Identification of Aim-1 as the underwhite mouse mutant and its transcriptional regulation by MITF.

Animal pigmentation mutants have provided rich models for the identification of genes modulating pathways from melanocyte development to melanoma. One mouse model is the underwhite locus, alleles of which manifest altered pigmentation of both eye and fur, sometimes in an age-dependent fashion. Here we show that the mouse homolog of a recently identified gene whose mutation produces Japanese gold-colored fish, medaka b, maps to the mouse underwhite locus. We identify distinct mutations of this gene, known as Aim-1, in three underwhite mouse alleles and find that structure/function differences correlate with recessive versus dominant inheritance. The human ortholog of AIM-1 was originally identified as a melanocyte-restricted antigen that is recognized by autologous T cells from a patient with melanoma. We also provide evidence that AIM-1 is transcriptionally modulated by MITF, a melanocyte-specific transcription factor essential to pigmentation and a clinical diagnostic marker in human melanoma. Although AIM-1 appears to reside downstream of MITF, chromatin immunoprecipitations do not reveal binding of MITF to a 5'-flanking region containing histone 3 acetylation, indicating that MITF either acts indirectly on AIM-1 or it binds to a remote regulatory sequence. Nevertheless, MITF links AIM-1 expression and the underwhite phenotype to a transcriptional network central to pigmentation in mammals.

Animal pigmentation mutants have provided rich models for the identification of genes modulating pathways from melanocyte development to melanoma. One mouse model is the underwhite locus, alleles of which manifest altered pigmentation of both eye and fur, sometimes in an age-dependent fashion. Here we show that the mouse homolog of a recently identified gene whose mutation produces Japanese gold-colored fish, medaka b, maps to the mouse underwhite locus. We identify distinct mutations of this gene, known as Aim-1, in three underwhite mouse alleles and find that structure/function differences correlate with recessive versus dominant inheritance. The human ortholog of AIM-1 was originally identified as a melanocyte-restricted antigen that is recognized by autologous T cells from a patient with melanoma. We also provide evidence that AIM-1 is transcriptionally modulated by MITF, a melanocyte-specific transcription factor essential to pigmentation and a clinical diagnostic marker in human melanoma. Although AIM-1 appears to reside downstream of MITF, chromatin immunoprecipitations do not reveal binding of MITF to a 5-flanking region containing histone 3 acetylation, indicating that MITF either acts indirectly on AIM-1 or it binds to a remote regulatory sequence. Nevertheless, MITF links AIM-1 expression and the underwhite phenotype to a transcriptional network central to pigmentation in mammals.
A recent study reported a novel transporter protein AIM-1 that is responsible for the pigment phenotype of medaka b gold-fish mutants (1). In this fish, mutations in Aim-1 produce albino-like depigmentation of black melanophores, leaving gold-red xanthophores, white leukophores, and silver iridophores as the remaining pigmented cells. The gene product is predicted to be a 12-transmembrane-transporter protein that shows homology to yeast and plant sucrose transporters (1). Human AIM-1 has been previously identified as a melanoma antigen that was recognized by autologous, patient-derived T cells (2). Its expression was shown to be restricted to the melanocyte lineage by Northern analysis (2). Murine Aim-1 is located on chromosome 15 in a region containing the pigmentation locus, underwhite. A series of mouse mutants has been described for this locus. All are characterized by various de-grees of pigment reduction (from partial to complete albinism) in the eyes and fur (3).
MITF is a tissue-restricted transcription factor essential to melanocyte development. It binds the canonical E-box sequence CACGTG as well as the non-palindromic sequence CACATG. The major pigmentation hormone, melanocyte-stimulating hormone, up-regulates MITF expression through cAMP signaling followed by cAMP-response element-binding protein phosphorylation and activation of the melanocyte-specific MITF promoter and may modulate multiple pigmentation genes through up-regulating MITF expression (4,5). The three major pigmentation enzymes tyrosinase, TYRP1, and DCT, all contain consensus MITF DNA binding elements that are conserved across species and are thought to be transcriptional targets of MITF (6 -8). In humans, germline heterozygous MITF mutation produces the pigmentation-deafness condition Waardenburg Syndrome IIA (9) and Tietz syndrome (10,11), manifesting pigmentation disturbances and deafness due to inner ear melanocyte deficiency (12). Interestingly, MITF expression is usually (if not always) maintained in human melanoma specimens, and it is increasingly used as a histopathologic marker for melanoma diagnosis (13)(14)(15)(16)(17).
We BLASTed the human AIM-1 mRNA sequence (accession number AF172849) against the human genome and located the gene to chromosome 5. Examination of the homologous region in the mouse genome revealed a previously described hypopigmentation locus, underwhite. Here experiments were carried out to explore a potential link between Aim-1 and the underwhite mutants. Distinct mutations in Aim-1 were found in three underwhite alleles, including a frameshift, as compared with wild-type controls. In addition, evidence is presented that suggests that the global transcriptional regulator of pigmentation, MITF, also resides upstream of Aim-1, linking this gene to the major pigmentation pathway in melanocytes.

EXPERIMENTAL PROCEDURES
BLAST-Human AIM-1 mRNA sequence (accession number AF172849) was BLASTed against the Human Genome. Mouse Aim-1 mRNA sequence (accession number AF360357) was BLASTed against the mouse Trace Archive.
Genomic DNA PCR-Genomic DNA from C57BL/6J and homozygous uw, uwd, and Uw Dbr mice was obtained from the Jackson Laboratory (Bar Harbor, ME). Genomic DNA from adult male BALB/c kidney tissue was purchased from CLONTECH. Primer pairs spanning each exon were designed as follows. Chromatin Immunoprecipitation-Chromatin immunoprecipitation assay (ChIPs) 1 was performed in human primary melanocytes (provided by Dr. Ruth Halaban, Yale University), SKMEL5, or IMR90 cells (ATCC) grown in logarithmic phase. Cells were harvested by scraping, homogenized in a hypotonic buffer (10 mM Tris-HCl, pH 7.4, 15 mM NaCl, 60 mM KCl, 1 mM EDTA, 0.1% Nonidet P40, 5% sucrose, 1ϫ Complete TM proteinase inhibitor mixture (Roche)) on ice using a Dounce homogenizer. The nuclei were isolated by centrifugation onto a 10% sucrose pad and then cross-linked with 1% formaldehyde in phosphate-buffered saline for 20 min at room temperature with gentle shaking. Nuclei were then spun down and resuspended in ChIPs buffer (10 mM Tris-HCl pH 7.4, 100 mM NaCl, 60 mM KCl, 0.1% Nonidet P40, 1ϫ proteinase inhibitors) and sonicated by two 1-min pulses using a Fisher dismembranator fitted with a micro-tip on ice. Antibodies against USF1 (C-20, Santa Cruz), USF2 (N-18, Santa Cruz), c-Myc (N-262X, Santa Cruz), polyclonal rabbit anti-MITF, or acetylated Histone H3 (Upstate) were then added to a 10-fold ChIPs buffer diluted sample and incubated on a nutator for 3 h at room temperature. Subsequently, Ultralink protein A/G beads (Pierce) were added to the sample and a control sample and incubated for an additional hour. Immunoprecipitates were then washed twice with ChIPs buffer, twice with 500 mM NaCl ChIPs buffer and once with TE, pH 8. The immunoprecipitates were released from the beads by incubating at 65°C for 20 min in 1% SDS/TE, and proteins were digested by proteinase K treatment side-by-side with an additional unprecipitated sample as input control. Cross-links were released by heating at 70°C for 10 h, DNA recovered by extraction with phenol and chloroform at high salt (0.6 M sodium acetate, pH 8), and then ethanol precipitated. Semi-quantitative PCR was then performed on samples to amplify fragments spanning the 5Ј-or 3Ј-adjacent region to the E box repeat (see Fig. 2a). The forward and reverse primers for the 5Ј-region are 5Ј-TGT TAA GTA CCA CGA GGA GAA ATA-3Ј and 5Ј-ACA TGG CCG TGA GGT AAT AAA G-3Ј, respectively, (annealing temperature 59°C). The primers for the 3Ј-region are 5Ј-CAG CAC CAC CCC CTC CCT CTC ATC ATA AC-3Ј and 5Ј-CCA CAG AGT CAA AGG GGC CAT CAT CAG C-3Ј (annealing temperature 65°C). The primers for the MITF-positive ChIP control region (intron 1 of PMEL17) are 5Ј-CAT AAG ATA CCC CAT TCT TTC TCC ACT T-3Ј and 5Ј-GAG AAT GTG GTA TTG GGT AAG AAC AC-3Ј (annealing temperature 57°C). PCR was carried out for 45 s at 94°C, 45 s at annealing temperatures as indicated for each primer set, and 90 s at 72°C for a total 35 cycles with Taq polymerase (Fisher).
Adenovirus Infection and RNA Preparation-Null adenovirus was purchased from Q-BIOgene. Adenoviruses encoding wild-type MITF, dominant-negative mutant MITF, or vector control encoding a fusion of green fluorescence protein-Wee1 for nuclear localization, were generated as previously described (18). Briefly, 10 6 human SKMEL5 melanoma cells were plated per 100-mm plate. On the second day, cells were overlaid with 2 ml of serum-free F10 media containing 10 mM MgCl 2 , and concentrated adenovirus was added at multiplicity of infection of 100 for each virus. The cells were incubated at 37°C for half an hour after which virus was removed and fresh full media was added. Total

RESULTS
Aim-1/B Is Mutated in uw, uwd, and Uw Dbr Mice-We used human map viewer (ncbi.nlm.nih.gov/) to determine that the human ortholog of Aim-1 is located on chromosome 5p. We identified a region homologous to the human Aim-1 locus on mouse chromosome 15 and noted that this site maps to the murine underwhite locus (ncbi.nlm.nih.gov/Homology/and Ref. 19). The complete mouse Aim-1 cDNA sequence (accession number AF360357) was BLASTed against the mouse Trace Archive to identify the exon/intron boundaries (Table I). Primer pairs located in the intron regions were designed to span individual exons. Each exon was amplified from the genomic DNA of uw, uw d and Uw Dbr mice as well as the control C57BL/6J and BALB/c strains. All PCR products matched the predicted lengths (data not shown). DNA sequencing identified mutations in all three mutant alleles, but none of their wild-type controls (Fig. 1a). The uw mutant was found to harbor a 7-base pair deletion in Exon 3 of Aim-1 (Fig. 1a), which results in a 43-amino acid frameshift followed by a premature stop at codon 308 ( Fig. 1, b and c). The uw allele thus encodes a protein that lacks the C-terminal 6 transmembrane domains of the predicted protein. The uw d mutant harbors a single nucleotide mutation from T to C in Exon 6 of Aim-1 (Fig. 1a), which leads to a mutation from a conserved serine to proline in the tenth transmembrane domain (Fig. 1, b and c). The dominant allele Uw Dbr contains a nucleotide alteration from G to A in the second exon (Fig. 1a), which causes a missense point mutation of D153N (conserved residue from medaka to human) in the fourth transmembrane domain (Fig. 1, b and c). These studies thus identify Aim-1 as the mutant gene in underwhite mice.
MITF Regulates Endogenous AIM-1 mRNA Levels in Human Melanoma Cells-Through recognition of E-1 box DNA elements (CA(C/T)GTG) MITF is thought to modulate expression of many major pigmentation genes including tyrosinase, 1 The abbreviations used are: ChIPs, chromatin immunoprecipitation assay; RACE, rapid amplification of cDNA ends.  Fig. 2, wild-type MITF stimulated AIM-1 expression significantly over baseline control while dominant-negative MITF repressed its expression. Similar results were also observed in human primary melanocytes (data not shown). These effects were consistently seen at multiple time points examined. The data suggest that MITF can modulate the expression of AIM-1 in human melanoma cells.

An E-Box Repeat Is Present in the Putative Promoter
Region of the Human Aim-1 Gene-Examination of the upstream region of the human Aim-1 gene revealed a 1.2-kb repetitive sequence containing 55 E boxes (CACGTG or CATGTG). Fig.  3a depicts representative smaller repeat units seen in this region. It is unclear whether this promoter structure (or specific sequence) is conserved in the mouse. A CENSOR (20) search on this repeat against RepBase (21)(22)(23) showed no homology to known human repetitive sequences. Also, a BLAST search against the human genome revealed only a few shorter copies of this repeat distributed in non-coding regions. Repeats can arise very quickly during genome propagation, especially if they by chance contain short transcription, recombination, or

Aim-1 Mutations in underwhite Mouse
replication signals. One common mechanism for rapid generation of repeats within one or a few generations is via short repeat expansion, replication, and/or recombination errors. It appears plausible, for example, that the hexanucleotide repeat (CATGTG) might be derived from a dinucleotide repeat (CA ϭ TG). 2 MITF Does Not Occupy AIM-1 5Ј-Flanking Sequences by ChIP-To examine occupancy by MITF and other E box binding factors such as c-Myc, USF1, and USF2 on the AIM-1 5Ј-flanking region, chromatin immunoprecipitation was carried out on human primary melanocytes, human melanoma cells (SKMEL5), and IMR90 cells (human fibroblast). Primers were designed that span the 5Ј-or 3Ј-region of the repeat (Fig.  3a). In primary melanocytes, but not fibroblasts (Fig. 3b), PCR products were amplified in samples immunoprecipitated with an antibody directed against acetylated histone H3, a component of transcriptionally active chromatin. In addition, quantitative PCR were performed on total RNA from all three cell types. AIM-1 transcripts were detected in primary human melanocytes, SKMEL5 melanoma cells but not IMR90 fibroblasts (Fig. 3c). These results are consistent with melanocyterestricted expression of AIM-1 (2). However, the various E-boxbinding proteins tested, including MITF, do not measurably interact with this region despite the consensus E-boxes. Positive control ChIPs for MITF are shown for a different E-boxcontaining genomic fragment (Fig. 3a). Similar results were also obtained from SKMEL5 melanoma cells, except detection of acetylated histone H3 did not extend as far in the 5Ј-region, perhaps reflecting differences in transcriptional regulation of AIM-1 in this unpigmented melanoma line. These results suggest that MITF or other E-box-recognizing proteins do not directly bind the E-box repeat in the melanocyte lineage. Correspondingly, luciferase reporter assays from this region (Ϫ1689 to ϩ28 and Ϫ264 to ϩ28) demonstrated substantially increased basal activity in melanocytes versus non-melanocytes, but no significant responsiveness to MITF overexpression (data not shown). These data suggest that MITF does not modulate AIM-1 transcription via direct interactions at the proximal promoter, but rather likely utilizes a more remote binding element, or operates through a different transcriptional intermediate. It is also formally possible that the altered RNA levels of AIM-1 by MITF could reflect changes in RNA stability, although such an activity has not previously been described for MITF.

DISCUSSION
The studies described here identify the Aim-1 gene as the locus responsible for the mouse underwhite mutations. Truncation of more than 40% of the protein in the recessive uw mutant leads to complete loss of pigmentation and exhibits the greatest pigment deficiency of the three alleles in uw/uw mice. In contrast, the S435P mutation in uw d allele produces a mild coat color dilution in homozygotes. Thus, the uw mutation is a strong loss-of-function mutant, which might behave as a functional null while the uw d C-terminal mutation implies a partial loss of function that does not measurably interfere with the product of the remaining wild-type allele. In addition, uw heterozygotes appear normally pigmented, suggesting that haploinsufficiency does not occur for this gene, a finding consistent with other pigment loci, such as albino (19), although quantitative measurements of pigmentation might reveal subtle differences that are not appreciated visually.
The dominant inheritance pattern of Uw Dbr strongly suggests that Aim-1 functions in a molecular complex with other species, possibly as a homo-or hetero-oligomer. Aim-1 bears remote sequence homology to the sucrose transporter family (1) whose oligomerization properties are not well studied although dimer or tetramer formation has been reported for other transporter families and genetic data even suggest interactions between different transporters in yeast (24). Since the Uw Dbr mutation occurs within a putative transmembrane motif, it is likely that the altered charge of the asparagine side chain (versus aspartate) disrupts a salt bridge or other necessary secondary structure that contributes to function in a manner that does not ablate intermolecular interactions. However, despite its dominant genetic behavior, this mutation produces a milder phenotype in homozygotes as compared with the frameshift allele (uw), indicating that even in homozygous form the mutant protein retains some wild-type function.
The putative transcriptional start described in this study is 2 G. Church, personal communication.

FIG. 3. E-box repeats in the putative promoter of human
Aim-1 and chromatin immunoprecipitation. a, diagram of the putative promoter of human Aim-1. Representative repeat units in the human E box-containing element are presented. b, chromatin immunoprecipitation of the putative human AIM-1 promoter region. Neither of two regions in the AIM-1 promoter (termed 5Ј-and 3Јfragments on promoter map) was bound by MITF or the other E box-binding proteins USF1, USF2, and c-Myc. A positive control chromatin immunoprecipitation from a separate genomic location is shown for MITF and USF1. The experiment was repeated three times. c, quantitative PCR on endogenous AIM-1 mRNA levels in human primary melanocytes, SKMEL5 melanoma cells, and IMR90 fibroblasts. AIM-1 mRNA levels were plotted after normalization to GAPDH. based on published data using 5Ј-RACE (1). However, more definitive methods (such as primer extension) will be needed to ascertain the transcriptional start site in mammals. This is of particular importance because the medaka AIM-1 mRNA contains an N-terminal region that is not present in the currently defined human and mouse sequences. Furthermore, no inframe stop codon is present prior to the putative initiation ATG for human or mouse, suggesting that the current sequence may not contain the true 5Ј-untranslated region.
Identification of a large E-box containing insertion within the human 5Ј-flanking region did not correlate with binding by the bHLHzip protein MITF, despite the fact that MITF could regulate endogenous AIM-1 expression levels. Thus, MITF likely regulates AIM-1 expression via an indirect mechanism (such as an intermediate transcription factor) or via an enhancer element located at a distance from this region. It is also possible that the reported transcriptional start site is incorrect (as discussed above), in which case a different promoter region may be subject to direct MITF binding and regulation. Nonetheless, AIM-1's position downstream of MITF places it into the company of multiple other genes of known importance in the pigmentation response, several of which are important as melanoma antigens.
Finally, identification of AIM-1 as a rodent pigmentation gene suggests the possibility that human mutations in the same factor might occur. The uw allele displays age-dependent improvement in the eye pigmentation phenotype and complete absence of pigment in the fur (25,26). Interestingly, Minimal Pigment Oculocutaneous Albinism patients (affected gene unknown) display similar phenotypes: no skin or eye pigment at birth and development of detectable eye pigment during the first decade of life (27). In addition, while this article was under review, M. Brilliant and colleagues published the identification of two of the underwhite alleles described in this study and further reported the assignment of the underwhite locus to human oculocutaneous albinism type 4 (OCA4) (28). Thus, underwhite mice appear to provide a model for this or other types of oculocutaneous albinism in humans.