Functional magnetic resonance imaging and c-Fos mapping in rats following a glucoprivic dose of 2-deoxy-d-glucose

Authors


Address correspondence and reprint requests to Simon M. Luckman, Faculty of Life Sciences, AV Hill Building, University of Manchester, Oxford Road, Manchester M13 9PT, UK. E-mail: simon.luckman@manchester.ac.uk

Abstract

J. Neurochem. (2010) 113, 1123–1132.

Abstract

The glucose analogue, 2-deoxy-d-glucose (2-DG) is an inhibitor of glycolysis and, when administered systemically or centrally, induces glucoprivation leading to counter-regulatory responses, including increased feeding behaviour. Investigations into how the brain responds to glucoprivation could have important therapeutic potential, as disruptions or defects in the defence of the brain’s ‘glucostatic’ circuitry may be partly responsible for pathological conditions resulting from diabetes and obesity. To define the ‘glucostat’ brain circuitry further we have combined blood-oxygen-level-dependent pharmacological-challenge magnetic resonance imaging (phMRI) with whole-brain c-Fos functional activity mapping to characterise brain regions responsive to an orexigenic dose of 2-DG [200 mg/kg; subcutaneous (s.c.)]. For phMRI, rats were imaged using a T2*-weighted gradient echo in a 7T magnet for 60 min under α-chloralose anaesthesia, whereas animals for immunohistochemistry were unanaesthetised and freely behaving. These complementary methods demonstrated functional brain activity in a number of previously characterised glucose-sensing brain regions such as those in the hypothalamus and brainstem following administration of 2-DG compared with vehicle. As the study mapped whole-brain functional responses, it also identified the orbitofrontal cortex and striatum (nucleus accumbens and ventral pallidum) as novel 2-DG-responsive brain regions. These regions make up a corticostriatal connection with the hypothalamus, by which aspects of motivation, salience and reward can impinge on the hypothalamic control of feeding behaviour. This study, therefore, provides further evidence for a common integrated circuit involved in the induction of feeding behaviour, and illustrates the valuable potential of phMRI in investigating central pharmacological actions.

Abbreviations used:
2-DG

2-deoxy-d-glucose

AcbC

nucleus accumbens core

AcbSh

nucleus accumbens shell

Arc

arcuate nucleus of the hypothalamus

BNST

bed nucleus of the stria terminalis

BOLD

blood-oxygen-level-dependent

CeA

central amygdala

DMN

dorsomedial nucleus

IC

inferior colliculus

LC

locus coeruleus

LH

lateral hypothalamus

NTS

the nucleus of the tractus solitarius

PAG

periaqueductal grey

PB

phosphate buffer

PBN

parabrachial nucleus

phMRI

pharmacological-challenge magnetic resonance imaging

PRN

pontine reticular nucleus

PVA

thalamic paraventricular nucleus

PVN

paraventricular nucleus

SO

superior olive

SPM5

statistical parametric mapping

VMH

ventromedial hypothalamus

VMN

ventromedial nucleus

VP

ventral pallidum

Glucose is a vital component of cellular function, and the maintenance of an adequate supply of glucose is of major importance to both central and peripheral tissues to support life (Levin et al. 1999; Williams et al. 2001). The constant level of glucose in the blood is regulated homeostatically by the neural and endocrine interplay between the brain and the regulatory mechanisms it controls in peripheral organs, notably the liver and the pancreas (Anand et al. 1964; Oomura et al. 1964; Donovan et al. 1991; Hevener et al. 1997; Cryer et al. 2003). It is likely that disruption or defects in the defence of a brain’s glucostatic ‘set-point’ may contribute to a number of pathological conditions, including diabetes and obesity (Levin and Sullivan 1987; Levin et al. 1999). Further understanding of how the brain regulates this glucose set-point could, therefore, have important therapeutic potential.

Research over the past four decades has identified a subset of neurones within the brain that are able to modulate their firing activity in response to changes in extracellular glucose levels (Anand et al. 1964; Oomura et al. 1969; Oomura and Yoshimatsu 1984; Dunn-Meynell et al. 2002; Kang et al. 2004; Levin et al. 2004). These neurones can be split broadly into two populations: those that increase their firing rate in response to elevations in extracellular glucose concentration (termed glucose excited); and those which are activated by a decrease in extracellular glucose concentration or cellular glucoprivation (termed glucose inhibited) (Song et al. 2001; Routh 2002; Thorens 2003; Yang et al. 2004; Burdakov et al. 2005). Both types of neurone are widely distributed in the brain, but are particularly represented in the hypothalamus and the brainstem, regions involved in the control of energy homeostasis and food intake (Oomura et al. 1974; Mizuno and Oomura 1984; Kow and Pfaff 1985; Silver and Erecinska 1998).

Glucose-sensing neurones are thought to act as sentinels of the body’s glucostatic ‘set-point,’ so that in times of hypo- or hyperglycaemia, they initiate a constellation of integrated counter-regulatory responses, the most notable of which are modulation of the hypothalamo-pituitary-adrenal axis, glucagon/insulin secretion, and energy intake (DiRocco and Grill 1979; Borg et al. 1995; Hevener et al. 1997; Ritter et al. 2000). The physiological importance of central glucose sensing has been further highlighted in a number of studies whereby peripheral injection of gold thioglucose, which selectively destroys glucose-responsive neurones in brain regions with a restricted blood-brain barrier, results in impaired feeding regulation and the subsequent onset of obesity (Bergen et al. 1996; Homma et al. 2006).

To understand the brain mechanisms underlying glucose homeostasis, particularly in the context of glucoprivation, many studies have used the glucose anti-metabolite, 2-deoxy-d-glucose (2-DG). 2-DG is an inhibitor of glycolysis and, when administered systemically or centrally, induces glucoprivation leading to counter-regulatory responses, including increased feeding behaviour (Brown 1962; Novin et al. 1973; Miselis and Epstein 1975; Berthoud and Mogenson 1977; Marty et al. 2007). Importantly, the increase in feeding behaviour can be blocked by central, but not peripheral administration of glucose, suggesting that the metabolic receptors responsible for glucoprivation-induced feeding are located primarily within the brain (Singer and Ritter 1996; Burdakov et al. 2005).

Studies with cellular resolution (electrophysiology and c-Fos immunohistochemistry) have demonstrated the existence of 2-DG-responsive neurones in the arcuate (Arc), ventromedial (VMN), paraventricular (PVN), and lateral hypothalamic (LH) nuclei, the nucleus of the tractus solitarius (NTS), parabrachial nucleus (PBN), and regions of the basolateral medulla containing A1/C1 noradrenergic and adrenergic neurones (Ritter and Dinh 1994; Dallaporta et al. 1999; Briski and Sylvester 2001; Yang et al. 2004). However, these studies are restricted to focusing on the above-mentioned nuclei. In an attempt to characterise the whole-brain response to 2-DG-induced glucoprivation and potentially identify connections with the hypothalamus and the brainstem, the present study combines blood-oxygen level-dependent (BOLD) pharmacological-challenge magnetic resonance imaging (phMRI) with c-Fos protein functional activity mapping, providing an insight into the whole-brain response to systemic 2-DG administration and the co-ordination of counter-regulatory responses. The study also illustrates the valuable potential of BOLD phMRI in investigating central drug action.

Materials and methods

Animals

All experiments were carried out using adult male Sprague-Dawley rats (Charles River Laboratories, Inc., Sandwich, UK). Animals were group housed in The University of Manchester animal unit in a constant environment of 21 ± 2ºC and 45 ± 10% humidity, on a 12:12 hours light-dark cycle with the dark phase commencing at 20:00 hours. Rat chow (Beekay International, Hull, UK) and tap water were available ad libitum unless stated otherwise. All procedures conformed to the requirements of the UK Animals (Scientific Procedures) Act, 1986 and local ethical review.

Feeding and c-Fos protein immunohistochemistry

Rats (275 ± 20 g; n = 5/6 per group) were housed singly 5 days prior to the experiment and received food and water ad libitum. During this acclimatisation period, rats were handled daily and food intake monitored. Rats were assigned randomly to receive subcutaneous (s.c.) injections of either vehicle (0.9% NaCl) or 200 mg/kg body weight (b.w.) 2-DG (Sigma-Aldrich Corp. Ltd., Poole, UK) between the hours of 09:00 and 14:00. All injections were given in a volume of 1 mL/kg body weight. The dose of 2-DG was based on literature (Nonavinakere and Ritter 1983; Tepper and Kanarek 1984; Giraudo et al. 1998) and previous in-house experiments. Food was weighed just before injection and again 90 min later before the animals were culled. The animals were deeply anaesthetized with 5% isoflurane (Concord Pharmaceuticals Ltd., Dunmow, Essex, UK) in oxygen (1 L/min) and perfused transcardially with heparinised saline (10 000 units/L heparin in 0.9% NaCl), followed by 4% paraformaldehyde in phosphate buffer (PB, 0.1M, pH 7.3). The brains were post-fixed overnight and then kept for 2 days in 30% sucrose in 0.1 M PB to cryoprotect the tissue, before freezing on dry ice. Thirty micrometer sections (120 μm apart) were cut in the coronal plane throughout the entire rostrocaudal extent of the brain and incubated in 20% methanol, 0.2% Triton X-100, 1.5% hydrogen peroxide for 30 min to deactivate endogenous peroxidases. Sections were then incubated at 23°C for 1 h in the blocking buffer: 0.1 M PB, 0.3% Triton X-100, 1% normal sheep serum, and then overnight at 4ºC in rabbit anti-c-Fos antibody (Oncogene Science Inc., Bayer Healthcare, MA, USA) diluted to 1 : 10 000 in blocking buffer. After washing, the sections were incubated sequentially at 23°C for 1 h in goat anti-rabbit IgG-biotin complex (Vector Laboratories, CA, USA) diluted 1 : 500 in blocking buffer followed by avidin-biotin–peroxidase complex (GE Healthcare, Buckinghamshire, UK) diluted 1 : 500 in PB and, finally, visualized with nickel-intensified diaminobenzidine (Vector Laboratories, UK).

c-Fos immunoreactivity was first examined qualitatively to determine areas expressing c-Fos-positive neurones. Areas that showed qualitative changes in immunoreactivity were then analysed blind by counting c-Fos expressing nuclei in the areas of interest. The areas that showed qualitative changes were the nucleus accumbens core (AcbC), nucleus accumbens shell (AcbSh), ventral pallidum (VP), thalamic paraventricular nucleus (PVA), hypothalamic PVN, Arc, VMN, dorsomedial nucleus (DMN), LH, central amygdala (CeA), periaqueductal grey (PAG), and NTS. Brain areas were photographed using an Axiovison upright microscope (Zeiss, Hertfordshire, UK) and an Axiocam colour CCD camera (Zeiss). The number of c-Fos-expressing cell nuclei was quantified in areas defined according to a standard atlas (Paxinos and Watson, 1986). Results are presented as mean and SEM for food intake at 90 min and the number of c-Fos-immunoreactive cells per section in each brain area. Treatments were compared using the two-way unpaired t-test using the Prism statistical package (GraphPad Software Inc, San Diego, CA, USA).

Blood-oxygen-level-dependent fMRI

Fourteen rats (260 ± 25 g, n = 7) were assigned randomly to receive subcutaneous injections of vehicle (0.9% w/v NaCl) or 2-DG (200 mg/kg) (Sigma-Aldrich). Animals were anaesthetized with 2.5% isoflurane (Concord Pharmaceuticals) in oxygen (2 L/min) to allow cannulation of a tail vein and subsequent anaesthetic maintenance by intravenous (i.v.) α-chloralose-HBC (Sigma-Aldrich). A bolus of α-chloralose (60 mg/kg body weight; i.v.) was injected manually over a period of 5 min whilst the isoflurane and oxygen were turned off. Then α-chloralose was infused continuously at a rate of 30 mg/kg/h i.v. by infusion pump for the remainder of the experiment. For imaging, rats were secured into an in house-built cradle with a nose cone to minimize movement. Temperature (RS 51 K-type thermometer; RS Components Ltd, Northants, UK), respiration rate (MR10 respiration monitor; Graseby Medical Ltd, Hertfordshire, UK) and transcutaneous pCO2 and transcutaneous pO2 were monitored (see Appendix S1), whereas the rats were allowed to breathe spontaneously. Imaging was carried out using a 7-Tesla, horizontal-bore magnet (Magnex Scientific Ltd., Abingdon, UK) with a transmit/receive birdcage volume coil connected to a SMIS computer console (Surrey Medical Imaging Systems Limited, Guildford, UK). For anatomical reference images, a T2-weighted fast spin echo was used (repetition time = 2 s, flip angle = 90º, base echo time = 30 ms, effective echo time = 60 ms, number of samples = 256, number of views = 128, number of averages = 16). For functional images, a T2*-weighted gradient echo was used to measure BOLD signal (repetition time = 172 ms, echo time = 15 ms, number of samples = 128, number of views = 64, number of averages = 4, voxel size = 0.313 mm × 1 mm × 0.313 mm, each volume took 70 s to acquire). Eleven contiguous slices, each of 1 mm thickness were aligned horizontally through the brain (Paxinos and Watson, 1986). A total of 60 brain volumes over a period of 70 min were acquired in all. 2-DG or vehicle was administered during volume 12.

Data were analysed with Statistical Parametric Mapping (SPM5) programme using a random effects model (The Wellcome Trust Centre for Neuroimaging, London, UK; (http://www.fil.ion.ucl.ac.uk/spm/software/spm5/). Individual brains were re-aligned and co-registered to the first volume, spatially normalized and smoothed to a full width half maximum of 0.939 mm isotropic Gaussian kernel. In a first-level analysis, a series of contrasts were constructed between five successive time blocks each consisting of 12 consecutive volumes (14 min each). The contrasts compared time blocks following injection (4 time blocks, each 12 volumes) to that of the pre-infusion period (1 time block, 12 volumes). These images were combined in a second-level random effects analysis using a two sample t-test. T contrasts were then constructed to discern the positive and negative effects of 2-DG compared with vehicle. The resulting T2 contrast statistical parametric maps were overlaid onto a T2-weighted anatomical template image (Schwarz et al. 2006), with a threshold level of p < 0.05 uncorrected. For the unbiased (operator-independent) identification of the BOLD MRI data, regions of interest were delineated using a 3D digital reconstruction of the Paxinos and Watson rat brain atlas (Paxinos and Watson 1998), co-registered with the rat brain template (Pic atlas) (Schwarz et al. 2006). Only clusters within regions containing ≥ 3 voxels were considered for further analysis. To provide a measure of response in these areas, Z scores and mean percentage BOLD contrast changes for the maximally responding voxel in each cluster were obtained using SPM5.

Results

Effects 2-DG on food intake and c-Fos immunoreactivity in freely-behaving rats

An unpaired, two-tailed t-test revealed a significant increase in food intake in 2-DG-treated animals compared with controls, 90 min post-injection (= 0.0024, Fig. 1). As this dose gave a robust orexigenic response, it was also used in the subsequent phMRI experiment.

Figure 1.

 Bar graph illustrating total food intake at 90 min post-injection of animals treated with vehicle or 2-deoxy-d-glucose (2-DG) (200 mg/kg b.w.). Each treatment was administered subcutaneously. Error bars show SEM. Unpaired two-tailed t-test. **< 0.01 compared with vehicle (n = 5–6/group).

Quantitative analysis of the number of c-Fos-positive neurones in each of the brain areas of interest showing qualitative activity (Fig. 2) revealed a significant increase in counts following 2-DG administration compared with vehicle. Increased numbers of c-Fos-positive cells were seen in the AcbC (= 0.0098), AcbSh (< 0.0098, Fig. 3a and d), VP (< 0.0001), PVA (< 0.0001, Fig. 3b and e), PVN (< 0.0001, Fig. 3c and f), Arc (= 0.0134, Fig. 4a and d), DMN (< 0.0001, Fig. 4b and e), LH (< 0.0001), CeA (= 0.0019, Fig. 4c and f), PAG (< 0.0001) and NTS (< 0.0001). None of the brain areas analysed showed a significant decrease in c-Fos immunoreactivity following 2-DG administration. No statistically significant difference between groups was seen in the VMN (= 0.1681).

Figure 2.

 Graph illustrating the number of c-Fos-positive neurones per section in brain regions after subcutaneous administration of vehicle or 2-deoxy-d-glucose (2-DG) (200 mg/kg b.w.). Bars represent mean and SEM (n = 5/group). Data analysed using an unpaired two-tailed t-test. *< 0.05, **< 0.01, ***< 0.001 compared with vehicle. Abbreviations; nucleus accumbens core (AcbC), nucleus accumbens shell (AcbSh), ventral pallidum (VP), thalamic paraventricular nucleus (PVA), hypothalamic paraventricular nucleus (PVN), arcuate nucleus (Arc), ventromedial nucleus (VMN), dorsomedial nucleus (DMN), lateral hypothalamus (LH), central amygdala (CeA), periaqueductal grey (PAG), and the nucleus of the solitary tract (NTS).

Figure 3.

 Photomicrographs showing c-Fos-labelled nuclei in selected brain areas following administration of vehicle (a, b, c), or 2-deoxy-d-glucose (200 mg/kg b.w., d, e, f). Abbreviations; anterior commissure (ac), nucleus accumbens shell (AcbSh), hypothalamic paraventricular nuclei (PVN), and thalamic paraventricular nuclei (PVA). Scale bars = 200 μm.

Figure 4.

 Photomicrographs showing c-Fos-labelled nuclei in selected brain areas following administration of vehicle (a, b, c), or 2-deoxy-d-glucose (200 mg/kg b.w., d, e, f). Abbreviations; third ventricle (3V), central amygdala (CeA), hypothalamic arcuate (Arc), dorsomedial nucleus (DMN), and ventromedial nucleus (VMN). Scale bars = 200 μm.

phMRI: Effects of 2-DG on BOLD signal in α-chloralose-anaesthetized rats

In the present study we show that administration of 2-DG induced significant positive BOLD signal in the agranular insular cortex, somatosensory cortex, hippocampus, inferior colliculus (IC), olfactory tubercle, caudate putamen, septum, substantia innominata, Arc, VMN, mesencephalic region, locus coeruleus (LC), PBN, pontine reticular nucleus (PRN), and the superior olive (SO) (Table 1, Fig. 5).

Table 1.   Regions of significant blood-oxygen-level-dependent activation relative to vehicle following administration of 2-DG (200 mg/kg, s.c.) detected by pharmacological-challenge magnetic resonance imaging (n = 7). Columns show Z scores, for the peak-responding voxel in brain areas showing significant changes in blood-oxygen-level-dependent signal following 2-DG treatment (analysed using a two-sample t-test)
Brain RegionphMRI Z-Score
PositiveNegative
  1. BNST, bed nucleus of the stria terminalis; IPAC, interstitial nucleus of posterior limb of the anterior commissure; phMRI, pharmacological-challenge magnetic resonance imaging.

Cortex
 Agranular insular2.12
 Cingulate cortex2.62
 Entorhinal cortex2.183.74
 Medial pre-frontal cortex 2.41
 Somatosensory cortex2.39
 Ventral orbitofrontal cortex3.11
Hippocampus
 Hippocamopus medial3.15
 Hippocampus postereodorsal2.36
 Hippocampus subiculum3.4
 Hippocampus ventral3.23
 Hippocamus antereodorsal2.57
Olfactory region
 Olfactory nucles3.74
 Olfactory tubuicle2.77
Basal ganglia
 Accumbens core3.91
 Accumbens shell3.96
 BNST3.92
 Caudate putamen2.123.34
 Diagonal band3.71
 IPAC2.91
 Septum1.79
 Substantia innominata2.15
 Ventral pallidum3.34
Amygdala
 Central amygdala2.55
Thalamus
 Dorsolateral thalamus2.67
 Inferior colliculus3.89
 Midline thalamus2.37
 Zonalen certa3.05
Hypothalamus
 Arcuate nucleus3.15
 Dorsomedial nucleus1.9
 Lateral hypothalamus3.05
 Paraventricular nucleus3.15
 Ventromedial nucleus3.1
Midbrain
 Mesencephalic region3.06
 Periaqueductat grey2.58
pons
 Locus cerellious2.33
 Parabrachial nucleus2.17
 Pontine reticulate nucleus2.56
 Raphe nucleus2.58
 Superior olive3.36
Figure 5.

 Group (n = 7) statistical parametric maps showing changes in blood-oxygen-level-dependent (BOLD) contrast, with a significance threshold set to < 0.01 uncorrected, following acute administration of 2-deoxy-d-glucose (2-DG) (200 mg/kg, s.c.). BOLD blobs in red indicate regions of increased activity compared with vehicle, whereas blobs in blue are regions of decreased activity. The colour bar represents t-values. Values above images represent approximate distances from bregma (mm).The histogram represents the total area under the BOLD signal percentage change curve produced following treatment with vehicle or 2-DG for each of the corresponding brain regions. Bars represent mean and SEM. Arc, arcuate nucleus; CeA ,central amygdala; DMN, dorsomedial hypothalamic nucleus; IC, inferior colliculus, LC, locus coeruleus LH, lateral hypothalamus; AcbShell, nucleus accumbens shell; Orb, orbitofrontal cortex; PAG, periaqueductal grey; PBN, parabrachial nucleus; PRN, pontine reticular nucleus; PVN, paraventricular hypothalamic nucleus; SO, superior olive; VP, ventral pallidum; VMN, ventromedial nucleus.

In addition, administration of 2-DG induced significant negative BOLD signal in the cingulate cortex, entorhinal cortex, medial pre-frontal cortex, ventral orbitofrontal cortex, ventral hippocampus, olfactory nucleus, AcbC, AcbSh, caudate putamen, bed nucleus of the stria terminalis (BNST), diagonal band, nucleus of posterior limb of the anterior commissure, VP, CeA, thalamus, zona incerta, DMN, LH, PVN, PAG and raphé nucleus (Table 1, Fig. 5).

phMRI: blood gases

As detailed in Appendix S1, blood O2 and CO2 were measured transcutaneously throughout the imaging experiments. CO2 was very stable in all animals throughout the experiment (before and after infusion of 2-DG or vehicle, Figure S1a). Oxygen was more variable, but there was no evidence for systematic effects of time or infusion of either vehicle or 2-DG (Figure S1b). Further details are given in the Appendix S1.

Discussion

The central sites mediating 2-DG-induced responses are poorly understood outside the realm of the classic homeostatic centres of the hypothalamus and caudal brainstem. In the present study, we have shown ‘whole-brain’ responses to a behaviourally relevant glucoprivic dose of 2-DG, using the complementary techniques of BOLD phMRI and c-Fos protein functional activity mapping.

Concurring with previous studies, intraperitoneal administration of 200 mg/kg 2-DG resulted in acute hyperphagia (King et al. 1978; Tepper and Kanarek 1984; Ritter and Taylor 1989, 1990; Giraudo et al. 1998). Functional brain activity following 2-DG administration was compared with vehicle injections, using the neuronal activity marker protein, c-Fos. 2-DG increased c-Fos expression in the PVA, PVN, Arc, VMN, DMN, LH, CeA, and the NTS, supporting previous studies investigating 2-DG-induced c-Fos immunoreactivity (Ritter and Dinh 1994; Solomon et al. 2006). Previous studies, however, limit reporting to selected brain areas and do not provide whole-brain coverage. The current study provides a complete whole-brain analysis and demonstrates, additionally, increased c-Fos immunoreactivity in the striatum (VP, AcbC, AcbSh) and PAG following treatment with 2-DG. A clear deviation between this and previous studies is the lack of c-Fos induction in some regions of the pons and medulla oblongata. Other, similar studies have reported robust c-Fos immunoreactivity in the PBN, LC, SO and ventrolateral medulla (Ritter and Dinh 1994; Ritter et al. 1998). The latter have tended to use higher doses of 2-DG and to remove food from the animals after injection, which might affect the response. However, interestingly, although these areas were not detected in the present study using c-Fos immunohistochemistry, they were identified using phMRI, highlighting complementarity between the two techniques.

c-Fos immunohistochemistry provides high spatial resolution and whole-brain coverage, although c-Fos expression does not automatically follow neuronal activity (Luckman et al. 1994), which can lead to false negative results. In addition, c-Fos is seldom translated in response to reduced activity so, in these instances, neuronal inhibition can not be discerned (Hughes and Dragunow 1995). c-Fos protein induction occurs over a period of 30–90 min, thus limiting temporal resolution. For these reasons, it is essential to complement functional immunohistochemistry with other functional imaging techniques to provide additional data and to aid interpretation. Systemic administration of 2-DG produced significantly enhanced BOLD activity in several brain areas including the frontal cortices (orbitofrontal, cingulate, insular cortex), mesolimbic system (AcbC, AcbSh, VP, BNST), striatum (caudate putamen, globus pallidus), amygdala (CeA), thalamus (PVA, IC), midbrain (PAG), hypothalamus (Arc, DMN, LH, PVN, VMN) and pons (LC, PBN, PRN, SO).

In terms of functionality, the combination of BOLD and c-Fos immunohistochemistry highlights a number of regions that show changes in both measures following 2-DG. Positive BOLD signals are triggered by oxygen depletion in response to increased metabolic demand of neuronal firing and synaptic activity [for review see (Logothetis and Wandell 2004; Nair 2005)]. In fact a linier relationship between positive BOLD and neuronal activity has been described (Heeger et al. 2000; Rees et al. 2000). Therefore areas showing increased signal using both techniques following 2-DG administration in the present study (Arc and VMN) potentially reflect increased synaptic and cellular electrical activity.

The functional origins underlying the negative BOLD signal are far more speculative, as they are thought to arise from a complex interplay between decreased neuronal metabolic demand (Shmuel et al. 2002, 2006; Devor et al. 2007) and the ‘vascular-steel’ effect (Harel et al. 2002; Shmuel et al. 2002). Despite this, numerous studies show strong evidence to suggest that negative BOLD is indicative of a true suppression of neuronal activation (Shmuel et al. 2002; Stefanovic et al. 2004; Devor et al. 2007). Taken in this context, the brain areas depicted in the present study showing a decrease in BOLD signal and an increase in c-Fos immunoreactivity following 2-DG administration (AcbC, AcbSh, VP, DMN, PVN, LH, CeA, PVA, PAG) potentially reflect decreased synaptic activity and increased cellular activity; an effect which could be occurring because of disinhibition in these regions.

The present study does, however, illustrate clear evidence of dissociation between BOLD fMRI and c-Fos immunoreactivity in a number of brain regions. For example, the IC showed a robust change in BOLD activity but failed to show any changes in c-Fos immunoreactivity. The possible reasons for this dissociation are discussed in detail elsewhere (Stark et al. 2006; Preece et al. 2009).

In light of the widespread effects of 2-DG on metabolism, it is interesting to note the relatively small number of brain sites expressing significant c-Fos immunoreactivity and BOLD activity in response to 2-DG. In addition, previous studies have found that both 2-DG-induced hyperphagia and c-Fos immunoreactivity persist in many of the brain areas described above following subdiaphragmatic vagotomy, implying that these effects are a consequence of the direct central metabolic effect of 2-DG (Miselis and Epstein 1975; Ritter and Dinh 1994). The functional responses to 2-DG observed in the current study may represent direct actions of 2-DG on neurones in each of these brain areas, or may represent transynaptic activation of these sites following more selective activation of specific populations of specialized glucoreceptive neurones within the brain.

Interpretation of the current results in the context of feeding centres of the brain highlights the hypothalamus and hindbrain as potential sites of action underlying the hyperphagia. Studies attempting to establish the primary site coordinating to the counter-regulatory hyperphagic response initiated by 2-DG-induced hypoglycaemia, have resulted in the emergence of two separate views. A substantial body of evidence points to an integral role of the ventromedial hypothalamus (VMH, comprising both the Arc and the VMN). Breakthrough studies demonstrated that perfusion of glucose directly into the VMH of systemically-induced hypoglycaemic animals abolishes the normal counter-regulatory responses, whereas infusion of 2-DG into the VMH caused a prompt increase in plasma glucose, glucagon, and catecholamines (Borg et al. 1994, 1995), suggesting that the neurones sensing glucopenia may be localised in the VMH. A recent study found that manipulation of VMH glucosensing by blocking VMH glucokinase mRNA expression reduced glucoprivic feeding, but exerted no effect on spontaneous feeding behaviour, implying behavioural specificity in the role of the VMH in food intake (Dunn-Meynell et al. 2009). In light of this evidence, the significant functional activation of the VMH by 2-DG seen in the present study further illustrates that VMH glucosensing plays an important role in the counter-regulatory behavioural and neuroendocrine responses to glucoprivation.

A second body of evidence points to a primary role for the hindbrain (DiRocco and Grill 1979). Direct injection of 2-DG into the LH, VMH, amygdala or striatum has no effect on feeding behaviour, despite a marked response to intracerebroventricular administration (Berthoud and Mogenson 1977). This intracerebroventricular injection of 2-DG failed to activate feeding in the presence of an obstruction to the cerebral aqueduct (Ritter et al. 1981), suggesting that 2-DG-induced glucopenia may be sensed pre-dominantly by neuronal populations in the hindbrain. Food intake can be stimulated by direct injections of the glucose anti-metabolite 5-thioglucose into areas of the hindbrain, such as the ventrolateral and dorsomedial medulla (Ritter et al. 2000). These regions, along with other hindbrain nuclei highlighted by phMRI in the present study (LC, PBN, PRN), contain neurones that respond either directly or indirectly to a glucoprivic dose of 2-DG (Ritter and Dinh 1994; Ritter et al. 1998). Neurones in the NTS sense glucose both directly and indirectly via vagal afferents from peripheral glucosensors in the hepatic portal vein. Confirming previous reports (Ritter and Dinh 1994; Moriyama et al. 2003), the current study found a marked increase in c-Fos immunoreactivity in the NTS, but no significant change in BOLD signal. There are various explanations for this. First, there may be a drug/anaesthetic interaction in this area not seen elsewhere in the brain; second, there may only be a delayed response in these regions which were not detected maximally in the time frame examined; and third, the response in NTS neurones is transcriptional but not electrical. Either way, counter regulation in response to glucoprivation is likely to require an integrated brain response.

As the present study has mapped functional responses of the whole-brain, it has also identified a number of novel 2-DG-responsive brain regions. Areas of particular interest are the Orb (cortex), AcbSh, and VP (mesolimbic system). These regions make up a corticostriatal connection with the hypothalamus, by which aspects of motivation, salience and reward can impinge on the hypothalamic control of feeding behaviour (Swanson 2000; Fulton 2009). Furthermore, the hypothalamus can be viewed as the head of a brainstem control column which implements the feeding behaviour. A recent study investigating functional brain responses to an orexigenic dose of a cannabinoid CB1 receptor agonist, also highlights this corticostriatal-hypothalamic pathway (Dodd et al. 2009). Thus, we have provided evidence for a common integrated circuit involved in the induction of feeding behaviour to different types of stimulus.

In summary, we have characterised the whole-brain response to 2-DG-induced glucoprivation using c-Fos protein functional activity mapping and BOLD phMRI. By using the complementarity of two functional techniques, we have identified the well-characterised connections of the hypothalamus and brainstem, whilst highlighting areas of the frontal cortices (orbitofrontal, cingulate, insular cortex), mesolimbic system (AcbC, AcbSh, VP, BNST), striatum (caudate putamen, globus pallidus), amygdala (CeA) and thalamus (PVA, IC) as additional brain regions responding to 2-DG-induced glucoprivation. This study, therefore, provides an insight into how the whole-brain responds to 2-DG to co-ordinate complex counter-regulatory responses, whilst further illustrating the accuracy and valuable potential of phMRI in investigating central pharmacological activity.

Acknowledgements

The authors wish to thank the technical assistance of Ms Karen Davies for maintenance of the MRI magnet and console. Dr Shane McKie and Dr Jennifer Stark for advice, and help on using SPM5. Dr Adam Schwarz (Eli-Lilly) and GlaxoSmithKline for providing us with a copy of the rat brain template and PIC-atlas image software. GTD was supported by a Biotechnology and Biosciences Research Council Integrative Mammalian Physiology priority post-graduate studentship supplemented by the Integrative Pharmacology Fund. The authors declare no conflict of interest.

Ancillary