Reduced Expression of the Endothelin Receptor Type B Gene in Piebald Mice Caused by Insertion of a Retroposon-like Element in Intron 1*

Mice carrying the piebald mutation exhibit white coat spotting due to the regional absence of neural crest-derived melanocytes. We reported previously that the piebald locus encodes the Ednrb gene and that piebald mice express low levels of structurally intact Ednrb mRNA and EDNRB protein (Hosoda, K., Hammer, R. E., Richardson, J. A., Baynash, A. G., Cheung, J. C., Giaid, A., and Yanagisawa, M. (1994) Cell 79, 1267–1276). Here, we report that both the life span of the Ednrb mRNA and the promoter activity of the Ednrb gene are indistinguishable between wild-type and piebald mice. Introns 2–6 of the Ednrb gene in piebald mice were correctly excised with an efficiency indistinguishable from those in wild-type mice in exon trapping experiments. We found that the piebald allele of the Ednrb gene has a 5.5-kb retroposon-like element in intron 1 possessing canonical sequences of a polyadenylation signal and a splice acceptor site. Abnormal hybrid transcripts carrying exon 1 of the Ednrb gene and a portion of the 5.5-kb element are expressed in piebald mice. The insertion of the 5.5-kb element into a heterologous intron in a mammalian expression vector markedly reduced the expression of the reporter gene. Premature termination and aberrant splicing of the Ednrb transcript caused by the retroposon-like element in intron 1 lead to a reduced level of the normal Ednrb transcript, which is responsible for the partial loss-of-function phenotype of piebald mice.

Melanocytes are specialized melanin-producing cells responsible for coat pigmentation (1). They arise from neural crest cells that leave the apical ridge of the neural tube and migrate dorsally to the somites and through the mesenchymal layer beneath the ectoderm until they eventually enter the epidermis (2,3). A number of naturally occurring as well as targeted mutations that produce developmental defects in neural crest cell migration, differentiation, or survival have been reported in mice (4 -8). These include piebald (s/s) mice, which are one of the spotting mutants that lack neural crest-derived melanocytes in the coat (5). In addition to the coat pigment defect, piebald lethal (s l /s l ) mice, which carry a severe mutation at the s locus, manifest megacolon (9,10). This defect is caused by the absence of enteric ganglia, which are also derived from the neural crest, in the distal portion of the colon (11).
A large number of studies have established that the signaling mediated by endothelins plays an essential role in the development of neural crest-derived cell lineages (12)(13)(14). Endothelin was originally identified as a potent vasoconstrictive peptide synthesized by vascular endothelial cells (15). Three closely related endothelin peptides (EDN1, EDN2, and EDN3) composed of 21 amino acids have been reported (16 -18). The two endothelin receptors EDNRA (endothelin receptor type A) and EDNRB (endothelin receptor type B) belong to the G protein-coupled heptahelical superfamily (19 -21). EDNRA preferentially interacts with EDN1 and EDN2, but not with EDN3. EDNRB accepts all three isopeptides equally. EDN3/EDNRB interaction is essential for the development of two neural crest-derived cell lineages, epidermal melanocytes and enteric neurons (22)(23)(24)(25).
We demonstrated previously that the Ednrb gene is allelic to the s locus (22). A DNA segment encompassing all of the coding exons of the Ednrb gene is deleted in the s l genome (22). On the other hand, the s allele expresses reduced levels of structurally normal Ednrb mRNA and EDNRB protein (22). The primary genomic lesion in the s mutation has not been described previously.
Here, we report that the Ednrb s allele harbors a 5.5-kb retroposonlike element in intron 1. This leads to aberrant transcription of the Ednrb gene, resulting in a reduced level of the normal Ednrb transcript, which is responsible for the white coat spotting in s/s mice.

EXPERIMENTAL PROCEDURES
Mice-Wild-type mice (C57BL/6J) and piebald mice (s/s SSL/Le) were obtained from The Jackson Laboratory.
Isolation and Characterization of the Ednrb Genomic Clones-Genomic DNA prepared from the livers of s/s mice was partially digested with Sau3AI, and the first two nucleotides of the ends were filled in with dATP and dGTP. The DNA was ligated to XhoI-digested FIXII arms (Stratagene), the ends of which were filled in with dCTP and dTTP. Recombinant DNA was packaged in vitro, and phages were plated on Escherichia coli XL1-Blue MRA(P2) (Stratagene). A FIXII mouse 129/Sv genomic library was purchased from Stratagene. Approximately 2 ϫ 10 6 phage clones were screened with either the full-length sequence (positions 1-1958) of mouse Ednrb cDNA (GenBank TM accession number U32329) or a 2.5-kb PstI fragment sequence of mouse Ednrb intron 1 (see Fig. 3A). The DNA probe was 32 P-labeled using a random primer labeling kit (Roche Applied Science). Plaque hybridization and preparation of recombinant phage DNAs were carried out by standard procedures (26). Positive clones were purified and characterized by restriction endonuclease mapping and Southern blot analysis. DNA sequencing of specific restriction fragments subcloned into the pBluescript II vector (Stratagene) was performed by the dideoxy chain termination method using Sequenase Version 2.0 (U. S. Biochemical Corp.). In addition to M13 universal and reversal primers (U. S. Biochemical Corp.), sequence-specific oligonucleotides were used for sequencing. PCR was performed to determine the sizes of introns. The reaction was cycled 30 times in a cycle profile of 1 min at 94°C, 2 min at 55°C, and 3 min at 72°C. Amplified DNA fragments were analyzed by agarose gel electrophoresis.
Construction of Plasmids-To construct a reporter plasmid for the measurement of the promoter activity of the Ednrb gene, a 5-kb BamHI-BssHII fragment starting in the 5Ј-flanking region and ending in the 5Ј-noncoding region of exon 1 (171 bp upstream of the initiation codon of the Ednrb gene, at the BssHII site) was linked to the firefly luciferase gene (p5kb-luc). In addition, an 11-kb SphI-BssHII promoter fragment was linked to the luciferase gene (p11kb-luc). The sequence of the luciferase gene was obtained from the reporter plasmid pTK-luc (27). The pTK-luc plasmid was digested with BamHI and BglII to remove the herpes simplex virus thymidine kinase promoter and blunt end-ligated to a 1-kb SphI-BssHII genomic fragment from ϳ1.3 kb to 171 bp upstream of the initiation codon (p1kb-luc). To prepare p5kb-luc, the p1kb-luc plasmid was digested with BamHI, and the resulting larger fragment containing a 1-kb BamHI-BssHII fragment from ϳ1.1 kb to 171 bp upstream of the initiation codon was ligated to a 4-kb BamHI genomic fragment from ϳ5.1 to ϳ1.1 kb upstream of the initiation codon. To prepare p11kb-luc, a 10-kb SphI genomic fragment from ϳ11.3 to ϳ1.3 kb upstream of the initiation codon was blunt end-inserted into the unique SalI site of the p1kb-luc plasmid. An Ednrb exontrapping plasmid harboring sequences of introns 2-4 or introns 5 and 6 of the Ednrb gene was prepared as follows. A 3-kb PvuII Ednrb genomic fragment starting in intron 1 (ϳ0.1 kb upstream of exon 2) and extending to intron 5 (53 bp downstream of exon 5) was blunt end-inserted into the unique NotI site in the intron of the exon-trapping vector pSPL3 (introns 2-4) (Invitrogen). In addition, a 4-kb EcoRI-SacI Ednrb genomic fragment starting in intron 4 (ϳ0.4 kb upstream of exon 5) and extending to 401 bp in exon 7 was blunt end-ligated to the 4-kb AvaI fragment of pSPL3, in which 2077 bp of the intron 1 and 13 bp of exon 2 are eliminated (introns 5 and 6). The plasmid containing introns 5 and 6 includes the chimeric exon of the 5Ј-region of Ednrb exon 7 and the 3Ј-region of pSPL3 exon 2. To construct a reporter plasmid harboring the 5.5-kb retroposon-like element inserted heterologously into a vector intron (see Fig. 6A), the 2.5-kb XbaI-MscI fragment of the pSPL3 plasmid, which contains an intron, splice donor and acceptor sites, and some flanking exon sequences of the human immunodeficiency virus type 1 tat gene, was blunt end-inserted into the unique BglII site between the promoter and reporter regions in the pTK-luc plasmid (pTK-tat-luc). The 6.5-kb EcoRV-SalI fragment carrying part of intron 1 from the Ednrb gene and the 5.5-kb retroposon-like element was then blunt end-inserted into the unique NotI site of the tat intron in pTKtat-luc in the right orientation (pTK-tat␣-luc) or the reverse orientation (pTK-tat␤-luc).
Transfection-For transfection and the measurement of promoter activity, ROS17/2 or C6 cells were seeded into 6-well culture plates and cotransfected with either a control plasmid (pUC18) or a luciferase reporter plasmid (pTK-luc containing the herpes simplex virus thymi-dine kinase promoter, p5kb-luc, or p11kb-luc) at 1 g/well and with a lacZ expression plasmid at 1 g/well using Lipofectamine and serumfree Opti-MEM I (Invitrogen) following the protocols of the manufacturer. Optimal transfection conditions were 4 l/well Lipofectamine and a 5-h incubation under serum-free conditions. Forty-eight hours after transfection, the cells were used for the luciferase assay. For transfection in exon trapping experiments, the Ednrb exon-trapping plasmid (50 ng/well) was introduced into COS-7 cells in 6-well culture plates alone or in combination with the 64-bp control exon-containing pSPL3 plasmid (50 ng/well; Invitrogen) using 4 l/well Lipofectamine and a 12-h incubation under serum-free conditions. After 48 h, the cells were used for RNA isolation. For transfection of the reporter plasmid harboring the 5.5-kb retroposon-like element inserted heterologously into a vector intron, COS-7 cells in 6-well culture plates were cotransfected with either pUC18 or a luciferase reporter plasmid at 1 g/well and with the lacZ expression plasmid at 1 g/well by the DEAE-dextran method as described (21). After 48 h, the cells were used for the luciferase assay.
Luciferase Assay-Transfectants were harvested after 48 h, and extracts were prepared and assayed for enzyme activity as described previously (28,29). Luciferase and ␤-galactosidase activities were measured using a Dynatech microtiter plate Model ML3000 luminometer and Model MR5000 spectrophotometer, respectively. Luciferase values were normalized for transfection and harvesting efficiency by measuring ␤-galactosidase activity, and the results are reported as relative light units.
Preparation and Culture of Mouse Cerebral Astrocytes-Astrocytes were obtained from newborn (first day of life) mouse cerebrum using procedures slightly modified from those of Lim and Bosch (30) and Hertz et al. (31). The cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal calf serum in a humidified atmosphere of 5% CO 2 at 37°C. They were used for experiments after the third passage (21-30 days old). Astrocytes were then subcultured in 6-well plates for 3 days and allowed to grow into subconfluent monolayers. Forty-eight hours prior to the addition of an RNA synthesis inhibitor, astrocytes were changed to serum-free medium. The cells were treated with actinomycin D (10 g/ml) or 5,6-dichlorobenzimidazole riboside (60 M); harvested 2, 4, 8, and 16 h after the inhibitor addition; and used for RNA isolation.
Northern Blot Analysis and Reverse Transcription (RT) 3 -PCR-RNA was extracted from frozen tissues of 8-week-old mice or from cultured cells using RNA STAT-60 reagent (Tel-Test, Inc., Friendswood, TX). For Northern analysis, total RNA (10 g) was separated on 1.1% agarose gel, transferred to a Hybond-N ϩ membrane (Amersham Biosciences), and hybridized with random-primed 32 P-labeled probes. The mouse Ednrb cDNA sequence and the 5.5-kb retroposon-like element sequence were used as probes. Some probes were obtained by PCR. For RT-PCR, firststrand cDNA was synthesized from 2 g of total RNA with (dT) [12][13][14][15][16][17][18] primers using SuperScript reverse transcriptase (Invitrogen) as recommended by the manufacturer. The synthesized cDNA (0.5 g) was amplified using buffer A included in the PCR optimization kit from Invitrogen and appropriate primers for 35 cycles in a cycle profile of 1 min at 94°C, 1 min at 55°C, and 15 min at 72°C. The amplification products were subcloned into the pCRII cloning vector (Invitrogen) and sequenced using Sequenase Version 2.0. In exon trapping experiments, cDNA synthesis and primary and secondary PCR amplification were performed using an exon trapping system (Invitrogen) as recommended by the manufacturer. The amplification products were subcloned into the pAMP10 cloning vector (Invitrogen) and sequenced using Sequenase Version 2.0. The 3Ј-rapid amplification of cDNA ends (3Ј-RACE) was carried out according to the 3Ј-RACE system of Invitrogen. The amplification products were subcloned into the pAMP1 cloning vector (Invitrogen) and sequenced using Sequenase Version 2.0.

RESULTS
We have demonstrated previously that the Ednrb mRNA is structurally intact in terms of the overall length and coding region sequence in s/s mice, but that the expression of Ednrb mRNA in the lung is decreased to ϳ25% of the wild-type levels (22). Northern blot analysis of hearts, livers, kidneys, brains, and intestines revealed that the expression of intact Ednrb mRNA is uniformly reduced in all s/s tissues examined except for the liver (Fig. 1). These observations are consistent with our previous finding that s/s mice show ϳ27% of the EDNRB density in the kidney compared with wild-type mice as determined by radioligand binding assay (22).
First, we examined whether the decreased Ednrb mRNA level was caused by a shortened life span of the Ednrb s mRNA by monitoring the pattern of disappearance of the Ednrb mRNA after the addition of the RNA synthesis inhibitor actinomycin D or 5,6-dichlorobenzimidazole riboside to the culture of mouse cerebral astrocytes. Like s/s tissues, astrocytes from s/s mice exhibited a low level of the Ednrb mRNA with normal length. However, no appreciable difference between wild-type and s/s mice was observed in the pattern of the Ednrb mRNA disappearance from 0 to 16 h after the addition of actinomycin D or 5,6-dichlorobenzimidazole riboside, indicating that the life span of the Ednrb s mRNA is normal (data not shown). This excluded the possibility that a less stable mRNA accounts for the reduced steady-state expression of the Ednrb s mRNA.
Second, we examined the possibility of decreased promoter activity of the Ednrb s gene by constructing reporter plasmids in which the expression of the firefly luciferase gene was driven by Ednrb promoter fragments from s/s and wild-type (129/Sv) mice. We isolated phage clones containing DNA segments that included the 5Ј-flanking region and exon 1 of the Ednrb gene from both wild-type and s/s mice as demonstrated by restriction mapping, oligonucleotide hybridization, and partial sequencing. SphI, BamHI, and BssHII restriction sites were conserved between the Ednrb promoter regions of wild-type and s/s mice (in the 5Ј-flanking region, ϳ11.3 and ϳ1.3 kb upstream of the initiation codon for SphI; in the 5Ј-flanking region, ϳ5.1 and ϳ1.1 kb upstream of the initiation codon for BamHI; and in the 5Ј-noncoding region of exon 1, 171 bp upstream of the initiation codon for BssHII in the Ednrb gene of both mice). A 5-kb BamHI-BssHII fragment from ϳ5.1 kb to 171 bp upstream of the initiation codon or an 11-kb SphI-BssHII fragment from ϳ11.3 kb to 171 bp upstream of the initiation codon in both mice was used as the promoter in luciferase reporter plasmids (designated as p5kb-luc and p11kb-luc, respectively), and the reporter plasmids were introduced into C6 glioma and ROS17/2 osteosarcoma cells, which express endogenous EDNRB, for the luciferase assay. The luciferase expression directed by the 5-kb Ednrb fragment in both ROS17/2 and C6 cells was indistinguishable between wild-type and s/s mice (data not shown). The 11-kb Ednrb fragment-directed luciferase expression in both cells was also indistinguishable between the two genotypes (data not shown). Mock-transfected cells expressed luciferase levels that were not significantly different from the background of the luminometer. These results indicate that promoter activity in the 11-kb 5Ј-region of the Ednrb s gene is normal. Moreover, the nucleotide sequences of a 453-bp region upstream of the initiation codon of the Ednrb gene (containing the 221-bp 5Ј-noncoding region of the mouse Ednrb mRNA reference sequence (GenBank TM accession number NM_007904)) are identical in wild-type and s/s mice, suggesting that exon 1 and the proximal promoter of Ednrb s are structurally intact (data not shown). This excluded the possibility that an abnormal cis-element in the 11-kb 5Ј-region of the Ednrb s allele contributes to the reduced mRNA expression.
Third, we examined the intron structure of the Ednrb s gene. Eleven and 12 positive clones that covered all introns of the Ednrb gene as demonstrated by restriction mapping, Southern hybridization, and partial sequencing were isolated from genomic DNA libraries of wild-type (129/ Sv) and s/s mice, respectively. Of the 11 and 12 clones each, three overlapping clones were used for a detailed characterization of the two mouse genomic sequences. The nucleotide sequences of the exon-intron junctions of the Ednrb gene from both mice were determined (data not shown). The sizes of exons 2-6 in the Ednrb gene of both mice were 113, 205, 150, 134, and 109 bp, respectively, and are in complete concordance with those of the reported mouse Ednrb genomic structure (GenBank TM accession number NT_039609). The sizes of introns 1-6 (except for Ednrb s intron 1) in both mice were ϳ20.0, 0.2, 1.5, 0.6, 0.6, The pUC18 plasmid was used as a negative (neg.) control plasmid. RNA was prepared from the transfected cells, and single-stranded cDNA was synthesized using the prepared RNA. Primary and secondary PCR amplifications of the synthesized single-stranded cDNA were performed using the Invitrogen exon trapping system. The primer sets for both primary and secondary PCRs were designed to anneal to the first and second exons on the vector. The products were analyzed on 2.0% agarose gel and visualized by ethidium bromide staining. The lengths of the DNAs are shown. FIGURE 1. Northern blot analysis using an Ednrb cDNA probe in various tissues of wild-type (؉/؉) and piebald (s/s) mice. Total cellular RNAs were prepared from the lungs, hearts, livers, kidneys, brains, and intestines of both mice and electrophoresed on 1.1% agarose gel. After being transferred to nylon membranes, the blot was hybridized with a 32 P-labeled mouse Ednrb cDNA probe. We used DNA containing exons 1-6 (positions 48 -1370) of the mouse Ednrb cDNA (GenBank TM accession number U32329) as a probe. The DNA probe was prepared by PCR. Rehybridization with a ␤-actin probe is shown as an internal loading standard.  (GenBank TM accession number  NT_039609). Surprisingly, the size of intron 1 in s/s mice was 5.5 kb longer than that in wild-type mice (see below). All of the 5Ј-and 3Ј-splice sites in Ednrb introns 1-6 of both mice conformed to the GT-AG rule (32); and in addition, adenosine residues in presumed branch sites within introns 1-6 were conserved between the two genotypes (data not shown) (33).
Subsequently, to investigate the precision and efficiency of excision of Ednrb introns 2-6 during RNA splicing, we performed exon trapping experiments in which the RT-PCR-amplified product from COS-7 cells transfected with a genomic insert-containing exon-trapping vector represented a chimeric transcript produced by the pairing of vector (pSPL3) and genomic (Ednrb) exons (34). The Ednrb exon-trapping plasmid harboring the sequences of introns 2-4 or introns 5 and 6 of the Ednrb gene from wild-type and s/s mice was introduced into COS-7 cells alone or together with the 64-bp control exon-containing pSPL3 plasmid as a positive control. We detected RT-PCR-amplified products of the expected length (779 and 808 bp, respectively, when introns 2-4 and introns 5 and 6 are excised) in cells transfected with the wild-type and s/s Ednrb exon-trapping plasmids (Fig. 2). The ratio of the transcripts produced by the Ednrb exon-trapping plasmid to those produced by the cotransfected positive control plasmid (244 bp) was not significantly different between the two genotypes (Fig. 2). We detected no PCR product in mock-transfected cells. These findings indicate that introns 2-6 of the Ednrb gene in s/s mice are correctly excised with an efficiency indistinguishable from those in wild-type mice during splicing. Together with the results of the structural analysis of Ednrb s introns 2-6, the data suggest that the sequences of Ednrb s introns 2-6 do not have any detrimental effects on gene expression.
Finally, we focused on an abnormality in intron 1 of the Ednrb s gene. Comparison of intron 1 in wild-type and s/s mice indicated that a 5.5-kb DNA element is inserted ϳ15 kb downstream of exon 1 and ϳ5 kb upstream of exon 2 in the Ednrb s gene (Fig. 3A). Nucleotide sequences of the insert and its flanking regions were determined for the wild-type and s/s DNAs (Fig. 3B). The entire length of the insert DNA is 5056 bp, and the nucleotide sequence data have been submitted to the GenBank TM /EBI Data Bank with accession number AB242436. The first 507 bp at the 5Ј-and 3Ј-ends of the insert DNA have the same sequences and are highly homologous to the 526-bp direct terminal repeats in insertion sequences of the nonagouti (a) and the black-and-tan (a t ) alleles of the agouti locus (35). The 507-bp direct repeat carries sequences of a canonical AATAAA polyadenylation signal (36) and of a T cluster (TTTCTTTT) 49 nucleotides downstream of the signal (36). The T cluster is conserved in the areas for transcription termination and 3Ј-processing and usually lies between 5 and 30 nucleotides downstream of the polyadenylation site (36). Moreover, canonical sequences of a splice acceptor site (37) and of a branch site 54 nucleotides upstream of the acceptor site (33) are present. A search for the core sequence of the 5.5-kb element in the data base revealed that a 460-bp 3Ј-region (positions 3987-4446), a 273-bp internal region (positions 3098 -3370), and a 164-bp internal region (positions 2177-2340) in the core sequence of the 5.5-kb element (GenBank TM accession number AB242436) are highly homologous to positions 3187-3647 (Gen-Bank TM accession number AF030884) of the mouse early transposon (ETn) element near the Pafaha-ps2 gene (38), positions 345-616 (Gen-Bank TM accession number X57268) of the mouse ETn-related t haplotype-specific element within the H-2 complex (39), and a short stretch of a retroviral pol gene (positions 3696 -3860 of GenBank TM accession number AF246633) in the mouse ETn-related MusD element (40), respectively, showing only partial sequence similarity. At the 5Ј-and 3Ј-ends of the insert DNA, there is a 6-bp repeat sequence (AGAAAC) that is found once in the wild-type DNA. These findings suggest that the 5.5-kb DNA insert in intron 1 of the Ednrb s gene is an as yet uncharacterized retroposon-like element.
We detected decreased levels of intact Ednrb mRNA in s/s mice (ϳ25% of the levels in wild-type mice) by Northern blot hybridization using exons 1-6 of the Ednrb cDNA as a probe. An additional RNA larger than the intact RNA seemed to be detected in s/s mice by this Northern blot hybridization (Fig. 1). In addition, the data suggested that the promoter and exon 1 of the Ednrb s gene are intact, but that intron 1 has the 5.5-kb retroposon-like element carrying canonical sequences of a polyadenylation signal and a splice acceptor site. Thus, the expression of the Ednrb gene was re-examined using DNA probes containing exon 1 or exons 2-6 of the Ednrb cDNA. As shown in Fig. 4A, Northern blot hybridization using an exon 1 probe detected a 6.5-kb RNA (ϳ75% of the amount of the wild-type Ednrb mRNA) as well as an RNA of normal size (4.4 kb; ϳ25% of the amount of the wild-type Ednrb mRNA) in s/s mice. When a DNA probe containing exons 2-6 was used, the 4.4-kb RNA (ϳ25% of the amount of the wild-type Ednrb mRNA), but not the 6.5-kb species, was detected in s/s mice (Fig. 4B). In wild-type mice, both probes detected only the 4.4-kb RNA. These findings suggest that the Ednrb s mRNA is initiated normally in s/s mice, but that ϳ75% of the initiated mRNA is terminated before exon 2. Moreover, a DNA segment indicated in Fig. 3B in the retroposon-like element hybridized with the 6.5-kb RNA in s/s mice, but not in wild-type mice (Fig. 4C).
Therefore, we examined the possibility that the 6.5-kb RNA expressed in s/s mice is a hybrid transcript carrying Ednrb exon 1 and the retroposonlike element sequences. Sequence analysis of segments trapped as a exon in the exon-trapping vector pSPL3, into which the retroposon-like elementcontaining fragment was inserted, revealed that a splice acceptor site in the 5Ј-direct repeat was used for the pairing with the splice donor site of the vector (data not shown). RNAs from wild-type and s/s mice were then analyzed by RT-PCR. RNA from the s/s mice (but not from the wildtype mice) produced DNA fragments (220 or 1720 bp) that were derived from chimeric RNA resulting from the pairing of the splice donor site in Ednrb exon 1 and the splice acceptor site in the 5Ј-direct repeat of the retroposon-like element (Fig. 5A). Furthermore, 3Ј-RACE resulted in amplification of a fragment (553 bp) that was derived from RNA terminated at a polyadenylation site 28 nucleotides downstream of the polyadenylation signal in the 3Ј-direct repeat of the retroposon-like element in s/s mice, but not in wild-type mice (Fig. 5B). The larger fragment detected in both genotypes resulted from a nonspecific amplification as demonstrated by sequencing. These results indicate that there is a hybrid transcript carrying ϳ1-kb Ednrb exon 1 and the ϳ5.5-kb retroposon-like element portion in s/s mice, which is paired at a splice acceptor site in the 5Ј-direct repeat and ends at a polyadenylation site in the 3Ј-direct repeat. It is likely that ϳ75% of the transcripts normally initi- ating at exon 1 of the Ednrb gene in s/s mice are terminated in the 3Ј-direct repeat region (at the polyadenylation site) of the retroposonlike element inserted into intron 1. The prematurely terminated RNA is then spliced aberrantly at the splice donor and acceptor sites of Ednrb exon 1 and the retroposon-like element, respectively, to produce the 6.5-kb mRNA.
We examined the effect of insertion of the 5.5-kb retroposon-like element on the expression of heterologous genes. The pTK-tat-luc plasmid is a luciferase reporter construct that uses the firefly luciferase gene as a reporter and that carries an intron, splice donor and acceptor sites, and some flanking exon sequences of the human immunodeficiency virus type 1 tat gene from the pSPL3 plasmid. The 6.5-kb EcoRV-SalI fragment carrying the entire 5.5-kb retroposon-like element sequence was inserted into the unique NotI site of the intron in pTK-tat-luc in the right orientation (pTK-tat␣-luc) or the reverse orientation (pTK-tat␤luc) (Fig. 6A). The resulting plasmid DNAs were introduced into COS-7 cells, and the luciferase activities were assayed in the transfectants. As shown in Fig. 6B, the pTK-tat-luc plasmid produced luciferase at 1.63 Ϯ 0.52 relative light units. The 5.5-kb retroposon-like element inserted in a reverse orientation (pTK-tat␤-luc) had almost no effect on the production of luciferase. On the other hand, the pTK-tat␣-luc plasmid produced luciferase at only 0.45 Ϯ 0.13 relative light units, which was ϳ25% of that produced by pTK-tat-luc. This is consistent with the steady-state level of structurally intact Ednrb mRNA in s/s mice. Mock-transfected cells expressed luciferase levels that were not significantly different from the background of the luminometer. These findings indicate that the 5.5-kb retroposon-like element has the ability to reduce gene expression if it is inserted into the intron in the right orientation.

DISCUSSION
In this study, we have shown that the Ednrb gene in s/s mice carries an insertion of a 5.5-kb retroposon-like element that includes canonical sequences of a polyadenylation signal and a splice acceptor site. This insertion causes premature termination and aberrant splicing in ϳ75% of the normally initiated Ednrb mRNA in s/s mice, resulting in reduced levels of the structurally intact Ednrb transcript (ϳ25% of the wild-type levels) (Fig. 7). The levels of the intact Ednrb mRNA seen in s/s mice are compatible with both the reduced EDNRB density to ϳ27% of the wildtype levels and the mild anomaly of white coat spotting in ϳ20% of the body surface area in s/s mice (22). Furthermore, this is consistent with our finding that insertion of the 5.5-kb retroposon-like element into a heterologous intron in a reporter plasmid reduces luciferase expression to ϳ25% of the levels seen in the reporter plasmid without the insertion. The premature termination occurs at the polyadenylation site in the 3Ј-direct repeat of the retroposon-like element, and aberrant splicing occurs at the splice acceptor site in the 5Ј-direct repeat of the retroposon-like element (Fig. 7). Thus, as shown in Fig. 6, if the element is inserted into the intron in the reverse orientation, which destroys both sites, the retroposon-like element is unable to reduce expression of a heterologous gene.
The 5.5-kb retroposon-like element has a 507-bp direct repeat at both the 5Ј-and 3Ј-ends, which is highly homologous to the 526-bp direct terminal repeats in insertion sequences of the nonagouti (a) and blackand-tan (a t ) alleles of the agouti locus (35). It is regrettable that the 526-bp direct repeats have not been investigated in detail, except for the proposal that they are involved in homologous recombination for a-to-a t reverse mutation (35). The retroposon-like element also shows sequence similarity to ETn (38) or its relatives, the t haplotype-specific element (39) and the MusD element (40) both of which share homology with ETn in a 460-bp 3Ј-region and in 273-and 164-bp internal regions of the core sequence, respectively. The ETns are a family of murine retrotransposon-like elements possessing long terminal repeats, and ϳ1000 copies are present in the mouse genome (41). These elements are abundantly transcribed in early mouse embryogenesis, although they Total cellular RNAs were prepared from the lungs, brains, intestines, and hearts of both wildtype (ϩ/ϩ) and piebald (s/s) mice and electrophoresed on 1.1% agarose gel. After transfer to nylon membranes, the blot was hybridized with a 32 P-labeled probe. We used DNA containing exon 1 (A) or exons 2-6 (B) or the DNA segment indicated in Fig. 3B in the retroposon-like element (C) as a probe. The DNA probe were prepared by PCR. The DNA probe for exon 1 or exons 2-6 carries sequence 48 -660 or 711-1370 of the mouse Ednrb cDNA (GenBank TM accession number U32329), respectively. Rehybridization with a ␤-actin probe is shown as an internal loading standard. In RT-PCR analysis (A), single-stranded cDNA was synthesized using total cellular RNAs from the lungs and brains of both wild-type (ϩ/ϩ) and piebald (s/s) mice and (dT) [12][13][14][15][16][17][18] primers, and the cDNA containing the sequence of the aberrant Ednrb s transcript was amplified using combinations of primers MEB1 and MEBI9 or MEBI12 (indicated in Fig. 3B). In 3Ј-RACE analysis (B), single-stranded cDNA was synthesized using total cellular RNAs from the lungs and brains of both mice and adapter primers, and the 3Ј-end of the aberrant Ednrb s transcript was amplified using primer MEBI18 (indicated in Fig. 3B). The products were analyzed on 2.0% agarose gel and visualized by ethidium bromide staining. The lengths of the DNAs are shown.
contain mainly non-retroviral noncoding sequences of unknown origin (41), except for a short stretch of a retroviral pol gene (40). The significance of these partial sequence similarities among the 5.5-kb retroposon-like element, ETn, and its relatives is currently unclear, whereas the MusD elements, classified as long terminal repeat-containing and retrotransposon-like and as possessing gag, pro, and pol genes homologous to those in type D viruses or betaretroviruses (40), are suggested to provide the proteins necessary for ETn retrotransposition in trans (42). However, it is FIGURE 6. Effect of the retroposon-like element on expression of a heterologous gene. A, construction of the reporter plasmid carrying the 5.5-kb retroposon-like element. The luciferase (LUC) reporter construct pTK-tat-luc and its derivatives carrying the 5.5-kb DNA element in the right orientation (pTK-tat␣-luc) and in the reverse orientation (pTK-tat␤-luc) are shown schematically. Plasmids pTK-tat-luc, pTK-tat␣-luc, and pTK-tat␤luc contain the herpes simplex virus thymidine kinase (TK) promoter. HIV, human immunodeficiency virus. B, luciferase activity in the transfected COS-7 cells with the luciferase reporter plasmid. COS-7 cells were cotransfected with either the control plasmid (pUC18) or luciferase reporter plasmid (pTK-tat-luc, pTK-tat␣-luc, or pTK-tat␤luc) as indicated and with a lacZ expression plasmid. Cell lysates were prepared and then assayed for luciferase activity, which is expressed as relative light units (RLU) and represents the mean Ϯ S.E. of triplicate assays normalized to ␤-galactosidase activity as an internal control. interesting to note that ETn insertions occur in at least 19 mouse loci (43), such as the T locus (44), the Fas locus (45,46), and the Adcy1 locus (41), producing loss-of-function mutations in most of these cases.
Here, a search of the 507-bp direct repeat for canonical consensus sequences revealed that, in addition to sequences of a canonical polyadenylation signal, a T cluster 49 nucleotides downstream of the signal, a canonical splice acceptor site, and a canonical branch site 54 nucleotides upstream of the acceptor site, there exist a TATA box-like sequence (TATAATT) (47) 68 nucleotides upstream of the polyadenylation signal and a consensus sequence for AP-2 binding (CCCCCAAGGC) (48) 181 nucleotides upstream of the polyadenylation signal. There is no conventional CCAAT box (49) in the direct repeat. The location of the TATA box-like promoter element, AP-2-binding site, and polyadenylation signal in the direct repeat suggests that the retroposon-like element is transcribed through the direct repeat and that the resulting transcript has a terminal redundancy as in retrotransposon or retrovirus possessing retroviral long terminal repeats (50). Furthermore, there is a duplication of six nucleotides of host DNA at the insertion site, which has been recognized in the integration process of retrotransposon or retrovirus (50). In addition, the retroposon-like element reveals no long open reading frames and no significant homology to known retroviral genes, except for a short stretch of a retroviral pol gene, but contains a 17-bp sequence (TGGCGCCCAACGTGAGG) with 16 of 17 matches with the 3Ј-end of Lys-tRNA (GenBank TM accession number K00289), which presumably serves as the primer-binding site (50), just downstream of the 5Ј-direct repeat, and a 12-bp polypurine tract, the priming site for retroviral DNA (ϩ)-strand synthesis (50), just upstream of the 3Ј-direct repeat. There is also a potential stem-loop structure with an ACC motif within the loop, which may be involved in genomic packaging (51), in a 43-bp 5Ј-region (positions 561-603 of GenBank TM accession number AB242436) of the core sequence of the retroposon-like element. These findings suggest that the insertion of the retroposon-like element into the Ednrb gene occurred through the mechanism of retrotransposition, whereas a full-size transcript, which is transcribed through the 5Ј-and 3Ј-direct repeats and encodes the whole stretch of the core sequence of the retroposon-like element, could not be detected in the data base. Retrotransposition results in moderate reiteration of retrotransposon in host genomes. Indeed, Southern blot hybridization with the retroposon-like element as a probe revealed that the retroposon-like element is repetitive in the mouse genome, as are retrotransposon and retrovirus sequences (data not shown). A search of the mouse genome showed that 16 or 24 copies of a stretch of the sequences, showing a match over almost the entire or Ͼ50% of the albeit partial region of the retroposon-like element, are dispersed throughout the genome. Thus, the repetition of the retroposon-like element in the genome supports the idea that the retroposon-like element was inserted into the Ednrb gene through the mechanism of retrotransposition.
We have demonstrated that a hybrid transcript carrying ϳ1-kb Ednrb exon 1 and the ϳ5.5-kb retroposon-like element portion is expressed in s/s mice. It ends at the polyadenylation site in the 3Ј-direct repeat and is spliced at the splice acceptor site in the 5Ј-direct repeat. The first inframe stop codon is 28 bp downstream of exon 1 in the hybrid transcript; and as a result, the hybrid transcript encodes a putative truncated protein with the 281 residues encoded by exons 2-7 replaced with an arbitrary peptide of nine residues: VCGPPRTHE. This truncation results in a deletion of the first extracellular loop through the cytoplasmic C-terminal tail of EDNRB and then abrogates functional expression of EDNRB. It has already been shown that deletion of the first two transmembrane domains of EDNRB abrogates its functional expression in spotting lethal rats (52). It may be presumed that a fraction of the transcription of the Ednrb mRNA in s/s mice is terminated at the putative polyadenylation site in the 5Ј-direct repeat because the 5Ј-direct repeat also harbors both a potential polyadenylation signal and a T cluster, as does the 3Ј-direct repeat in the 5.5-kb retroposon-like element. However, 3Ј-RACE using a primer in intron 1 of the Ednrb gene flanking the 5.5-kb element in the 5Ј-region and an adapter primer that targets the poly(A) tail failed to amplify any sequence in s/s mice (data not shown). Furthermore, Northern analysis could not detect the band (ϳ16 kb) expected when the Ednrb s mRNA transcription ends in the 5Ј-direct repeat. We therefore suggest that the Ednrb s RNA transcription is not terminated at the putative polyadenylation site in the 5Ј-direct repeat. About 25% of the initiated Ednrb mRNA in s/s mice escapes termination in the 3Ј-direct repeat and is terminated at the same polyadenylation site as wild-type mRNA. Provided that the normally terminated mRNA is aberrantly spliced at the splice donor and acceptor sites of Ednrb exon 1 and the 5Ј-or 3Ј-direct repeat of the retroposon-like element, respectively, the resulting transcript should encode the same truncated protein as described above. We could not obtain any evidence of the existence of the normally terminated but aberrantly spliced Ednrb mRNA in s/s mice. Almost all of the normally terminated Ednrb mRNA in s/s mice is of normal size and at steady-state levels consistent with the EDNRB density levels in s/s mice. In other words, almost all of the normally terminated Ednrb s mRNA is normally spliced and then encodes structurally intact protein. Therefore, although the splice acceptor site in the retroposon-like element is activated to produce a 6.5-kb hybrid transcript in the absence of the authentic acceptor site of exon 2 in the prematurely terminated Ednrb s mRNA, it will not be activated in the presence of the authentic acceptor site.
We reported previously that s l /s l , s/s l , s/s, and s l /ϩ mice, in which the relative ratio of Ednrb expression is about 0:12.5:25:50%, show a "graded" coat color phenotype in the extent of white spotting, having a white coat in Ͼ95, 40 -50, ϳ20, and ϳ0% of the body surface area, respectively (22). This indicates that the extent of white spotting is precisely dependent on the dosage of Ednrb expression. In contrast, the megacolon phenotype occurs in the s l /s l mice, but almost never in the s/s l , s/s, or s l /ϩ mice (6,9). This is compatible with the idea that the two neural crest-derived cell lineages require different minimal threshold levels of Ednrb expression (22). However, with the s allele, Ednrb expression might be more severely impaired in the melanocyte lineage in a cell type-selective manner. Our present data demonstrate that the expression of structurally intact Ednrb mRNA is uniformly reduced in most of the s/s tissues, including the intestine. Furthermore, not only Ednrb gene expression, but also heterologous gene expression is reduced because of the insertion of the 5.5-kb retroposon-like element into the intron of the gene, reflecting the generalized ability of the 5.5-kb inserted element to reduce gene expression. These findings argue against the possibility that the Ednrb s mRNA is expressed in a cell typeselective manner. Thus, it is likely that ϳ12.5% of the wild-type levels of EDNRB density is sufficient for the normal development of myenteric ganglion neurons, whereas 25-50% of the wild-type levels is required for complete development of epidermal melanocytes. We could not detect reduced expression of structurally intact Ednrb mRNA in s/s liver. The premature termination and aberrant splicing of the Ednrb transcript are likely to occur in s/s liver because the 6.5-kb RNA seems to be detected. The insertion of the 5.5-kb retroposon-like element may enhance liverspecific Ednrb transcription.