Molecular Mechanisms for Activation of the Agouti-Related Protein and Stimulation of Appetite

PMI-5011 is a bioactive extract of Russian tarragon (Artemisia dracunculus L.) consisting of six potential components (22). It was prepared as previously described from hydroponically grown plants and purified as a resin (19,20,23).

The human adrenocortical carcinoma cell line NCI-h295R (American Type Culture Collection, Manassas, VA) and mouse neuronal N38 (24) and GT1-7 (25) cell lines were cultured as previously described (26). PMI-5011 was dissolved in 70% ethanol or DMSO and added to the media at the concentration of 40 μg/ml.

Two whole mouse brains were removed and transferred immediately into 7 ml of dissection media (500 ml Hank's buffered salt solution; Invitrogen, Carlsbad, CA), 1.19 g HEPES, 2.52 g glucose, 50 μl gentamicin (Invitrogen), 1.5 g BSA (Sigma, St. Louis, MO), 1.5 g Mg₂SO₄, and 111 mg CaCl₂. Cells were dissociated and gradually attached to the bottoms of culture plates reaching ∼90% confluency in 10–12 days.

A 1,107-bp AGRP promoter fragment (−796/+312) was amplified from human genomic DNA with Platinum Pfx polymerase (Invitrogen) using the PCR primers 5′-GTCAGCTAGCGTCTTAAACCCCTGGCCTTG-3′ and 5′-CAGTCTCGAGCCTTGGAGTCCCCTCTAGG-3′. Amplicons was digested with NheI/MluI and cloned into the pGL3basic luciferase vector (Promega, Madison, WI). Site-directed mutagenesis was performed as described in line n the supplementary data in the online appendix (available at http://diabetes.diabetesjournals.org/cgi/content/full/db10-0172/DC1). The numbering of DNA sequence cited in this manuscript (e.g., for the CACCC-box) was according to the sequence appearing in our previous publication (5) and GenBank accession no. AF314194.

Cells were transfected with Geneporter II (Gene Therapy Systems, San Diego, CA) or Fugene 6 according to the manufacturer's protocols. Firefly and renilla luciferase activities were measured using the Dual Luciferase Kit (Promega) and a single-tube FB12 luminometer (Berthold Detection Systems, Oak Ridge, TN). More details about the constructs are provided on line in the supplementary data.

Nuclear extracts from N38 cells were prepared as described elsewhere (27). A complete list of the oligonucleotides used is shown on line in the supplementary data.

NCI-h295R human adrenocortical cells were cultured in a 100-mm cell culture plate and transfected with KLF4-expressing vector or cotransfected with KLF4 and −796/+373 AGRP promoter construct for 16 h. The experimental procedure was according to the manufacturer's instructions (Upstate, Millipore, Billerica, MA). More details about the primers and conditions are provided on line in the supplementary data.

Total RNA was extracted from cultured cells or whole tissue using the RNeasy Mini Kit (Qiagen). Hypothalamic extracts were prepared as previously described (26). Quantitative PCR was performed using the TaqMan one-step RT-PCR core reagents kit (Applied Biosystems, Foster City, CA) as previously described (28). The primers and probes are described in detail on line in the supplementary data.

For Western blot analysis, total protein lysates (30–50 μg/lane) were separated on 15% or 10% SDS-PAGE and blotted to Immun-Blot polyvinylidene difluoride membranes (Bio-Rad, Hercules, CA). Antibodies for Agrp were from Alpha Diagnostic International (San Antonio, TX) and Klf4 from Santa Cruz Biotechnology (Santa Cruz, CA). For loading control anti–glyceraldehyde-3-phosphate dehydrogenase (AM4300; Ambion) or β-actin (Abcam, Cambridge, MA) antibody were used. The signal was detected with Amersham ECL Western Blotting Detection Reagents (GE Healthcare, Buckinghamshire, U.K.). Western blot images were quantified using Quantity One software (Bio-Rad).

GT1-7 cells were plated on 12-mm glass coverslips coated with poly-d-lysine. Cells were washed in PBS and fixed in 4% paraformaldehyde. Immunocytochemistry was performed using Klf4 and Agrp antibodies (dilution 1:200) in a cocktail. This was followed by three washes in the presence of a secondary Alexa Fluor 488 goat anti-rat antibody and Alexa Fluor 594 goat anti-rabbit antibody (Invitrogen, Eugene, OR). Finally, slides were mounted in Prolong Gold Antifade Reagent with 4′,6-diamidino-2-phenylindole (Invitrogen). Microphotographs of stained cells were taken with a Axioscop 2 microscope (Zeiss) at 40× magnification. Color images were captured with a Zeiss Axiocam HRc CCD camera at 1,300 × 1,030 resolution.

In vivo bioluminescence imaging was conducted on an IVIS 100 system (Xenogen, Alameda, CA) using LivingImaging acquisition and analysis software (Xenogen). The rationale for the production of these mice and a full description of the constructs used is provided elsewhere (6). The mice used in this experiment contained the Klf4-binding CACCC-box and additional enhancer elements from downstream regions. Mice were anesthetized with isoflurane and received an intraperitoneal injection of the substrate d-luciferin (100 mg/kg). An integration time of 3 min with a binning of 100 pixels was used for luminescence image acquisition. Signal intensity was quantified as the sum of all detected photon counts within the region of interest after subtraction of background luminescence. Mice were fed chow diet (Chow 5001, 12.5% Kcal, Purina lab diet; Framingham, MA) ad libitum and imaged the day before the start of feeding with PMI-5011 and imaged again 1 week later, at the end of the experiment.

The study protocols for all mouse feeding studies were approved by the institutional animal care and use committee of the Pennington Biomedical Research Center (Louisiana State University System, Baton Rouge, LA). Mice were maintained on a 12-h/12-h day/light cycle in a pathogen-free animal facility with lights coming on at 6:00 a.m. Indirect calorimetry was performed as we have previously described (29). More details are provided on line in the supplementary online data.

Fat-free body fluid and fat-free mass were measured with a nuclear magnetic resonance instrument (Bruker Optics, The Woodlands, TX) as we have previously reported (29). Fat-free mass (i.e., muscle mass) was calculated by subtracting the fat mass and fluid mass from total body weight. Data are shown as means of grams of fat or grams of lean mass.

Several mouse-feeding studies were performed using chow and high-fat diets (HFDs) with and without PMI-5011. Details about each study are provided on line in the supplementary data.

Statistical significance was evaluated using one-way ANOVA and the Student t test. The data were expressed as means ± SE and calculated using variance analysis and the Newman-Keuls test for multiple comparisons among groups. Bonferroni adjustments were made for multitesting. P values <0.05 were considered to be statistically significant. The same methods were used for testing for differences in circulating hormone levels. Data analysis was carried out on SAS (SAS version 9.1).

For the meta-analysis, the data from the three HFD studies were combined to examine whether there were statistically significant differences in body weight between the HFD- and PMI-5011–fed groups. The random effects ANOVA meta-analysis model was used. The results represent the unweighted ANOVA analysis. The ANOVA analysis in a bootstrap fashion was also used, and again the P value was nearly identical. Data analysis was carried out on SAS (SAS version 9.1).

The gene names used were according to standard nomenclature (http://www.genenames.org/ and http://www.informatics.lax.org/mglhome/nomen/gene.shtml).

Algorithmic analysis of the human AGRP promoter (Tess/Transfac and Alibaba2) revealed the presence of two conserved CACCC-boxes at position +163/+169 (position A or proximal position), and position −277/−283 that is a reverse CACCC-box/G-rich stretch (position B or distal position) (Fig. 1,A). CACCC-boxes are typical binding motifs for the transcription factor KLF4 (12,,–15), which became a candidate effector of AGRP. The direct effect of KLF4 on Agrp was tested by transiently transfecting GT1-7 cells with a KLF4 expression construct, which resulted in an increase of endogenous Agrp mRNA and protein levels (Fig. 1,B and D, respectively). This effect was recapitulated in mouse whole-brain primary cultures from two male mice (Fig. 1 C and E).

Immunocytochemistry was performed to determine the specificity of the effect of KLF4 on Agrp expression. Endogenous Klf4 protein was upregulated in dividing GT1-7 cells, which coincided with significant upregulation of Agrp (Fig. 1,F). Cells that had been successfully transfected with a KLF4 expression construct displayed significant accumulation of KLF4 in the nucleus and robust upregulation of Agrp (Fig. 1 G), suggesting that KLF4 directly enhances endogenous Agrp expression in hypothalamic cells.

We assessed the effects of overnight food deprivation on mouse hypothalamic Klf4 and Agrp and found that both were upregulated in mRNA preparations (Fig. 1 H and I).

The effect of KLF4 on the AGRP promoter was evaluated by site directed mutagenesis of the two candidate CACCC-boxes at positions A and B. The mouse clonal hypothalamic cell line (N38) and the human adrenocortical carcinoma cell line NCI-h295R were used for these experiments because they both express endogenous Agrp whereas the NCI-h295R cell line also represents a “human” environment because we are using the human AGRP promoter. Mutagenesis of the proximal (position A or mutA) but not the distal CACCC-box (mutB) diminished significantly the effect of KLF4 on AGRP promoter in the mouse N38 (Fig. 2,A) and completely abolished promoter activity in the human environment of the NCI-h295R cells (Fig. 2,B). Because KLF4 had been suggested to interact with the Sp1 family of transcription factors (30), we used a Sp1 expression construct but found that it had no effect on AGRP promoter activity, and neither did it enhance the effect of KLF4 (Fig. 2 C).

Electrophoretic mobility shift assays (EMSAs) revealed a complex binding pattern for vector control (P) and a specific band (arrow) in KLF4 (K)-transfected cells (Fig. 2,D). Preincubation with excess unlabeled wild-type probe competed out this band in the KLF4-transfected cells (lane 6) but not with the mutant probe (lane 7), further emphasizing the significance of the intact CACCC-box at position A. Importantly, the complex of interest could be completely competed away with a known KLF4 binding oligonucleotide from the cytochrome CYP1A1 gene (30) referred to in this figure as “BTE” (Fig. 2,D, lane 8). The radiolabeled mutant probe (mutA) did not result in the appearance of the additional band after incubation with the KLF4-transfected nuclear extracts (Fig. 2 D, lanes 9–15).

In an EMSA supershift experiment, specific antibodies were used to identify the band of interest (Fig. 2,E). Again, only KLF4 cotransfected extracts (K) resulted in the formation of an additional band (Fig. 2,E, lane 4). Binding was completely blocked with an anti-Klf4 antibody (Fig. 2,E, lane 5) but not with the negative control antibody directed against Sp1 (Fig. 2,E, lane 6). The mutant probe again failed to form any Klf4-specific DNA:Protein complexes (Fig. 2 E, lanes 7–12).

In chromatin immunoprecipitation (ChIP) experiments, the human NCI-h295R cells were transiently transfected first with the KLF4 expression construct, and in a second experiment they were cotransfected with the KLF4 expression construct plus the AGRP −796/+312 promoter construct. The second experiment represented a positive control because this promoter construct contains the KLF4-binding CACCC-box. The coimmunoprecipitate from the first experiment was higher in the KLF4-transfected cells and this effect was further enhanced in the promoter cotransfected cells from the second experiment that had been enriched for the CACCC-box (Fig. 2 F).

We tested the hypothesis that a botanical extract of tarragon, PMI-5011, could activate Agrp because tarragon has traditionally been used as an appetite stimulant. First, we tested the hypothesis that PMI-5011 activates Agrp at the promoter level in vitro. GT1-7 cells were transient transfection with a promoter construct containing the minimal promoter of AGRP and cells were treated overnight with PMI-5011. The activity of the AGRP promoter (−796/+373) was significantly increased by PMI-5011 (Fig. 3,A). We then tested the hypothesis that PMI-5011 also activates the AGRP promoter in vivo. We used luciferase-expressing transgenic mice whereby luciferase was driven by the minimal promoter of AGRP that encompasses the proximal CACCC-box at position +163/+169, as we have described elsewhere (6). Two mice from two different founders were imaged while consuming a chow diet (Fig. 3,B). Their diet was subsequently supplemented with PMI-5011 for 1 week, and the same mice were imaged again. Luciferase expression was increased sixfold by PMI-5011 in mice from both founders and at all sites of expression as illustrated by the ventral and dorsal views (Fig. 3 B).

In this experiment, we tested the hypothesis that Klf4 mediates the activation of Agrp by PMI-5011. First, we showed that PMI-5011 activated both endogenous Agrp and Klf4 in GT1-7 cells (Fig. 4,A and B, respectively). The specific requirement of Klf4 for the activation of Agrp by PMI-5011 was confirmed using antisense siRNA probes against Klf4. First, the efficacy of siRNA primers against Klf4 was confirmed in GT1-7 cells by blocking the upregulation of Klf4 in cells treated with PMI-5011 (Fig. 4,C), whereas the control (scrambled) primers had minimal effect. siRNA against Klf4 also abrogated the upregulation of Agrp by PMI-5011 (Fig. 4 D).

The in vivo effects of PMI-5011 on energy expenditure and body composition were evaluated by indirect calorimetry in metabolic chambers. Two groups of mice of equal mean body weights received chow or chow plus PMI-5011 diets, and metabolic parameters were measured for 4 days. Hypothalamic Klf4 and Agrp protein levels were significantly higher in the PMI-5011 group (Fig. 5,A). Body weight and body composition (fat and fat-free mass) were not affected by PMI-5011 (Fig. 5,B and C), but food intake was significantly higher (Fig. 5,D). Oxygen consumption and energy expenditure (heat production) were significantly lower in the PMI-5011–fed group (Fig. 5,E and F). The Respiratory exchange ratio (RER) and total activity were not affected by PMI-5011 (Fig. 5 G and H).

Because body weight was not increased as it would be expected given the increased food intake and reduced energy expenditure, another study was performed to examine the long-term (24-day) effects of PMI-5011 on food intake still under a chow diet. Hypothalamic Klf4 and Agrp protein levels, as well as total food intake, were significantly higher in the PMI-5011–fed group of mice (Fig. 6,A and B), but body weight was again unaffected (data not shown). We then examined the hypothesis that a higher fat content may be required to be present in the diet for effects of PMI-5011 to become evident on body weight. We examined the data from three previously performed studies that had used PMI-5011 in a HFD. We found that body weight of the mice receiving the PMI-5011 was enhanced in all three studies and both Agrp and Klf4 were significantly elevated (Fig. 6,C). A meta-analysis was then performed using the data from all three studies that showed a statistically significant increase for the overall body weight of mice consuming HFD supplemented with PMI-5011 (Fig. 6 D).

We measured the effects of PMI-5011 on the circulating levels of hormones known to affect Agrp expression and energy balance: insulin, leptin, ghrelin (active), peptide tyrosine tyrosine (PYY) (total), and gastric inhibitory polypeptide or glucose-dependent insulinotropic peptide (GIP) (total). In this experiment, C57BL/6 male mice were fed ad libitum HFD or HFD plus PMI-5011 for 11 days. PMI-5011 was provided by gavage. Insulin and leptin were significantly lower in the PMI-5011 group (Fig. 7,A and B) but ghrelin, PYY, and GIP were not significantly affected (Fig. 7 C–E).

In addition to Agrp, the expression levels of other orexigenic and anorexigenic neuropeptides were measured in the hypothalami of control and PMI-5011 chow-fed mice that had been used in the metabolic chambers. Overall, hypothalamic Agrp mRNA levels were increased in the PMI-5011 group (Fig. 8,A), and this effect was more pronounced in four of seven mice (i.e., high responders P < 0.05, not shown here). Neuropeptide Y (Npy) mRNA levels were not significantly affected by PMI-5011 (Fig. 8,B), but the mRNA levels of the other orexigenic peptides, Orexin and Mch, were significantly higher in the PMI-5011 group (Fig. 8,C and D). The mRNA levels of the anorexigenic peptides Pomc and Cart were not significantly affected by PMI-5011 (Fig. 8 E and F).

In the present study, we investigated the effects of a bioactive extract of tarragon, PMI-5011, on appetite stimulation. Agrp was our primary target because of it is a well-characterized appetite effector (10,11,31). We found that PMI-5011 significantly upregulated the promoter of the human AGRP gene and increased its expression levels in neuroblastoma cells as well as in mouse hypothalamic primary cell cultures. A landmark feature of hypothalamic Agrp is its upregulation by food deprivation (3,26,32). We found that Klf4 was also upregulated by overnight food deprivation. This suggests that the two genes may be coregulated. We do not know if Klf4 upregulation precedes that of Agrp, which would enhance the notion that it may be involved in the upregulation of Agrp by food deprivation in vivo. Further experiments would be required to confirm this possibility.

Using bioinformatics tools, KLF4 was identified as a candidate transcription factor for the activation of AGRP because of the presence of conserved CACCC-boxes along its promoter, which are typical binding motifs for KLF4 (33,34). PMI-5011 was found to augment both Klf4 and Agrp protein levels, whereas transient transfections of cells with a KLF4 expression construct confirmed the direct stimulation of Agrp. Further experiments using siRNA showed that Klf4 is required for the upregulation of Agrp by PMI-5011.

Two candidate CACCC-boxes at positions +163/+169 and −277/−283 on the AGRP promoter were studied, but only the proximal box was found to be a functional binding site for KLF4. Subsequent EMSA, supershifts, and ChIP experiments established the significance of the proximal CACCC-box for the binding of Klf4. Luciferase-expressing transgenic mice with luciferase driven by the AGRP promoter encompassing the KLF4-binding CACCC-box at position +163/+169 confirmed in vivo the activation of the AGRP promoter by PMI-5011.

Feeding studies were performed to evaluate the effects of PMI-5011 on Klf4/Agrp and other appetite-regulating neuropeptides and to assess its effects on overall metabolic and physiological parameters. In a study using indirect calorimetry, PMI-5011 increased hypothalamic Klf4/Agrp and stimulated food intake. The respiratory exchange ratio and total locomotor activity were not affected, but oxygen consumption and energy expenditure were significantly lower in the PMI-5011–fed group. Based on these outcomes, one would expect the PMI-5011–fed mice to gain body weight, but this was not the case probably because of the short period of feeding and/or because of the type of diet used (chow). Thus, a 24-day feeding experiment was performed to test the hypothesis that PMI-5011 may require additional time to confer body-weight gain. Food intake and hypothalamic Klf4/Agrp protein levels were again significantly higher in the PMI-5011–fed group, but body weight was again not affected. We therefore examined the effects of PMI-5011 in three previous studies that had used HFD to determine whether a high fat content was required for PMI-5011 to affect body weight. In all three studies, body weight was increased consistently and data meta-analysis combining the three studies showed statistical significance. Food intake had not been measured in these mice because these studies had been set up to measure the effects of PMI-5011 on insulin sensitivity under severe obesigenic conditions (35). These data show that hypothalamic increases in Klf4 and Agrp by PMI-5011 can lead to increased food intake but body weight is affected only under HFD.

We also found that PMI-5011 reduced the circulating levels of insulin and leptin that are known inhibitors of hypothalamic Agrp (36). The gut-secreted glucagon-like peptide 1 (GLP-1) is also affected by tarragon (20), whereas PMI-5011 has antidiabetic effects (20), and, importantly, it attenuates expression of protein tyrosine phosphatase 1B (PTP-1B) (35). Attenuation or deletion of PTP-1B confers resistance to body-weight gain (37), and we speculate that our mice did not gain significantly in body weight by PMI-5011 because of peripheral reduction in PTP-1B (that was not measured in our experiments). These multifaceted effects of PMI-5011 could be because of one or more of its six components (22). Each component (or several of them in concert) could thus have specific or pleiotropic effects on metabolic processes, perhaps by crossing the blood brain barrier, and/or by modulating peripheral hormones like insulin and leptin.

The complex effects of PMI-5011 are further underscored by the fact that Orexin and Mch were also upregulated, whereas the anorexigenic Pomc and Cart were not significantly affected. It merits further investigation to determine whether Klf4 is the mediating transcription factor that activates Orexin and Mch that express in different neurons. By algorithmic analysis, we found that Orexin has a CACCC-box in its minimal promoter, but we do not know if it is functional. We did not find a putative CACCC-box on the Mch promoter, but Klf4 interacts with SP1 (30), and it could use SP1 binding sites on these two orexigenic peptides. Importantly, these data show that PMI-5011 affects the expression levels of multiple orexigenic neuropeptides in a complex fashion that may extend beyond Klf4 and involve other transcription factors and perhaps the hypothalamic receptors of peripheral hormones that are also affected by PMI-5011 (e.g., insulin, leptin).

The data presented here identify Klf4 as a major transactivator of Agrp. In addition, we report that the tarragon extract PMI-5011 has complex metabolic effects that include an increase in hypothalamic Klf4 and food intake, activation of multiple components of the orexigenic circuitry (i.e., Agrp, Mch, Orexin), and reduction of insulin and leptin circulating levels. Klf4 is typically downregulated in colorectal cancers (13,17,38,39) and has anticarcinogenic properties when normally expressed (21). PMI-5011 could thus be used for enhancing Klf4 and appetite in conditions like anorexia and cancer cachexia.

This research was supported by grants R01 DK062156 (to G.A.); RO1 DK 45278 and Utah Science, Technology, and Research (to D.A.Y.); 5P50 National Institutes of Health (NIH)/National Center for Complementary and Alternative Medicine (NCCAM) AT002776 (to W.T.C.); and 5P50 NIH/NCCAM AT002776-9001 and RO1 DK064071 (to A.Z.).

No potential conflicts of interest relevant to this article were reported.

O.I. performed the experiments shown in Figs. 3, 6, 7, and 8. A.M.S. performed the experiments shown in Figs. 2 and 4. M.-J.P.-Y. and D.A.Y. performed the experiments shown in Fig. 1. D.M.R. and W.T.C. provided the tarragon extract PMI-5011. A.Z. performed parts of the experiments shown in Fig. 8. G.A. performed the experiment shown in Fig. 5 and wrote the manuscript.

We thank Dr. Fulu Bai (Pennington Biomedical Research Center, Baton Rouge, LA) for help with the promoter analysis, Dr. Zhanguo Gao (Pennington Biomedical Research Center, Baton Rouge, LA) for help with the ChIP assays, and Dr. Ron Horswell (Pennington Biomedical Research Center, Baton Rouge, LA) for the meta-analysis of the mouse feeding studies. We also thank Dr. Denise Belsham (University of Toronto, ON, Canada) for the kind gift of the N38 neuronal cells, and the Botanical Research Center at the Pennington Biomedical Research Center (Louisiana State University System, Baton Rouge, LA).