Rescue From Oculocutaneous Albinism Type 4 Using Medaka slc45a2 cDNA Driven by Its Own Promoter

Patients and vertebrate mutants with oculocutaneous albinism type 4 (OCA4) have mutations in the solute carrier family 45 member 2 (slc45a2) gene. However, there is no empirical evidence for this gene–phenotype relationship. There is a unique OCA4 mutant in medaka (b) that exhibits albinism only in the skin, but the mechanism underlying this phenotype is also unknown. In this study, we rescued medaka OCA4 phenotypes, in both the eyes and the skin, by micro-injection of an slc45a2-containing genomic fragment or slc45a2 cDNA driven by its own 0.9-kb promoter. We also identified a spontaneous nucleotide change of 339 bp in the promoter as the b mutation. There are multiple transcription start sites in medaka slc45a2, as in its human ortholog, and only the shortest and eye-specific mRNA is transcribed with the b mutation. Interestingly, we further revealed a conserved pyrimidine (Py)-rich sequence of ∼10 bp in the promoter by medaka–pufferfish comparative genomics and verified that it plays an indispensable role for expression of slc45a2 in the skin. Further studies of the 0.9-kb promoter identified in this study should provide insights into the cis/trans-regulatory mechanisms underlying the ocular and cutaneous expression of slc45a2.

T HE colors and patterns on animal body surfaces are often important for visual communication in the wild and are determined primarily by pigment cells (chromatophores) in vertebrates. The chromatophores are distributed in the skin, and their types, sizes, densities, and physiological activities affect these colors and patterns. Although mouse mutants have contributed greatly to our knowledge of skin-and coat-color formation (see Coat Color Genes, http://www.espcr.org/ micemut/), mammals possess only one type of chromatophore, the melanocyte. In fish, up to six chromatophore types (melano-, leuco-, erythro-, xantho-, irido-, and cyanophores) have been identified, and there are two distinctive model species to which molecular genetics can be feasibly applied, the zebrafish and the medaka. Chromatophore studies in these species have successfully provided novel clues to the development, regulation, and interaction of these chromatophores (e.g., Parichy et al. 2000;Fukamachi et al. 2004a;Watanabe et al. 2006).
We have previously reported that the gene slc45a2 of solute carrier family 45, member 2, which was formally called antigen isolated from immuno-selected melanoma 1 (aim-1) or membrane-associated transporter protein (matp), is mutated in b-locus mutants of medaka (Fukamachi et al. 2001). This gene is also mutated in human patients with oculocutaneous albinism type 4 (OCA4), underwhite mice, ''cream'' horses, and silver chickens (Newton et al. 2001;Mariat et al. 2003;Gunnarsson et al. 2007). The phenotypes in these different species are quite similar to one another in that melanin deposition is severely suppressed in the skin, whereas the eyes become slightly pigmented during maturation. Population genetic studies have indicated that polymorphisms of human SLC45A2 are also related to population differences in human skin color (Nakayama et al. 2002;Soejima et al. 2006;Yuasa et al. 2006;Graf et al. 2007). slc45a2 mRNA is expressed in melanocyte/ melanophore precursors during embryonic development (Fukamachi et al. 2001;Baxter and Pavan 2002), and the activity of tyrosinase (melanin-synthesizing enzyme) is suppressed in underwhite melanocytes (Costin et al. 2003), but not in b melanophores (Shimada et al. 2002;Fukamachi et al. 2004b). Although these results strongly suggest that a deficiency in slc45a2 function causes the OCA4 phenotype (possibly, via the reduced activity of tyrosinase), evidence that directly supports this gene-phenotype correlation has not yet been obtained, such as knockout/knockdown of slc45a2 or rescue from the mutant phenotype by the wild-type allele.
There is a unique OCA4 mutant in medaka; it has amelanotic skin, but its eyes are melanized. The fish is 1 traditionally called an orange-red variant and has the ''b'' allele at the b locus (Aida 1921;Matsumoto and Hirose 1993;Shimada et al. 2002). Whereas other b-locus mutants with the typical OCA4 phenotype (b c1 , b d2 , b d4 , b d8 , b g8 , b g21 , and b p ) have mutations in the proteincoding region of slc45a2, a mutation has not been identified from the b allele. Considering that the b embryo does not transcribe slc45a2 in the skin (Fukamachi et al. 2001), the b mutation probably occurs in a generegulating region that specifically controls the cutaneous expression of slc45a2.
In this study, we provide empirical in vivo evidence for the slc45a2-OCA4 causal relationship. Further, we identify (1) a promoter sequence that is sufficient for the oculocutaneous expression of slc45a2, (2) multiple transcription start sites that are used tissue specifically, and (3) the b mutation that causes the skin-specific albinism. We also report an intriguing instance in which comparative genomics successfully pinpoint a functional motif in the promoter.

MATERIALS AND METHODS
Construct preparation: Cosmid (HC19): We screened a bacterial artificial chromosome (BAC) library (Kondo et al. 2002) using the AlkPhos direct labeling and detection system (Amersham Pharmacia, Piscataway, NJ) and isolated a clone that contained slc45a2. The clone was subcloned into the SuperCos 1 cosmid vector (Stratagene, La Jolla, CA) after Sau3AI partial digestion using MaxPlax Lambda Packaging Extract (Epicenter Technologies, Madison, WI). We screened for it with colony PCR and isolated a cosmid that contained all seven exons of slc45a2 (HC19; Figure 1A).
Promoter-cDNA fusion constructs (B and b constructs): We first amplified the slc45a2 open reading frame (ORF) of the wildtype ''B'' allele (HNI inbred) by RT-PCR and cloned the products into the pCR4-TOPO vector (Invitrogen, San Diego). Then, the AgeI-XbaI interval of the pEGFP-1 vector (Clontech, Palo Alto, CA), which contains the entire enchanced green fluorescent protein (EGFP) ORF, was substituted with the AgeI-SpeI interval of the slc45a2 ORF plasmid (the AgeI site is within the first exon of slc45a2 and the SpeI site is on the pCR4-TOPO vector). We then amplified the 59 promoter region together with the first exon of the wild-type B allele (Sakura) or the mutant b allele (AA2 inbred) by genomic PCR. The products were inserted between AgeI (in the multicloning site of the pEGFP-1 vector) and AgeI (within the first exon of slc45a2) of the slc45a2 plasmid ( Figure 1B). The dam-methylase-sensitive XbaI site in the pEGFP-1 vector was demethylated in Escherichia coli strain ER2925 (New England Biolabs, Beverly, MA) for digestion.
For the ligation reactions, we resolved the digested inserts/ vectors by electrophoresis on 1-2% agarose gels and recovered the fragments using the QIAquick gel extraction kit (QIA-GEN, Valencia, CA). The vectors were dephosphorylated with shrimp alkaline phosphatase (United States Biochemical, Cleveland), when necessary. After further purification by phenol-chloroform extraction and isopropanol precipitation, the fragments were ligated using a DNA ligation kit Ver. 2 (Takara, Berkeley, CA).
Mutated promoter-cDNA fusion constructs (B-mut and B-del constructs): We PCR amplified two fragments from the wildtype ''B construct'' using four primers (one of them containing the mutation to be introduced; Figure 1C), which mutually overlapped by a 20-bp sequence, 59-TTTCTTAAAAACACCCG CCC-39, immediately upstream of the mutation. These fragments were joined by a second PCR using the 59-most and 39-most primers and were used to replace the EcoRI-EcoRI interval of the B construct.
Micro-injection: All the injection constructs were purified using the Plasmid Midi kit (QIAGEN), dissolved in doubledistilled water to a final concentration of 10-20 ng/ml, and used for micro-injection. The injected embryos were incubated in 0.001% methylene blue water at 25°until hatching.
Sequencing the 59 upstream region of the medaka slc45a2 locus: We isolated genomic DNA from the tail fins of adult fish. The primers for genomic PCRs were designed on the basis of the amacr-slc45a2 intergenic sequence, which we had determined previously (Fukamachi et al. 2001), with melting points 60°, calculated with the CPrimer software (http:/ /iubio.bio. indiana.edu:7131/soft/molbio/Listings.html). All PCRs were performed using the following parameters: 94°for 1 min; 30 cycles of 98°for 20 sec, 60°for 1 min, 72°for 1-10 min; with a final extension at 72°for 10 min. We used a presequencing kit (United States Biochemical) for the preparation of the sequencing templates and sequenced the products using a BigDye terminator cycle sequencing kit (Applied Biosystems, Foster City, CA). Sequences were assembled and compared with SeqMan II software (DNAStar, Madison, WI).
RT-PCR: We used Isogen (Nippon Gene) to isolate the total RNA from embryos and the eyes of adult fish. First-strand cDNA was synthesized with ReverTra Ace reverse transcriptase (Toyobo) following the manufacturer's protocol. All PCRs were performed with the conditions described above.
59 RNA-ligase-mediated RACE: We used the FirstChoice 59 RNA-ligase-mediated RACE (RLM-RACE) kit (Ambion, Austin, TX). Total RNAs were extracted from the eyes and skin of wildtype (color interfere; Fukamachi et al. 2004a) and b (Hd-rR) fish using TRIzol reagent (Invitrogen). Nested PCR products were resolved by electrophoresis on a 1% agarose gel. Four major bands of two different sizes (see Figure 4B) were excised, and the DNAs were recovered as described above and cloned into the pCRII-TOPO vector (Invitrogen). A total of 61 positive clones (i.e., 13-16 clones from each band) were PCR amplified and sequenced.
Electrophoretic mobility shift assay: The mouse B16 melanoma cell line and the mouse L929 fibroblast cell line were cultured at 37°with 10 mm HEPES-buffered L-15 medium containing 15% fetal calf serum. For the experimental use of logarithmically growing cells, 5 3 10 5 cells were inoculated into a 600-ml plastic flask and incubated for 2 days.
Complementary oligonucleotide probes (see Figure 6A for the sequence of one strand) for electrophoretic mobility shift assay (EMSA) were annealed, their 59 ends were labeled with ½g-32 PATP using T4 polynucleotide kinase (Toyobo), and they were purified using the QIAquick nucleotide removal kit (QIAGEN). For the binding reaction, 4 mg of nuclear extract were used. The binding reaction was carried out in 20 ml of a mixture containing 15 mm Tris-HCl (pH 7.4), 60 mm KCl, 0.5 mm DTT, 7.5% glycerol, 0.25 mg/ml bovine serum albumin, 0.1 mg/ml poly(dI-dC), and 100 fmol of the 32 P-labeled probe. In a competition assay, 1 or 10 pmol (10-fold or 100-fold molar excess, respectively) of nonradiolabeled probe DNA was added to the standard mixture. After incubation at room temperature for 1 hr, the samples were loaded onto 0.5% polyacrylamide gels in 0.53 Tris-borate-EDTA buffer and resolved electrophoretically at 10 V/cm for 2 hr at room temperature. The gels were then dried, and images were obtained with a BAS-5000 imaging analyzer (Fuji Film).

RESULTS
An slc45a2-containing genomic fragment rescued melanin synthesis in the OCA4 medaka: We microinjected into medaka OCA4 mutant (b g8 ) eggs a cosmid (HC19) containing all the exons/introns of medaka slc45a2 and 9 and 2 kb of the 59-and 39-flanking regions, respectively ( Figure 1A). Among the b-locus mutants, melanin synthesis is suppressed most severely in the b g8 mutant (Shimada et al. 2002). We observed that 36 of 140 injected embryos exhibited chimeric deposition of melanin in the eye and/or skin (eye, 3; skin, 24; both, 9; Figure 2, A and B). Some fish retained the recovered skin melanophores until the adult stage, suggesting stable gene expression from HC19 (Figure 2, C and D). None of the mature fish successfully passed the transgene to their offspring, although one of the pairs occasionally (10%) produced melanized but severely malformed and lethal embryos (data not shown). Therefore, HC19 must contain a gene and its regulatory regions that can rescue the medaka OCA4 phenotype.
slc45a2 cDNA driven by its 0.9-kb promoter rescued melanin synthesis in the OCA4 medaka: The rescuing gene is most likely slc45a2, but another gene might exist in HC19. A series of our published and preliminary experiments indicated that the transcription of slc45a2 is not very strong (difficult to be detected) in medaka, e.g., faint signals on whole-mount in situ hybridization (Fukamachi et al. 2001), no signal from embryos microinjected with an slc45a2 promoter-EGFP plasmid (see below), no band on Northern hybridization, and no knockdown by morpholino injection (data not shown). The weak expression was similarly reported from mouse (Baxter and Pavan 2002), and it has also recently been reported that strong expression of SLC45A2 may suppress melanin deposition in humans (Graf et al. 2007). Therefore, we undertook to rescue the b g8 phenotype with slc45a2 cDNA driven by its own promoter, rather than with a promoter for its ectopic expression.
We first sequenced the 59-flanking region (4.4 kb) of the medaka slc45a2 locus and compared the sequences between two genetically divergent wild-type strains, HNI (northern inbred) and Sakura (southern noninbred). Besides many single-nucleotide polymorphisms (SNPs) and small insertions/deletions, we detected a large insertion of 5 kb in the HNI allele 0.9 kb upstream from the translation initiation codon (Figure 3, A and B). Because the insertion of a fragment of several kilobases (e.g., a transposon) into a promoter often abolishes the ordinary transcription of the gene (e.g., Iida et al. 2004), we expected that the region upstream from this insert may be dispensable for slc45a2 expression.
Therefore, we constructed a plasmid in which 922 bp of the 59-promoter region, which should also contain the 59-untranslated region (UTR), were connected to the slc45a2 ORF (1728 bp) and the 36 bp of the 39-UTR (see Figure 1B). Micro-injection of the construct (B construct) successfully rescued the b g8 phenotype. From 127 injected eggs, nine embryos exhibited chimeric deposition of melanin in the eye and/or skin (eyes, two; skin, three; both, four; Figure 2, E and F). Therefore, the medaka OCA4 phenotype is caused by a defect in slc45a2 and not in a neighboring gene in HC19. This result also shows that the 0.9-kb promoter can drive slc45a2 expression at a level sufficient for melanin production in both the eyes and the skin.
The b mutation occurs in the promoter of slc45a2: Because the 0.9-kb promoter rescued both the ocular and the cutaneous phenotypes, we expected the b mutation (see Introduction) to occur in the promoter. Indeed, we found a unique nucleotide change in the promoter of the b mutant, an inversion of 167 bp, an insertion of 48 bp, and a deletion of 172 bp (Figure 3; see Figure 4C for details). The origin of the inserted 48 nucleotides is unknown, because the sequence cannot be aligned to neighboring sequences. This allele (inversion/insertion/deletion, inv/ins/del) was not found in at least 11 wild-type fish, which had been collected from various rivers and ponds in Japan and South Korea ( Figure 3C; see Takehana et al. 2004 for references).
To assess the function of the b promoter, we substituted the 922-bp promoter of the B construct with the 795-bp promoter of the b mutant (b construct; Figure 1B) and micro-injected it into b g8 or SK 2 eggs (SK 2 is a triple mutant of the b g8 , leucophore-free, and guanineless loci recovered in this study; see Figure 2F). Fourteen of 284 injected embryos exhibited chimeric deposition of melanin, but only in the eyes ½note that only 2/9 embryos rescued by the B construct had pigment only in the eye (P , 0.001, chi-square test). Therefore, we concluded that the b promoter is responsible for the mutant phenotype and that it is likely to be the inv/ins/del sequence.
Multiple transcription start sites in medaka slc45a2: Interestingly, we found that slc45a2 mRNA in the b mutant lacks part of the 59-UTR ( Figure 4A). This indicates the existence of multiple transcription start sites (i.e., the b mutation selectively abolishes the transcription of the full-length mRNA) or utilization of a weaker transcription start signal in the b mutant. Our results from RLM-RACE supported the former scenario. The wild-type medaka transcribe at least two major variants of slc45a2 mRNA: the longer form in both the skin and the eyes and the shorter form only in the eyes ( Figure  4B). The longer form was not detectable in the b mutant (even after 60 cycles of nested PCR), whereas the shorter form seemed to be expressed as usual. Thus, the b mutation specifically suppresses the transcription of the longer mRNA, which is necessary for melanization in the skin but not in the eyes, whereas the shorter mRNA is transcribed independently of the b mutation.
Medaka-pufferfish comparative genomics pinpoint a functional motif in the 0.9-kb promoter necessary for the cutaneous expression of slc45a2: The 339-bp region disrupted by the b mutation probably contains a functional motif, such as a transcription-factor-binding site (TFBS), that regulates the transcription of the longer form of slc45a2 mRNA. However, we could not efficiently predict such motifs using online software, such as TFBIND (http:/ /tfbind.ims.u-tokyo.ac.jp/) or TFSEARCH (http:/ / molsun1.cbrc.aist.go.jp/research/db/TFSEARCH.html), because .100 TFBSs were identified within the 0.9-kb promoter. We also failed to make a prediction based on comparative genomics (phylogenetic footprinting), using human, mouse, rat, zebrafish, Takifugu, and Tetraodon genomes ( Figure 5, A-C; other data not shown).
Intriguingly, however, a pairwise alignment of the medaka and pufferfish (Takifugu or Tetraodon) promoters using the mVISTA program (http://genome.lbl.gov/ vista/index.shtml) revealed up to four conserved putative motifs in the 0.9-kb promoter ( Figure 5C). This result allowed us to identify a pyrimidine (Py)-rich motif (motif D in Figure 5), which is located in very close proximity to the 59-most transcription start site of slc45a2, at least in medaka ( Figure 4A), zebrafish ½expressed sequence tag (EST) CF266172 in GenBank, mouse (EST BB858311), and human (Graf et al. 2007). Although the sequences of the Py-rich motif are not strictly conserved among species, its functional importance in the skin is further suggested by two observations: (1) the Py-rich motif is interrupted by the 39 edge of the b mutation in medaka ( Figure 4A), and (2) cultured mouse B16 melanoma cells express nuclear proteins that specifically bind to the Py-rich motif ( Figure 6).
To assess this hypothesis, we micro-injected promoter-cDNA constructs with a mutation exclusively in the (E) and SK 2 (F) embryos rescued by the B construct (see Figure 1B). Note that the pink-and silver-colored pigment cells (leucophores and iridophores, respectively) in b g8 are eliminated in SK 2 by the leucophore-free and guanineless genes (see text), which allows clearer detection of the rescued melanophores. The skin of adult SK 2 fish is also transparent, which allows the observation of the intact internal organs, as in the see-through strain (Wakamatsu et al. 2001).
Py-rich motif: the B-mut construct, in which the Py-rich motif (59-CTTTCTCTCTTTCCTCTTTACT-39) was substituted with purines (59-GAAAGAGAGAAGGGAGAAA TGA-39), and the B-del construct, in which the Py-rich motif was deleted ( Figure 1C). The B-mut or the B-del construct rescued the OCA4 phenotype in 5 of 64 or 7 of 76 b g8 or SK 2 embryos, respectively, but only in the eyes ½5/5 or 7/7 significantly (P , 0.001, chi-square test) differs from 2/9. Thus, the Py-rich motif plays an indispensable role in melanin production in the skin, facilitating the transcription of the longer form of slc45a2 mRNA in melanophores.

DISCUSSION
Vertebrate OCA4 mutants caused by mutations in slc45a2: Our results presented here provide robust in vivo evidence of an slc45a2-OCA4 causal relationship. We believe that the OCA4 phenotypes of other vertebrates are also caused by the loss of slc45a2 function and will be rescued by wild-type slc45a2 nucleotides. However, this might not be the case in the underwhite dominant brown (uw Dbr ) mutant mouse, which has a missense point mutation in the coding region of slc45a2 (Newton et al. 2001;Du and Fisher 2002). Dominantnegative slc45a2 protein has been proposed to explain the dominant inheritance of the uw Dbr phenotype. However, the detailed function of slc45a2 protein in melanin synthesis is as yet not well understood (but see Costin et al. 2003).
There is another intriguing OCA4 mutant, B9 medaka, which exhibits ''variegated'' deposition of melanin in the adult skin (Aida 1921). We found that the B9 mutant has no mutation in the coding region of slc45a2, but carries the inv/ins/del sequence (i.e., the b mutation) in the promoter (our unpublished observation). The indistinguishable phenotypes of the b and B9 embryos can be explained by this finding. The mechanism for the later and incomplete enhancement of melanin deposition in the B9 skin remains unknown. We assume that other genomic regions in addition to the 0.9-kb promoter control the expression of medaka slc45a2. This is because we achieved a lower rescue frequency when using the B construct (9/127) than when using the cosmid HC19 (36/140; P , 0.001, chi-square test). Sequence comparisons of the entire slc45a2 locus (i.e., 59/39 genomic regions and introns contained in HC19; see Figure 1A) between the b and B9 alleles may reveal additional  Figure 1A) of three medaka strains ½two wild types (HNI and Sakura) and the b mutant were compared. Shaded boxes on the right indicate the translated region of the first exon. Up to 108 SNPs and 96 allele-specific nucleotides (i.e., insertions and deletions of 1-23 bp) were detected, which are indicated as asterisks, triangles, and inverted triangles. Additional deletions of 326 bp in HNI and 207 bp in Sakura are indicated by broken lines. The position of the large HNI-specific insert is indicated at the bottom. Solid and open boxes represent the inv/ins/del sequence found in the b allele (see text). Arrows show the approximate positions of the PCR primers used in B and C. (B) Electropherogram of genomic PCR products amplified by the primers indicated in A. Template DNAs are indicated at the top. Much longer products were amplified from HNI and the HNI-derived cosmid (HC19) and slightly shorter products were amplified from AA2 (b/b inbred) and the Hd-rR (b/b inbred)-derived cosmid (C27), relative to those amplified from Sakura and b g8 (b g8 was recovered from g-irradiation mutagenesis of Sakura). (C) Electropherogram of genomic PCR products amplified by the same primers used in B. Template DNAs are indicated at the top. These fish are from the southern Japanese population, except for Yamagata and Sokcho, which belong to the northern Japanese and the eastern Korean populations, respectively. All the products were sequenced directly, but the inv/ins/del sequence occurred only in the b allele. Polymorphism information is summarized in Figure 4C. mutations and provide new insight into the cutaneous expression of slc45a2 in adult fish.
Medaka promoter for oculocutaneous expression of slc45a2: The 0.9-kb promoter isolated in this study can drive slc45a2 transcription at a sufficient level for melanin production in both the eyes and the skin of medaka. This system might be useful in assessing the functional significance of the polymorphisms/mutations in slc45a2 in vivo; i.e., if mammalian orthologs can similarly rescue the b g8 phenotype, we should be able to functionally assess the polymorphism/mutation of interest (see Introduction) by preparing constructs for micro-injection.
The promoter sequence should also be useful in investigating the molecular mechanisms that control slc45a2 expression. Microphthalmia-associated transcription factor RT-PCR using various forward primers (f1-f6 in C) and a reverse primer (r2) designed to bind to the second exon (sequence not shown). Template cDNA was prepared from the eyes of wild-type adult medaka (HNI). Identical results were obtained using cDNAs from 1-, 2-, and 4-day HNI embryos (data not shown). The forward primers (f1-f6) and another reverse primer (r1) designed to bind to the first exon of slc45a2 (sequence not shown) were used to perform the genomic PCR. Comparison of the RT-PCR and genomic PCR results indicates that slc45a2 is transcribed from the region between f2 and f3. Right: the same analyses using cDNA and genomic DNA of the b mutant. Primers f1 and f2 did not amplify any fragment from either template, because their complementary sequences were destroyed by the b mutation (C). Primer f3 amplified products from the genomic DNA but not from the cDNA, indicating that slc45a2 is transcribed from the region between f3 and f4 in the b mutant. Identical results were obtained using cDNAs from 1-, 2-, and 4-day b embryos (data not shown). (B) Electropherograms of 59 RLM-RACE products. Two major bands of different sizes were obtained from the wild-type eyes. The longer mRNA was also transcribed in the skin, but the shorter mRNA was not. The b mutant lacked the longer product in both the eyes and the skin. (C) The b mutation. The 0.9-kb promoter sequence is shown. A solid or shaded arrowhead on top shows the 59 end of the 0.9-kb promoter in the promoter-cDNA constructs (Figure 1) or the position of the large northern-population-specific insert (Figure 3), respectively. Residues highlighted by solid and shaded boxes are inverted and deleted sequences in the b allele, respectively. Underlined residues are polymorphic among the populations analyzed in Figure 3C. Residues highlighted by solid or shaded circles are transcription start sites in the skin or eyes, respectively, determined by 59 RLM-RACE. The longer transcript is transcribed from different transcription start sites in the eyes and skin. The transcription start site of the shorter form seems not to be strictly controlled, possibly because of the repetitive sequence, (CTTT) n . A transcription start site predicted by the Neural Network Promoter Prediction program (http:/ /www.fruitfly.org/seq_tools/promoter.html) is also shown with an open arrowhead. Solid arrows indicate the positions of the PCR primers used in A. The closest E-box to the promoter (see discussion), the Py-rich motif D (see Figure 5), and the translation initiation codon are boxed.  (Mayor et al. 2000). The translation start codon is underlined. Transcription start sites are indicated by arrowheads. Conserved nucleotides are indicated by shaded boxes. Note that many nucleotides are conserved not only in the translated region but also in the promoter region. Similar results were obtained from mouse-rat and Takifugu-Tetraodon comparisons (see C). The motif D, which was revealed by the medakapufferfish comparison (see C), is boxed. (B) The zebrafish sequence was only poorly aligned to any of the human, mouse, rat, Takifugu, Tetraodon, and medaka sequences (data not shown). Two of the four motifs predicted by the medaka-pufferfish comparison exist in zebrafish. (C) The medaka-pufferfish comparison. The mVISTA program aligned an identical medaka sequence differently to Takifugu and Tetraodon sequences, which are shown as Medaka1 and Medaka2, respectively. The upper regions (not shown here) rarely contain conserved nucleotides, even between the pufferfish genera. Nucleotides conserved only between the pufferfish genera are indicated by shaded boxes. Solid boxes show nucleotides conserved among the four sequences. (mitf) has been suggested to control slc45a2 expression (Du and Fisher 2002). Indeed, mitf upregulates slc45a2 transcription in both the mouse and medaka (Du and Fisher 2002;Béjar et al. 2003), and mitf mutant mice show reduced expression of slc45a2 (Baxter and Pavan 2002). However, the 0.9-kb promoter, which does not possess the mitf-binding motif called E-box (CATGTG), rescued the OCA4 phenotype (see Figure 4C). Moreover, an immunoprecipitation assay has revealed that MITF does not bind to E-boxes 0.3-1.5 kb upstream from the translation start codon of human SLC45A2 (Du and Fisher 2002). These observations suggest that mitf does not interact with the promoter of slc45a2. mitf upregulates slc45a2 expression, probably by binding to other gene-regulating regions (as discussed above), or only indirectly by regulating the expression of other transcription factors that bind to the slc45a2 promoter.
Medaka-pufferfish comparative genomics reveal an indispensable Py-rich motif: As an initial step in such a functional dissection of the 0.9-kb promoter, we used comparative genomics to predict TFBSs. Although phylogenetic footprinting is a good method with which to identify conserved and therefore functionally important sequences, the prediction of TFBSs in a promoter is often difficult. This is because (1) TFBSs are generally small (,10 bp); (2) sequences surrounding TFBSs are highly divergent, even in length, which disturbs the alignment of the TFBSs; (3) the TFBS itself is not strictly conserved among species; and (4) only limited numbers of genomic sequences are available, especially for vertebrates (see Boffelli et al. 2004;Wasserman and Sandelin 2004).
Interestingly, however, a comparison of the medaka and pufferfish genomes exceptionally predicted a functional sequence that plays an essential role, the Py-rich motif ( Figure 5C). Although we do not believe that this medaka-pufferfish method will reveal a functional motif in all promoters, the recent publication of the medaka whole-genome sequence (Kasahara et al. 2007) should provide expanded opportunities to characterize promoter structures in silico before in vitro/in vivo experiments. However, the Py-rich motif should not be the only functional motif in the medaka slc45a2 promoter; another motif that controls the shorter mRNA in the eye ( Figure 4A) must also exist. To obtain a more complete understanding of the transcriptional regulation of slc45a2, further investigations will be necessary, including comparative genomic analyses of more species and in vitro screening for functional regions using deletion constructs.
We detected a nuclear protein that binds to the Pyrich motif in mouse melanoma cells ( Figure 6). Identification of this protein (Stead and Mcdowall 2007) may provide new insight into trans-regulatory mechanisms for the cutaneous expression of slc45a2. However, we obtained an EMSA result identical to that shown in Figure 6 using mouse L929 fibroblast cells, which do not transcribe slc45a2 or produce melanin (data not shown). Considering that the Py-rich motif is located very close to the transcription start site ( Figure 5), it is possible that the protein that binds to the Py-rich motif is a general transcription factor that initiates transcription. To explain the tissue-specific transcription of slc45a2, other mechanisms must be considered, such as melanocyte-specific cofactors that activate the binding protein, another motif on the promoter to which a melanocytespecific transcription factor binds, etc.
In summary, we rescued the medaka OCA4 phenotype with slc45a2, isolated the promoter sufficient for the oculocutaneous expression of slc45a2, revealed multiple mRNA variants that are tissue-specifically transcribed, identified the b mutation as the cause of the skin-specific albinism, and identified the Py-rich motif, which is essential for melanogenesis in the skin. Further studies of the 0.9-kb promoter sequence, including the Py-rich motif, will provide a more detailed understanding of the cis/trans regulatory mechanisms underlying the oculocutaneous expression of slc45a2.  Figure 5A) and nuclear extracts from cultured mouse melanoma cells. (A) Probe sequences. Part of the promoter sequence of mouse slc45a2 is shown at the top. Residues are numbered with the translation start site as 11. A solid box indicates the Py-rich motif. Substituted residues in probes 1-4 are underlined. (B) EMSA results. Labeled probes and nonlabeled competitors are indicated at the top. The amount of competitor is 103 or 1003 the labeled probe. Probes 1 and 4, but not 2 and 3, inhibited the binding of the labeled probe to the nuclear proteins.