Antagonism of Central Melanocortin Receptors in Vitro and in Vivo by Agouti-Related Protein

See allHide authors and affiliations

Science  03 Oct 1997:
Vol. 278, Issue 5335, pp. 135-138
DOI: 10.1126/science.278.5335.135


Expression of Agouti protein is normally limited to the skin where it affects pigmentation, but ubiquitous expression causes obesity. An expressed sequence tag was identified that encodes Agouti-related protein, whose RNA is normally expressed in the hypothalamus and whose levels were increased eightfold in ob/ob mice. Recombinant Agouti-related protein was a potent, selective antagonist of Mc3r and Mc4r, melanocortin receptor subtypes implicated in weight regulation. Ubiquitous expression of human AGRP complementary DNA in transgenic mice caused obesity without altering pigmentation. Thus, Agouti-related protein is a neuropeptide implicated in the normal control of body weight downstream of leptin signaling.

Analysis of mouse obesity mutations has helped define regulatory circuits that govern energy expenditure (1). In mice carrying certain alleles of theAgouti coat color gene such as lethal yellow(Ay ) or viable yellow(Avy ), pleiotropic effects including a yellow coat, obesity, and increased body length are caused by ubiquitous expression of chimeric transcripts encoding a normal Agouti protein (2-4). Agouti is a paracrine signaling molecule (5) that affects pigmentation by antagonism of the melanocortin 1 receptor (Mc1r) (6, 7), one of five related heterotrimeric GTP-binding protein–coupled receptors named for their ability to respond to α-melanocyte stimulating hormone (α-MSH) and adrenocorticotrophic hormone (ACTH) (8). Expression and action of Agouti is normally limited to the skin (3, 5), but recombinant Agouti protein will also antagonize Mc2r and Mc4r (6, 9), expressed primarily in the adrenal gland and the central nervous system (CNS), respectively (8,10).

Using a characteristic pattern of cysteine spacing from the COOH-terminal region of Agouti to search an expressed sequence tag database, we isolated a gene from 129/sv mice and from humans that encodes a protein nearly identical in size and genomic structure toAgouti that we named Agouti-related protein(Agrp) (Fig.1A). The same gene was recently described by Shutter et al. as Agouti-related transcript(11). Reverse transcriptase–polymerase chain reaction (RT-PCR) and Northern (RNA) hybridization experiments demonstrated thatAgrp RNA was expressed primarily in the adrenal gland and the hypothalamus (Fig. 1, B and C). To investigate functional overlap between Agrp andAgouti, we examined whether the steady-state level ofAgrp RNA would be altered by ectopic expression of Agouti in Ay/a animals. Northern hybridization analysis of hypothalamic and adrenal gland RNA fromAy/a or coisogenic a/a animals revealed an ≈fivefold reduction of AgrpRNA in the hypothalamus of Ay/a animals (Fig.1C). We also measured the levels of hypothalamicAgrp RNA in ob/ob animals and found an ≈eightfold increase relative to coisogenic controls. In the adrenal gland, levels of Agrp RNA inAy/a and nonmutant animals were below the level of detection, but could easily be detected in ob/ob animals.

Figure 1

Structure and expression of Agrp.(A) Comparison of mouse and human AGRP sequences with that of mouse Agouti. Arrowheads indicate signal sequence cleavage sites; arrows indicate different forms of recombinant AGRP. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. Cysteine residues are in bold. (B) RT-PCR assay for mouse Agrp RNA (23). (C) Northern hybridization assay for mouse Agrp RNA in brain tissues (9 μg, left) or in hypothalamus and adrenal glands ofAy/a and ob/ob mice (5 μg, right); each lane represents RNA from a single animal. Relative ratios of Agrp RNA determined with a PhosphorImager were based on signal from an exon 4 cDNA probe compared with Actb andGapd control probes hybridized to the same blot.

To determine whether AGRP antagonizes melanocortin signaling, we used the baculovirus expression system to produce conditioned media containing recombinant human AGRP, and measured antagonist activity using a Xenopus melanophore cell line developed by Lerner and colleagues (12). Melanophores provide a rapid and sensitive bioassay for melanocortin agonists and antagonists because pigment granule dispersion induced by α-MSH can be measured in microtiter plates as a change in optical density (12). Using the ability of conditioned media to inhibit α-MSH–induced pigment dispersion, we partially purified multiple forms of AGRP that cofractionate with α-MSH antagonist activity by Blue Sepharose and anion-exchange chromatography (Fig. 2). One peak of α-MSH antagonist activity contained mature AGRP with the signal sequence removed (form A) and a mixture of three AGRP fragments cleaved after residues 46, 48, or 50 (form B). The second major peak contained AGRP fragments cleaved after residues 69 or 71 (form C).

Figure 2

AGRP activity in Xenopusmelanophores. (A) One microgram of protein from serial steps in the purification procedure analyzed by silver-stained 4 to 20% SDS–polyacrylamide gel electrophoresis. Two liters of conditioned media (lane 1) was applied to a Blue Sepharose Fast Flow Column, then eluted with 40 mM CAPS (pH 10.8), 2.5 M NaCl. Fractions with α-MSH antagonist activity (lane 2) were concentrated (Centriprep 3), buffer-exchanged into 40 mM CAPS (pH 10.8), 20 mM NaCl, applied to a HiTrap Q anion-exchange column, then eluted with a 20 to 800 mM NaCl gradient in 40 mM CAPS (pH 10.8). (B) Major peaks of α-MSH antagonist activity in the flow-through (fractions 2 to 13, lane 3) and in fractions 20 to 26 (lane 4) were dialyzed into storage buffer [20 mM Pipes (pH 6.8), 50 mM NaCl]. NH2-terminal sequencing of the two predominant bands in each peak revealed mature AGRP and two heterogeneous smaller forms as indicated. AGRP purity, estimated by densitometry of a 10-μg sample loaded on a 10% Tricine gel stained with ProBlue, was used to calculate effective concentrations of 40 and 11 μM for form A+B (lane 4) and form C (lane 3), respectively. (C) Quantitative α-MSH dose-response analysis of different Agrp forms measured at equilibrium conditions 180 min after addition of AGRP and α-MSH. In the 96-well melanophore assay (12), pigment dispersion is calculated as (A 650 final −A 650 initial)/A 650final, where A 650 is the absorbance at 650 nm. Data points represent mean ± SEM of triplicate samples.

Partially purified AGRP forms A+B and form C are potent, specific antagonists of α-MSH–induced pigment dispersion inXenopus melanophores, with calculated antagonist dissociation constant (K B) values of 7.0 and 1.2 nM, respectively (Fig. 2C). AGRP did not inhibit pigment granule dispersion in the absence of α-MSH and did not inhibit pigment granule dispersion induced by forskolin, a direct activator of adenylate cyclase (13).

To examine the selectivity of AGRP for human melanocortin receptors, we added various concentrations of AGRP form A+B to cell lines that had been stably transfected with each of the five different receptor subtypes, then measured the ability of α-MSH or ACTH to induce adenosine 3′,5′-monophosphate (cAMP) accumulation. At concentrations up to 100 nM, AGRP had no effect on hMC1R or hMC2R, and only slightly inhibited hMC5R (Fig. 3). By contrast, AGRP concentrations of 1 nM or more caused a dose-dependent inhibition of α-MSH–induced cAMP accumulation mediated by hMC3R and hMC4R.

Figure 3

Effects of AGRP on human melanocortin receptors. 293 cells (hMC1R, hMC3R, hMC4R, or hMC5R) or OS3 cells (hMC2R) stably transfected with the indicated receptor were preincubated with the indicated amounts of AGRP for 30 min; various amounts of α-MSH or ACTH were added for 30 min in the presence of 0.2 mM isobutylmethylxanthine, and total cAMP accumulation was determined on duplicate wells (9). Data points represent the mean ± SEM of 2 to 3 independent experiments. As a control for proteins other than AGRP, conditioned media from insect cells infected with an unrelated baculovirus were loaded and eluted from a Blue Sepharose column with conditions identical to those used for AGRP, dialyzed into storage buffer (490 μg/ml), then used at a dilution identical to that used to prepare 100 nM AGRP form A+B. Nanomolar concentrations of AGRP form A+B antagonize the hMC3R and hMC4R but do not meet criteria for competitive antagonism (15); therefore,K B values cannot be calculated.

Because AGRP form C can antagonize α-MSH in melanophores (Fig. 2) or in Mc4r-transfected cells (13), the COOH-terminal cysteine-rich region is probably sufficient for biologic activity, as is the case for Agouti (14). Sequence comparison of AGRP and Agouti highlights a short region of similarity beginning with the third Cys residue, CCDPCAXCXCRFF, that may contain determinants required for melanocortin antagonism (Fig. 1A). Nonetheless, the exact biochemical mechanism by which these proteins act is not clear. Most evidence favors competitive antagonism, whereby α-MSH and Agouti (or AGRP) bind to mutually exclusive sites on melanocortin receptors (14). The effects of AGRP on melanophores are consistent with competitive antagonism, because increasing amounts produced a proportionate and parallel displacement of the α-MSH dose-response curve without affecting maximal signaling (Fig. 2C). For the hMC4R, however, AGRP concentrations of 10 and 100 nM produced a decrease in basal levels of cAMP accumulation, as well as a decrease in the maximal level of α-MSH–induced cAMP accumulation (Fig. 3), neither of which is consistent with competitive antagonism (15). It has been proposed that some effects of Agouti are mediated by alterations in calcium flux (16), an intriguing finding given the similarity in cysteine spacing between Agouti, AGRP, and certain calcium channel antagonists (4). It is possible that Agouti or AGRP does not bind directly to melanocortin receptors or binds to more than one cell surface protein, uncertainties that may be resolved by studies of Agouti and AGRP binding.

To determine whether or not Agouti and AGRP have comparable effects in vivo, we constructed transgenic mice in which the human AGRPcDNA was controlled by the ubiquitously expressed β-actin promoter. Weight gain of six independent transgenic founders was significantly increased over nontransgenic littermates (17), and a transgenic line was established. Among F1 animals carrying the β-actin AGRP transgene, increased weight gain was detectable at 4 weeks of age, reached levels 100 or 70% above that of nontransgenic females or males, respectively, and was nearly indistinguishable from that caused by a β-actin Agouti transgene (Fig.4). Body length and food consumption were also increased by the AGRP transgene (17). By contrast, none of 15 animals carrying the AGRP transgene exhibited a difference in coat color from their nontransgenic littermates (Fig. 4). Thus, although AGRP mimics the effect of Agouti on weight gain, body length, and food consumption, it has no effect on pigmentation.

Figure 4

Effects of AGRP or human Agouti in transgenic mice. A β-actin human AGRP cDNA construct nearly identical to one described previously for human Agouti (24) was injected into F2 (C57BL/6J × CBA/J) embryos. One of six F0 founders (17) was bred to C57BL/6J animals. At 11 weeks of age, F1 transgenic animals (females: n = 4; males: n = 3) weighed significantly more than nontransgenic littermates (females:n = 6, P = 0.00005; males: n = 6;P = 0.00002). The time of onset and level of weight gain caused by the AGRP transgene (A) were similar to those caused by the Agouti transgene (B), but the AGRP transgene had no effect on pigmentation (C). Data points represent the mean ± SEM.

Given the eightfold increase of hypothalamic expression inob/ob mice, we propose that Agrpnormally regulates body weight via central melanocortin receptors, analogous to the relation between Agouti and the Mc1r for regulation of pigmentation. Huszar et al. (18) have shown that Mc4r-deficient animals develop obesity and metabolic derangements that mimic those in Ay/− mice, which suggests that obesity caused by ubiquitously expressedAgrp or Agouti is mediated largely by Mc4r. However, AGRP may also be a physiologic ligand of Mc3r (Fig. 3), which has been implicated in Agouti-induced obesity (19) and whose CNS expression (20) more closely matches that of Agrp. In the brain,Agrp RNA is localized primarily to the arcuate nucleus and median eminence (11), but unlike Agouti, which has a very small sphere of action in vivo (3), AGRP may diffuse more widely, particularly if it is processed to a smaller COOH-terminal form in vivo. AGRP is also produced by the adrenal gland but does not affect hMC2R and therefore may act at a more distant site.

What advantages do endogenous receptor antagonists such as Agouti or AGRP offer for homeostatic regulation? In the case of melanocortins, which activate five receptors to varying extents, an antagonist limited in its tissue distribution or biochemical specificity, or both, allows individual regulation of receptor subtype signaling. Melanocortin receptors were identified on the basis of their response to the agonist α-MSH (6), but physiologic signaling via Mc1r is regulated mainly by alterations in levels of the antagonist, Agouti. Similarly, regulation of Mc3r or Mc4r signaling could be mediated primarily by changes in Agrp expression rather than proopiomelanocortin, the precursor of α-MSH and ACTH. Agouti or AGRP, or both, may also transduce a signal via melanocortin receptors independent of melanocortin binding (21), consistent with the effects we observed on basal levels of cAMP accumulation.

Leptin deficiency lies upstream ofAgrp expression, directly or indirectly, but other signaling systems implicated in energy balance (22) may also regulate Agrp expression. Additional studies based on gene targeting may help to placeAgrp in a genetic pathway for feeding behavior, which should be useful in understanding and developing treatments for disorders of body weight regulation.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed at Beckman Center B271A, Stanford University School of Medicine, Stanford, CA 94305–5323, USA. E-mail: gbarsh{at}


View Abstract

Navigate This Article