brocard et al., 1996), was removed in Xαaf1o mutants. Therefore, the RXRα A/B domain deletion should also result in a null mutation for RXRα2 and α3. This was confirmed by western blot analysis (data not shown). Histological analysis of the testes of Xαaf1o adults (n=4), which are fertile, did not reveal abnormalities, whereas Xαaf1o/Xβ/Xγ adults (n=2), which are sterile because of the RXRβ-null mutation, did not display any testis defect in addition to those exhibited by Xβ-null mutants (Kastner et al., 1996). Thus, RXRα2 and RXRα3 isoforms are not necessary for male fertility.

The relative contributions of AF-1 and AF-2 activities to the functions of RXRα in placentation (Sapin et al., 1997; Wendling et al., 1999; Barak et al., 1999) were assessed by comparing wild-type, Xαaf1o, Xαaf2o, Xαaf1o/Xβ/Xγ and Xαaf2o/Xβ/Xγ placental morphologies. Placentas from E18.5 Xαaf1o (n=3) and E14.5 Xαaf1o/Xβ/Xγ (n=6) mutants were macroscopically and histologically indistinguishable from wild-type placentas (data not shown). Histological defects observed at E14.5 in Xαaf2o (n=5) and Xαaf2o/Xβ/Xγ (n=5) placentas consisted mainly in a thickening of labyrinthine trabeculae, and a lack of clear frontier between the spongiotrophoblastic and the labyrinthine zones. They were similar to those previously detected in RXRα-null placentas (Sapin et al., 1997; data not shown).

The outcome of these placental abnormalities was assessed at E18.5: Xαaf2o placentas were macroscopically distinguishable from wild-type and Xαaf2/+ placentas by their pale appearance, which reflects a deficiency in red blood cells within the labyrinthine zone (Fig. 7a,b; data not shown). Histological analysis of E18.5 Xαaf2o placentas (n=4) showed (1) an abnormal, wavy aspect of the interface between spongiotrophoblast and labyrinth (Fig. 7c,d); and (2) a thickening of the labyrinthine trabeculae separating fetal capillaries and maternal blood sinuses (Fig. 7e-h). Labyrinthine trabeculae represent the placental barrier across which nutrient and gas exchanges between fetal and maternal blood occur (reviewed by Cross, 2000). Limitation of the rate of fetal-maternal exchanges caused by a thickened placental barrier is probably responsible for in utero growth retardation of Xαaf2o mutants (Mascrez et al., 1998).

The RXRαaf1o mutation was introduced into RAR(α, β or γ)-null genetic backgrounds to assess the functional contribution of the RXRα AF-1 domain A/B in RAR/RXR heterodimers. Steady-state levels of RXRαΔA/B in E12.5 Xαaf1o/A(α, β or γ) mutant embryos were comparable with those of RXRα in Aα-, Aβ- and Aγ-null embryos, respectively (data not shown). Therefore, the abnormal phenotypes seen in Xαaf1o/A(α, β or γ) mutants most probably reflect the lack of the RXRα A/B domain.

Aα, Aβ and Aγ mutants are viable (Kastner et al., 1995; Ghyselinck et al., 1997). In contrast, no living Xαaf1o/Aα mutants were recovered from Xαaf1/⁺/Aα^+/− intercrosses, even though they could be collected at Mendelian ratio at E18.5. Similarly, only few Xαaf1o/Aβ and Xαaf1o/Aγ adults were viable (two and one obtained versus ten and seven expected, respectively). The lethality of Xαaf1o/A(α, β or γ) mutants clearly show that the RXRα AF-1 domain A/B is essential, at least in some RAR genetic settings.

Xαaf1o/A(α, β or γ) compound mutants were analyzed by histology at E14.5 and E18.5. Xαaf1o/Aα mutants displayed, with a low penetrance, the majority of the cardiovascular, respiratory and urogenital defects previously observed in Xα/Aα mutants (Tables 3, 4; Kastner et al., 1994; Kastner et al., 1997a). Similarly, Xαaf1o/Aβ and Xαaf1o/Aγ mutants reproduced a milder form of the ocular defects observed in Xα/Aβ and Xα/Aγ fetuses (Table 2; Kastner et al., 1994; Kastner et al., 1997a), including closer eyelid folds (E in Fig. 6a,c,d), thickening of the ventral portion of the cornea (C), shortening of the ventral retina (V), ventral rotation of the lens (L), coloboma of the optic disc (green arrowheads in Fig. 6d) and absence of the sclera (SC, compare Fig. 6f with 6g). Thus, in a genetic background null for either RARα, RARβ or RARγ, the RXRα A/B domain becomes indispensable for a large subset of RA-dependent functions involved in morphogenesis.

It is interesting to note that the defects observed in Xαaf1o/A(α, β or γ) mutants were less penetrant or less severe than those observed in Xαaf2o/A(α, β or γ) mutants (Mascrez et al., 1998). For example, lung hypoplasia was present in about two thirds of Xαaf1o/Aα mutants, whereas it is completely penetrant in Xαaf2o/Aα mutants. One third of Xαaf1o/Aβ and Xαaf1o/Aγ mutants displayed a coloboma of the optic disc, an abnormality observed in all Xαaf2o/Aβ and Xαaf2o/Aγ mutants. Ocular abnormalities found with the same penetrance in Xαaf1o/RAR-null and in Xαaf2o/RAR-null mutants, were less severe in Xαaf1o/RAR-null mutants. For example, shortening of the ventral retina, ventral rotation of the lens, thickening of the corneal stroma and closer eyelid folds, are less severe in Xαaf1o/Aβ and Xαaf1o/Aγ mutants than in Xαaf2o/Aβ and Xαaf2o/Aγ mutants (Table 2; Fig. 6c,d; for additional examples, compare Fig. 6h and 6j with 6i and 6k). Altogether, these observations indicate that, for a large fraction of the RAR-dependent events, the functions of the RXRα AF-1 domain A/B are less crucial than those of the AF-2 activity.

We also investigated the possible involvement of the RXRα AF-1 domain A/B in the regulation of a transgene, whose expression is under the control of the RARβ2 promoter that contains a RA-reponse element (Mendelsohn et al., 1991). To this end, the RARβ2 promoter-lacZ reporter transgene was introduced into the RXRαaf1o genetic background. At E13.5, lacZ expression was similar in Xαaf1o (n=9) and wild-type (n=13) fetuses (data not shown). In order to eliminate a functional compensation by RXRβ (Mascrez et al., 1998; Wendling et al., 1999), lacZ expression was also studied in Xαaf1o/Xβ mutants. At E13.5, expression of lacZ was similar in wild-type (n=13) and Xαaf1o/Xβ (n=9) fetuses, including the interdigital regions (data not shown). Altogether, these data indicate that the RXRα AF-1 domain A/B, in contrast to RXRα AF-2 (Mascrez et al., 1998), is dispensable for in vivo transactivation by RXRα, at least in the context of the RARβ2 promoter.

The RXRα deletion created in the present study encompasses most of the N-terminal A/B domain. This region contains the autonomous ligand-independent transcriptional activation function, AF-1, which exhibits some specificity in transfected cells in vitro, depending upon the cell-type and promoter context (Nagpal et al., 1992; Nagpal et al., 1993). Little is known about the molecular mechanisms through which RXRα AF-1 domain A/B exerts its transcriptional activity. For other members of the nuclear receptor superfamily, e.g. PPARγ, the A/B domain has been shown to interact directly with co-activators, e.g. PGC-2 (Castillo et al., 1999). Along the same lines, phosphorylation of the A/B domain may modulate either the interaction of receptors with co-activators or their affinity for ligands. For example, phosphorylation of ERβ A/B domain promotes recruitment of the SRC-1 co-activator (Tremblay et al., 1999), whereas phosphorylation of the A/B domain reduces the ligand-binding affinity of PPARγ, thus negatively regulating the transcriptional activity of PPARγ (Shao et al., 1998). In F9 cells, phosphorylation of the RARγ A/B domain is required not only for induction of several RA-target genes (Taneja et al., 1997; Rochette-Egly et al., 1997; Bastien et al., 2000), but also for receptor degradation by the ubiquitin-proteasome pathway (Kopf et al., 2000). At least some of these features could probably be extended to RXRα, which is phosphorylated at several serine and threonine residues in the A/B domain (Adam-Stitah et al., 1999). The present investigation was designed to establish the developmental role of the RXRα A/B region, as a prerequisite to further genetic and molecular studies aimed at elucidating the molecular mechanisms that underlie physiological events in vivo.

Involution of the primary vitreous body and of the hindlimb interdigital mesenchyme, and cardiomyocyte differentiation and morphogenesis of some cranial skeletal elements and vertebrae are impaired in Xαaf1o mutants (this study), in RAR(α, β and γ) mutants (Lohnes et al., 1993; Ghyselinck et al., 1997; Kastner et al., 1997b) and in animals with vitamin A and RA deficiency (Wilson et al., 1953; Lussier et al., 1993). It appears therefore that in RXRα/RAR heterodimers, the AF-1 domain A/B of RXRα participates in the transduction of retinoid signals normally required for the completion of these developmental events. The observation that the developmental defects exhibited by the Xαaf1o/Xβ/Xγ mutants can all be attributed to abnormalities in the RAR/RXR signaling pathway suggest that RXR partners other than RAR do not exert any developmental function that require the RXRα A/B region.

During development, RXRα is the most important RXR, as Xα fetuses die in utero and display severe cardiac and ocular defects, whereas Xβ/Xγ mutants are viable and do not exhibit any congenital defects (Kastner et al., 1994; Krezel et al., 1996). However, a lack of RXRα or its AF-2 activity can be functionally compensated by RXRβ and RXRγ in mouse embryos, as the spectrum of defects observed in Xα/Xβ and Xαaf2o/Xβ/Xγ mutants is much broader than that of Xα mutants (Mascrez et al., 1998; Wendling et al., 1999). RXRβ and RXRγ also compensate for the deletion of RXRα A/B domain, as the PHPV and the hindlimb interdigital webbing become completely penetrant and eventually more severe in Xαaf1o/Xβ/Xγ mutants; however, no additional abnormalities are generated in these latter mutants, except for single cases of myocardial hypoplasia and conotruncal septum agenesis. Thus, the functional compensation by RXRβ or RXRγ cannot account for the ‘weak’ phenotype of the Xαaf1o mutants.

Two scenarios could account for the paucity of the developmental defects in Xαaf1o and Xαaf1o/Xβ/Xγ mutants. First, the A/B domain might in fact be dispensable for the majority of the developmental events that require RXRα. Given the phenotypes of the various Xαaf1o/RAR(α, β or γ) mutants (see below and Tables 2-4), this possibilty is unlikely. Alternatively, the RXRα A/B domain might be necessary for optimal transactivation of several developmental target genes, through RXRα/RAR heterodimers. According to this second scenario, the contribution of the RXRα A/B domain to the transactivation potential of the heterodimers might not be detectable in the protected environment of the animal facility, unless the phenotyping techniques are sophisticated enough, or the examined sample size becomes very large (Brookfield, 1992; Thomas, 1993). Our findings that one out of nine Xαaf1o/Xβ/Xγ mutants displays an hypoplasia of the ventricular myocardium and that another one of these nine has an agenesis of the conotruncal septum, defects that are characteristic of the RXRα-null phenotype but have never been observed in the several dozens of wild-type embryos that we have analyzed over the past 10 years, support the second scenario (Kastner et al., 1994; Sucov et al., 1994). The requirement of RXRα A/B domain for myocardial growth or fusion of conotruncal ridges in some fetuses might then reflect stochastic inter-individual variations in the expression of other synergistic factors involved in RA-dependent cardiac morphogenesis. This hypothesis predicts that more VAD-related developmental abnormalities might be detected if a larger sample of Xαaf1o/Xβ/Xγ fetuses were examined.

Other evidence supporting the conclusion that the RXRα AF-1 domain A/B could be necessary for transactivation by the various RXRα/RAR heterodimers is suggested by the observation that the corresponding Xαaf1o/RAR(α, β or γ) mutants display most of the VAD-related defects observed in the cognate Xα/RAR(α, β or γ) double mutants (Kastner et al., 1994; Kastner et al., 1997a). In fact, in the absence of a given RAR, the RXRα A/B domain becomes essential for enabling the remaining RARs to functionally replace the missing one (discussed by Kastner et al., 1997a; Chiba et al., 1997a; Chiba et al., 1997b; Mascrez et al., 1998; see Tables 2-4). Altogether, these data favor the involvement of RXRα AF-1 domain A/B in most of the developmental events that are mediated through the various RAR/RXRα heterodimers, under conditions where either RAR and/or retinoid levels become limited.

Thickening of the labyrinthine trabeculae and lack of definite frontier between the labyrinth and spongiotrophoblast are hallmark features of RXRα^−/− placentas, and probably reflect a failure of a late step of trophoblast cells differentiation (Sapin et al., 1997). Our data show that the RXR A/B domain appears largely unnecessary for the differentiation of trophoblast cells. In contrast, the AF-2 of RXRα is clearly essential for this differentiation process, as (1) Xαaf2o placentas are histologically indistinguishable from placentas that lack RXRα; and (2) defects of Xαaf2o/Xβ/Xγ placentas are not more severe than in Xαaf2o. Furthermore, as Xαaf2o/RAR(α, β or γ) mutants are obtained at the expected Mendelian ratio at E18.5, and are not more growth deficient than Xαaf2o mutants (B. M., M. M., N. B. G. and P. C., unpublished), it is likely that terminal differentiation of labyrinthine trophoblast cells involves heterodimers in which RXRα is transcriptionally active, but distinct from RXRα/RAR heterodimers. On the other hand, the formation of the labyrinthine zone of the placenta is totally inhibited in Xα/Xβ mutants, and severely impaired in PPARγ mutants, leading to embryonic death before E10.5 (Wendling et al., 1999; Barak et al., 1999). Thus, RXRα and RXRβ, probably in the form of PPARγ/RXR heterodimers, have an early role in placentogenesis. We have shown that Xαaf1o/Xβ/Xγ and Xαaf2o/Xβ/Xγ mutants do not display these early defects. Thus, neither the RXRα AF-1 domain A/B nor the AF-2 activity is required on their own in early placentogenesis, suggesting that RXRs could act in this process as silent partners with PPARγ. Studies with RXRα mutants that lack both AF-1 and AF-2 are in progress to investigate this possibility further.

Among the various RA-dependent morphogenetic events involving the RXRα AF-1 domain A/B, separation of the digits is the only one that specifically requires this region, as Xαaf1o and Xαaf1o/Xβ/Xγ mutants display soft tissue syndactyly of the hindlimbs, whereas Xαaf2o and Xαaf2o/Xβ/Xγ mutants never display this defect (Mascrez et al., 1998). Phosphorylation of RXRα at a specific serine residue located in the A/B domain is necessary for the antiproliferative response of F9 teratocarcinoma cells to RA (Rochette-Egly and Chambon, 2001). It is noteworthy that RA-dependent involution of the interdigital mesenchyme results from arrest of cell proliferation, as well as from apoptosis (Dupé et al., 1999). Therefore, the possibility exists that phosphorylation of the RXRα A/B domain might have important functions in the cascade of molecular events that, in vivo, leads to the normal disappearance of the interdigital mesenchyme.

AF-1 and AF-2 activities do not have the same importance in the transcriptional activity of RXRα during embryonic development: (1) Xαaf2o/Xβ/Xγ fetuses display a large array of congenital defects not found in Xαaf1o/Xβ/Xγ fetuses; and (2) Xαaf2o/RAR(α, β or γ) compound fetuses are more severely affected than Xαaf1o/RAR(α, β or γ) fetuses (Mascrez et al., 1998; see Tables 2-4). Moreover, the AF-2, but not the AF-1 of RXRα is crucial for the transcription of the RARβ2-lacZ transgene (Mascrez et al., 1998; this study). At the molecular level, little is known about the mechanism through which AF-1 activates transcription. In the case of ERα, AF-1 and AF-2 appear to interact simultaneously in vitro with distinct interfaces of the TIF2 co-activator, leading to synergistic activation of transcription (Benecke et al., 2000). Our present data showing that similar abnormalities are exhibited by Xαaf1o/RAR(α, β or γ) and Xαaf2o/RAR(α, β or γ) compound mutants, although more severe in the latter case, not only suggest that RXRα AF-1 and AF-2 may synergize in vivo as they do in vitro (Nagpal et al., 1993), but also that AF-2 is more important than AF-1 for most of the developmental events that require RXRα/RAR heterodimers.

We thank C. Birling-Ziegler, I. Michel, B. Weber, C. Dennenfeld and B. Féret; the staff of IGBMC common services, ES cell culture and animal facility for their excellent technical assistance; P. Kastner for the gift of plasmids and helpful discussions; J. Brocard for CMV-Cre mice; C. Rochette-Egly for gift of the antibodies; M. C. Rio for stromelysine-3 probe; N. Messaddeq for transmission miscroscopy; C. Werlé and the secretarial staff for their help in the preparation of this manuscript. This work was supported by funds from the Centre National de la Recherche Scientifique (CNRS), the Institut National de la Santé et de la Recherche Médicale (INSERM), the Hôpital Universitaire de Strabourg, the Collège de France, the Institut Universitaire de France, the Association pour la Recherche sur le Cancer (ARC) and Bristol-Myers Squibb. B. M. was the recipient of fellowships from the French Ministry of Research, the Ligue Nationale contre le Cancer, the Association pour la Recherche contre le Cancer and the Université Louis Pasteur.