Developmental regulation of the orphan receptor COUP-TF II gene in spinal motor neurons

Transcription factors, belonging to the steroid/thyroid hormone receptor superfamily, play important roles in the control of the development of vertebrates and invertebrates (reviewed by Leid et al., 1992; Linney, 1992; Oro et al., 1992). Among the best characterized members of the superfamily are the receptors for steroids, retinoids, thyroid hormones, vitamin D, and ecdysone. Characteristic structural features of hormone receptors are the DNA binding domain (DBD) and the ligand binding domain (LBD) (reviewed by Green and Chambon, 1988; Evans, 1988; Beato, 1989; Petkovich, 1992). Upon ligand binding, receptors are capable of regulating the expression of specific target genes. In addition to the well characterized hormone receptors, numerous other members of the superfamily have been isolated, but their ligands have not been identified, and hence they are termed ‘orphan’ receptors. Examples of orphan receptors are the chicken ovalbumin upstream promoter-transcription factors (COUP-TFs). Originally, COUP-TFs were purified from HeLa cells as proteins capable of binding to a specific DNA element (the COUP element) of the chicken ovalbumin promoter (Sagami et al., 1986). Subsequently, two closely related genes for COUP-TFs were isolated in human: hCOUP-TF I (Wang et al., 1989; Richie et al., 1990) also known as ear3 (Miyajima et al., 1988), and hCOUP-TF II (Richie et al., 1990; Wang et al., 1991) which is identical to ARP-1 (Ladias and Karathanasis, 1991). Mouse (M.-J. Tsai, unpublished data), sea urchin (Chan et al., 1992), Xenopus (Matharu and Sweeney, 1992), zebrafish (Fjose et al., 1993), and Drosophila (Mlodzik et al., 1990) COUP-TF homologs were also isolated.

Studies with tissue culture cells have shown that COUP-TFs have promiscuous binding activity to hormone response elements of the vitamin D receptor (VDR), thyroid hormone receptors (TRs) and retinoid receptors (RARs and RXRs) and can repress hormonal induction of target genes (Cooney et al., 1992; Kliewer et al., 1992; Tran et al., 1992; Widom et al., 1992). Several different mechanisms are implicated in this repression including a direct competition of COUP-TFs for the hormone response elements and heterodimerization with RXRs (Cooney et al., 1992, 1993; Tran et al., 1992; Kliewer et al., 1992; Segars et al., 1993). It is thought that heterodimerization reduces the concentration of free RXRs that are required as cofactors for effective binding of VDR, TRs, and RARs to their own response element. Collectively, these studies show that the regulation of gene expression by hormone receptors does not only depend on the receptors and their ligands, but also on the interaction of these receptors with activators and repressors.

Although these molecular studies shed light on the bio-chemical mechanism of action of COUP-TFs, the role of COUP-TFs in development is virtually unexplored, except for Drosophila. A genetic analysis of the fly COUP-TF cognate seven-up has shown that this gene is required in a specific subset of photoreceptor neurons during eye development (Mlodzik et al., 1990). To begin to investigate the function of COUP-TFs in vertebrates, we have cloned the gene encoding a chicken COUP-TF (cCOUP-TF II). Sequence comparison with human COUP-TF sequences identifies this avian gene as cCOUP-TF II. Characterization of the expression pattern by in situ hybridization revealed that cCOUP-TF II is expressed at high levels and transiently in the developing ventral spinal cord in regions where motor neurons form. Since the differentiation of ventral cell types is dependent on inductive signals from the notochord (van Straaten et al., 1985; Hirano et al., 1991; Yamada et al., 1991), we grafted an additional notochord adjacent to the neural tube in order to induce the development of ectopic motor neurons. We found ectopic expression of cCOUP-TF II mRNA in such induced motor neurons. Both the pattern of expression and the induction of cCOUP-TF II in ectopic motor neurons show that this receptor is a marker for spinal motor neurons and, furthermore, might be involved in motor neuron development.

To generate a specific probe to clone a chicken homolog of human COUP-TFs, polymerase chain reaction (PCR) was employed. Two oligonucleotides were synthesized for use as primers to amplify a 270 bp fragment of the LBD of hCOUP-TF I (Wang et al., 1989; the 5′ primer is position 628-647, the 3′ primer is position 890-871). The 5′ oligonucleotide included a BamHI restriction site: 5′-ATGGATCCG- GCATCGAGAACATCTGCGA-3′. The 3′ oligonucleotide included an EcoRI restriction site 5′-ATGAATTCAAGATGCGGATGTG- GTCCAT-3′. Purified DNA (1 μg) from a chick embryonic λZAPII cDNA library (stages 14-17, Hamburger and Hamilton, 1951) was used as a template. The following thermal cycling program was used to amplify the 270 bp DNA fragment. Denaturation at 96°C for 1 minute, annealing at 50°C for 2 minutes and extension at 72°C for 3 minutes for 35 cycles. The amplified DNA fragment was gel purified and digested with BamHI and EcoRI and subcloned into a similarly digested pGEM–7Zf(+) vector (Promega). The plasmid was named pCOUP-270. The PCR fragment was sequenced on both strands using SP6 and T7 primers and the 7-deaza-dGTP Sequenase kit (USB). Sequence analysis confirmed the fragment as part of an authentic chicken COUP-TF gene.

The subcloned PCR fragment was isolated and ³²P-labeled using random primer labeling (Boehringer Mannheim) and subsequently used to screen 5×10⁵ pfu of the chick embryonic λZAPII cDNA library (stages 14-17). Several positive clones were identified, isolated, and sequenced. Sequence analysis of two of these cDNA clones showed that they were overlapping homologs of the hCOUP- TF II. Comparison of these clones with pCOUP-270 revealed that pCOUP-270 has three mismatches which are contributed by the PCR primers derived from hCOUP-TF I.

Total RNA from freshly isolated embryonic tissues was extracted with isothiocyanate and acidic phenol using the Stratagene kit. 8 μg of total RNA or 1 μg of poly(A)⁺ RNA was electrophoresed on a 1% agarose/formaldehyde gel and capillary-blotted overnight with 20× SSC (1× SSC is 150 mM NaCl, 15 mM sodium citrate, pH 7.4) onto Hybond-N membrane (Amersham). After UV crosslinking for 5 minutes, the membrane was prehybridized in hybridization solution (50% formaldehyde, 0.75 M NaCl, 0.1 M Tris-HCl pH 7.8, 5 mM EDTA, 0.1% SDS, 100 μg/ml herring sperm DNA, 100 μg/ml yeast tRNA) for 24 hours at 42°C in a roller bottle. Hybridization was carried out overnight at 42°C in hybridization solution containing 500,000 cts/minute per ml of ³²P-labeled probe. The membrane was washed three times in 2× SSC, 0.1% SDS at room temperature for 10 minutes, and twice in 2× SSC, 0.1% SDS at 55°C for 45 minutes. Exposure to X-ray film was done with an intensifying screen at −80°C for 80 hours. The probe for hybridization was produced by asymmetric PCR. Conditions for the 50 μl PCR reaction were 35 cycles (1 minutes at 94°C; 1 minute at 50°C; 2 minutes at 72°C) with Stratagene reaction buffer and Taq polymerase, and with 25 μM dATP, 25 μM dTTP, 25 μM dGTP, 0.5 μM dCTP, 70 μCi [α-³²P]dCTP, 0.4 μM T7 primer and 80 ng of BamHI-linearized plasmid pCOUP-270. This procedure produced an antisense ssDNA encompassing part of the LBD. RNA size marker (BRL) was used to determine the length of the cCOUP-TF transcripts.

EMSA was generally performed as previously described (Cooney et al., 1992). Oligonucleotides (5′-AGAAGTTTGACCTTTGACAC- CAT-3′ and 5′-TTCTATGGTGTCAAAGGTCAAAC-3′) corresponding to the authentic COUP element, a DR1 element (a direct repeat of AGGTCA with one base pair spacing), were synthesized and annealed. The double stranded oligonucleotides were end-labeled using Klenow enzyme and 100 μCi each of [α-³²P]dATP and [α- ³²P]dTTP, and used as probe for EMSA. 200 pg of probe was used per reaction. Nuclear extracts were prepared as previously described (Gossett et al., 1989) from 7 stage 25 chick embryos (1.3 g) and from 40 dissected stage 27/28 spinal cords (230 mg, some included hindbrain). A 20-fold molar excess of unlabeled DR1 or of unlabeled progesterone response element derived from the tyrosine aminotransferase gene (Strähle et al., 1987) was used for the competition analyses. Anti-COUP-TF antibodies (1.4 μg; affinity purified against hCOUP-TF I; see below) were used for the supershift experiments and were preincubated with the samples for 5 minutes. The products of the reactions were resolved by native 5% polyacrylamide gel elec- trophoresis. The full-length cCOUP-TF II cDNA was subcloned into the XhoI/EcoRI sites of pGEM-7Zf(+). In vitro synthesized protein was generated in the presence of [³⁵S]methionine using the TNT T7 coupled reticulocyte lysate system (Promega).

Antiserum against proteins binding to the authentic COUP element of the chicken ovalbumin promoter was previously raised (Wang et al.,1989). For EMSA antibodies specific for COUP-TFs were affinity purified from this antiserum. To prepare large quantities of purified COUP-TF for affinity chromatography, hCOUP-TF I was overexpressed in E. coli using the glutathione S-transferase (GST) system (Smith and Johnson, 1988). The entire coding region of the hCOUP- TF I cDNA was cloned in frame into the BamHI/EcoRI sites of the IPTG inducible E. coli expression vector pGEX-2T (Pharmacia). GST-hCOUP-TF I was batch purified from whole cell extracts using glutathione Sepharose 4B beads (Pharmacia) and was eluted from the beads with 100 mM glutathione. The purified protein was conjugated to a CNBr-activated Sepharose 4B column (Pharmacia) to form a GST-hCOUP-TF I affinity column. The anti-COUP-TF serum was passed over the column and washed. Bound COUP-TF-specific antibodies were eluted with 1 M NaCl, 0.1 M glycine (pH 2.3) and neutralized with equal volumes of 0.1 M Tris-HCl (pH 11). To test the specificity of the purified antibodies, both in vitro produced hCOUP- TF I and hCOUP-TF II were used in EMSA, and the addition of these antibodies resulted in supershifted complexes. Using HeLa cell nuclear extracts in EMSA the purified antibodies only supershifted the retarded complex which comigrated with the complex containing the recombinant hCOUP-TFs, and they did not recognize any of the other COUP element binding factors such as RXRs (data not shown).

Embryo collection, sectioning and in situ hybridization were performed as previously described (Sundin et al., 1990). Antisense RNA probes labeled with [α-³⁵S]UTP (1000 Ci/mmol, Amersham) were produced with T7 RNA polymerase and BamHI-linearized pCOUP-270. Hybridization was done overnight at 50°C with a probe concentration of 0.09 ng/μl. Posthybridization treatments were as follows. (i) Two washes in 50% formamide, 2× SSC, 20 mM β-mercaptoethanol (FSM) at 63.5°C for 30 minutes, (ii) digestion with 10

μg/ml RNase A in 4× SSC, 20 mM Tris-HCl (pH 7.6), 1 mM EDTA at 37°C for 30 minutes, and (iii) two washes in FSM at 63.5°C for 45 minutes. Slides were dipped in Kodak NTB-2 emulsion and exposed for 10 days. Sections were analyzed with a Leitz Diaplan microscope. The micrographs shown are double exposures, the red color represents the in situ hybridization signal using dark-field optics, and the blue color is the cell nuclei stained with Hoechst 33258, viewed under epifluorescence optics.

Stage-10 to −12 chick embryos were dissected with sharpened tungsten needles in 0.9% NaCl containing 1 μg/ml dispase (Boehringer Mannheim). After removal of mesenchymal cells, the notochord fragment was washed briefly in 1% BSA, 0.9% NaCl. A window was opened on the shell of the host egg (stages 10–11), and black Indian ink (Rapidograph 3080-F, Bloomsbury, N.J.; diluted 1:3 in 0.9% NaCl) was injected into the subgerminal cavity to visualize the embryo. At the posterior trunk level, a slit was made with the tungsten needle between the neural tube and the segmental plate, where the notochord graft was inserted at various depths. The middorsal insertion was achieved by transplanting the graft into the closing neural tube. These notochord transplantation experiments involved a total of 49 embryos. The embryos were incubated, and 28 of them survived until about stage 21 (about 48 hours after the surgery). They were fixed overnight in 4% paraformaldehyde, PBS at 4°C (PBS is 137 mM NaCl, 2.7 mM KCl, 12 mM Na₂HPO₄, pH 7.4).

After briefly washing with sterile PBS, the embryos were transferred to 0.5 M sucrose, PBS (treated with 0.1% diethylpyrocarbonate for 4 hours and autoclaved for 10 minutes) and gently shaken on the platform for 1–3 hours at 4°C. The embryos were embedded in OCT (Bright Instrument Company) on a tracing paper placed on dry ice. Sections from 18 embryos were cut at 13 μm on a cryostat and alternately collected onto poly-L-lysine coated slides as used for in situ hybridization. One set of sections representing 10 embryos was subjected to in situ hybridization, and the other set was stained for SC1 and/or Islet-1 immunolocalization.

For SC1 immunolocalization, sections were treated with 1% periodic acid for 10 minutes at room temperature and washed with 20 mM Tris-HCl (pH 8), 150 mM NaCl containing 0.1% Triton X-100 (TST). After blocking with centrifuged TST containing 5% (w/v) nonfat powdered milk (TSTM), the sections were incubated with SC1 (undiluted hybridoma supernatant, kindly provided by Drs Tom Jessell and Hideaki Tanaka) for 3 hours at room temperature. Horseradish peroxidase-conjugated anti-mouse IgG (Southern Biotechnology Associates) at 2 μg/ml was used as the secondary antibody. The peroxidase reaction was performed at room temperature for 15 minutes in 150 mM NaCl, 20 mM Tris-HCl (pH 8) containing 500 μg/ml of 3-3′-diaminobenzidine and 0.03% H₂O₂. For Islet-1 immunolocalization, aldehyde groups were blocked with 0.34% L- lysine and 0.05% NaIO₄ in PBS for 30 minutes. Sections were rinsed in PBS and then incubated in TSTM for 30 minutes. Incubations of primary and secondary antibodies were done in TSTM for 2 hours. Primary antibody was a 1:100 dilution of hybridoma supernatant of anti-Islet-1 antibody (kindly provided by Dr Tom Jessell), and the secondary antibody was fluorescein isothiocyanate-conjugated goatanti-mouse IgG (H+L) (Southern Biotechnology Associates) at 10 μg/ml. Washes after antibody incubation were three times in TST. SC1-immunostained sections were mounted in Canada balsam/xylene (1:1) and analyzed with bright-field optics, and for Islet-1-immunostained sections a solution containing 6 g glycerol, 2.4 g Moviol-488 (Calbiochem), 6 ml H₂O, 12 ml 0.2 M Tris-HCl pH 8.5, and 2.5% (w/v) 1,4-diazobicyclo-[2.2.2.]-octane (Aldrich) was used, and analysis was performed under epifluorescence optics.

Chicken COUP-TF II cDNAs were isolated using a 270 bp probe obtained from an embryonic chick cDNA library by PCR amplification. Two primers spanning a 270 bp region of the LBD of hCOUP-TF I were selected. This region of the LBD is conserved between hCOUP-TF I and hCOUP-TF II, but is divergent between other hormone receptors, thus reducing the probability of amplifying a DNA fragment from the cDNA of another receptor subfamily. A search of Genbank revealed that the subcloned 270 bp DNA fragment was 86% homologous to hCOUP-TF I and 88% homologous to hCOUP-TF II. At the amino acid level, these homologies translated into 98% and 100% identity with hCOUP-TF I and hCOUP-TF II, respectively. The 270 bp fragment was subsequently used to screen the embryonic chick cDNA library at high stringency. Overlapping cDNA clones were characterized and their sequences were assembled. The resulting nucleotide sequence and corresponding open reading frame is shown in Fig. 1A. Alignments of the amino acid sequences of the cCOUP-TF II, hCOUP-TF I and hCOUP-TF II (Fig. 1B) show that the overall identity of the avian protein with hCOUP-TF I and hCOUP-TF II is 89% and 95%. DBD identities are 98% (hCOUP-TF I) and 100% (hCOUP-TF II), and in the LBD, identities of 96% (hCOUP- TF I) and 99% (hCOUP-TF II) are noted. The amino-terminal region shares 44% and 68% identity with hCOUP-TF I and hCOUP-TF II, respectively. Taken together, these sequence comparisons suggest that the chicken COUP-TF shown in Fig. 1 is COUP-TF II, designated as cCOUP-TF II. Further support for this assignment comes from the high homology between the murine COUP-TF II sequence (unpublished results, M.-J. T.) and cCOUP-TF II.

To characterize the size of cCOUP-TF II transcripts, northern blots were hybridized with the 270 bp probe derived from the cCOUP-TF II LBD (Fig. 1A). Analysis of poly(A)⁺ RNA from stage 17 embryos (Hamburger and Hamilton, 1951) reveals one major RNA species of 3.8 kb in length and two minor transcripts of 2.8 kb and 5.4 kb (Fig. 2A, lane 10). The minor species could be due to cross-hybridization to a related mRNA species, such as transcripts from another chicken COUP-TF gene, although it is presently not known whether the chicken genome contains more than one COUP-TF gene. Alternatively, the presence of several transcripts could be caused by alternative splicing of the cCOUP-TF II pre- mRNA. The developmental profile of cCOUP-TF II

expression reveals that mRNA levels are relatively low during gastrulation, neurulation and early somitogenesis (stages 4 –8), but from stage 8-9 onward expression gradually increases to reach maximum levels around stages 20 –25 (Fig. 2A). Later in development, expression in all tissues examined is lower than in younger whole embryos (Fig. 2B). The notable exception is in the pancreas, which has levels of cCOUP-TF II mRNA as high as those detected in embryos of stages 15 –28. rRNA quantitation of the samples analyzed on these northern blots revealed that equal amounts of RNA were loaded onto the gels (data not shown). The absolute amount of cCOUP-TF II mRNA has not been determined, but the fact that there is a weak but distinct band at stages 4-5, although only 8 μg of total RNA were used, indicates that there are significant levels of cCOUP-TF II transcripts during gastrulation. Taken together, cCOUP-TF II mRNAs are detectable at all stages and in all tissues analyzed, yet the amount of mRNA varies greatly during development.

COUP-TFs were originally identified by their ability to bind to an upstream element in the chicken ovalbumin promoter, the COUP element (Sagami et al., 1986; Wang et al., 1989). This element is a DR1 steroid response element, i.e., an AGGTCA direct repeat with a 1 bp spacing. In vitro binding studies demonstrated that DR1 elements are one of the DNA sequences to which COUP-TFs bind with high affinity, and they are likely to be one of the primary response elements through which the action of COUP-TFs is mediated (Cooney et al., 1992). To analyze whether the cCOUP-TF II mRNA observed in embryonic tissues (Fig. 2) generates functional protein capable of binding to its response element, electrophoretic mobility shift assays (EMSA) were performed (Fig. 3)Nuclear extracts from whole stage-25 chick embryos were prepared, and 1 μg of protein was incubated with ³²P-labeled DR1 (Fig. 3A, lane 5). Several retarded protein-DNA complexes of different mobilities were observed. A rapid and a slow mobility complex and a complex of intermediate mobility probably consisting of two closely migrating bands were detected. Each complex could be specifically competed with a 20-fold molar excess of unlabeled DR1, but not with unlabeled progesterone response element (PRE) (Fig. 3A, lanes 6 and 7). Thus, the formation of each complex is dependent on the presence of a DR1. To identify which of these complexes contained COUP-TFs, immunoanalysis was performed with affinity purified antibodies that react with hCOUP-TF I and II proteins (see Materials and methods). Preincubation of the nuclear extract with these antibodies selectively retarded the mobility of the upper band of the intermediate complex indicating that this band contained cCOUP- TF protein (Fig. 3A, lane 8). In contrast, the other bands were not supershifted suggesting that they do not contain cCOUP- TF protein. We conclude that in whole stage-25 embryos COUP-TF protein is functionally expressed and competes with a number of different factors for binding to DR1 elements. If there were several cCOUP-TF proteins in chicken, EMSA might not distinguish between them and thus the supershift seen in Fig. 3A may involve more than one COUP-TFs.

In situ hybridization showed that cCOUP-TF II is highly expressed in the developing spinal motor neurons (see below, and Fig. 4). To assay for the presence of cCOUP-TF, nuclear extracts were prepared from dissected spinal cords of stage-27 to −28 chick embryos, and 2 μg of nuclear protein was incubated with ³²P-labeled DR1. In these extracts one predominant retarded complex of intermediate mobility was observed (Fig. 3A, lane 1). This complex formed in a DR1- dependent manner and was supershifted by anti-COUP-TF antibodies (Fig. 3A, lanes 2 –4). This indicates that cCOUP-TF is present in this complex and is the predominant DR1 binding activity in the spinal cord.

To examine this point further, we compared the mobility of this complex with that formed by in vitro synthesized cCOUP- TF II and DR1. A cCOUP-TF II cDNA, comprising the entire open reading frame, was subcloned into pGEM-7Zf(+), and the cCOUP-TF II protein was synthesized by coupled in vitro transcription-translation. The resulting cCOUP-TF II protein formed a single complex which had the same mobility as the predominant COUP-TF complex seen with chick spinal cord extracts (Fig. 3B, compare lanes 9 and 13). Binding of the in vitro synthesized cCOUP-TF II was also DR1-dependent (Fig. 3B, lanes 10-11), and the complex was supershifted to the same position as the chick spinal cord complex when anti-COUP-TF antibodies were added (Fig. 3, compare lanes 4 and 12). We conclude that in the developing spinal cord, COUP-TF is likely to be the predominant factor complexing with the DR1 response elements, an element characteristic of COUP-TF target genes.

In an effort to explore the role of cCOUP-TF II during embryonic development, we determined its spatiotemporal expression pattern by in situ hybridization using the same probe as employed for northern blot analyses. Antisense riboprobe gave specific signals (Fig. 4), whereas the sense riboprobe produced only background levels of silver grain density (not shown). It is possible that the chick genome contains two COUP-TF genes corresponding to human and murine COUP-TF I and II (Richie et al., 1990; and M.-J. T., unpublished data). Using murine COUP-TF I and II probes similar to our avian probe, resulted in non-overlapping expression patterns in mouse embryos (M.-J. T., unpublished data) suggesting that our chicken probe is specific for the cCOUP-TF II transcripts. cCOUP-TF II displays a complex expression pattern in chick embryos. The gene is expressed in dorsal root ganglia (Fig. 4B), in all cranial sensory ganglia (e.g. Fig. 6) and in derivatives of the lateral and intermediate mesoderm such as mesonephros, mesentery, gut wall, and limb bud mesenchyme (data not shown). In the central nervous system, cCOUP-TF II transcripts were found in various parts of the brain (see below, and Fig. 6) and in the spinal cord during the period when spinal motor neurons are born. The following description will focus on cCOUP-TF II expression in the developing spinal motor neurons.

The time course of cCOUP-TF II expression in the spinal cord was analyzed in detail at the upper thoracic level (somites 13 –20) because at this level somatic as well as sympathetic motor neurons (column of Terni) develop (Levi-Montalcini, 1950). At stage 15, when the earliest motor neurons in the brachial spinal cord complete their terminal mitosis (Hollyday and Hamburger, 1977; Prasad and Hollyday, 1991), cCOUP- TF II expression in the spinal cord is uniform, but very low (not shown). A localized pattern of expression is first detected in the ventral spinal cord at stage 18 (Fig. 4A). At this stage, the intermediate zone has not yet developed, and the cCOUP- TF II expression domain coincides with the area where motor neurons are developing (Ericson et al., 1992; and references therein). By stage 22, cCOUP-TF II expression is highly upregulated in the ventral portion of the intermediate zone where motor neurons are located (Fig. 4B). In contrast, ventricular zone, floor plate, and dorsal intermediate zone express significantly lower levels of cCOUP-TF II.

To demonstrate that the cCOUP-TF II-positive area in the ventral spinal cord comprises motor neurons, adjacent transverse sections were analyzed either by in situ hybridization using a cCOUP-TF II probe, or with antibodies against SC1 or Islet-1 proteins. In combination, these two markers define motor neurons in the spinal cord (Tanaka and Obata, 1984; Ericson et al., 1992). SC1 is localized on the cell surface and is a member of the immunoglobulin superfamily (Tanaka et al., 1991). Islet-1 is a homeodomain protein of the LIM family (Karlsson et al., 1990; Ericson et al., 1992). Immunolocalization of SC1 (Fig. 5B) and of Islet-1 (Fig. 5C) in the ventral spinal cord is coextensive with the region that expresses cCOUP-TF II mRNA (Fig. 5A). This result indicates that cCOUP-TF II-positive cells are motor neurons.

By stage 25, cCOUP-TF II expression in the lateral motor column reaches the highest level (Fig. 4C). Around stage 26, a few small cCOUP-TF II-positive cell clusters separate dorsally from the lateral motor column (data not shown). By stage 28, a second cCOUP-TF II-positive expression domain can be discerned (Fig. 4D, arrowhead). At stage 29 (Fig. 4E), both domains are completely separated, and it becomes clear that the region adjacent to the central canal coincides with the preganglionic sympathetic neurons of the column of Terni, and the region in the ventral horn coincides with the somatic motor column. The appearance of these two cCOUP-TF II expression domains closely reflects the establishment of these two distinct motor neuron populations as described by Prasad and Hollyday (1991). By stage 32, cCOUP-TF II expression in somatic motor neurons is greatly reduced, while expression in the column of Terni persists (Fig. 4F). By stage 37, expression in somatic motor neurons has virtually ceased, whereas motor neurons in the column of Terni are still slightly cCOUP-TF II- positive (data not shown). By day 18 (stage 44), shortly before hatching, cCOUP-TF II expression throughout the spinal cord is at background levels (data not shown).

In summary, cCOUP-TF II is transiently expressed in somatic and in sympathetic motor neurons. Around stage 18, when motor neurons begin to develop, cCOUP-TF II expression is rather low but begins to be localized. Expression is strongly upregulated when motor neurons undergo differentiation and migrate to their final locations. Subsequently, cCOUP-TF II expression begins to decrease first in the somatic motor column and then in the column of Terni, and finally, the gene ceases to be expressed by embryonic day 18.

Recent studies using motor neuron markers suggest that the differentiation of motor neurons depends on inductive signals emanating from the notochord and/or the floor plate (Yamada et al., 1991, 1993; Ericson et al., 1992; Goulding et al., 1993). Our finding that cCOUP-TF II is expressed in developing motor neurons led us to ask whether notochord grafts can change the expression of cCOUP-TF II. We therefore analyzed the expression domains of cCOUP-TF II in spinal cords of embryos that had received an additional notochord at stages 10 –11. 48 hours after grafting, when the embryos had reached about stage 21, they were analyzed by immunostaining and in situ hybridization. Fig. 5A shows an example in which the ectopic notochord was placed dorsally. Apart from the two cCOUP-TF II expression domains in the ventral region, representing the normal lateral motor columns, an ectopic domain of cCOUP-TF II expression is located dorsally (Fig. 5A, arrow) in a region where only very low levels of cCOUP-TF II expression are seen in the normal embryo (compare also Fig. 4B). Note, cCOUP-TF II transcripts are not found in neuroepithelium immediately next to the grafted notochord, but at a certain distance as has already been observed for SC1 by Yamada et al. (1991). This distance is similar to that seen between the native notochord and the native lateral motor column (Fig. 5A). Sections adjacent to that shown in Fig. 5A were immunostained with monoclonal antibodies against the motor neuron markers SC1 or Islet-1. The ectopic cCOUP-TF II expression domain in the dorsal spinal cord (Fig. 5A) are coextensive with the ectopic domain of SC1 (Fig. 5B, arrow) and of Islet-1 (Fig. 5C, arrow). Nine out of ten embryos analyzed showed coextensive ectopic SC1 and cCOUP-TF II expression. In three embryos the ectopic cCOUP-TF II- positive area was localized to the dorsal spinal cord and was completely separated from the ventral lateral motor column. In four other specimens the ectopic region of expression fused to the native lateral motor column resulting in an enlarged lateral motor column. In two cases, both an expansion of the native cCOUP-TF II-positive domain and the formation of a separated cCOUP-TF II-positive domain was observed. Except for one case in which the in situ hybridization analysis was not successful, cCOUP-TF- and SC1-positive domains were always coextensive.

In addition to the spinal motor neurons, cCOUP-TF II was also expressed in other regions of the developing nervous system.

In the brain, cCOUP-TF II expression was not restricted to the intermediate zone as in the spinal cord (Fig. 4), but was also seen in the ventricular zone of several structures (Fig. 6). The floor plate was cCOUP-TF II-negative in younger embryos (Fig. 6B,C,E; see also Fig. 4), but the floor plate of the hindbrain of older embryos was positive (Fig. 6F).

Expression in the brain was seen as several isolated domains. In the forebrain, the gene was expressed in a contiguous domain spanning between diencephalon and telencephalon at stage 19 (Fig. 6A). At stage 23 this region comprises the presumptive thalamus, hypothalamus, and the optic stalk (Fig. 6B). After stage 25, this domain extended into the striatum of the telencephalon (data not shown). In stage 30 embryos (data not shown), the posterior border of this expression domain was found to correspond to the D1/D2 boundary of the diencephalic neuromeres (Figdor and Stern, 1993). In the midbrain, the oculomotor nucleus, the anteriormost somatic motor element, expressed cCOUP-TF II at stage 22 (Fig. 6C) and expression was highly upregulated by stage 28 (Fig. 6D) exhibiting a complex pattern which reflects the intricate organization of this nucleus (Sarnat and Netsky, 1981). cCOUP-TF II was strongly expressed in cells that align along the oculomotor nerve (Fig. 6D, arrows). The resolution of the in situ hybridization data did not allow the identification of these cells, but they might be Schwann cells that had delaminated from the ventral neural tube (Lunn et al. 1987; Loring et al., 1988). The hindbrain showed widespread expression including the presumptive cerebellum (Fig. 6E,F). Expression was not restricted to the motor component, but small spots of higher cCOUP-TF II expression were recognized in the intermediate zone which colocalized with developing nuclei (Fig. 6F). In the peripheral nervous system, cranial sensory ganglia are positive (Fig. 6B,E,F) and an interesting pattern of cCOUP- TF II mRNA was seen in the otocyst (Fig. 6F) where expression was highly upregulated in the dorsomedial aspect, the area where the primary sensory neurons arise and where the ganglion primordium will make contact. In summary, expression of cCOUP-TF II outside the spinal cord is complex and spatially restricted; cCOUP-TF II transcripts are not only detected in the motor component nor were they restricted to the intermediate zone. In the diencephalon, cCOUP-TF II expression seems to reflect the segmental organization of this part of the brain (Figdor and Stern, 1993). Such diversity in cCOUP-TF II expression implies that the gene may be required in various developmental and local contexts.

During embryonic development a large variety of neuronal cell types are generated from undifferentiated neuronal precursor cells. After a phase of extensive proliferation, neuronal differentiation begins which entails cell migration, axonal growth, establishment of synaptic connections and cell death. Little is known about which genes are required for these different processes. Our study raises the possibility that the chicken orphan receptor cCOUP-TF II could be one of the factors required for the differentiation of spinal motor neurons. We found that by stage 18 expression of cCOUP-TF II starts to be upregulated in the ventral spinal cord in a region that gives rise to motor neurons, and expression reaches highest levels in the lateral motor column at the time the majority of motor neurons are postmitotic (Fig. 4C). Expression gradually declines (Fig. 4F). Shortly before hatching, cCOUP-TF II transcripts are no longer detectable in spinal motor neurons by in situ hybridization, which underlines the fact that in the spinal cord cCOUP- TF II is a transiently expressed gene. That the cCOUP-TF II- positive neurons in the spinal cord are motor neurons is supported by several lines of evidence. First, the regions of cCOUP-TF II expression are coextensive with histologically identifiable domains of the spinal cord known to contain somatic and preganglionic sympathetic motor neurons (Levi-Montalcini, 1950; Prasad and Hollyday, 1991). Second, cCOUP-TF II mRNA is found in cells that express motor neuron markers previously identified, such as Islet-1 (Ericson et al., 1992) and SC-1 (Tanaka and Obata, 1984). Third, when ectopic motor neurons are induced by notochord grafting, cCOUP-TF II expression is also induced.

The biochemical function of COUP-TFs has been studied using transfection assays and electrophoretic mobility shift assays (EMSA) (Cooney et al., 1992, 1993; and references therein). Characterization of COUP-TF response elements showed that COUP-TFs interact with direct repeats of AGGTCA with variable nucleotide spacings, but preferably to DR1. The DR1 element also binds RXRs (Hamada et al., 1989; Mangelsdorf et al., 1992) and peroxisome proliferator activated receptor (Tugwood et al., 1992) with high affinity. EMSA of the DR1 element with chick spinal cord nuclear extracts revealed one major band. This band contains COUP- TF, since addition of anti-COUP-TF antibodies produced a supershift. EMSA with in vitro generated cCOUP-TF II results in a band that comigrates with that observed in spinal cord extract (Fig. 3). This argues for the possibility that COUP-TFs are the predominant proteins in the spinal cord capable of interacting with DR1 elements and that the complex formed with spinal cord nuclear extracts possibly is a homodimer of COUP- TFs because EMSA using COUP-TFs from spinal cord and in vitro synthesized cCOUP-TF II resulted in similar mobilities. One could argue that the complex formed might be a COUP- TF/RXR heterodimer. However, it has previously been shown that such a complex would have a mobility different from that of COUP-TF homodimers (Kliewer et al., 1992; Cooney et al., 1993). It remains to be determined whether other elements with different spacings and base compositions also bind COUP-TFs from spinal cord nuclear extracts and whether the corresponding complexes are homo- or heterodimers. EMSA using whole stage-25 chick embryo extracts also reveals a DR1-dependent protein-DNA complex that supershifts with anti-COUP-TF antibodies and, hence, is likely to be identical to the complex seen in the spinal cord. The identity of the other complexes seen in whole embryo extracts is presently not known, but it is conceivable that they contain other members of the steroid/thyroid receptor superfamily.

Transient transfection assays and EMSA showed that COUP-TFs can act as repressors of various steroid/thyroid hormone pathways in a dose-dependent manner (Cooney et al., 1992, 1993; and references therein). Evidence that COUP-TFs might function as a repressor in vivo comes from studies on the regulation of the oxytocin gene in ovarian granulosa cells. Wehrenberg et al. (1992) showed that COUP-TF binds to a COUP element in the oxytocin promoter under conditions where the oxytocin gene is switched off. The oxytocin promoter has been shown to be responsive to estrogen, retinoic acid and thyroid hormone, acting through their cognate receptors (Richard and Zingg, 1990, 1991; Burbach et al., 1990; Adan et al., 1992, 1993). In transient transfection experiments, COUP-TFs can repress activation of the oxytocin promoter by each of these receptors, and the repression of the estrogen receptor was shown to be due to direct competition between COUP-TF and the estrogen receptor for binding to key response elements in the oxytocin promoters (S.Y. T., unpublished results). Thus, in vivo and in vitro studies of the repressor activity of COUP-TFs suggest that high expression levels of COUP-TF and of genes regulated through DR elements are mutually exclusive. Intriguingly, our study shows that cCOUP-TF II is highly expressed in motor neurons at the time when genes are expressed that are known to be positively regulated by DR elements. These genes include RARβ (Smith and Eichele, 1991; Muto et al., 1991; Ruberte et al., 1993; de Thé et al., 1990; Sucov et al., 1990), cellular retinol-binding protein I (CRBPI) (Maden et al., 1990; Ruberte et al., 1993; Smith et al., 1991), and cellular retinoic acid-binding protein II (CRABPII) (Ruberte et al., 1993; Durand et al., 1992). The simultaneous expression of these genes and cCOUP-TF II raises the question as to whether COUP-TF is acting as a negative or positive factor in spinal motor neurons. It has been shown that repression by COUP-TFs is dependent upon the ratio of COUP-TF and the other factors such as RAR or RXR which bind to a common element. However, the level of expression of cCOUP-TF II in motor neurons might not be sufficient to significantly repress these genes (Cooney et al., 1993; Tran et al., 1992; Widom et al., 1992). Alternatively, COUP- TF II may sensitize these promoters to 9-cis retinoic acid or all-trans retinoic acid as has been proposed for the apolipoprotein AI promoter by Widom et al. (1992). Finally, cCOUP-TF II might be converted to a positive regulator upon binding to an as yet unknown COUP ligand present in spinal motor neurons.

Other transcription factors expressed in the spinal cord include Hox genes (reviewed by McGinnis and Krumlauf, 1992), Pax genes (reviewed by Gruss and Walther, 1992) and Islet-1 (Ericson et al., 1992). Similar to cCOUP-TF II, the expression patterns of Pax genes (Goulding et al., 1993) and of Islet-1 (Ericson et al., 1992) are affected by grafting of an ectopic notochord. While Hox and Pax gene products are implicated in patterning of the spinal cord (Graham et al., 1991; Gruss and Walther, 1992), the spatiotemporal expression pattern of Islet-1 argues more for a role of this protein in specifying the fate of motor neurons (Ericson et al., 1992). Additional support for this view comes from the fact that Islet-1 is related to cell fate determining homeodomain proteins encoded by the Lin-11 and Mec-3 genes of Caenorhabditis elegans (Ericson et al., 1992). The high degree of homology between vertebrate COUP-TFs and the Drosophila seven-up (svp) gene led to the suggestion that the insect and the vertebrate orthologs are functionally conserved (Mlodzik et al., 1990; Fjose et al., 1993). A well studied function of svp entails the control of cell fate of photoreceptor neurons in the developing compound eye (Mlodzik et al., 1990). By analogy to svp, cCOUP-TF II could prevent motor neurons from acquiring alternative fates. However, spinal motor neurons begin to be specified in chick by stage 10 (Yamada et al., 1993) and at that time cCOUP-TF II is not yet expressed in a localized fashion. This would argue against the possibility that cCOUP-TF II is cell fate determining factor for spinal motor neurons. More likely, as our studies suggest, cCOUP-TF II may be involved in certain aspects of the later development of already committed spinal motor neurons.

We thank Drs T. Jessell and T. Yamada for technical advice, discussion and comments on the manuscript and acknowledge Bernard Allan, Gloria Montarfur-Garcia and Maya Dajee for their excellent technical support. The antisera for SC-1 and Islet-1 were kindly provided by Drs T. Jessell and H. Tanaka. We also thank Dr Michael Figdor for valuble discussion. This study was supported by grants from the NIH (HD28999, G. E. and DK-45641, M.-J. T.) and from The McKnight Foundation (G. E.). B. L. is supported by a fellowship from the Swiss Science Foundation, and S. K. is supported by The Muscular Dystrophy Association.