Hypothalamic-Specific Manipulation of Fto, the Ortholog of the Human Obesity Gene FTO, Affects Food Intake in Rats

Sequence variants in the first intron of FTO are strongly associated with human obesity and human carriers of the risk alleles show evidence for increased appetite and food intake. Mice globally lacking Fto display a complex phenotype characterised by both increased energy expenditure and increased food intake. The site of action of FTO on energy balance is unclear. Fasting reduces levels of Fto mRNA in the arcuate nucleus (ARC) of the hypothalamus, a site where Fto expression is particularly high. In this study, we have extended this nutritional link by demonstrating that consumption of a high fat diet (45%) results in a 2.5 fold increase in Arc Fto expression. We have further explored the role of hypothalamic Fto in the control of food intake by using stereotactic injections coupled with AAV technology to bi-directionally modulate Fto expression. An over expression of Fto protein by 2.5-fold in the ARC results in a 14% decrease in average daily food intake in the first week. In contrast, knocking down Arc Fto expression by 40% increases food intake by 16%. mRNA levels of Agrp, Pomc and Npy, ARC-expressed genes classically associated with the control of food intake, were not affected by the manipulation of Fto expression. However, over expression of Fto resulted in a 4-fold increase in the mRNA levels of Stat3, a signalling molecule critical for leptin receptor signalling, suggesting a possible candidate for the mediation of Fto's actions. These data provide further support for the notion that FTO itself can influence key components of energy balance, and is therefore a strong candidate for the mediation of the robust association between FTO intronic variants and adiposity. Importantly, this provide the first indication that selective alteration of FTO levels in the hypothalamus can influence food intake, a finding consistent with the reported effects of FTO alleles on appetite and food intake in man.


Introduction
Mutations leading to highly penetrant forms of human obesity frequently disrupt the central control of appetite and lead to increased food intake [1]. It will be important to establish whether common polymorphisms associated with obesity act similarly. Common variants within intron 1 of the fat mass and obesity associated gene (FTO) are strongly and consistently associated with human adiposity [2,3]. Ten subsequent studies have examined the influence of FTO variants on measures of appetite, food intake or energy expenditure [4][5][6][7][8][9][10][11][12][13]. While the obesity risk alleles are not associated with reduced energy expenditure, nine of these studies reported an association with increased appetite or measured ad libitum food intake [4,[6][7][8][9][10][11][12][13]. As yet, there is no evidence that the risk alleles influence FTO expression and the variants could conceivably influence adiposity through effects on more distant genes. The global ablation of Fto expression in mice [14] results in high early mortality but the surviving mice are small, with reduced lean and fat mass. The absolute food intake of Fto +/+ and Fto 2/2 mice is identical but when corrected for lean mass, Fto 2/2 have an increased food intake. Similarly Fto 2/2 mice have increased oxygen consumption when corrected for lean mass. More recent findings reported in a mouse with a missense loss of function ENU induced mutation in Fto [15] strongly suggest that FTO is likely to be involved in the control of energy balance.
FTO is widely expressed across multiple tissues, but is most highly expressed in brain and especially the hypothalamus [2,16]. We have previously shown that murine Fto mRNA levels are decreased by fasting and increased by feeding in the arcuate nucleus (ARC) of the hypothalamus [16]. In this study, we have extended this observation by demonstrating that consumption of a high fat diet results in an increase in ARC Fto expression. We then examined whether manipulating Fto levels in the ARC might have any effects on spontaneous food intake in rats. In parallel, we examined the effects of manipulating Fto levels in the nearby paraventricular nucleus (PVN), a site known to be important for the control of energy balance but where Fto expression was not altered by nutritional state [16]. We demonstrate that manipulation of Fto expression in the arcuate nucleus bi-directionally influences acute food intake in rats, implicating the brain and specifically the ARC in mediating at least some of the effects of FTO on energy balance.

Endogenous Fto Expression in the ARC
We have previously reported that Fto expression in the ARC is decreased with nutritional deprivation. We examined the effect of high fat feeding on the expression of Fto in hypothalamic nuclei. After 10 weeks of exposure to a high fat (45%) diet, body weight (HFD 549.5622.6g vs. chow 426.8633.3g; p = 0.01) and fat mass (retroperitoneal fat: HFD 1.360.2 vs. chow 0.760.1% of body weight; p = 0.03) were significantly increased in the high fat fed group without an increase in energy intake (HFD 9963.7 vs. chow 10662.7Kcal). These physiological changes was accompanied by an increase in the Fto mRNA levels in the ARC by 2.5 fold compared to rats provided with regular chow (Fig 1A). In order to determine whether Fto is co-expressed with any specific ARC neuronal population, we performed co-localisation studies of Fto with the classical ARC expressed anorexigenic pro-opiomelanocortin (Pomc) using double in situ hybridisation. We confirmed abundant expression of Fto in the ARC, and although its expression overlaps with that of Pomc (Fig 1B), Fto appears to be diffusely expressed throughout the ARC and is therefore likely to be present in many neuronal and non-neuronal populations. Thus, rather than manipulating Fto expression in specific ARC neuronal populations, we chose a strategy employing AAV mediated transfer of Fto cDNA and shRNA coupled with stereotactic injection in order to manipulate Fto expression within the ARC (Fig 1C).

Manipulation of Fto Expression within the ARC and PVN
In order to determine the optimum AAV serotype and titre, as well as to confirm the accuracy of intranuclear injections, AAV-2/ 5,2/7,2/8 and 2/9 with GFP were initially produced. AAV-2/7 provided most effective and specific transduction for the hypothalamic nuclei ( Fig 1C) and the time course for GFP expression was similar to that reported for other AAV serotypes reported such that the protein was expressed by day 3 following injection and lasted for at least 2 months. Thus all the experimental AAVs used for the following studies were generated using serotype 2/7.
To modulate Fto expression levels in the ARC and PVN, we generated four AAV 2/7vectors: i.) AAV mediated transfer of full length Fto cDNA was used for overexpression; ii.) shRNA targeting Fto mRNA for knockdown of gene expression and iii.) two different AAV-controls (empty vector for AAV-Fto and containing a scrambled shRNA sequence for AAV-shRNA). The rat Fto specific shRNA that we generated effectively inhibited the expression of endogenous Fto in hypothalamic GT1-7 cells by 90%, whereas transfection with scrambled control did not affect the production of the Fto protein (Fig 2A). 20610 12 vg/ml of AAV-Fto, AAV-shRNA and their respective controls were stereotactically injected bilaterally to transduce either the ARC or PVN (Fig 1C) of rats. To assess the specificity of AAV delivery, at the end of the study, 3 weeks after AAV injections, micro-punches from both the ARC or PVN were obtained for all experimental animal and tested for changes in Fto expression using real-time quantitative RT-PCR ( Fig 2B). From the same dissections, protein expression was measured by semiquantitative western blotting analysis ( Fig 2C). For the over expression studies, at the protein level, we achieved approximately 2.5 fold increase in total Fto levels in the ARC and a 1.5 fold increase in the PVN. Whilst for the knockdown experiments, Fto expression was decreased by 40% in both the ARC and the PVN compared to rats injected with scrambled AAV.

Effects of AAV-Fto and AAV-shRNA Microinjection in Hypothalamic ARC
We investigated the effect of Fto on feeding behaviour in rats. At the end of the experiment, we check for Fto expression as an indication for correct site of injection. Three animals (2 from the ARC study and 1 from the PVN study) were excluded from further analysis as expression of both Fto RNA and protein levels in these animals were not altered at least 5% as compared to controls. Over expression of Fto in the ARC resulted in a decrease in food intake, compared to rats injected with empty vector (Fig 3A, B). Conversely, a 40% reduction in Fto protein expression resulted in an increase in food intake (Fig 3C, D). Expressing the data as average daily food intake, the peak of the effect was reached within the first week: 14% reduction in the Fto over expressing group and 16% increase in food intake with knockdown ( Fig 3E). However, despite having an impact on food intake, there was no significant difference between the body weight of rats that received injection of the AAV-Fto and AAV-shRNA with respective controls (Fig 3G,  H). There was also no significant difference in fat mass ( Fig 3F).

Effects of AAV-Fto and AAV-shRNA Microinjection PVN
As a comparator, we also perturbed expression of Fto in the PVN. Interestingly, although over-expression of Fto in the PVN also resulted in a reduction of food intake (Fig 4A, B, D), a similar degree of Fto knockdown in the PVN had no effect on food intake ( Fig 4C, 4D).

Effects of Manipulating ARC Fto on Genes Involved in Energy Homeostasis
Using real-time quantitative RT-PCR we determined that the ARC-expressed genes classically associated with the control of food intake, namely, Agrp, NPY and Pomc were not affected by altered Fto expression (Fig 5). However, a 4-fold increase in the expression of the transcription factor Stat3, and a 5-fold reduction in Tyrosine Hydroxylase, was observed with Fto over expression in the ARC, suggest possible candidates for the mediation of Fto's action.

Discussion
Previously, we have demonstrated that Fto is highly expressed in hypothalamic nuclei critical for regulating energy homeostasis, and that its expression within the ARC is reduced following a 48hr fast. Here we report that a high fat diet results in an increase of ARC Fto expression, demonstrating that Fto levels can be bidirectionally regulated depending on nutritional status. Due to the diffuse nature of Fto expression within the ARC, we chose to manipulate Fto expression throughout the nuclei resulting in a bidirectional impact on food intake.
Lack of FTO from conception, in both humans [17] and mice [14], results in severe growth retardation and high levels of early mortality, accompanied in humans, but not mice, by multiple developmental anomalies. We have reported that FTO is a nuclear-localised 2-oxoglutarate and Fe 2+ dependent dioxygenase which is capable of demethylating 3-methyl thymine in nucleic acids [16] and that heterozygous, loss-of-function mutations in human FTO exist but are found in both lean and obese subjects [18]. Thus, FTO's true, physiological substrate(s) and function remains unclear.
However, evidence up-to-date suggests that the FTO variant confers a predisposition to obesity to be involved in the regulation of food intake rather than in the regulation of energy expenditure. As of the ten studies examining the influence of FTO variants, except for one based on questionnaires [5], all others reported an association with increased appetite whist three studies on energy expenditure showed no effects [4,6,8]. Despite this, the effects of global reduction in Fto function on energy homeostasis in mice are complex with Fto null and hypomorphic mice having low adipose stores with increased energy expenditure. Our findings demonstrate that reduction in Fto expression selectively in the ARC results in increased food intake, a finding consistent with the hyperphagia seen in mice lacking Fto and congruent with the fact that fasting (a strong stimulus to eating) reduces Fto mRNA levels in the ARC [16]. Over-expression of Fto in either ARC or PVN caused a reciprocal reduction in food intake. Our paradigm of stereotactic injection of AAV into the ARC or PVN does not permit correlations between physiological end points and levels of gene expression, because only a subset of the total neurons within each nuclei were infected. However, it is noteworthy that reduction in Fto expression in only a limited number of ARC cells is sufficient to affect food intake. mRNA levels of Agrp, Pomc and Npy, ARC-expressed genes classically associated with the control of food intake, were not affected by the manipulation of Fto expression. However, over expression of Fto resulted in a 4-fold increase in the mRNA levels of Stat3, suggesting a possible candidate for the mediation of Fto's actions within the ARC. Stat3 is a ubiquitous transcription factor  that is indispensable during early embryogenesis. In adult tissue, in addition to its critical role in leptin receptor signalling [19], Stat3 has also been shown to play crucial roles in a variety of biological functions including cell growth, apoptosis and cell motility [20]. The increase in Stat3 expression associated with overexpression of Fto and reduced food intake is consistent with its potential role in mediating leptin's anorectic effects. Tyrosine hydroxylase (TH) expression in the ARC was inversely modulated by Fto. It is tempting to speculate that Fto's effect on TH, may be relevant to the energy expenditure phenotype seen in Fto null mice by acting on the norepinephrein pathways or on the dopaminergic reward pathways for effects on food intake. However, the physiological relevance of these findings require further exploration.
In conclusion, our findings provide further support for the notion that FTO itself can influence key components of energy balance, most likely acting through central action on appetite, and is therefore a strong candidate for the mediation of the robust association between FTO intronic variants and adiposity. Importantly they provide the first indication that selective alteration of Fto levels in the hypothalamus can influence food intake, a finding consistent with the reported effects of FTO alleles on appetite and food intake in man.

Ethics Statement
All animal procedures were carried out in strict accordance with regulations and guidelines of the United Kingdom Home Office.

Animals
Adult male Wister rats (250-300g, Charles River) were single housed throughout the study. All rats were maintained under controlled temperature (22uC) and on a 12-h light, 12-h dark schedule (lights on 7:00-19:00). Standard chow (Special Diet Services, SDS) and water were available ad libitum. For the high fat  Response of each gene was measured 3 weeks following intra-ARC injection, data normalized to B-actin and expressed in terms of fold induction over its expression in controls. P-value was calculated using a two-tailed distribution unpaired Student's t-test. Data is represented as the mean 6S.E.M of at least 6 independent rats per group; *p,0.05. doi:10.1371/journal.pone.0008771.g005 diet study, groups of male Wister around 200g were put on a 45% fat diet (D12451, SDS) for 10 weeks. Prior to all procedures animals were allowed to acclimatise for at least one week.

Design and Construction of AAV Vectors
Six pairs of oligos of 19 nucleotides within the Fto coding region of the rat were designed using Ambion software and the efficacy of these sequences in knocking down the Fto transcript was tested in cultured GT1-7 cells. The two most effective pair of sequences were cloned into a hairpin structure. Briefly, the olionucleotide, containing either one of these 19 nucleotide sequences, separated by a short spacer from the reverse complement of the same sequence and a five thymidine termination signal, was cloned into the pSilencer expression vector system (Ambion) downstream of an U6 RNA polymerase III promoter. The U6 promoter, Fto hairpin sequence and terminator sequences were subcloned into the multicloning site of the pAAV-MCS backbone (Stratagene) to generate the corresponding AAV-shRNA vectors. An AAV vector expressing a scrambled shRNA sequence was used as a control for the knock-down study. To generate Fto expressing AAV vectors, the cDNA of murine Fto was cloned into an AAV backbone plasmid under the control of the ubiquitous hybrid promoter CAG that contains the chicken ß-actin promoter and CMV enhancer. Empty AAV vectors expressing no proteins were used as a control for the overexpression study. AAV-GFP was also produced to initially test for optimum serotype and titre, as well as accuracy of intranuclear injections.

Viral Production and Purification
Vectors were generated by helper virus-free transfection of HEK293 cells using three plasmids [21] with modifications [22]. Cells were cultured to 70% confluence in roller bottles (RB) (Corning) in DMEM supplemented with 10% FBS and then cotransfected with: 1) a plasmid carrying the expression cassette flanked by the viral ITRs (described above); 2) a helper plasmid carrying the AAV rep and cap genes; and 3) a plasmid carrying the adenovirus helper functions. In this study we initially used plasmids carrying serotype 5,7,8,9 cap genes to generate AAV-2/5,2/7,2/8, and 2/9-GFPs to obtain optimal viral infection for the hypothalamic nuclei. GFP immunohistochemistry was performed 7 days after injections. AAV-2/7 seemed best for the hypothalamus, thus all the experimental virus were generate using serotype 7 cap genes. Plasmids carrying the adenovirus helper functions and cap genes were kindly provided by K.A. High, Children's Hospital of Philadelphia. Vectors were purified by two consecutives cesium chloride gradients using an optimized method [23], dialyzed against PBS, filtered, titred by qPCR and stored at 280uC until use.

Stereotactic Surgery
Weight-matched rats were anesthetized with 40mg/kg ketamine and 20mg/kg xylazine, injected intramuscularly. Under anesthesia, a temporary 31 gauge needle (Hamilton) was inserted into the ARC or PVN according to the stereotactic coordinates of Paxinos and Watson [24]. The stereotactic coordinates for injection into the ARC were 2.8mm caudal to bregma, 0.4mm lateral to the midline and 10mm below the surface of the skull; for PVN 1.8mm caudal to bregma, 0.5mm lateral to the midline and 8mm below the surface of the skull. With a 10ul Hamilton micro-syringe, 1ul of 2610 12 vg/ml vector preparations were injected into the nuclei over a 10min period. After vector delivery, the needle was left in place for 5 min to prevent reflux. Food intake and body weight were measured daily for two weeks following AAV microinjection, and were analyzed beginning on day 4 to allow time for postoperative recovery and for adenoviral gene to be expressed. At the end of the study, 3 weeks after AAV injection, whole brains were rapidly isolated and frozen on dry ice. Mesenteric and retroperitoneal fat pads were also dissected and weighted to the nearest 2 decimal place.

Histological Analysis
For double-labelling of tissue sections with both isotopic and non-isotopic cRNA probes, a modification of our previously described protocol to detect Fto expression [16] was employed. Briefly, brains were sliced at 14um thickness and mounted on poly-L-lysine-coated slides (Fisher Scientific), dried overnight and then stored at 280uC. Slides were fixed in 4% paraformaldehyde followed by a microwave pretreatment with 10mM sodium citrate buffer, pH6.0 and dehydrated using a graded concentrations of ethanol (50-100%). 35 S-labelled Fto cRNA was prepared as previously described [16], in addition, as a counterprobe, a digoxigenin (DIG)-labelled Pomc cRNA probe was prepared. Diglabelled probes for sense and antisense Pomc mRNA were generated from the cDNA template corresponding to exon3 of Pomc in a pGem-T vector (Promega) and transcribed by SP6 or T7 RNA polymerase (Progema). 35 S-labeled Fto probes and DIGlabelled Pomc probes were synthesized in parallel with either 35 Slabeled UTP (GE Healthcare) or DIG-labelling mixture containing 10 mM ATP, CTP, GTP; 6.5 mM UTP; 3.5 mM DIGlabelled UTP (Boehringer Mannheim). Both probes were added to the hybridization buffer at the same concentration and sections were hybridized overnight at 57uC. After hybridization, slides were RNase treated followed by immunohistochemical detection of DIG. This was achieved by incubation of the brain section in blocking buffer containing 1:500 dilution of AP conjugated anti-DIG antibody (Boehringer Mannheim) follow by the incubation in the detection buffer containing NBT, BCIP and levamisole. Slides were then covered with K5D emulsion (Ilford) and exposed for at least 2 weeks for the detection of 35 S signal. Cells were considered positive for a message (for either DIG or 35 S) if the signal was at least twice the background level. To assess differential sensitivity between isotopic and non-isotopic probes, the experiment was run with reversed labelled probe (i.e. Fto cRNA labelled with DIG and Pomc labelled with 35 S).
To validate injection site and to establish a time course of protein expression for AAV2/7 serotype, at 3, 7 and 60 days after AAV-GFP injection, rats were anesthetized and perfused with 4% paraformaldehyde. Brains were fixed with 4% paraformaldehyde by transcardial perfusion and processed to produce 35um coronal cryostat sections. Successful injection and optimal virus titer were determined by the presence of GFP producing cells in the ARC and the PVN.

RT-PCR
Micro-punches of ARC and PVN from individual rats were obtained based on the method originally described by Palkovits [25] with the shape of the basal hypothalamus, the ventricles and the major tracts serving as landmarks for localization during the process of removal. Accuracy of the micro-punches was confirmed with cresyl violet staining on the cryostat sections after nuclei removal. RNA was isolated from the micro-punch using the PARIS TM system (Ambion) and 200ng of purified RNA were used in a random-primed first strand cDNA synthesis reaction, using Superscript III reverse transcriptase (Invitrogen). Quantitative PCR reactions for Fto were performed in triplicate on an ABI 7900HT (Applied Biosystems) and using ABI PCR master mix, according to manufacturer's protocols. Arcuate expression of Agrp, Pomc and Npy was also quantified by real time PCR in these same samples, but the RNA had to be amplified before being used as template. Briefly, 5ng total RNA was DNase treated (TURBO DNase TM , Ambion) and subjected to the WT-OVATION TM Pico RNA Amplification System (NuGen Technologies, Inc) according to manufacturer's instructions. For the real time quantitative PCR data, all statistics were done on delta Ct values normalized to Bactin and expressed as ratio of control.

Western Blot Analysis
Lysates for western blots were prepared using the PARIS TM kit with a protease inhibitor cocktail (Sigma). Western blots were probed with mouse monoclonal antibodies directed against the Cterminal FTO (1:500; Abcam). B-actin (1:1000; Abcam) was used as a loading control. Bands were visualized and semi-quantified using Chemoluminescence (Syngene).

Statistical Analysis
All values are expressed as mean 6S.E.M. For each variable, statistical comparison between control and treated groups for each site were performed by two-tailed unpaired Student's t-test. For the analysis of food intake and body weight over time, two-way repeated measures ANOVA was used with time and treatment as variables for comparison. Statistical analysis was performed using Graph Pad Prism software (GraphPad Prism).