Molecular and Phenotypic Analysis of 25 Recessive, Homozygous-Viable Alleles at the Mouse agouti Locus

Corresponding author: Edward J. Michaud, Oak Ridge National Laboratory, Bldg. 1061, MS 6445, P.O. Box 2008, Oak Ridge, TN 37831-6445., michaudejiii{at}ornl.gov (E-mail)

Communicating editor: N. A. JENKINS

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Agouti is a paracrine-acting, transient antagonist of melanocortin 1 receptors that specifies the subapical band of yellow on otherwise black hairs of the wild-type coat. To better understand both agouti structure/function and the germline damage caused by chemicals and radiation, an allelic series of 25 recessive, homozygous-viable agouti mutations generated in specific-locus tests were characterized. Visual inspection of fur, augmented by quantifiable chemical analysis of hair melanins, suggested four phenotypic categories (mild, moderate, umbrous-like, severe) for the 18 hypomorphs and a single category for the 7 amorphs (null). Molecular analysis indicated protein-coding alterations in 8 hypomorphs and 6 amorphs, with mild-moderate phenotypes correlating with signal peptide or basic domain mutations, and more devastating phenotypes resulting from C-terminal lesions. Ten hypomorphs and one null demonstrated wild-type coding potential, suggesting that they contain mutations elsewhere in the 125-kb agouti locus that either reduce the level or alter the temporal/spatial distribution of agouti transcripts. Beyond the notable contributions to the field of mouse germ cell mutagenesis, analysis of this allelic series illustrates that complete abrogation of agouti function in vivo occurs most often through protein-coding lesions, whereas partial loss of function occurs slightly more frequently at the level of gene expression control.

IN mice, as in many other mammals, the wild-type pigmentation pattern of the fur is called agouti. Individual hairs of an agouti coat are black, with a narrow ring of yellow pigment just below the tip; this alternative synthesis of eumelanin (black-brown pigments) vs. pheomelanin (yellow-red pigments) is regulated in a paracrine manner by the agouti locus in mice (BULTMAN et al. 1992 ). A short pulse of agouti expression at days 3–5 during the hair-growth cycle (BULTMAN et al. 1992 ) leads to a controlled wave of agouti protein secretion from the dermal papillae (MILLAR et al. 1995 ; MATSUNAGA et al. 2000 ). At nearby hair-bulb melanocytes, the agouti protein binds the melanocortin 1 receptor (Mc1r), thereby inhibiting binding by the agonist, -melanocyte stimulating hormone (-MSH; reviewed in DINULESCU and CONE 2000 ). Signaling by -MSH through Mc1r promotes eumelanin synthesis, whereas agouti binding to Mc1r downregulates this signaling cascade and induces a transient switch from eumelanin to pheomelanin synthesis within hair-bulb melanocytes. Thus, recessive mutations at the agouti (a) locus, which impair agouti protein activity or reduce the level of agouti mRNA synthesis, result in a darker, less-yellow coat due to reduced pheomelanin banding of individual hairs. Conversely, dominant mutations, in which deregulated agouti mRNA synthesis leads to greater than normal agouti activity in the skin, give rise to increased yellow pigmentation of the fur. Most dominant agouti mutations produce ectopic agouti in tissues additional to the skin, resulting in a pleiotropic, maturity-onset obesity syndrome that is largely due to constitutive antagonism of related melanocortin receptors (Mcrs) in the central nervous system/hypothalamus (BULTMAN et al. 1992 ; LU et al. 1994 ; MICHAUD et al. 1994A ; HUSZAR et al. 1997 ; OLLMANN et al. 1997 ). Facilitating agouti-mediated signaling through Mcrs in the skin and brain (MILLER et al. 1997 ; GUNN et al. 1999 ; NAGLE et al. 1999 ) are two other genes, mahogany and mahoganoid. Although the cellular function of mahoganoid is currently unknown, recent evidence suggests that mahogany is a low-affinity receptor for agouti on the surface of melanocytes (HE et al. 2001 ).

Five protein structural domains are shared among the highly conserved mammalian homologs of agouti (KWON et al. 1994 ; LEEB et al. 2000 ). Biological function of the various domains has been ascribed largely through analysis of mutations in the 131-amino-acid (aa) mouse agouti protein in vivo and through melanocortin (Mc) binding inhibition assays in vitro. The N-terminal 22 aa comprise a cleavable signal peptide that is indispensable in vivo (HUSTAD et al. 1995 ; PERRY et al. 1996 ), as it provides entry into the secretory pathway (BLOBEL and DOBBERSTEIN 1975 ). The cysteine-rich C terminus (40 aa), which is believed to form a highly ordered, disulfide-bonded structure (PALLAGHY et al. 1994 ), is both necessary and sufficient for Mc-binding inhibition at Mcrs in vitro (WILLARD et al. 1995 ; KIEFER et al. 1997 ), suggesting that the C-terminal domain binds directly to Mcrs in vivo. The mature agouti N terminus (34 aa) contains a conserved and presumably glycosylated arginine that is important for optimal activity in vivo (PERRY et al. 1996 ). Similarly, the central, 29-aa basic domain is also required for full activity in vivo (PERRY et al. 1996 ; MILTENBERGER et al. 1999 ). Although the biological role(s) of these domains remain uncertain, the mature N terminus and/or basic domain may mediate a low-affinity interaction with mahogany on the surface of melanocytes (HE et al. 2001 ) and potentially promote a high-affinity interaction between the C terminus and adjacent Mc1rs. Sandwiched between the basic domain and the cysteine-rich region are several proline residues that contribute mildly to Mcr affinity and selectivity for the agouti protein in vitro (KIEFER et al. 1997 ). The prolines may provide a flexible "hinge" between the cysteine-rich C terminus and the rest of the protein, thereby better accommodating the possibly disparate protein-protein interactions mediated by these two portions of agouti in vivo.

To investigate further agouti structure/function relationships within its native context (i.e., when synthesized from its normal chromosomal location and acting at its normal target, the hair-bulb melanocyte), we have characterized an allelic series of germline-induced, homozygous-viable, and recessive mutations, using both phenotypic and molecular tools. The 25 alleles analyzed here represent a relatively high percentage of all a-locus mutations that have been recovered in the morphological specific-locus test (SLT) at the Oak Ridge National Laboratory (ORNL) over several decades (RUSSELL 1951 , RUSSELL 2002 ). The mutant mice exhibit either hypomorphic (reduced function) or amorphic (no function) phenotypes ranging from extremely mild to null, with 56% (8 of the 18 hypomorphs, 6 of the 7 amorphs) encoding aberrant agouti proteins and 44% (10 hypomorphs, 1 amorph) being apparent regulatory mutations. The severity of associated phenotypes indicated the degree to which agouti activity was inhibited at the cellular level, and the visual interpretation of coat phenotypes was verified with a quantitative, chemical analysis of fur melanins. In addition to these data specific to agouti function, correlating the types of heritable DNA damage with various chemicals and radiation utilized in SLTs significantly augments growing toxicological databases and our current understanding of the mutability of mammalian germ cells.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mice:
All mice originated and were maintained at the Oak Ridge National Laboratory and were fed Purina Laboratory Chow. The C3H/Rl (A/A), 101/Rl (A^w/A^w), and C57BL/10Rl (a/a) strains have been maintained by brother-sister inbreeding for >100 generations. All a-locus alleles analyzed arose in SLTs in which (101/Rl x C3H/Rl)F₁, and rarely (C3H/Rl x 101/Rl)F₁, males or females were treated with a mutagen (or were used as untreated controls; see Table 1) and then bred at selected intervals to noninbred T-stock females that were homozygous for recessive alleles (a, Tyrp1^b, Tyr^c-ch, p, Myo5a^d, Bmp^se, Ednrb^s) at seven visibly marked loci (RUSSELL 1951 ). The a-locus mutations (A^w*, A*) were therefore originally recovered as G₁ progeny (A^w*/a, A*/a) carrying one copy of the severe regulatory allele, nonagouti (BULTMAN et al. 1992 ). Recessive agouti mutations were identified visually on the basis of reduced pheomelanin in the coat compared to nonmutated, agouti-pigmented littermates (A^w/a, A/a). Because nonagouti is dominant to null alleles, G₁ progeny carrying new amorphic mutations resembled homozygous a/a mice (black dorsal and ventral fur except for a few yellow-pigmented hairs on and behind the ears and around the perineum and mammae). In contrast, hypomorphs are dominant to a, and the primary hypomorphic mutants displayed a range of phenotypes that were intermediate between wild-type agouti (A^w/A^w, A/A) and nonagouti. Presumed primary a-locus mutants were allelism tested (by crossing to a^e/a^e) and were subsequently recovered for propagation and additional genetic tests. Mice carrying hypomorphic alleles were outcrossed to C57BL/10Rl for six or more generations to segregate away induced mutations at other loci and then intercrossed to determine homozygous viability and phenotype. Homozygous, hypomorphic phenotypes for each allele were generally less severe (more yellow) than in mice heterozygous for the same allele (A^w*/a, A*/a). Primary mutants that appeared nonagouti were outcrossed to (101/Rl x C3H/Rl)F₁, and 12 homozygous lines (A^w*/A^w*, A*/A*, or the marker a/a) were established, revealing whether the mutation was amorphic or a-like. Amorphic a-locus mutations were identified by having uniformly "jet-black" fur, even in areas where nonagouti mice show some pheomelanin. Because all the alleles selected for this study were homozygous-viable (to exclude multi-locus mutagenic events), all were originally propagated as a-locus homozygous stocks. Despite avoidance of brother-sister matings, the small size of the stocks eventually caused decline in vigor in some of them, and six stocks were subsequently bred to be congenic with C57BL/10Rl, yielding a-locus heterozygotes (Table 1). One of these (a^2R) was successfully restored to homozygous breeding status. Although the 20 a-locus-homozygous stocks do not have identical genetic backgrounds, all of the hypomorphs have considerable contribution from C57BL/10Rl (due to the initial six or more outcrosses), and those propagating amorphs contain 101/Rl and C3H/Rl contributions.

Chemical analysis of melanins:
Hair melanins were analyzed by direct chemical degradation of small hair samples, as described previously (ITO and FUJITA 1985 ; OZEKI et al. 1995 ). The method is based on the formation of the specific degradation products, aminohydroxyphenylalanine (AHP) from pheomelanin and pyrrole-2,3,5-tricarboxylic acid (PTCA) from eumelanin. Ten to 20 mg of hair was homogenized in water (10 mg hair/ml), using a Ten-Brocke glass homogenizer. Hydrolysis of the hair with hydriodic acid produced AHP (yield 20%) that was then quantified by HPLC with electrochemical detection. Permanganate oxidation of hair samples produced PTCA (yield 2%) that was quantified by HPLC with UV detection. Due to the consistent yield of these degradation products, 1 ng AHP/mg hair corresponds to 5 ng pheomelanin/mg hair. Likewise, 1 ng PTCA/mg hair corresponds to 50 ng eumelanin/mg hair. Four to six hair samples were plucked from the dorsal midline of a single animal for each allele. Analysis of variance with a 95% confidence level determined that the variance was not equal; therefore, data were analyzed using a two-sample Student's t-test, assuming equal variance.

RNA preparation:
Total RNA was prepared from the skin of 4-day-old neonatal mice using standard guanidine isothiocyanate procedures (AUSUBEL et al. 1989 ) or the rapid total RNA isolation kit (5 Prime 3 Prime). Skin was obtained from the dorsal and/or ventral body region, excluding the head, legs, and tail. For homozygous mutant mice, up to three neonates were used per RNA preparation. For mutant stocks maintained as heterozygotes, no fewer than five neonates were used since the phenotype of a^R/a and a/a littermates was not readily apparent at day 4. In the case of a^6R, however, a single heterozygous a^6R/a neonate was analyzed from an outcross between a homozygous mutant and C57BL/10Rl mate.

RT-PCR analysis:
Total RNA (2 µg) derived from dorsal and ventral skin was reverse transcribed using random hexamer primers (Pharmacia, Piscataway, NJ) and MMLV-reverse transcriptase (Promega, Madison, WI). The RT-PCR products in Fig 3 were obtained by using 1 µl of ventral skin cDNA and one of two forward primers (exon 1A-specific 5'-CACCAGTCTGAGTCCTTGAGCC-3'; exon 2-specifc 5'-GACGCTTGGAGATGACAGGAGTCTG-3') with a common reverse primer (exon 4-specific 5'-AAACGGCACTGGCAGGAGGC-3'). The entire agouti coding region was amplified by PCR using 1–10 µl of dorsal and vental skin cDNA and primers specific for the very 5' end of exon 2 (5'-GCTTCTCAGGATGGATGTCA-3') and the untranslated region of exon 4 (5'-CAATCACCCGTTCCCGAAGC-3'). The PCR products were cloned into the TA vector (Invitrogen, San Diego) and multiple, individual clones were sequenced for each mutant allele.

View larger version (59K): In this window In a new window Download PPT slide   Figure 1. Coat-color phenotypes of recessive agouti mutations. Representative mice from each mutant category are shown, from the mildest category (mild) at the top to the most severe (null) at the bottom. A wild-type mouse (101/Rl; Aw/Aw) is shown above or to the right of the mutant in each photo. The specific alleles used from each phenotypic category are as follows: mild, a1R; moderate, a11R; umbrous-like, a13R; severe, a17R; null, a21R. All mice were 8 weeks of age and homozygous for the mutant or wild-type alleles.

View larger version (58K): In this window In a new window Download PPT slide   Figure 2. Chemical analysis of hair pigments. AHP (derived from pheomelanin) and PTCA (derived from eumelanin) are indicated by open and solid bars, respectively. Data represent mean AHP or PTCA values (nanograms per millogram of hair) analyzed from a single mouse for each allele. Error bars indicate the variation between separate AHP and PTCA measurements (n 2) from a single chemical degradation experiment. Alleles analyzed are indicated on the y-axis and are grouped by phenotypic category, from mildest at the top to most severe (least amount of pheomelanin) at the bottom. The genotypes and ages of adult mice were as follows: 101/Rl (Aw/Aw), 22 weeks; a1R/a1R, 31 weeks; a2R/a2R, 26 weeks; a3R/a, 21 weeks; a5R/a5R, 25 weeks; a6R/a, 18 weeks; a7R/a, 29 weeks; a8R/a8R, 24 weeks; a9R/a9R, 24 weeks; a11R/a11R, 25 weeks; a12R/a, 18 weeks; a13R/a13R, 22 weeks; a14R/a14R, 18 weeks; a15R/a15R, 26 weeks; a16R/a16R, 27 weeks; a17R/a17R, 28 weeks; a18R/a, 26 weeks; a19R/a19R, 22 weeks; a20R/a20R, 31 weeks; and C57BL/10Rl (a/a), 22 weeks.

View larger version (41K): In this window In a new window Download PPT slide   Figure 3. Schematic of the agouti gene and RT-PCR analysis. (A) Alternative promoters drive expression of multiple agouti isoforms in wild-type skin. The horizontal line represents genomic DNA spanning 125 kb at the agouti locus and the boxes represent exons (not drawn to scale). Noncoding exons (1A, 1A', 1B, 1C, and parts of 2 and 4) are indicated by open boxes. Solid boxes represent protein-coding exons (2–4). In B, primers used for RT-PCR analysis are indicated by arrows (above exons 1A and 2; below exon 4). Two promoter regions (ventral-specific and hair cycle-specific) are separated by 100 kb, as indicated by a break in the line. Each promoter utilizes two sets of alternative 5' exons that splice into the same set of coding exons, as indicated by V-shaped lines joining the exons. In mice carrying either the A (C3H/Rl) or the Aw (101/Rl) allele, the hair cycle-specific promoter utilizes two transcription initiation sites (at 1B or 1C), yielding two transcripts of approximately the same size in both the dorsal and ventral skin. In mice carrying the Aw allele, a single transcription initiation event (at 1A) from the ventral-specific promoter produces two alternative transcripts, which differ by the inclusion or exclusion of exon 1A'. (B) RT-PCR analysis of agouti transcripts in wild-type and representative mutant alleles. Mutant alleles are indicated by their superscript symbol (e.g., 1R for a1R). Two primer sets, indicated by Ex 2–4 or Ex 1A–4 at the bottom, were used to amplify differentially expressed agouti transcripts. All PCR reactions were performed simultaneously from reverse-transcribed RNA prepared from the ventral skin of 4-day-old homozygous neonatal mice. To the left, the arrow labeled A, Aw marks the exon 2–4 product (267 bp) that is present in mutants derived from either a parental A (C3H/Rl) or a Aw (101/Rl) allele. The arrow labeled Aw indicates the exon 1A–4 products (492 or 446 bp) that are unique to Aw-derived alleles. DNA size standards (100-bp ladder) are shown in the first and last lanes, with the 200 and 500 bp bands indicated on the right.

Northern analysis:
Selection of poly(A)⁺ mRNA was performed using the Oligotex mRNA mini kit (QIAGEN, Valencia, CA) and total RNA (250 µg) derived from dorsal neonatal skin. For a^6R, RNA was derived from the whole skin of a single a^6R/a neonate, rather than from the dorsal skin only. A Northern blot made by formaldehyde agarose gel electrophoresis was hybridized with a ³²P-labeled agouti cDNA probe (CHURCH and GILBERT 1984 ; SAMBROOK et al. 1989 ) and exposed to a phosphorimager (Fujix BAS system; Fuji, Stamford, CT) plate overnight to quantify the intensity of the agouti band and then to X-ray film for 3 days. The blot was subsequently hybridized with a low specific-activity probe for Gapdh and exposed to a phosphorimager plate 4 hr for quantification and then to X-ray film overnight.

Genomic DNA analysis:
Genomic DNA was obtained from tail biopsy and Southern analysis was performed using standard procedures (CHURCH and GILBERT 1984 ; SAMBROOK et al. 1989 ). Individual exons of agouti including splice junctions were amplified by PCR using the following primers: exon 2 forward (5'-ATCCCTTACCACCATCTTCT-3') and reverse (5'-AGGAAGCACCATAGCTCTG-3'); exon 3 forward (5'-CTGGCTTTGCTCCTTTCT-3') and reverse (5'-GTTTCCTGGAGGCCAGGAAG-3'); exon 4 forward (5'-AGAGGTCTCTGTCCTGACCT-3') and the reverse primer described above that binds within the untranslated region of exon 4. The entire 5.5-kb genomic region containing all three coding exons was amplified using the Expand Long Template PCR system (Boehringer Mannheim, Indianapolis) and the exon 2 forward and exon 4 reverse primers just described. To amplify the remainder of the 3' untranslated region (310 bp total) that was not included in previous RT-PCR or genomic clones and that includes the poly(A) signal, the forward primer 5'-GCTTCGGGAACGGGTGATTG-3' and reverse primer 5'-TTTCCCTATGCAAGAGTGGC-3' were used. The hair cycle-specific promoter region and 5' untranslated exons 1B and 1C (715 bp total) were amplified using the forward primer 5'-GGAGAGCCGCAGCCTAATCC-3' and reverse primer 5'-CTGAAAGGGAACCATACAGA-3'. All PCR products were then cloned into the TA or TOPO TA vectors (Invitrogen) and multiple individual clones were sequenced.

DNA sequence analysis:
DNA samples from individual clones were sequenced using the ABI dye terminator ready reaction mix (Perkin-Elmer, Norwalk, CT) or the Big Dye terminator sequencing kit (Perkin-Elmer) and an ABI 377 automated DNA sequencer. The DNA sequence was then analyzed using the following software: Sequence Navigator (version 1.0.0; Applied Biosystems, Foster City, CA), Factura (version 1.2.0r6; Applied Biosystems), Auto Assembler (version 1.4.0; Applied Biosystems), and MacVector (version 6.5; Oxford Molecular, Palo Alto, CA).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Phenotypic analysis:
Eighteen hypomorphic alleles were grouped into four phenotypic categories (see below) according to the amount of pheomelanin reduction that is visually apparent in the hair, with primary weight given to the appearance of fur along the dorsal midline. Seven alleles with jet-black fur comprised a separate, fifth group. Individual alleles are listed in Table 1, each designated by its superscript symbol. The order of hypomorphic alleles (a^1R through a^18R), both between and within the four phenotypic categories, corresponds to the severity of coat phenotypes in adult mice of the indicated genotype (note that some are heterozygotes; see MATERIALS AND METHODS). The mutagens employed, year when each allele was originally identified, and original ORNL stock designation are also shown. The germ-cell stage in which each mutation arose is indicated also, on the basis of the interval between mutagen exposure and conception of the mutant and utilizing existing germ-cell development data (OAKBERG 1984 ).

Fig 1 shows representative homozygous mutant mice from each phenotypic group compared with a wild-type agouti mouse in each panel. Mild hypomorphs exhibit a subtle diminution in adult dorsal fur pheomelanin compared to the wild-type agouti strains of origin C3H/Rl (A/A) or 101/Rl (A^w/A^w). Other body regions, such as the sides and ventrum, are not noticeably different from wild type. Moderate hypomorphs comprise the largest phenotype group in which the darkened agouti coat displays reduced pheomelanin in all body regions, although the effect is generally more apparent in dorsal fur. The third category (umbrous-like) is unusual due to preferential darkening of fur along the dorsal midline from nose to tail. In the ventrum and at the dorsal/ventral boundary, however, nearly wild-type levels of yellow-banded fur are apparent, with gradual attenuation in pheomelanin along the sides of the animal toward the dorsal midline. Severe agouti hypomorphs exhibit an obvious reduction in fur pheomelanin in all body regions such that the fur appears very dark but not completely black. The fur of homozygous null agouti mice is entirely black, even in areas where nonagouti mice typically display a few yellow hairs (ears, perineum, and mammae). The single exception in this category is a^19R, which is not a true null, as it displays some yellow pigmentation around the mammae and perineum, but no yellow hairs on or behind the ears (or in any other region).

To augment visual description of these mutant phenotypes and to gain a quantifiable measure of agouti gene loss of function, we directly analyzed the amount of pheomelanin and eumelanin in the fur of recessive agouti alleles. In this relatively simple yet sensitive procedure (ITO and FUJITA 1985 ; OZEKI et al. 1995 ; ITO 1998 ), chemical treatment of small fur samples yields specific degradation products that are readily detectable by HPLC. The amount of AHP isomers and PTCA is directly proportional to the amount of pheomelanin and eumelanin, respectively, in the original fur sample. This type of analysis has been useful in determining even subtle variations in the distribution of hair melanins induced by various coat color mutations in the mouse (GUNN et al. 2001 ; LAMOREUX et al. 2001 ).

Fig 2 shows the chemical analysis of melanins from dorsal midline fur of adult agouti mutant mice. AHP values varied over a 10-fold range and generally correlated with the severity of visual coat phenotypes in the alleles tested, although a few exceptions were noted (a^12R, a^16R, and a^20R). As a group, mild mutations contained about the same level (102%) of AHP in dorsal fur as did the wild-type control (101/Rl); moderate alleles contained 71%; umbrous-like, 46%; severe, 48%; nulls, 13%; and nonagouti (C57BL/10Rl), 9%. Mean AHP levels for individual mutations compared to the wild-type control were significantly lower (P < 0.05) for the darker mutations (nulls, nonagouti, and hypomorphs a^11R, a^13R, a^14R, a^15R, a^16R, a^17R, and a^18R), but not for the more temperate hypomorphs (a^1R, a^2R, a^3R, a^5R a^6R, a^7R a^8R, a^9R, and a^12R). Heterozygous status of five alleles (a^3R/a, a^6R/a, a^7R/a, a^12R/a, and a^18R/a) and mild heterogeneity in the genetic backgrounds of homozygous alleles did not alter the overall trend, nor did age variation among adult mice tested (18–31 weeks) contribute significantly to the variance in mean AHP or PTCA determinations, although age generally enhances the level of visible pheomelanin in the coat. With respect to PTCA levels, significant differences (P < 0.05) were observed for most alleles (except a^19R) compared to either the 101/Rl (wild type) or the C57BL/10Rl (nonagouti black) control; however, the variation was <2-fold and did not follow a consistent trend with respect to the agouti allelic series. Slight heterogeneity in genetic background among the various mutant stocks could explain this low yet significant variation in eumelanin content. Altogether, these data demonstrate a fair (albeit imperfect) correlation between chemically determined pheomelanin content and visual determination of the severity of agouti loss of function in the various alleles, suggesting that this method provides a semiquantitative measure of the effect of agouti mutations on the level of pheomelanin synthesis in vivo.

Molecular analysis:
The 25 recessive alleles were generated in (101/Rl x C3H/Rl)F₁ hybrid mice (see MATERIALS AND METHODS), so the mutations arose on either a C3H/Rl (A) or a 101/Rl (A^w) chromosome. The wild-type, parental alleles A and A^w differentially express four alternatively spliced transcripts (BULTMAN et al. 1992 , BULTMAN et al. 1994 ; VRIELING et al. 1994 ) that encode a common protein sequence (Fig 3A). Whereas both A and A^w express the temporally restricted transcripts that give rise to the characteristic subapical band of pheomelanin in dorsal and ventral hairs, only A^w expresses the ventral-specific transcripts that uniquely generate a predominantly yellow or cream-colored ventrum in A^w-derived alleles. To determine the parental allele of origin for the 25 recessive mutations, and to help explain ventral fur phenotypes among the various alleles, we employed a reverse transcriptase-polymerase chain reaction (RT-PCR) strategy using ventral skin from neonatal mice and two PCR primer sets (Fig 3B). One set of primers (Ex 2–4) amplified the coding region that is common to all transcripts in both A- and A^w-derived alleles. The other set (Ex 1A–4) is unique to ventral-specific transcripts in A^w-derived alleles only and yields a larger-sized product. Findings from this analysis were then confirmed by Southern analysis (CHEN et al. 1996 ) using an Exon 1A'-specific probe (data not shown). As summarized in Table 2, 11 hypomorphic and 4 amorphic alleles arose on the C3H/Rl (A) chromosome, whereas 7 hypomorphs and 3 amorphs were derived from 101/Rl (A^w). Although each of the smaller phenotype groups contained exclusively A- or A^w-derived alleles, the distribution appeared random in the larger phenotypic categories (moderate, null), suggesting no significant correlation between phenotype category and parental allele mutated.

To identify the precise nature of DNA damage that produced the agouti coat phenoytpes, we analyzed the agouti locus at both the genomic and mRNA expression levels. The entire coding region (exons 2–4) of agouti was amplified by RT-PCR from both dorsal and ventral skin of neonatal mice carrying each mutant allele (data not shown; see MATERIALS AND METHODS). The RT-PCR products were cloned, and several clones from each allele were sequenced to identify potential mutations within the agouti protein-coding region. The affected exon(s) were then isolated from genomic DNA of the appropriate mutant by PCR, cloned, and sequenced to verify the initial RT-PCR results. Further independent verification of these sequence alterations was gained through RNase protection assay, using the entire agouti coding region as a probe (data not shown). Table 2 summarizes the results of these molecular analyses.

Of the 25 recessive alleles characterized, 14 alleles with phenotypes ranging from mild to null demonstrated aberrant protein-coding potential. Point mutations and small intragenic deletions were found in several domains of the agouti protein that have been shown previously to be important for agouti function in vivo (HUSTAD et al. 1995 ; PERRY et al. 1996 ; MILTENBERGER et al. 1999 ). Two mutations were identified in the signal peptide (a^8R and a^9R, moderate), four within the basic domain (a^3R, mild; a^10R, moderate; a^12R and a^13R, umbrous-like), and six in the cysteine-rich C terminus (a^1R, mild; a^14R umbrous-like; a^19R, a^20R, a^21R, and a^22R, null). Two additional null alleles (a^23R and a^25R) contained deletions that induced reading-frame shifts and/or premature termination signals such that the protein sequence of the entire basic region and C terminus was either deleted or completely modified. Overall, these results indicate that lesions in the N-terminal and central regions of the agouti protein tend to generate hypomorphic phenotypes, whereas C-terminal alterations generally induce complete loss-of-function phenotypes.

In the remaining 11 alleles, several lines of evidence indicated that they have wild-type protein-coding potential, suggesting that these alleles probably contain mutations that affect proper regulation of agouti expression in the skin. Further sequence analysis was performed to examine the introns and splice junctions between exons 2 and 4, the 3' noncoding sequence, the 5' untranslated exons 1B and 1C, and several hundred base pairs around the hair cycle-specific promoter (Fig 3A). No DNA alterations were found in any of these regions of the 11 putative regulatory alleles, indicating that the lesions most likely lie elsewhere in the 125-kb agouti locus (VRIELING et al. 1994 ). The ventral-specific promoter and its adjacent 5' untranslated exons were not analyzed in any of the mutants because the hypomorphic phenotypes were most evident in dorsal banded fur.

To ascertain whether the level of mRNA expression was altered in any of the mutant mice, Northern analysis was performed (Fig 4). Using Gapdh as a loading control, 5 of the 11 putative regulatory alleles (right side: a^4R, a^5R, and a^6R, moderate; a^16R and a^17R, severe) demonstrated diminished levels of steady-state agouti mRNA expression compared to either of the wild-type controls or to the previously characterized protein-coding mutations (the majority are in the left of Fig 4). However, one protein-coding mutation, a^10R, demonstrated slightly reduced levels of mRNA expression due to a splicing defect. A single-base change at the normal exon 4 splice-acceptor site prompts usage of a surrogate splice acceptor within the exon 4 coding region, which is apparently less efficient, as indicated by a larger RNA product in the a^10R lane. Northern analysis also indicated that 2 (a^11R, moderate; a^15R, umbrous-like) of the 11 putative regulatory alleles express normal to high levels of agouti message, which is wild type in sequence. In the case of alleles maintained as heterozygotes (a^2R, a^3R, a^7R, a^12R, and a^18R), quantification of RNA levels was not possible since a^R/a individuals could not be distinguished from a/a littermates. In addition, RNA analysis was not performed for the a^24R allele due to poor stock viability. However, extensive sequence analysis of genomic DNA from a^24R adults clearly indicated wild-type sequence in the protein-coding region, 3' untranslated region, and 1 kb surrounding the hair cycle-specific promoter and hair cycle-specific 5' untranslated exons.

View larger version (48K): In this window In a new window Download PPT slide   Figure 4. Northern analysis. Each lane contains poly(A)+ mRNA (2.5 µg) prepared from the dorsal skin of 4-day-old neonatal mice, whose superscript allele designation is indicated at the top (e.g., 1R for a1R). Arrows to the right indicate the agouti-specific band and the internal control, Gapdh. Left, gene expression in the C3H/Rl (A/A) control vs. alleles known to carry agouti protein-coding alterations. Right, gene expression in the C3H/Rl (A/A), 101/Rl (Aw/Aw), and C57BL/10Rl (a/a) controls vs. other alleles that carry either protein-coding or regulatory mutations. Homozygous mice were used where available. For the 6R allele, RNA was prepared from a single a6R/a neonate. For all other heterozygous alleles (a2R, a3R, a7R, a12R, a18R), RNA was prepared from an unknown mixture (aR/a and a/a) of five or more neonates.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Agouti structure and function:
One of the most challenging aspects of this study was accurate correlation of the severity of agouti mutant phenotypes with the molecular changes in the locus (summarized in Fig 5A). On the basis of earlier studies (BULTMAN et al. 1991 ; PERRY et al. 1996 ; KIEFER et al. 1997 , KIEFER et al. 1998 ), we expected that the more severe mutations would map to the C-terminal region of agouti that is believed to bind Mcrs. In fact, all null mutations except one (a^24R) altered the coding potential of the C terminus and thus ablated agouti activity at the cellular level, although the a^19R allele was nearly null with respect to ventral pigmentation. The hypomorphic mutations, in contrast, mapped to various regions throughout the agouti protein or appeared to be regulatory in nature and generated a range of phenotypes, some of which were extremely subtle. Assortments of apparent regulatory mutations, point mutations, and small deletions were found in each phenotype group, with the exception of the severe hypomorphs. Overall, the C-terminal mutations were more devastating than the basic-region mutations, and the signal-peptide lesions were comparatively modest. Exceptions to these general rules are noted, however (a^1R, a^3R, a^10R). Among the putative regulatory mutations, five alleles reflected lower mRNA levels in skin relative to the wild-type controls, although the absolute decrease in mRNA expression did not necessarily correlate with severity of the phenotype (e.g., a^4R, a^5R vs. a^16R, a^17R). On the basis of variation among the homozygous coding-region mutations and the wild-type controls, mild heterogeneity in genetic backgrounds accounts for no >20% fluctuation in mRNA levels. These data suggest that potentially significant differences in agouti transcriptional/translational control are likely to play a major role both temporally and spatially in determining how mutant phenotypes correlate with agouti mRNA expression patterns in some of the regulatory alleles. Further analysis using in situ hybridization of the skin and hair follicles and perhaps immunolocalization of the agouti protein should clarify the basis for many of the subtle hypomorphic phenotypes.

View larger version (31K): In this window In a new window Download PPT slide   Figure 5. Summary of molecular analysis. (A) Schematic of the 125-kb agouti locus. The thick horizontal line represents the genomic DNA, with solid boxes representing protein-coding exons, open boxes representing noncoding exons (not drawn to scale), and arrows above the noncoding exons representing transcriptional start sites. Two promoter regions are separated by 100 kb, as indicated by a break in the line. Location of regulatory elements is postulated by the double-headed arrows above the gene, both 5' (BULTMAN et al. 1992 , BULTMAN et al. 1994 ; VRIELING et al. 1994 ) and 3' (MILLER et al. 1994 ) to the coding exons. Specific alleles believed to carry regulatory mutations are indicated at left, beneath the transcriptional promoters. Allele designations (e.g., 2R for a2R) for the presumed regulatory mutations are aligned in columns that refer to the phenotype categories, from the mildest at the left to the most severe at the right. Dotted lines extending downward from the coding exons show an expanded view of the agouti protein, with the various structural domains indicated. The positions of hypomorphic (1R–14R) and amorphic (19R–25R) coding mutations are indicated below the protein schematic. The position of the 25R mutation also corresponds to the exon 2/3 boundary, and the 10R mutation corresponds to the exon 3/4 boundary. (B) Alignment of the cysteine-rich C-terminal sequence of human AGRP and mouse agouti proteins. Amino acids 87–132 of AGRP and 92–131 of agouti are shown. Conserved cysteines are highlighted in yellow, and the Mcr binding triplet, RFF, is shown in red. The pattern of disulfide bonding is indicated by brackets, with linkages 1–3 forming the "core" of the three-dimensional structure, as determined by BOLIN et al. 1999 , while bonds 4 and 5 appear to present and stabilize the RFF-containing loop. Asterisks below mouse agouti residues Cys110, Ser112, Cys115, and Arg118 correspond to the sites of mutations in alleles 20R, 14R, 19R, and 21–22R, respectively.

Visual phenotype analysis: In addition to agouti gene structure and expression patterns, other factors influence the appearance of agouti coat phenotypes. The expression level of wild-type agouti is determined not only by the degree of Mcr antagonism in vivo, but also by the dosage of partially functional agouti proteins. For example, mice homozygous for any given hypomorphic allele generally exhibited milder coat phenotypes (slightly more pheomelanogenic) than were apparent in mice heterozygous (a^R/a) for the same allele. Factors acting independently of agouti also influence agouti coat phenotypes. Chief among these is the effect of age on the spatial distribution of pheomelanin along the dorsal/ventral axis. These age-related changes are most striking between weaning and adulthood (3–12 weeks) and are particularly evident in mice carrying hypomorphic alleles. For example, those in the mild category appear nearly wild type as adults, but have a darker, umbrous dorsum at weaning age. Mice with more severe hypomorphic mutations (severe or umbrous-like as adults) typically exhibit nonagouti fur over the entire dorsum and agouti fur in the ventrum at weaning age. Reduced pheomelanin-banded hairs become apparent in adults on the sides and in the dorsum, depending on the phenotypic category. In contrast, amorphic mutations show black fur in both young and adult mice. For these reasons, we deliberately chose to analyze adult fur phenotypes (and primarily dorsal fur) rather than juvenile phenotypes, as the former were generally more stable and better represented each allele. Investigators in the past have consistently observed that pheomelanin is lost from dorsal fur prior to its loss from ventral fur as one progresses down the agouti dominance hierarchy (SILVERS 1979 ). Moreover, this effect is not entirely explained by the differential expression of ventral-specific agouti transcripts (BULTMAN et al. 1994 ; VRIELING et al. 1994 ), as equivalent expression of agouti variants in the dorsum and ventrum of transgenic mice consistently generates a more pheomelanic ventrum (PERRY et al. 1995 , PERRY et al. 1996 ; MILTENBERGER et al. 1999 ). Our observation that the dorsal midline is most affected by pheomelanin loss resulting from a-locus mutation further suggests that this is the region in which melanocytes are most susceptible to mild variations in agouti activity. One possibility is that hair-bulb melanocytes along the dorsal/ventral axis may express variable densities of Mcrs and/or the mahogany protein on their cell surfaces. The effective threshold concentration of agouti required by the melanocyte to switch from eumelanin to pheomelanin synthesis would be greater or more stringent in the dorsum than in the ventrum. Collectively these observations indicate that, in addition to the quality and quantity of agouti gene activity, independent and possibly dynamic factors within the hair follicle microenvironment contribute to agouti coat phenotypes.

Chemical analysis of melanins: The chemical method utilized for directly quantifying pheomelanin and eumelanin content of the fur allowed an objective interpretation of agouti loss of function in the various alleles. In general agreement with the visibly scored phenotypes, the amounts of the pheomelanin-specific degradation product AHP steadily declined over an 10-fold range as one progresses down the allelic series analyzed here. However, because absolute AHP levels were not always in perfect agreement with the perceived severity of some agouti mutations, the visual phenotype may be a more sensitive discriminator of agouti activity in these cases than the chemical analysis. Seemingly random variation within a <2-fold range was observed for the eumelanin-derived degradation product PTCA in dorsal fur, indicating that, as anticipated, the primary effect of agouti mutation was at the level of pheomelanin, not eumelanin, synthesis. A surprising result from this analysis, however, is that the range of adult agouti phenotypes from wild type to jet black actually reflects only a minor change (2.3%) in the total melanin distribution. By converting dorsal fur AHP and PTCA values to pheomelanin and eumelanin content, respectively (ITO and FUJITA 1985 ), dorsal midline fur of wild-type agouti mice contains 97.3% eumelanin and 2.7% pheomelanin. Moderate hypomorphic alleles, such as a^11R, contain half as much pheomelanin or 1.4% (and 98.6% eumelanin), and visibly black fur, as in the amorphic allele a^20R, contains only 0.4% pheomelanin (and 99.6% eumelanin). The reflective qualities of pheomelanin granules and the manner in which individual hairs overlap along the coat probably augment these small differences in pheomelanin content such that even subtle changes in band width or pigment intensity are maximally exposed and readily detectable by the unaided eye.

Regulatory mutations: Using several levels of molecular analysis, we determined that 11 of the 25 agouti alleles (10 hypomorphs, 1 amorph) contained wild-type protein-coding sequence, suggesting that regulatory mutations somewhere in the locus interfere with proper and efficient temporal/spatial expression of the wild-type gene products in these mice. The precise genetic alterations remain unidentified, however, as very little is currently known about the elements that contribute to elaborate transcriptional control within the large agouti locus. Five of the noncoding hypomorphs that exhibited moderate (a^4R, a^5R, a^6R) to severe (a^16R, a^17R) phenotypes expressed reduced levels of agouti mRNA in neonatal skin compared to wild-type controls. This finding probably accounts for their hypomorphic status, since a strong correlation has been established previously between the level of steady-state RNA synthesis in dominant agouti alleles and the amount of pheomelanin in the coat (DUHL et al. 1994 ; MICHAUD et al. 1994B ; YEN et al. 1994 ; HUSTAD et al. 1995 ; KLEBIG et al. 1995 ; ZEMEL et al. 1995 ; ARGESON et al. 1996 ; MILTENBERGER et al. 1999 ). Although, as discussed above, imperfect correlation was observed between the absolute level of agouti expression and the severity of phenotypes in some alleles, our data nonetheless suggest that approximately threefold or greater decrease in agouti expression levels are sufficient to induce hypomorphic phenotypes in vivo. This finding probably reflects the delicate threshold of agouti activity required for effective Mcr antagonism.

Two hypomorphic alleles for which no DNA alterations were identified (a^11R, moderate; a^15R, umbrous-like) may represent a different class of regulatory mutation, as these mice expressed wild-type agouti message at normal to high levels in neonatal skin. Although the timing of agouti expression (day 4) was apparently normal in these mice, mutational alterations in spatial control elements could misdirect agouti expression to anomalous cell type(s) in the skin, leading to less optimal diffusion and localization of the wild-type agouti protein. Normally, the agouti protein is expressed by specialized cells in the dermal papillae (MILLAR et al. 1995 ) and has a limited diffusion distance in vivo, as suggested by skin transplantation (SILVERS and RUSSELL 1955 ) and transgenic experiments (KUCERA et al. 1996 ). In situ hybridization or immunolocalization studies will be needed to determine if agouti is indeed mislocalized in the a^11R and a^15R alleles. Three other hypomorphs (a^2R, mild; a^7R, moderate; a^18R, severe) and the amorphic allele a^24R also exhibited no coding alterations and likewise may have acquired devastating lesions in critical regulatory element(s) that either reduce/eliminate agouti expression or sufficiently relocate its site of synthesis in vivo such that subthreshold levels of functional agouti protein reach hair-bulb melanocytes. As more data become available from mouse genome sequencing efforts, previously uncharacterized regions of the 125-kb agouti locus should become more amenable to mutation screening, and the lesions in these 11 noncoding alleles more readily identifiable. In the meantime, these alleles will provide a rich mutant resource for future investigation into the poorly understood, yet complex, arena of agouti transcriptional control.

Signal peptide mutations: Two hypomorphic mutations (a^8R, a^9R) with moderate coat phenotypes were found to alter the coding potential of the agouti signal peptide. The a^8R allele carries an in-frame deletion of 9 bp that eliminates Phe₁₄-Cys₁₆; the slightly more severe a^9R allele contains a single missense mutation (T A) that substitutes Asp for the conserved Val₃. Requirement for an intact signal peptide in vivo has been demonstrated previously (HUSTAD et al. 1995 ; PERRY et al. 1996 ), in keeping with early seminal studies that indicated a paracrine role for agouti in pigmentation (SILVERS and RUSSELL 1955 ). Analyses by two neural networks, TargetP (EMANUELSSON et al. 2000 ) and SignalP (NIELSEN et al. 1997 ), provide clues to the cell biological defects caused by these mutations. Reduced hydrophobic-core length in the a^8R signal peptide and the net negative charge at the a^9R N terminus may interfere with targeting of the nascent preproteins to the endoplasmic reticulum (ER; TargetP score: wt 0.961, a^8R 0.900, a^9R 0.955). In addition, reduced efficiency of signal peptide processing at an alternative site (Ala₂₂) in a^8R and at the wild-type site (Ser₂₂) in a^9R may also play a role (SignalP Y score: wt 0.743, a^8R 0.662, a^9R 0.708). Although the functional impact of an altered N terminus (HisLeuAla) in the mature a^8R protein is not known, reduced ER targeting efficiency/fidelity of the a^8R preprotein is probably sufficient to induce a loss-of-function phenotype in vivo. Potential ramifications include reduced secretion rate and delayed diffusion of the mature protein to target cells. Because the a^9R allele is predicted to generate a wild-type mature protein, these analyses also suggest that reduced quantity, not quality, of mature secreted a^9R protein is responsible for reduced agouti function in these mice.

Basic region mutations: All coding-region mutations were expressed at approximately wild-type levels, with the exception of the hypomorphic allele a^10R. This allele contained a single base change at the splice-acceptor site in exon 4, prompting usage of an internal AG in the fourth exon that is apparently less efficient than the wild-type site. Abnormal splicing in a^10R provides an in-frame message for translation and simultaneously deletes two lysine residues (Lys₇₆Lys₇₇) in the basic domain. Although the approximately threefold reduction in mRNA levels may contribute to the mutant phenotype in a^10R, deletion of the two basic residues is probably of greater significance, since the regulatory mutations, such as a^4R and a^5R, exhibit greater reductions in mRNA levels yet display a less severe coat phenotype.

The central basic domain of agouti is highly conserved and important for activity in vivo, yet its precise biological role is poorly defined (PERRY et al. 1996 ; MILTENBERGER et al. 1999 ). More than one-half of the 29 aa in this domain are basic and often arranged in pairs, suggesting that endoproteolytic processing may generate smaller functional peptides in vivo. Western analysis of agouti in mouse tissues has since eliminated this early hypothesis (OLLMANN and BARSH 1999 ), leaving alternative models, such as the following: (1) interaction with mahogany and/or mahoganoid (DINULESCU and CONE 2000 ; GUNN and BARSH 2000 ; HE et al. 2001 ); (2) facilitation of intracellular trafficking/biogenesis of the mature agouti protein; (3) promoting diffusion to cellular targets following secretion; and/or (4) direct interaction with Mc1r (VIRADOR et al. 2000 ). A common theme that emerged from characterizing the three point mutations (a^3R, a^12R, a^13R) and one small deletion (a^10R) that map to the agouti basic region is that net charge may be important for biological activity. In wild-type agouti, 16 basic (12 Lys, 4 Arg) and 2 acidic (Glu) residues generate a net +14 charge over the middle of the protein. Deletion of Lys₇₆Lys₇₇ in a^10R reduces the net positive charge to +12, generating a moderate dark-agouti phenotype. A single base change (G A) in the mild allele, a^3R, converts 1 of the 2 acidic residues (Glu₆₆) to a basic aa (Lys), thereby increasing the net charge to +16. The second acidic aa (Glu₆₈), which is absolutely conserved in all homologs, was also modified to Lys in two more severe alleles, a^12R (CG AA) and a^13R (G A). Comparison of the phenotypes of a^3R (mild) vs. a^12R and a^13R (umbrous-like) suggests that the position of charged residues, rather than simply the net charge, influences agouti activity. Although the biological role of the basic domain remains unclear, the four mutations identified here could serve as useful tools for testing various hypotheses. Potential protein-protein interactions could be directly addressed in vivo by epistasis or double-mutant studies. In addition, although wild-type agouti is not processed beyond the signal peptide in vivo, the basic region mutations identified here could introduce new cleavage site(s) for proteolytic convertases, thereby yielding aberrantly processed forms of agouti in vivo.

C-terminal mutations: The final group of coding-region mutations maps to the C terminus of agouti, with most alleles being null except for a^1R and a^14R. Spacing of the 10 cysteine residues in the C terminus is absolutely conserved in all agouti homologs, as are 9 out of the 10 cysteines in the C terminus of the agouti-related protein (AGRP, Fig 5B) that naturally antagonizes Mc3r and Mc4r in the central nervous system (HUSZAR et al. 1997 ; SHUTTER et al. 1997 ). This arrangement of cysteines bears resemblance to that of the disulfide-bonded -conotoxins and -agatoxins (OLIVERA et al. 1994 ; PALLAGHY et al. 1994 ), suggesting that agouti and AGRP fold into an "inhibitor cysteine-knot" motif that presents the conserved residues Arg₁₁₆Phe₁₁₇Phe₁₁₈ (Fig 5B) that mediate direct interaction with Mcrs (KIEFER et al. 1998 ). Indeed, both mouse agouti (WILLARD et al. 1995 ) and AGRP (BURES et al. 1998 ) disulfide bond in the same pattern found in the -agatoxins; however, agouti and AGRP both contain an additional disulfide linkage (disulfide 5 in Fig 5B). Based on NMR studies of a chemically synthesized and biologically active form of AGRP, three disulfide bridges (bonds 1–3 in Fig 5B corresponding to Cys₉₂-Cys₁₀₇, Cys₉₉-Cys₁₁₃, and Cys₁₀₆-Cys₁₂₄ in mouse agouti) build the core of the C-terminal domain structure (BOLIN et al. 1999 ). These core linkages are located at the base of the structure to anchor the bottoms of loops that present Mcr-interacting motifs. The fourth disulfide linkage in AGRP, which is analogous to Cys₁₁₅-Cys₁₂₂ in mouse agouti (disulfide 4 in Fig 5B), appears to stabilize the relatively rigid "active loop" that contains the Arg₁₁₆Phe₁₁₇Phe₁₁₈ binding determinants. The fifth disulfide (Cys₁₁₀-Cys₁₃₁ in mouse agouti; bond 5 in Fig 5B), which is not present in the -agatoxins, is highly reducible in vitro (BURES et al. 1998 ), suggesting that it may be more exposed on the exterior of the three-dimensional structure.

Using transgenic mice, PERRY et al. 1996 determined that individual cysteines are indeed critical for the activity of mouse agouti in vivo. Individual cysteines were changed to serine, the mutant cDNAs were expressed under the control of the ß-actin promoter, and founder mice were assayed for yellow-pigmented fur and maturity-onset obesity (PERRY et al. 1996 ). Mutation of four of the five disulfide bonding pairs (Cys₉₂-Cys₁₀₇, Cys₉₉-Cys₁₁₃, Cys₁₀₆-Cys₁₂₄, and Cys₁₁₅-Cys₁₂₂; bonds 1–4 in Fig 5B) completely abolished activity. In contrast, mutation at Cys₁₁₀ or Cys₁₃₁, which together form the fifth disulfide connection, resulted in partial loss of function, particularly a tendency toward black pigmentation in the dorsum and yellow pigmentation in the ventrum (PERRY et al. 1996 ). Interestingly, this disulfide bond, which has no analog in the -agatoxins, forms the most flexible peptide loop (BOLIN et al. 1999 ) and was found to be more labile to reduction in vitro (BURES et al. 1998 ). Collectively, the in vitro and in vivo data suggest that the loop formed by Cys₁₁₀-Cys₁₃₁ (disulfide 5 in Fig 5B) probably does not form the core structure of the agouti C terminus as originally suggested by BOLIN et al. 1999 , but instead serves a secondary role such as stabilizing or facilitating presentation of the Arg₁₁₆-Phe₁₁₇Phe₁₁₈-containing loop.

In the present investigations, two alleles were found to disrupt C-terminal cysteine residues (a^20R Cys₁₁₀ stop; a^19R Cys₁₁₅ Ser); however, only the former was a true null. In a^20R, a C A change at Cys₁₁₀ introduces a stop codon, indicating no translation of the Mcr-binding determinants and elimination of three disulfide bonds, hence the null phenotype. The T A mutation in a^19R, however, results in a nearly null phenotype in which residual agouti activity is detectable as pheomelanin around the mammae and perineum, but not on or behind the ears, thus placing this allele between nonagouti and true nulls in the agouti dominance hierarchy. The a^19R mutation prevents formation of the fourth disulfide pair (Cys₁₁₅-Cys₁₂₂) that in AGRP stabilizes the Arg₁₁₆Phe₁₁₇Phe₁₁₈-containing loop, but does not actually form the core of the C-terminal structure (BOLIN et al. 1999 ). The nearly null phenotype of a^19R mice is consistent with these structural data. Furthermore, this phenotype suggests that the ventral hair follicle microenvironment around the mammae and perineum is extremely sensitive to minimal levels of agouti activity. Surprisingly, however, this allele is analogous to the Cys₁₁₅Ser mutation studied by PERRY et al. 1996 in which the transgenic mice exhibited no agouti activity whatsoever. Discrepancy between the phenotype of a^19R mice with the findings of PERRY et al. 1996 may be explained by their use of transgenic founder animals that may not uniformly express the transgene in all cell types.

Two other null mutations were shown to affect the putative Mcr-binding determinant at Phe₁₁₈ (a^21R Phe₁₁₈-Ala₁₂₁; a^22R Phe₁₁₈ Ser). The a^21R mutation deletes the last aa in the triplet binding determinant for Mcrs, in addition to two other C-terminal aa, thus inducing a reading-frame shift that eliminates the disulfide bonding partners for three of the five disulfide connections. Because the a^22R allele is a single missense mutation at Phe₁₁₈, it provides the first direct evidence that this aromatic residue is indeed critical for agouti activity in vivo. An alternative explanation for the eumelanic phenotype of a^22R, in particular, is that the point mutation of Phe₁₁₈ may convert agouti into a potent agonist of Mc1r (OLLMANN et al. 1998 ), rather than into a crippled antagonist. Although our genetic tests show no indication that a^22R behaves differently from other jet-black agouti mutations analyzed here (i.e., A/ a^22R looks identical to other null alleles balanced with A or A^w), this hypothesis could be formally addressed by direct Mc1r binding studies with labeled mutant agouti protein in vitro.

Considering the complex structural constraints guiding the successful folding needed for building a functional agouti C terminus, it is not surprising that deletions that alter the entire coding potential of this region also induced amorphic phenotypes. In a^23R, a 3-bp deletion/2-bp insertion event at the end of exon 2 introduces six missense aa at positions 49–54, followed by premature termination just inside exon 3, thereby eliminating the entire basic and C-terminal domains. A 359-bp deletion was identified in a^25R, encompassing all of exon 3 and resulting in direct splicing of exon 2 to exon 4. A similar effect on the agouti mRNA has been described for the a^5MNURg allele that resulted from deletion of a larger genomic fragment (5 kb) in this vicinity (BULTMAN et al. 1991 ).

Surprisingly, the mildest allele characterized in this study also carries a lesion within the agouti C terminus. A single base change at the stop codon (G T) in a^1R results in an additional 27 aa at the very C terminus of an otherwise normal agouti protein, producing apparently little impact on normal protein folding or on the interactions with Mc1r. This finding is consistent with the observation that the C-terminal end of AGRP exhibits very little structure in solution and is not involved in stabilizing or presenting the Arg₁₁₆Phe₁₁₇Phe₁₁₈-containing loop (BOLIN et al. 1999 ). In contrast, the other hypomorphic mutation in the agouti C terminus (a^14R, T C) converts Ser₁₁₂ Pro and induces a more severe, umbrous-like phenotype. Due to the helix-disrupting or turn potential of prolines and the proximity of this mutation to the closely spaced, conserved cysteines (Fig 5B), this mutation may alter the pairing of one or more disulfide bonds. Furthermore, lower rates of agouti secretion could occur in a^14R due to ER retention and retrograde translocation of improperly folded proteins to the proteasome (WIERTZ et al. 1996 ).

Toxicology:
In addition to new insights into the structure and function of agouti, analysis of these multiple a-locus alleles provides useful toxicological data for better understanding the molecular mechanisms by which the various mutagens impact the mammalian germline. Analysis of SLT mutations at seven loci in the mouse have shown that the germ-cell stage is the chief determinant of whether mutational lesions are multi or single locus (RUSSELL 1991 , RUSSELL 2002 ). The proportion of the latter type, which is generally compatible with homozygous viability, is considerably higher for mutations induced in spermatogonia than for those induced in postspermatogonial stages or oocytes. For the present investigations, only homozygous viable a-locus alleles were selected, and thus it is not surprising that 21 of the 23 (91%) that originated in mutagenized germ cells (a^4R and a^22R were spontaneous) were induced in spermatogonia. The remaining 2 were induced by X rays, 1 in oocytes (a^2R) and 1 in early spermatids (a^8R).

Overall, differences were observed among the various treatment groups in the ratio of coding vs. regulatory mutations. Of eight mutations induced by chemicals in spermatogonia, six (86%) were coding-region alleles, and these were point mutations, not deletions. By contrast, fewer than one-half (6 out of 15 or 40%) of mutations induced by radiations in spermatogonia were coding-region alleles, with 3 of the 6 being deletions. Of the two mutations not induced in spermatogonial stages, one occurred in the protein-coding region. Other information derived from SLT-induced mutations (RUSSELL 2002 ) indicated that mutational lesions induced in postspermatogonial stages are more extensive than those induced in spermatogonia and that, among mutations induced in spermatogonia, those induced by radiations are more extensive than those induced by chemicals. The proportion of a-locus alleles that are not in the coding region follows the same relative order. It might therefore be suggested that these alleles may turn out to include a relatively small percentage of single-base-pair changes.

Of the five mutations generated with N-ethyl-N-nitrosourea (ENU) in spermatogonial stem cells, only one was an uncharacterized regulatory mutation (a^15R), and four were single-base substitutions. The A/T T/A transversions in a^9R and a^19R and the A/T G/C transition in a^14R typify 80% of the ENU-induced mouse germ-cell mutations identified in other studies (NOVEROSKE et al. 2000 ). The fourth ENU mutation (a^20R), a G/C A/T transversion, is a type that had been found only rarely in the past. The N-methyl-N-nitrosourea (MNU)-induced allele, a^3R, was shown to carry a G/C A/T transition, as is characteristic of this methylating agent in mouse germ cells (VOGEL and NATARAJAN 1995 ; FAVOR 1999 ). Although the mechanism by which ethylene oxide induces mutations in the mammalian germline remains unclear (DELLARCO et al. 1990 ), the A/T T/A mutation identified in a^10R is consistent with N-alkylation damage (VOGEL and NATARAJAN 1995 ). Finally, synthetic fuel induced an unidentified regulatory mutation in a^7R; however, this lesion could be a deletion like the previously characterized Bmp5^3SYNg allele that was also generated with synthetic fuel in SLTs (MARKER et al. 1997 ).

Of the seven loci screened in SLTs, the a locus is one of the two least mutable for either radiation or chemical treatment, representing only 0–7% of all mutants recovered (RUSSELL 2002 ). Spontaneous mutations arising in mitotic stages of gametogenesis are also extremely rare at the a locus. In fact, among the 66 such mutations reported for SLTs by three laboratories, none has been at the a locus (RUSSELL and RUSSELL 1996 ). On the other hand, among spontaneous mutations arising during the perigametic interval (RUSSELL and RUSSELL 1996 ; RUSSELL 1999 ) over one-third involved the a locus. Two of these perigametic a-locus mutations were analyzed in the present study. One (a^22R) was determined to be a single-base-pair change of a type (A/T G/C) that has been reported to be common after ENU mutagenesis (see above); the other (a^4R) was found to be outside the coding regions, and the DNA alteration remains to be identified.

Concluding remarks:
The extensive and diverse set of recessive, homozygous-viable, and intragenic mutations in the agouti gene generated in SLTs at ORNL has permitted us to perform detailed structure/function analysis of the agouti protein. As was previously shown for other SLT loci (Tyrp1, SARANGARAJAN et al. 2000 ; Tyr, YOKOYAMA et al. 1990 and RINCHIK et al. 1993 ; Myo5a, HUANG et al. 1998