Compared to lean subjects, obese men have less activation in the dorsolateral prefrontal cortex, a brain area implicated in the inhibition of inappropriate behavior, satiety, and meal termination. Whether this deficit precedes weight gain or is an acquired feature of obesity remains unknown. An adult animal model of obesity may provide insight to this question since brain imaging can be performed in lean vs. obese conditions in a controlled study. Seven diet-induced obese adult minipigs were compared to nine lean adult minipigs housed in the same conditions. Brain activation after an overnight fasting was mapped in lean and obese subjects by single photon emission computed tomography. Cerebral blood flow, a marker of brain activity, was measured in isoflurane-anesthetized animals after the intravenous injection of 99mTc-HMPAO (750 MBq). Statistical analysis was performed using statistical parametric mapping (SPM) software and cerebral blood flow differences were determined using co-registered T1 magnetic resonance imaging (MRI) and histological atlases. Deactivations were observed in the dorsolateral and anterior prefrontal cortices in obese compared to lean subjects. They were also observed in several other structures, including the ventral tegmental area, the nucleus accumbens, and nucleus pontis. On the contrary, activations were found in four different regions, including the ventral posterior nucleus of the thalamus and middle temporal gyrus. Moreover, the anterior and dorsolateral prefrontal cortices as well as the insular cortex activity was negatively associated with the body weight. We suggested that the reduced activation of prefrontal cortex observed in obese humans is probably an acquired feature of obesity since it is also found in minipigs with a diet-induced obesity.
Neuroimaging studies suggest that obesity might be associated with abnormal neuronal activity in certain brain regions (1). Del Parigi et al. (2,3,4,5) postulated that the activation of the prefrontal cortex is an important component of the central response aimed at promoting the termination of a feeding episode, notably because the prefrontal cortex has efferent inhibitory projections to the central orexigenic network. Several research groups consistently found that obesity is associated with a decreased activity of the prefrontal cortex (6,7). Because Le et al. (8) showed that this abnormality disappears in formerly obese women with successful weight loss, we can hypothesize that less activation in the prefrontal cortex is not “hard-wired” in disease-prone subjects but rather depends on the environmental and/or nutritional context. Because controlled studies are difficult to perform in obese individuals, we cannot state whether this abnormal neuronal activity in the prefrontal cortex is pre-existent or an acquired feature of obesity.
In a recent review, our group stood up for the use of the pig model in brain imaging and neurosurgery (9). Aside from having similar brain structure to humans, the pig is emerging as a model of predilection in biomedical research because of its metabolic features and ability to develop disorders observed in the humans (e.g., excessive fat deposition, diabetes, atherosclerosis, hypertension, etc (10,11)). The number of studies using minipigs as a model of adult human obesity has increased in the recent years (12,13,14,15). Because our hypothesis considers that the environmental and/or nutritional context is responsible for this brain anomaly, we decided to use a diet-induced model of obesity. Over-consumption of a high-fat and/or high-carbohydrates diet can be used in animal models to provoke obesity (16). As reminded by Miesel et al. (17), the interest is to produce a diet-induced obese animal model that mimics all the symptoms of human metabolic syndrome, but whose behavioral and metabolic perturbations are caused by nutrition rather than genetic modifications. We previously described the establishment of diet-induced obesity in the minipig (14,15). This model is consequently a unique opportunity to determine whether the less prefrontal cortex activation in obese adults is an acquired feature of obesity.
The aim of our study, using single photon emission computed tomography (SPECT) to investigate the brain activity, was threefold. First, we compared the prefrontal cortex basal cerebral blood flow of lean and diet-induced obese minipigs to state whether or not a decreased neuronal activity in this particular brain region is an acquired feature of obesity. Second, we identified all the other brain areas of which the activity differed between lean and obese animals, and compared them with the data obtained in humans. Third, we investigated whether the brain activity was related to the body weight and/or the amount of food consumed the day before brain imaging.
Methods and Procedures
Animals and housing
The experiment presented in this paper was conducted in accordance with the current ethical standards of the European Community (Directive 86/609/EEC), agreement no. A35-622 and authorization no. 01894. Trained staff members were provided for the care and management of animals under the supervision of a veterinarian. Seventeen adult male Göttingen miniature pigs (41.9 ± 1.3 kg at the beginning of the experiment) were used in this study (Ellengard Göttingen Minipigs ApS, Dalmose, Denmark). All animals were housed in individual pens under controlled conditions (temperature maintained between 22 and 23 °C) with a 12:12-h light-dark cycle and free access to water. Group housing was avoided to prevent fights between animals.
Before the onset of the experiment the animals were fed with a standard diet (2,463 kcal/kg) with a formula and rations (41.5 g/kg0.75 of live weight once a day at 0900 hours) designed to maintain a lean phenotype in the Göttingen minipigs. The standard diet was composed of 33% barley, 25% wheat bran, 12% soy shell, 10% wheat, 10% sunflower meal, 6% soy meal, and other minor components. Fat provided 2.17% of the total nutritional value. Two groups were then constituted. Nine animals were still fed once a day, a ration calculated to maintain a stable body weight (102 kcal/kg0.75), and eight animals were fed with a Western diet (WD) (3,473 kcal/kg) enriched with carbohydrates and lipids offered ad libitum during 5 months (one ration offered at 0900 hours and calculated to exceed the daily consumption of the animals). The WD was composed of 32.65% wheat, 15% soy meal, 12% wheat bran, 10% barley, 10% sunflower oil, 10% cornstarch, 5% saccharose, and other minor components. Fat provided 22.74% of the total nutritional value.
The brain imaging modality used to investigate brain activity was the SPECT of technetium-99m hexamethyl-propylene-amine-oxime (HMPAO, Ceretec, GE Healthcare, Velizy, France). Lean minipigs were compared to obese minipigs.
Animal anesthesia and radiolabel administration. Animals were subjected to an overnight fasting before brain imaging. Preanesthesia was induced by ketamine (5 mg/kg intramuscularly, Rhône Merieux, Lyon, France). Suppression of pharyngotracheal reflex was obtained by inhalation of isoflurane (3–5% v/v, Baxter, France) immediately before intubation. A surgical level of anesthesia was obtained with isoflurane (2–3% v/v) delivered by a mechanical ventilator. Oxygen fraction (FiO2) and tidal volume were adjusted so that spO2 measured by pulse oxymetry (Ohmeda, GE Healthcare Clinical Systems, Limonest, France) was 98% or more and spO2 measured by IR capnometer (Armstrong; Armstrong capnometer, Gambo Engström, Bromma, Sweden) was <5%. A venous catheter was inserted into the animals' left ear in order to inject the radiolabel compound. The eyes and ears of the animals were concealed respectively with surgical tape and cotton wool to minimize sensory stimulations, and a 20-min resting period was observed before injection of the radiolabel (Tc-99m, 740 MBq, CHU Pontchaillou, Rennes, France; HMPAO, CIS Bio International, Bagnols/Ceze, France).
Image acquisition. SPECT imaging was performed with a γ-camera (APEX SP-6, Elscint, Tel-Aviv, Israel) fitted with a fan-beam collimator (50 cm focus). Sixty images with a 120 s exposition were acquired at different projection angles (6 degrees per step). Transaxial images (128 × 128 matrix) were reconstructed with a filtered backprojection using a Metz filter (power parameter q = 3). Spatial resolution of the final images was 0.6 mm per pixel for x and y directions and 1.47 mm per pixel in z direction. The size of the images was multiplied by 4 in all directions to create a brain volume about identical to that of a human so that statistical parametric mapping (SPM) software could process a smaller-sized brain (J. Ashburner, SPM diffusion list).
Image processing. SPM2, (Wellcome Department of cognitive Neurology, London, UK) software implemented in MATLAB 7.1 (Mathworks, Sherborn, MA) was used for spatial preprocessing and statistical analysis. SPM software was adapted to the characteristics of the pig brain. Briefly, templates, a priori, and render images were replaced by analyze files obtained in our laboratory using 16 animals different from those used in this experiment. SPECT imaging, T1 magnetic resonance imaging (MRI) and computed tomography scans were performed in series on these animals during an anesthesia identical to that used for image acquisition in the current study. SPECT images were acquired and processed as described above. The T1 MRI imaging was performed on a Siemens Syngo MR B 15 (Siemens Healthcare, Saint-Denis, France). Three-dimensional imaging sequence was performed in the transverse direction using a matrix size of 384 pixels with a spatial resolution of 0.4 mm per pixel in x and y directions and 0.8 mm per pixel in the z axis. Image sequence characteristics were repetition time 2,060 ms, echo time 3.71 ms and inversion time 1,100 ms. These images were manually segmented using ImageJ software to strip the skull while preserving the brain matter. All these volumes were co-registered using the build-in function of SPM. They were also referenced to the commissura anterior-commissura posterior plane with the origin set at commissura posterior according to the stereotactic reference selected by Felix et al. (18) for the pig brain stereotactic atlas. The MRI images were also used, after co-registration, to calculate the three-dimensional rendering used by SPM. The final image was filtered using a Gaussian filter (5 mm kernel). These images were used as templates for SPM calculation. Finally, the bounding box used for normalization was recalculated to preserve matrix size similar to that of the template (J. Ashburner, SPM diffusion list, 1999). SPECT images obtained in our experiment were spatially normalized and masked before statistical comparisons. Spatial normalization was restricted to linear 12-parameters affine transformations, in order to minimize deformations of the original images. Spatially normalized images were then re-sliced using tri-linear interpolation to a final voxel size of 1.8 × 1.8 × 8 mm3, and smoothed using 8 mm Gaussian kernel (2 mm real size).
Statistical analysis. With the purpose of accounting for interindividual differences in global cerebral blood flow, the regional Tc-99m HMPAO uptake was standardized to the mean global uptake using proportional scaling. To reduce the number of statistical comparisons, voxels with signal intensities above 50% of the mean global value only were entered in the group comparisons. The lean and obese SPECT brain images were compared together using a two-sample t-test (two conditions, 16 subjects, degrees of freedom (1.0; 14.0)). The SPECT images of one obese animal were not usable. Consequently, we had seven scans for the obese condition and nine scans for the lean condition. The a priori analysis on the prefrontal cortex was performed using the SPM small volume correction applied to a volume of interest (Figure 1a) generated upon a three-dimensional digital pig brain atlas developed in our laboratory (Saikali et al., 2010) and restricted to the prefrontal cortex (including the anterior prefrontal, dorsolateral and orbitofrontal cortices). For the small volume correction analysis, a value of P = 0.005 uncorrected was set as the threshold. A global analysis was then performed on the entire brain volume to identify other brain regions of differential activity between groups. A value of P = 0.05 uncorrected was set as the threshold (extent threshold of 5 voxels), and clusters comprising a minimum of 50 contiguous voxels were considered significant. Because this statistical threshold for the global analysis was very liberal, due to the limited number of subjects, we decided to perform a second-level analysis with the MarsBaR region of interest tool (19). Each brain structure identified during the first-level analysis was considered as a region of interest to compare their average responses between the two groups (two-sample t-test with two conditions and 16 subjects, degrees of freedom (1.0; 14.0)). For this second-level analysis, a corrected value of P = 0.05 was chosen. A trend was considered for P values comprised between 0.05 and 0.10. Finally, the body weight and the food consumed the day before brain imaging were considered as covariates and their influence on brain activity was evaluated via simple regressions (correlation) and F-tests.
Identification of brain regions. The statistical analyses with SPM produced listings of voxels of which the activation differed between treatments or according to the covariates. Each voxel was associated with a set of coordinates (x, y, z) corresponding to its spatial location in the commissura anterior-commissura posterior plane with the origin set at commissura posterior. The brain regions of differential activity were identified with a three-dimensional digital pig brain atlas developed in our laboratory (20). All the cerebral region of interest used for the MarsBaR second-level analysis were constructed upon the anatomical volumes created for this atlas. The only structure that was not described in the digital atlas was the ventral tegmental area. Consequently, we localized it and constructed its region of interest upon the basis of the atlas published by Felix et al. (18).
Feeding pattern and body weight
Before the onset of the WD, all the animals consumed their food in only one meal, when fed on the morning. Thereafter, the nine rationed minipigs (102 kcal/kg0.75 per day) continued to do so. Interestingly, some time after the onset of WD, and despite the ad libitum provision of food, the eight animals of the other group consumed their food in one quick hyperphagic meal. Along time, the number of daily meals in obese animals increased to reach in average 4–5 meals a day at the moment of the brain imaging sessions. Alternation of hyperphagic periods and spontaneous fasting (during one or several days) were also frequently observed in obese animals. At the moment of the brain imaging sessions, lean minipigs weighed in average 38.0 ± 1.9 kg. Obese minipigs weighed in average 67.1 ± 3.7 kg and consumed ∼198 kcal/kg0.75 and per day (P < 0.001 for body weight and food consumption compared to lean minipigs). It took ∼10 weeks for the obese minipigs after the onset of the WD to reach their final body weight. The day before brain imaging, lean animals ingested in average 1,560 ± 21 kcal and obese animals ingested in average 2,183 ± 146 kcal.
The a priori analysis performed on the prefrontal cortex revealed a decreased brain activity in three regions in obese minipigs compared to lean minipigs: the left and right dorsolateral prefrontal cortices, and the left anterior prefrontal cortex (Figure 1b). The first-level analysis performed on the whole brain volume also revealed a significant difference of activity in several other brain regions, of which some were deactivated and others were activated in obese minipigs compared to lean minipigs (Table 1). Although the significance threshold was initially set at P = 0.05 uncorrected for multiple comparisons, half of the differences were significant below a P = 0.003 threshold. The results from the second-level analysis are summarized in Table 2, notably highlighting a significant deactivation of the left dorsolateral prefrontal cortex, the left anterior prefrontal cortex, and the ventral tegmental area, as well as an activation of the ventral posterior nucleus of the thalamus. Some other regions had a tendency to be differently activated. All these brain structures are represented in Figure 2. The regression analyses performed on the body weight and food consumption revealed several brain regions of which the activity was positively or negatively correlated to these covariates (Table 3). The insular and prefrontal cortices for which an inverse association with body weight was found are illustrated in Figure 3.
Table 1. Brain regions for which differential cerebral blood flow was found in obese vs. lean adult minipigs (R, right; L, left)
Table 2. Brain regions identified during the first-level analysis and for which a second-level analysis was performed with the MarsBaR region of interest tool
Table 3. Brain regions of which the cerebral blood flow is correlated to the body weight or the quantity of food (kcal) consumed during the day before brain imaging (R, right; L, left)
Our study showed that the establishment of a diet-induced obesity in minipigs was accompanied by a modification of the basal brain metabolism that is probably an acquired and specific feature of the disease. Several brain regions were found deactivated in comparison to lean subjects, such as the dorsolateral and anterior prefrontal cortices, the ventral tegmental area, the nucleus pontis, and nucleus accumbens. Moreover, the dorsolateral and anterior prefrontal cortices as well as the insular cortex activity was negatively correlated to the body weight. Other brain regions were found activated in obese minipigs such as part of the thalamus and temporal gyrus. These results demonstrated that some brain functional features described in obese humans can also be found in diet-induced obese minipigs, which suggests that the nutritional context is of prime importance in the modulation of the basal brain metabolism and the possible emergence of cerebral functional abnormalities.
Different authors consistently brought forward evidence that obesity is associated with particular brain functional features and notably less activation of the prefrontal cortex (6,7,8). Tataranni's group (1,3) also found a greater neuronal activation of the prefrontal cortex and greater neuronal deactivation of the limbic and paralimbic areas in response to satiation. In addition, the same group demonstrated that obese subjects showed a greater decreased activity in the orbitofrontal cortex in response to the sensory aspects of a meal after a prolonged fast (2). This highlights the fact that the prefrontal cortex has a particular role in the treatment of feeding signals as well as hunger and/or satiety states. It seems that neuronal abnormality in the vicinity of the prefrontal cortex as well as limbic/paralimbic areas is a specific feature of obesity. Our own data are consistent with the results of Volkow et al. (7), who found an inverse association between BMI and basal prefrontal metabolic activity. Indeed, obese minipigs had less activation of the anterior prefrontal cortex and the dorsolateral prefrontal cortex than lean minipigs. Moreover, the right dorsolateral and left anterior prefrontal cortices activity was negatively correlated with body weight, similarly to that of the left insular cortex. Interestingly, Bragulat et al. (21) found that the perception of food-related odors during hunger induce less activation of the posterior insula in obese subjects compared to lean controls.
Del Parigi et al. (1) postulated that the activation of the prefrontal cortex is an important component of the central response aimed at promoting the termination of a feeding episode because of the inhibiting effects it exerts on the orexigenic network composed of the hypothalamus, thalamus, limbic/paralimbic areas, and basal ganglia. The prefrontal cortex in the Göttingen minipig is particularly well developed (24% of the total neocortex and 10% of the total brain volume) and has reciprocal connections with several thalamic nuclei (22). Manganese-enhanced MRI also revealed segregated circuits that unite the prefrontal cortex, the thalamus and the basal ganglia, with further projections directed toward the ventral tegmentum for example (22). Interestingly, the ventral tegmental area and the ventral posterior nucleus of the thalamus (a trend was also found for the nucleus accumbens) are some of the brain regions of which the global activity differed in obese minipigs compared to lean minipigs. The activated thalamus is in accordance with the hypothesis of a decreased inhibition operated by the frontal cortex. Another structure, the middle temporal gyrus, was found activated in our obese minipigs (trend after the second-level analysis). Previous studies observed a modification of the posterior middle temporal gyrus activity in obese human subjects in response to satiation or food visual stimuli in comparison to hunger or nonfood stimuli (23). Rosenbaum et al. (24) also showed a leptin-reversible increase in the neural activity of this structure in response to visual food cues. These scientific arguments support the hypothesis that all the aforementioned cortical structures are important for the regulation of hunger/satiety and in the executive and decision-making functions related to food.
The ventral tegmental area and the nucleus accumbens were found deactivated like the prefrontal cortex, which seems to oppose Del Parigi's inhibition hypothesis. An alternative explanation for the deactivated ventral tegmental area and nucleus accumbens can be proposed. These two areas are widely implicated in the reward circuit of the brain, cognition, motivation and addiction (25). A deactivation in these regions, as observed in our study, would suggest a deficit in the reward system related to food and explain the hyperphagic behavior and altered meal patterns observed in obese animals and humans. In a recent review focused on appetite and reward, Fulton (26) pulled together research describing the neuronal mechanisms regulating motivation for food, and especially the importance of the midbrain dopamine neurons and corticolimbic nuclei that encode emotional and cognitive aspects of feeding. Studies in rodent models of obesity reported an inverse relationship between weight and D2 receptors (27,28). Interestingly, Wang et al. (29) showed that the availability of dopamine D2 receptor was decreased in obese humans in proportion to their BMI, and more recently Volkow et al. (30) demonstrated that the low level of dopamine striatal D2 receptors is associated with low prefrontal metabolism in obese humans, a specific brain feature previously described in drug-addicted subjects. Addiction to food, especially through the exposure to a high-carbohydrates, high-fat and high-caloric diet, might induce some kind of habituation and consequently enhance the level of food stimulation required to fulfill the hedonic requirements of the organism. This hypothesis would be in agreement with the hyperphagic behavior observed in the obese minipigs, as well as their tendency to increase the number of meals per day.
Whether the alteration of the brain dopamine system and prefrontal cortex metabolism is a cause or consequence of obesity is still unknown. The answer brought by our study is that less activation of the prefrontal cortex is definitely an acquired anomaly related to obesity, and not a “hard-wired” feature. Nevertheless, in addition to the inverse association between prefrontal cortex activation and body weight, it is interesting to notice that in lean minipigs, there was a high interindividual variability of the prefrontal cortex activation, which might suggest the existence of a predisposition factor. Le et al. (8) previously showed that the decreased prefrontal cortex activity can be corrected since it disappeared in formerly obese women with successful weight loss. Such a brain feature consequently depends on the environmental and/or nutritional context of the subjects. Since only the nutritional parameters were modified between the lean and obese minipigs, we can suggest that plethoric and deleterious nutrition is at the origin of the specific brain metabolism of the obese subjects. Moreover, the prefrontal cortex activity was not related to the quantity of food ingested the day before brain imaging, which means that the prefrontal cortex basal metabolism depends on long-term nutritional factors rather than short-term intake differences. Two hypothetical mechanisms can be proposed. First, altered prefrontal cortex metabolism might be a “collateral damage” of obesity and appear only once the disease is completely declared. Hypertension is known to decrease the blood flow in the prefrontal lobe (31), and decreased gray matter in the prefrontal cortex was found in subjects with type 2 diabetes (32). We previously demonstrated that the obese minipigs used in the current study suffered from insulin resistance and probably from a prediabetic or diabetic condition (15). Second, the neuronal abnormality might be a direct consequence of the ad libitum provision of a highly palatable and caloric food, and promote the emergence of feeding disorders and obesity. Our study compared the brain metabolism of lean subjects vs. subjects submitted to 5 months of deleterious diet, i.e., the equivalent in humans of a well-established morbid obesity (14,15). In a recent review, Velloso (33) reminded that high-fat diets induce an inflammatory response in the hypothalamic areas involved in the control of feeding and thermogenesis. This inflammatory process would damage the neuronal circuitries that maintain the homeostatic control of the body's energy stores, therefore favoring body mass gain and obesity (33). Such an inflammatory process might also occur in the prefrontal lobe but some proofs of this phenomenon are still to be produced.
Though very promising, because it highlighted some brain metabolism abnormalities similar to what has been observed in obese humans, our study in minipigs still has some limitations that further works should overcome. First, the relatively small number of animals limited the statistical power of this preliminary study, and some metabolic differences that were identified did not resist to a correction for multiple comparisons and were not found significant in the second-level analysis. Second, the lack of baseline SPECT scans for the obese animals and/or a washout period during which feeding conditions would have been reversed in both groups, prevented us to conclude with certainty that the observed brain anomalies were definitely the consequence of the deleterious diet. A replicate study, improving the experimental design and increasing the number of subjects, would allow us to confirm our hypotheses.
The tight relationship between prefrontal cortex, reward system, and food addiction still needs to be explored. For example, it has been shown in the rat that electrical stimulation of the prefrontal cortex increases dopamine release in the nucleus accumbens (34). The same effects were observed in the caudate nucleus after repetitive transcranial magnetic stimulation of the human prefrontal cortex (35). More interesting is the fact that similar treatment led to decreased food craving in humans (36). We demonstrated in this study that the less prefrontal cortex basal metabolism in minipigs was an acquired feature of obesity. In addition, we found that this deactivation was negatively associated with body weight and accompanied by a decreased activity in the brain reward circuit. The minipig model consequently represents a real asset to investigate the causes and consequences of a decreased activity in the prefrontal cortex and other regions of the brain reward circuit on the eating behavior and weight control.
This study was supported by a grant from the CRITT Santé Bretagne. We gratefully acknowledge the efforts and cooperation of the scientific and technical staff that participated closely or by far in this experiment, and especially Alain Chauvin, Benoît Janson and Jihane Vitre-Boubaker.