The agouti (a ) gene normally functions to control the differential production of melanin pigments in the skin that gives rise to the true wild-type coat color of mice. The gene derives its name from the South American mammals, Agouti paca and Agouti taczanowskii, which have the same grizzled coat pigmentation pattern as that conferred by the agouti gene in mice. Unlike most genes that influence coat color, agouti does not function directly within the melanocyte. Instead, agouti acts in a non-cell-autonomous manner (as a paracrine factor), as first demonstrated by a series of classic skin transplantation experiments (Silvers and Russell 1955).

Because of its role in regulating coat color, agouti has served as an important model of gene action and interaction for nearly a century. Over 25 different dominant and recessive agouti locus alleles have been identified (Green 1989, Silvers 1979, Siracusa 1991), in which phaeomelanin (yellow-red pigment) synthesis is generally dominant over eumelanin (black-brown pigment) synthesis. At the top of the dominance hierarchy and perhaps the most notable of all agouti alleles are the lethal yellow (A^y) and viable yellow (A^vy) mutations that develop a dominant pleiotropic syndrome including obesity, insulin resistance, increased linear growth, increased susceptibility to certain neoplasms (reviewed in Wolff 1987), as well as yellow fur (reviewed in Bray and York 1979, Silvers 1979, Wolff et al. 1986). These "yellow obese" mice have served as useful experimental models for obesity and diabetes research (Bray and York 1979, Fan et al. 1997, Huszar et al. 1997, Mynatt et al. 1997, Wolff et al. 1986, Yen et al. 1994).

AGOUTI-INDUCED OBESITY

Body mass in both lean and obese individuals is generally determined by the balance between energy intake and expenditure, which are controlled by the central nervous system (reviewed in Bray and York 1979, Spiegelman and Flier 1996, Weigle and Kuijper 1996). Excessive food intake promotes progressive weight gain if it is not accompanied by a compensating increase in energy expenditure. This is largely what happens in animals that carry the Lep^ob or Lepr^db mutations (reviewed in Spiegelman and Flier 1996); both of these mutants exhibit an uncontrolled feeding behavior, a positive energy balance and early-onset obesity. Conversely, although the yellow agouti mutants have a stronger than normal motivation to consume food [they eat 10-36% more than their lean littermates (Frigeri et al. 1988, Yen et al. 1976, Yen et al. 1984)], their satiety mechanisms remain intact (Bray and York 1979). Also, neither the moderate hyperphagia nor the decreased thermogenesis (Yen et al. 1984) observed in the A^y and A^vy mutants can account for the obesity (reviewed in Yen et al. 1994). Instead, it has been proposed that a major determinant of obesity in the yellow obese mutants is the enhanced efficiency with which they utilize calories (Yen et al. 1994). In other words, the dominant yellow agouti mutants seem more proficient at storing their consumed calories as fat rather than utilizing those calories for physical activity or for maintaining body heat. Measurement of mean body weight gain vs. number of calories consumed by female yellow A^vy mice was used to verify a three- to fourfold higher caloric efficiency (Frigeri et al. 1988). This level of increased caloric efficiency is consistent with the degree of the adiposity observed in the mutant animals (Yen et al. 1976).

Hyperinsulinemia is evident in A^vy mice at ~6 wk of age (compared with 10 and 15 d for Lepr^db and Lep^ob mutant mice, respectively) (Frigeri et al. 1983), and can become as high as 20-fold over lean controls by 6 mo of age (Gill and Yen 1991). Because insulin promotes nutrient partitioning into adipose tissue and stimulates adipocyte growth and development (reviewed in Bray 1996), the hyperinsulinemia in the yellow agouti mutants may contribute to their obesity and possibly to other traits of the pleiotropic syndrome (Wolff et al. 1986). A positive temporal relationship has been established between the hyperinsulinemia and the activity of hepatic lipogenic enzymes in yellow A^vy/a mice (Yen et al. 1976), and pancreatic -cell hyperplasia is evident in A^vy males at 21 d of age before any detectable weight gain or changes in insulin or glucagon levels (Warbritton et al. 1994). However, the relationships among hyperinsulinemia, insulin resistance and obesity are complex; currently, it is not possible to attribute the obesity in the yellow agouti mutants to the hyperinsulinemia alone (reviewed in Yen et al. 1994).

Endocrine abnormalities are commonly observed in many of the different mouse lines exhibiting obesity, with changes in adrenal corticoids being particularly noteworthy (reviewed in Bray 1996). Adrenal corticoid levels are elevated in Lep^ob and Lepr^db homozygotes compared with their lean controls (Coleman and Burkart 1977, Dubuc et al. 1975) but remain unchanged in the yellow agouti mutants (Wolff and Flack 1971). Nevertheless, the adrenals are necessary for the full expression of the obesity syndrome in the yellow agouti mutants. Adrenalectomy normalizes hyperglycemia in A^vy/a mice (Shimizu 1989) and reduces fat deposition in both yellow and lean agouti mice, but does not completely prevent the relative obesity of yellow vs. lean littermates (Jackson et al. 1976).

In addition, the pituitary gland is required for complete manifestation of the yellow obese syndrome. Hypophysectomy offsets the hyperinsulinemia that normally develops in A^vy mice (Salem et al. 1989) but only reduces and does not prevent the excess fat deposition or the unique "anabolic" effects of the dominant yellow agouti mutations (Plocher and Powley 1976, Salem et al. 1989). Yellow obese mice typically exhibit mild increases (~10%) in skeletal length, lean muscle mass and fat-free dry weight, even in castrated males that lack endogenous testosterone (reviewed in Wolff et al. 1986). These observations have led to the hypothesis that agouti may somehow mimic the effects of growth hormone. Interestingly, the Lep^ob/Lep^ob mice exhibit just the opposite effect on skeletal growth (Heston and Vlahakis 1962), and it has been shown that the obese (fa/fa) Zucker rat expresses reduced levels of both growth hormone and growth hormone-releasing hormone (Ahmad et al. 1993 and 1989). A causal relationship between growth hormone and obesity was ruled out over two decades ago when it was demonstrated that introducing a genetic deficiency of growth hormone does not prevent the excess adiposity and relative weight gain in A^y mutants (Wolff 1965).

Parabiosis experiments have been particularly useful in dissecting the hormonal contributions to obesity in several genetic models (reviewed in Yen et al. 1994). Surgical union of obese A^y/a and nonobese a/a mice had no effect on either partner's body weight or fat content, suggesting that circulating hormones are not directly involved in the development of agouti-induced obesity (Wolff 1963). Because the agouti protein is normally secreted from the cell (see below), the results of the parabiosis experiments may simply indicate that the agouti protein is not sufficiently stable to circulate between parabiotic partners, or that agouti acts in a localized manner.

MOLECULAR CHARACTERIZATION OF THE AGOUTI GENE AS IT RELATES TO OBESITY

A full length agouti cDNA was cloned from a neonatal skin cDNA library of A/A (C3H inbred strain) mice, and its pattern of expression was analyzed (Bultman et al. 1992). By Northern blot analysis, the agouti messenger RNA is not expressed in adult tissues (except testis), but is expressed in neonatal skin in a manner that correlates nicely with its role in pigmentation. In situ hybridization was used to determine that cells of the dermal papilla (just below the hair bulb) are the major site of agouti expression, with minor levels of agouti messenger RNA detectable throughout the epidermis as well (Millar et al. 1995). These data corroborate the earlier findings from tissue recombination experiments that suggested the agouti signal originates from the dermis, rather than the epidermis, for most agouti alleles (Mayer and Fihsbane 1972, Poole 1974 and 1975, Poole and Silvers 1976).

The primary amino acid sequence of the agouti protein was predicted from the nucleotide sequence of the cDNA. The agouti protein is 131 amino acids in length and contains four noteworthy features: 1 ) an amino-terminal signal presequence that is important for entry into the secretory pathway, 2 ) a central region where 16 out of 29 amino acids are basic arginine or lysine residues, 3 ) a poly-proline stretch that follows the basic region, and 4 ) a cysteine-rich carboxy-terminal domain. The carboxy-terminal domain was shown to be as biologically active as the full-length protein in cell-based in vitro assays (Kiefer et al. 1997, Willard et al. 1995). The most remarkable feature of the carboxy-terminus is that the 10 cysteine residues are spaced similarly to the conserved ordering of cysteines in a large group of neurotoxins found in the venom of the primitive hunting spiders and cone snails (reviewed in Olivera et al. 1994). Direct analysis of the cysteine oxidation state in recombinant agouti protein indicates that all 10 cysteines are disulfide bonded in a pattern (Willard et al. 1995) that is consistent with the pattern of disulfides in the agatoxins (Olivera et al. 1994).

The common, unifying molecular feature of all of the yellow obese agouti mutations is the ubiquitous and strong expression of the wild-type agouti coding sequences from another transcriptional promoter (Argeson et al. 1996, Bultman et al. 1992, Duhl et al. 1994a and 1994b, Michaud et al. 1994a and 1994b, Miller et al 1993). Analysis of the A^y allele revealed that a 170-kb deletion upstream of the agouti gene deletes all but the promoter and first noncoding exon of a ubiquitously expressed gene called Raly (Michaud et al. 1993 and 1994a). The deletion causes the agouti coding exons to come under the transcriptional control of the Raly promoter, which leads to the the ubiquitous expression of the agouti protein (Bultman et al. 1992, Michaud et al. 1994a). The embryonic lethality of A^y in the homozygous condition is most likely caused by the large deletion and is unrelated to the ectopic expression of agouti (Michaud et al. 1993). Molecular analysis of several of the nonlethal yellow alleles of agouti, including a new allele called A^iapy, revealed that in all cases, like A^y, the agouti protein is ectopically expressed in a ubiquitous manner. The common molecular lesion in these dominant, homozygous-viable alleles is the insertion of a cryptic promoter element, such as an intracisternal A particle (Argeson et al. 1996, Duhl et al. 1994b, Michaud et al. 1994b), directly upstream of the agouti coding exons.

Elucidation of the molecular nature of the dominant agouti mutations prompted the working hypothesis that the ectopic expression of the agouti gene causes the metabolic dysfunction that leads to obesity. However, because each of the dominant agouti mutations had a genomic rearrangement, it was possible that the structural changes in the DNA caused a deletion and/or dysregulation of a second gene in the vicinity of agouti that was itself responsible for the phenotype of the agouti mutants. To rule out this possibility, transgenic mice were prepared with an expression construct that caused the ubiquitous expression of the wild-type agouti gene. For this purpose, the -actin or phosphoglycerate kinase promoters (Klebig et al. 1995) were used to drive the expression of the wild-type agouti cDNA (BAPa and PGKPa, respectively). Examination of several lines prepared with each construct indicated that the transgene was expressed in multiple tissues at levels that were equal to or exceeded that observed in the A^y mutant mice. When the BAPa or PGKPa transgene was crossed onto the C57BLl/6J inbred line, the transgenics developed yellow fur, and, as the animals aged, became obese. Transgenic males became 30-40% heavier and females 60-70% heavier than nontransgenic controls (Klebig et al. 1995). The analysis of fat pad masses indicated that ~80% of the body weight differences between transgenic and nontransgenic mice was attributable to increases in dissectable fat depots (R. L. Mynatt, unpublished data). A 2-wk feeding study in one line, BAPa20, indicated that hyperphagia is not necessary for the relative weight gain of transgenic vs. nontransgenic littermates (R. L. Mynatt, unpublished data). These findings lend direct support to the increased caloric efficiency hypothesis described in the previous section. Furthermore, the basal core temperature was measured in one line, BAPa20, and was found to be significantly depressed by 0.81°C (P < 0.0005) compared with nontransgenic controls (Kim et al. 1996), indicating that decreased thermogenesis contributes to a positive energy balance in these mice. With respect to insulin and glucose levels, both males (both BAPa and PGKPa) and females (BAPa only) become hyperinsulinemic within 12 wk of age, whereas only males developed overt hyperglycemia (Klebig et al. 1995). The ratio of insulin to glucose in the BAPa20 transgenic mice was twice that of the nontransgenic controls. This finding suggests that the mice produce higher levels of insulin to remain normoglycemic, which is a hallmark of noninsulin-dependent diabetes. Thus, the unambiguous parallels between the phenotype of these transgenic mice and the dominant yellow agouti mutants firmly established that the ectopic expression of agouti in a ubiquitous manner is sufficient to induce both obesity and diabetes.

Coupling agouti mutation analysis with ubiquitous expression of these mutations in transgenic mice revealed that the same structural features of agouti are generally important for both the production of yellow pigment and the development of obesity (Hustad et al. 1995, Perry et al. 1995 and 1996). That agouti acts extracellularly to induce obesity was suggested by deletion of 10 residues of the hydrophobic core of the amino-terminal signal sequence. When expressed in transgenic founder mice (C57BLl/6J genetic background) under the control of the -actin promoter, this signal peptide mutation resulted in completely non-yellow mice that remained lean. Expression of agouti sequences containing an ENU-induced point mutation in the signal sequence (Hustad et al. 1995, Perry et al. 1995), or a mutation in the putative N-linked glycosylation site, resulted in only patchy yellow fur over the ventral surface of the animal but no obesity (Perry et al. 1996). Deletion of approximately half of the central basic region of agouti did not significantly impair the development of yellow pigmentation or obesity, suggesting that this region of agouti may be dispensable for either activity. In contrast, substitution of individual cysteines with serine residues at some positions in the agouti carboxy-terminus completely eliminated the potential for both yellow pigmentation and obesity. A mutation at cysteine 110 or 131, however, was only partially disabling because some mice expressing these mutations produced yellow fur (in the ventrum only) and a few became obese. It has been proposed that cysteine 110 and 131 form a disulfide bonded pair that is at least partially dispensable for either biological activity (Perry et al. 1996).

MOLECULAR MODELS OF AGOUTI-INDUCED OBESITY

At present, it is not known whether melanocortin receptor antagonism in peripheral tissues contributes to the obesity, insulin resistance or other pleiotropic effects of the yellow obese syndrome. Although the tissue distribution of MC4-R has not yet been reported for the mouse, expression of MC4 receptors in the rat seems to be restricted to the brain (Mountjoy et al. 1994). Agouti antagonism of MC1-R is not significant with respect to obesity because chronic agouti expression in the skin results in yellow fur but no alterations in body weight or hyperglycemia (Kucera et al. 1996). The ubiquitously expressed MC5-R (Gantz et al. 1994, Griffon et al. 1994, Labbe et al. 1994) poses an unlikely target for the agouti protein because physiologically relevant concentrations of agouti do not antagonize this receptor in cell-based assays (Kiefer et al. 1997, Lu et al. 1994). The MC3-R and MC2-R receptors remain potential targets for the agouti protein in peripheral tissues. In addition to its expression in the limbic system and hypothalamus, MC3-R is expressed in the placenta and gut (Gantz et al. 1993, Roselliu-Rehfuss et al. 1993), and agouti has been shown to be a high affinity antagonist of human, although not rat, MC3-R (Kiefer et al. 1997, Lu et al. 1994). No data is yet available for agouti antagonism of MC2-R, although it is known that functional MC2 receptors are expressed in mouse adipocytes (Boston and Cone 1996) as well as in the adrenal cortex (Mountjoy et al. 1992).

To test whether adipose tissue is a direct target for agouti action in vivo, transgenic mice were generated that express agouti from the transcriptional promoter of aP2 (aP2a) (Mynatt et al. 1997), a gene that encodes the adipocyte fatty acid binding protein. Compared with the BAPa or PGKPa transgenic mice discussed above, the aP2a transgenic mice expressed extremely high levels of agouti in both white and brown adipose tissue, with negligible expression in other peripheral tissues or the brain. Even at a late age, the aP2a transgenic mice did not become overweight or hyperinsulinemic (Mynatt et al. 1997). This finding indicated that agouti expression in adipocytes alone is not sufficient to induce the metabolic changes that cause obesity and/or diabetes. However, when the aP2a mice were given daily subcutaneous insulin injections for 1 wk, the transgenic mice gained significantly more weight (1.7-fold) than their nontransgenic controls (Mynatt et al. 1997). This finding strongly suggests that agouti and insulin act synergistically to promote weight gain in vivo, perhaps due to the combination of their similar lipogenic and antilipolytic effects in the animal (reviewed in Bray 1996, Yen et al. 1994). Moreover, these findings establish a physiologically relevant role for agouti in adipose tissue in the yellow obese mutants.

Additional analysis of the yellow agouti mutants revealed that the expression of two key enzymes involved in de novo fatty acid synthesis and desaturation, fatty acid synthetase (FAS) and stearoyl-CoA desaturase were elevated in the liver and adipose tissue (Jones et al. 1996). Similarly, recombinant agouti protein stimulates both the expression and activity of FAS and causes an increase in triglyceride accumulation in 3T3-L1 adipocytes. This effect can either be completely prevented by Ca²⁺ channel blockade (Jones et al. 1996) or mimicked with Ca²⁺ agonists (Zemel et al. 1995a), suggesting that alterations in Ca²⁺ influx, without complementary alterations in Ca²⁺ efflux, may mediate these lipogenic effects of agouti. The soleus muscle of A^vy mice also exhibited an elevation in steady-state levels of intracellular free Ca²⁺ and increases in Ca²⁺ influx rate (Zemel et al. 1995b). A similar effect has been observed in both primary and cultured skeletal myocytes after treatment with recombinant agouti protein. These findings suggest that perturbations in calcium signaling and calcium homeostasis by extracellular agouti protein may contribute substantially to the insulin resistance and lipogenic bias in the yellow obese mice. Possible mechanisms include a G-protein-mediated coupling between melanocortin receptors and Ca²⁺ channels, direct stimulation of Ca²⁺ channels by extracellular agouti protein, or an indirect effect on Ca²⁺ channels by blocking voltage-gated or ATP-gated potassium channels.

All of this work on the characterization of the molecular etiology of the yellow obese syndrome is potentially useful for studying human obesity. Humans have a closely related homologue of the mouse agouti gene that is 80% identical overall and 87% identical within the cysteine-rich carboxy-terminus (Kwon et al. 1994, Wilson et al. 1995). Although it is presently not clear whether any mutations in the human agouti gene are associated with an obesity-related phenotype, it is noteworthy that the wild-type human agouti gene is expressed in adipose tissue. This finding, coupled with the results that adipose tissue-specific expression of agouti in transgenic mice is physiologically active, suggests that the human agouti gene may normally function in the regulation of lipid metabolism. Additional experiments that utilize the molecular and genetic reagents that have become available through the study of obesity genes in the mouse will likely continue to provide insights into the molecular basis of obesity in humans.