Functional retinoid receptors in budding ascidians

Authors

  • Mika Kamimura,

    1. Laboratory of Molecular and Cellular Biotechnology, Faculty of Science, Kochi University, Kochi 780-8520, Japan.
    Search for more papers by this author
    • †Present address: Molecular and Developmental Biology, Nara Institute of Science and Technology, Graduate School of Biological Sciences, 8916-5 Takayama, Ikoma, Nara 630-01, Japan.

  • Shigeki Fujiwara,

    1. Laboratory of Molecular and Cellular Biotechnology, Faculty of Science, Kochi University, Kochi 780-8520, Japan.
    Search for more papers by this author
  • Kazuo Kawamura,

    1. Laboratory of Molecular and Cellular Biotechnology, Faculty of Science, Kochi University, Kochi 780-8520, Japan.
    Search for more papers by this author
  • Toshitsugu Yubisui

    1. Laboratory of Molecular and Cellular Biotechnology, Faculty of Science, Kochi University, Kochi 780-8520, Japan.
    Search for more papers by this author

Abstract

A homolog of retinoid X receptors (RXR), named PmRXR, was cloned from the budding ascidian, Polyandrocarpa misakiensis. Gel-shift assays revealed that PmRXR and a previously identified P. misakiensis retinoic acid receptor (PmRAR) formed a complex to bind vertebrate-type retinoic acid response element (RARE). Transfection assays were carried out using a reporter gene containing a RARE upstream of lacZ. Two chimeric effector genes were constructed by placing PmRXR and PmRAR cDNA fragments (containing the DNA-binding, ligand-binding and ligand-dependent transactivation domains) downstream of the human RXRα and RARα cDNA (covering the N-terminal coding region), respectively. Each chimeric cDNA was ligated to a notochord-specific enhancer. In case the embryos were transfected with all three transgenes and treated with retinoic acid (RA), the reporter gene was activated in the notochord cells. The result suggests that the PmRXR/PmRAR complex functions as an RA-dependent transcriptional activator. The PmRXR mRNA was detected in a mesenchymal cell type, called glomerulocyte, in the developing Polyandrocarpa bud. As this cell type has been shown to express PmRAR mRNA, it seems possible that the PmRXR/PmRAR complex mediates RA signaling in this cell type to induce the expression of genes involved in the morphogenesis of the developing bud.

Introduction

Ascidians belong to the phylum Chordata. However, unlike vertebrates, about half of the ascidian species can proliferate asexually by budding. The growing bud of the ascidian Polyandrocarpa misakiensis is formed as an outgrowth of the parental body wall. After separation from the parent, the gut primordium is formed from the atrial epithelium in the most proximal region of the bud. As the atrial epithelium is a differentiated pigmented epithelium, the process involves transdifferentiation ( Fujiwara & Kawamura 1992; Kawamura & Fujiwara 1994). Retinoic acid (RA) can induce a secondary anteroposterior axis in the developing bud ( Hara et al. 1992 ). The amount of retinal, a precursor of RA, is greatly reduced during bud development ( Kawamura et al. 1993 ). In addition, the activity of aldehyde dehydrogenase (ALDH), a potential RA synthase, is observed in the proximal region of the bud ( Kawamura et al. 1993 ; Harafuji et al. 1996 ). These observations suggest that RA is an endogenous inducer of transdifferentiation in the developing bud.

In vertebrates, the RA signal is mediated by two distinct types of nuclear receptors, called retinoic acid receptors (RAR) and retinoid X receptors (RXR; Mangelsdorf & Evans 1992). The RAR bind their target DNA sequence, called retinoic acid response element (RARE), as a heterodimer with an RXR ( Mangelsdorf & Evans 1995). The RXR can bind another DNA sequence named RXRE as a homodimer ( Pfhal et al. 1994 ; Mangelsdorf & Evans 1995). Although RA affects morphogenesis in protochordates ( Hara et al. 1992 ; Katsuyama et al. 1995 ; Holland & Holland 1996), echinoderms ( Sciarrino & Matranga 1995) and crustaceans ( Hopkins & Durica 1995), the only invertebrate RAR homolog was reported in P. misakiensis ( Hisata et al. 1998 ).

In the present study, we isolated cDNA clones encoding P. misakiensis RXR (PmRXR). To demonstrate that PmRXR and PmRAR are functional retinoid receptors, we carried out gel-shift and transactivation assays. In situ hybridization was performed to examine the localization of PmRXR mRNA in the developing bud.

Materials and Methods

Animals

Polyandrocarpa misakiensis was cultured on glass slides in culture boxes, settled in Uranouchi Inlet near the Usa Marine Biological Institute of Kochi University. Ciona savignyi and Ciona intestinalis were kindly provided by Z. Imoto of Kochi University, T. Nishikata of Konan University and W. Godoh of Okayama University.

Isolation and sequence analysis of cDNA clones

A cDNA fragment corresponding to the DNA- binding domain was amplified by polymerase chain reaction (PCR) using primers 5′-GAATTCTG(TC)(TG)C(TCAG)AT(TCA)TG(TC)GG(TCAG)GA-3′ and 5′-GT(AG)AA(AG)AC(TC)AT(TCAG)G(CA)(TC)GT(TC)AT(AG)AC-3′. Longer cDNA were isolated by the rapid amplification of cDNA ends (RACE) method ( Frohman et al. 1988 ). The cDNA sequence was determined using automatic sequence analyzing system 373A (Applied Biosystems, Foster City, CA, USA). The phylogenetic tree was obtained using the maximum-likelihood method ( Strimmer & von Haeseler 1997), based on the alignment of amino acid sequences obtained using ClustalW program in the MacVector Software (Teijin, Tokyo, Japan).

Plasmid constructs

Human/ascidian chimeric retinoid receptor cDNA were constructed as follows. A DNA fragment containing human RXRα A/B region was amplified by PCR, using primers 5′-AATACGACTCACTATAG-3′ and 5′-GCAGCTGTATACTCCATAGT-3′, and pCMX-hRXRα ( Mangelsdorf et al. 1990 ) as a template. The PCR product was digested with HindIII and BstZ17I. The PmRXR cDNA fragment was excised with BstZ17I and MfeI. These two fragments were ligated and used to replace the hRXRα cDNA in the pCMX-hRXRα. The resulting plasmid was named pCMX-(h + Pm)RXR. A human RARα A/B region cDNA fragment was amplified using primers 5′-TAAACTGCCCACTTGGCA-3′ and 5′-AGGCTTGTATACGCGGGGTAGA-3′, and pCMX-hRARα ( Umesono et al. 1991 ) as a template. The PCR fragment was digested with NdeI and BstZ17I. The PmRAR cDNA fragment was excised with BstZ17I and BamHI. These two fragments were ligated and used to replace the hRARα cDNA in the pCMX-hRARα. The resulting plasmid was named pCMX-(h + Pm)RAR. The lacZ fragment in the – 3.5 kb Ci-Bra/lacZ ( Corbo et al. 1997 ) was replaced with the (h + Pm)RXR or (h + Pm)RAR. The plasmids were named Ci-Bra/(h + Pm)RXR and Ci-Bra/(h + Pm)RAR, respectively. The RARE-containing reporter gene, named DR5/Z, was constructed based on – 251 bp Ci-Bra/lacZ ( Corbo et al. 1997 ). Two Suppressor of Hairless-binding sites were replaced with an RARE via site-directed mutagenesis ( Kunkel 1985) using an oligonucleotide primer 5′-ACCAAGTTTCAACTTAGGTCACCAGGAGGTCAGAAAGTAACACGTCAC-3′.

Gel-shift assays

The (h + Pm)RXR and (h + Pm)RAR proteins were produced using pCMX-(h + Pm)RXR and pCMX-(h + Pm)RAR, respectively, as templates, according to the protocol for the TnT coupled wheat germ extract kit (Promega, Madison, WI, USA). The double-stranded DNA probe containing RARE was prepared by annealing the complementary oligonucleotides 5′-GACAGTCAGGTCACCAGGAGGTCAGAGTCA-3′ and 5′-ACTGACTCTGACCTCCTGGTGACCTGACTG-3′. The 5′ end-labeling of the probe using [γ-32P] adenosine triphosphate (ATP), binding reaction and electrophoresis were carried out according to the protocol for the gel-shift assay system (Promega). For control and competition experiments, AP2 oligonucleotide supplied in the gel-shift kit (Promega) was used. The autoradiogram was obtained using Molecular Imager and Molecular Analyst version 2.1 (Bio-Rad, Richmond, CA, USA).

Electroporation and in situ hybridization

Electroporations were carried out as described in Corbo et al. (1997) . Thirty micrograms of a lacZ reporter gene was used for each electroporation. Co-electroporations involved the use of 30 μg each of Ci-Bra/ (h + Pm)RXR and Ci-Bra/(h + Pm)RAR. The Ciona embryos were electroporated 25–30 min after fertilization. All-trans RA stock solutions (0.8 m M in dimethylsulfoxide (DMSO)) were added to the filtered sea water (FSW) containing electroporated embryos to a final concentration of 1 μM RA (and 0.125% DMSO). The embryos were allowed to develop until the tail- bud stage, fixed with formaldehyde and stained for β-galactosidase activity, as described by Corbo et al. (1997) . Whole-mount in situ hybridization was carried out with digoxigenin-labeled RNA probes as described by Hisata et al. (1998) . The stained specimens were embedded in JB-4 plastic resin (Polyscience Inc., Warrington, PA, USA), and sectioned at a thickness of 2–3 μm.

Results

Isolation and sequence analysis of an ascidian RXR homolog

Partial cDNA clones encoding an ascidian RXR homolog were obtained by PCR and 3′ RACE. The longest open reading frame encodes a polypeptide of 363 amino acids, and lacks a part of the N-terminal transactivation domain (A/B region; Fig. 1A). The amino acid sequence of PmRXR was about 80, 50 and 30% identical to that of vertebrate RXR, insect RXR homologs and other nuclear receptors, respectively. The cysteine residues responsible for zinc-finger formation, the P-box and D-box that determine the target DNA sequence ( Danielsen et al. 1989 ; Mader et al. 1989 ; Umesono & Evans 1989), and the T-box that is also important for DNA binding ( Zilliacus et al. 1995 ) were highly conserved among vertebrate RXR and PmRXR ( Fig. 1B,C). The ligand-binding domain (E/F region) of PmRXR is more than 70% identical to that of vertebrate RXR, and about 40% identical to the Drosophila RXR homolog named Ultraspiracle (USP; Fig. 1D). The core sequence of the ligand-dependent transactivation (AF-2) subdomain, located within this region ( Folkers et al. 1993 ) was also highly conserved among chordate RXR ( Fig. 1D).

Figure 1.

Comparison of the amino acid sequences of Polyandrocarpa misakiensis retinoid X receptor (PmRXR) and other nuclear receptors. The sequences of human RXRα, mouse RXRβ, chick RXRα, Drosophila Ultraspiracle (USP), human RARα, mouse thyroid hormone receptor α (TRα), and human estrogen receptor α (ERα) were obtained from the Swissprot database (P19793, P28704, P28701 P20153, P10276, P16416 and P03372, respectively). Those of Xenopus RXRβ, zebrafish RXRγ, RXRδ, and RXRε, and PmRAR were obtained from the EMBL/Genbank/ DDBJ database (S73269, U29894, U29941, U29942 and D86615, respectively). (A) The domain structure of PmRXR. The cDNA lacks the N-terminal region. (B) The C region. The amino acid residues conserved in all the RXR and Drosophila USP are indicated by asterisks. The P-box and D-box are indicated by boxes. The cysteine residues responsible for zinc-finger formation are indicated by arrowheads. (C) The D region. The T-box is indicated by a box. The amino acid residues conserved in all the RXR and Drosophila USP are indicated by asterisks. (D) The E/F region. The position of 11 α-helices are shown by bars. Zebrafish RXRδ and RXRε, and Drosophila USP contain a long insertion between helix 5 and helix 6. These RXR homologs do not bind 9-cis retinoic acid (RA; Oro et al. 1990 ; Jones et al. 1995 ). The core sequence of the transcriptional activation function 2 (AF-2) subdomain is indicated by a box.

A phylogenetic tree was constructed by the maximum-likelihood method ( Strimmer & von Haeseler 1997). Partial amino acid sequences covering the C, D and E/F regions of various nuclear receptors were compared ( Fig. 2). PmRXR, vertebrate RXR and insect RXR homologs formed a monophyletic cluster. Within the cluster, insect RXR homologs diverged first, and then PmRXR formed a separate branch from all of the vertebrate RXR subtypes. The bootstrap value for all of these branchings was 100. Hisata et al. (1998) showed that PmRAR clustered with vertebrate RAR, but was on a separate branch from all of the vertebrate RAR subtypes. The branching pattern obtained in the present study coincides with the previous result ( Fig. 2).

Figure 2.

A phylogenetic tree constructed by the maximum-likelihood method from the amino acid sequences of Polyandrocarpa misakiensis retinoid X receptor (PmRXR), Polyandrocarpa misakiensis retinoic acid receptor (PmRAR) and other nuclear receptors. Bootstrap values are indicated at each branching point. The amino acid sequences of human ERα, mouse TRα, Xenopus RARγ, human RARα, chick RARβ, Drosophila Ultraspiracle (USP), silkworm RXR homolog, mouse RXRβ and RXRγ, chick RXRα, and human RXRα were obtained from the Swissprot database (P03372, P16416, P28699, P10276, P22448, P20153, P49700, P28704, P28705, P28701 and P19793, respectively). Those of PmRAR, zebrafish RXRγ, RXRδ and RXRε, and Xenopus RXRβ were obtained from the EMBL/GenBank/DDBJ database (D86615, U29894, U29941, U29942 and S73269, respectively).

DNA-binding activity of PmRXR/PmRAR complex

Human/ascidian chimeric retinoid receptor cDNA, named (h + Pm)RXR and (h + Pm)RAR, were constructed. In both cases, the domains responsible for DNA-binding, ligand-binding, dimerization and ligand-dependent transactivation were encoded by ascidian cDNA. In vitro synthesized (h + Pm)RXR and (h + Pm)RAR proteins were used for gel-shift assays to test the DNA-binding activity of PmRXR and PmRAR. An oligonucleotide probe was designed to contain a vertebrate-type RARE, called DR-5 ( Leid et al. 1992 ). The DR-5 probe contains two core motifs (AGGTCA) with a five nucleotide-spacer.

A DNA/protein complex was formed when the DR-5 was added to the reaction mixture containing (h + Pm)RXR and (h + Pm)RAR ( Fig. 3). No shifted band was obvious when the DR-5 was mixed with either (h + Pm)RXR or (h + Pm)RAR only ( Fig. 3). An oligonucleotide probe containing an AP2-binding sequence did not form a complex with the mixture of (h + Pm)RXR and (h + Pm)RAR ( Fig. 3; for AP2-binding sequences see Williams et al. 1988 ). The complex formation was inhibited with 50- or 150-fold molar excess of unlabeled DR-5 oligonucleotide, but not with unlabeled AP2 oligonucleotide ( Fig. 3). The same set of experiments was carried out with an oligonucleotide probe containing another RARE, called DR-2, and similar results were obtained (the DR-2 contained two core motifs with two nucleotide-spacer; data not shown).

Figure 3.

The DNA-binding activity of the PmRXR/PmRAR complex. Neither the (h + Pm)RAR protein nor (h + Pm)RXR protein alone formed a complex with the DR-5 probe (lanes 1 and 2). A mixture of (h + Pm)RXR and (h + Pm)RAR formed a complex with the DR-5 (lane 4), but not with the AP2 probe (lane 3). Unlabeled competitor DNA were added to the reaction mixture containing the DR-5 probe, and the (h + Pm)RXR and (h + Pm)RAR proteins (lanes 5–8). The level of the shifted complex was greatly reduced in case 50- and 150-fold molar excesses of unlabeled DR-5 competitor were used (lanes 5 and 6, respectively). There was no inhibition in the formation of the protein/DNA complex by addition of a 50- and 150-fold molar excess of unlabeled AP2 competitor (lanes 7 and 8).

Transactivation assays

The RA-dependent transactivation activity of PmRXR and PmRAR was tested by transfection assays. A lacZ reporter gene containing a 251 bp upstream enhancer region of Brachyury gene in the ascidian Ciona intestinalis (– 251 bp Ci-Bra/lacZ) is activated in the notochord, muscle and mesenchyme cells in the embryo ( Corbo et al. 1997 ). Two closely linked binding sites for Suppressor of Hairless (Su(H)) in the enhancer are necessary for expression in the notochord ( Corbo et al. 1997 , 1998). These Su(H)-binding sites in the – 251 bp Ci-Bra/lacZ were replaced with DR-5. Ciona embryos were electroporated at the one-cell stage with this reporter gene, named DR5/Z, and stained for β-galactosidase at the tail-bud stage. The lacZ expression is detectable in muscle and mesenchyme cells, but not in notochord cells ( Fig. 4A). The embryos, electroporated with DR5/Z, were then treated with 1 μM all-trans RA continuously from the gastrula stage. The embryos show a moderate defect in neural tube formation at the tail-bud stage ( Fig. 4B, arrowhead). The staining pattern obtained with these embryos is indistinguishable with the untreated embryos ( Fig. 4B).

Figure 4.

Retinoic acid (RA)-dependent transactivation of the DR5/Z reporter gene by Polyandrocarpa misakiensis retinoid X receptor (PmRXR) and Polyandrocarpa misakiensis retinoic acid receptor (PmRAR). (A) The embryos were electroporated with DR5/Z only. The mesenchyme (me) and muscle (mu) in the tail-bud embryo exhibit β-galactosidase activity. (B) The embryos, electroporated with DR5/Z, were treated with 1 μM all-trans RA from the gastrula stage. Retinoic acid affected neural tube formation (arrowhead). However, the staining pattern was not affected. (C) The embryos were co-electroporated with DR5/Z and the Ci-Bra/(h + Pm)RXR and Ci-Bra/(h + Pm) RAR effector genes. The effector genes affected neither the staining pattern nor morphogenesis. (D) The co-electroporated embryos were treated with 1 μM all-trans RA. Tail morphogenesis was affected and ectopic staining was observed in a cell mass in the tail region (n*). (E) The embryos, co-electroporated with the – 3.5 kb Ci-Bra/ lacZ reporter gene and the two effector genes, were treated with 1 μM all-trans RA. The notochord cells (n), whose arrangement was abnormal, were weakly stained.

Two effector genes, named Ci-Bra/(h + Pm)RXR and Ci-Bra/(h + Pm)RAR, were constructed by placing the (h + Pm)RXR and (h + Pm)RAR cDNA downstream of the notochord-specific – 3.5 kb Ci-Bra enhancer ( Corbo et al. 1997 ). Co-electroporation of the embryos with DR5/Z and the two effector genes affected neither the staining pattern nor morphogenesis ( Fig. 4C). However, treatment of the co-electroporated embryos with 1 μM all-trans RA caused a severe mutant phenotype, whereby the tail became short and stout ( Fig. 4D). In these embryos, ectopic staining is obvious in the cell mass located between two rows of tail muscle cells ( Fig. 4D). Co-electroporation was then carried out using the notochord-specific reporter gene – 3.5 kb Ci-Bra/lacZ ( Corbo et al. 1997 ). The RA-treatment of these embryos caused a similar phenotype and weak but specific staining of the cell mass in the tail region ( Fig. 4E). In normal embryos, the notochord cells align in a single row between the tail muscle cells. However, alignment of the notochord cells shown in Fig. 4(E) is disorganized and similar to that of the ectopically stained cell mass in Fig. 4(D).

Localization of PmRXR mRNA in the developing bud

In situ hybridization was carried out to localize PmRXR mRNA in the developing bud ( Fig. 5). Whole-mount specimens of developing buds, 48 h after separation from the parent, were examined. The atrial epithelium starts dedifferentiation from this stage. The PmRXR mRNA was detected in a mesenchymal cell type, called glomerulocyte in the proximal region ( Fig. 5A,C,D). Glomerulocytes are a few times larger in diameter than the other mesenchyme cells and have a donut-shaped cytoplasm around the nucleus ( Fig. 5D,E). They are known to differentiate in the epidermal sheet and come out into the mesenchymal space ( Hirose & Mukai 1992). No signal was detected in any cell type in the lateral or distal region ( Fig. 5B,C,E).

Figure 5.

Expression of Polyandrocarpa misakiensis retinoid X receptor (PmRXR) mRNA in the developing bud of Polyandrocarpa misakiensis. (A) The proximal region of the bud was examined with an antisense PmRXR RNA probe. The PmRXR mRNA is detectable in the glomerulocytes within the epidermis or in the mesenchymal space (arrowheads). (B) All of the cell types including the glomerulocyte (arrowhead) in the lateral region were not stained, even with the antisense probe. (C) Schematic illustration of the developing bud. (D) High magnification view of the boxed region in (A). (E) High magnification view of the boxed region in (B). Glomerulocytes can be easily distinguished from the other cell types by a large donut-like morphology. ae, atrial epithelium. ep, epidermis.

Discussion

In the present study, we isolated a cDNA, named PmRXR, encoding an ascidian homolog of RXR. From the sequence conservation, PmRXR is suggested to function as a heterodimerization partner of PmRAR that had been identified as an ascidian homolog of RAR ( Hisata et al. 1998 ). As PmRXR and PmRAR cDNA lack the 5′ end encoding a part of the A/B region, we supplemented them with the corresponding region of human RXRα and human RARα cDNA, respectively. As contribution of the A/B region for DNA-binding, ligand-binding, dimerization, and ligand-dependent transactivation is not critical ( Chen et al. 1994 ), we used these chimeric cDNA, named (h + Pm)RXR and (h + Pm)RAR, to test the function of PmRXR and PmRAR. Gel-shift assays showed that (h + Pm)RXR and (h + Pm)RAR bound vertebrate-type RARE ( Fig. 3). The formation of shifted complexes requires both proteins, suggesting that physical interaction (probably heterodimerization) between (h + Pm)RXR and (h + Pm)RAR is required for specific DNA binding. Electroporation experiments demonstrated the RA-dependent transactivation mediated by (h + Pm)RXR and (h + Pm)RAR. In case the embryos were electroporated with the RARE-containing reporter gene (DR5/Z) and two effector genes (Ci-Bra/(h + Pm)RXR and Ci-Bra/(h + Pm)RAR), and then treated with RA, ectopic staining was detectable in the tail region. The position and arrangement of the ectopically stained cells resembled those stained with a notochord-specific reporter gene (– 3.5 kb Ci-Bra/lacZ), suggesting that it is the notochord that expressed lacZ depending on RA and the effector genes. These functional analyses suggest that the PmRXR/PmRAR complex functions as a retinoic acid-dependent transcriptional activator.

Previous reports showed that RA affected morphogenesis mainly in the anterior neural tissues ( Denucé 1991; Katsuyama et al. 1995 ). The embryos, electroporated with the reporter gene only, showed a similar phenotype upon treatment with RA. The RA does not affect tail morphogenesis, suggesting that the notochord lacks endogenous RXR and/or RAR in ascidians ( Katsuyama et al. 1995 ; Katsuyama & Saiga 1998). However, electroporation of the two effector genes caused a severe mutant phenotype in the tail region. It is possible that RA treatment of the embryos expressing (h + Pm)RXR and (h + Pm)RAR in the notochord caused activation of a developmental program that was otherwise silent in the notochord. This supports the conclusion that the PmRXR/PmRAR complex functions as a RA-dependent transcriptional activator.

There has been no evidence suggesting that insects have a retinoic acid signaling pathway. No RAR homolog has been reported in insects. Drosophila USP does not bind 9-cis RA ( Oro et al. 1990 ). The USP functions as a heterodimerization partner of the ecdysone receptor ( Thomas et al. 1993 ; Yao et al. 1993 ), or a receptor for juvenile hormones ( Jones & Sharp 1997). However, in the crustacean Uca pugilator, RA affects limb regeneration ( Hopkins & Durica 1995). The RXR homolog, cloned from this species, is more similar to vertebrate RXR than to insect USP ( Chung et al. 1998 ). The RA signaling (possibly mediated by RXR) might have occurred in the common ancestor of chordates and crustaceans at an early stage of animal evolution. The PmRXR is on a separate branch from all the subtypes of vertebrate RXR. This suggests that vertebrate RXR subtypes arose after the divergence of urochordates and vertebrates. This is consistent with the idea that two sets of genome duplication events have occurred during vertebrate evolution ( Holland et al. 1994 ). Hisata et al. (1998) showed that PmRAR was also on a separate branch from all the vertebrate RAR subtypes. A similar result was obtained in the present study ( Fig. 2).

Implantation of RA-treated mesenchyme cells into the Polyandrocarpa bud caused secondary axis formation (K. Hara, pers. comm., 1992). In addition, the conditioned medium of the RA-treated mesenchyme cells affects expression of the specific antigen of the atrial epithelium explants (K. Kawamura, unpubl. data, 1994). From these observations, mesenchyme cells were thought to be the target of RA action ( Kawamura & Fujiwara 1995). In the present study, PmRXR mRNA was detected in a mesenchymal cell type, called glomerulocyte, in the proximal region of the developing bud ( Fig. 5). The PmRAR mRNA was also detected in this cell type in the same region of the bud at a similar stage of development ( Hisata et al. 1998 ). These results suggest that glomerulocytes receive RA and secrete protein factors that directly induce transdifferentiation of the atrial epithelium. A possible cascade is shown in Fig. 6. Retinoic acid synthesis is triggered in the proximal region of the developing bud by separation from the parental body wall. As ALDH activity is primarily detected in the epidermis of the proximal region, RA is thought to be synthesized in the epidermis or glomerulocytes anchored within the epidermis ( Kawamura et al. 1993 ). It is suggested that RA synthesis, activation of the PmRAR/PmRXR complex, and up-regulation of ALDH and PmRAR form a positive feedback loop ( Kawamura et al. 1993 ; Kawamura & Fujiwara 1995; Hisata et al. 1998 ). The PmRXR/PmRAR complex within the glomerulocytes activates a set of genes involved in the induction of transdifferentiation of the atrial epithelium. Many glomerulocytes are found not only in the mesenchymal space but also within the epidermal layer ( Fig. 5A), suggesting that the production of transdifferentiation factors may start before the glomerulocytes come out into the mesenchymal space.

Figure 6.

A model for the mechanism of retinoic acid (RA) action that induces transdifferentiation of the atrial epithelium in the developing bud of Polyandrocarpa misakiensis. The proximal region of the bud is schematically drawn. Aldehyde dehydrogenase (ALDH) is activated and synthesizes RA in the epidermis at the proximal end of the bud. The RA is incorporated into various types of cells including the glomerulocytes. In response to RA, the PmRXR/PmRAR complex activates target genes. The target genes may include PmRAR ( Hisata et al. 1998 ), ALDH ( Kawamura et al. 1993 ), and secreted protein factors that stimulate transdifferentiation of the atrial epithelium.

Acknowledgements

We thank D. J. Mangelsdorf, R. M. Evans, N. Moghal and K. Umesono for mammalian RXR and RAR plasmids, J. C. Corbo and M. Levine for Ciona transgenes, Z. Imoto, T. Nishikata and W. Godoh for animals, and T. Suzuki for construction of phylogenetic trees. Thanks are also owed to T. Matsuoka, N. Iwasaki, S. Furuyoshi, M. Ashiuchi and S. Nagata for allowing us to use their machines. This work was supported in part by a Grant-in-Aid from the Ministry of Education, Culture, Sports and Science, Japan. S. F. is also supported by funds from the Kato Memorial Bioscience Foundation.

Ancillary