Food Reward, Appetite, Satiety, and Obesity Edmund T. Rolls Oxford Centre for Computational Neuroscience and University of Warwick, UK L Cranach 1528 Uffizi, Florence What Reward Decision/Action Oxford University Press 2014 Neuroimaging: Computational Neuroscience F. Grabenhorst S.M. Stringer C. McCabe G. Deco (Barcelona) I. Araujo A. Treves (SISSA,Trieste) B. Parris J. Feng (Warwick) T. Webb (Warwick) Neuropsychology: G. Perry J. Hornak L. Franco (Malaga) H. Berlin Orbitofrontal Neurophysiology: Inferior Temporal Neurophysiology J.V.Verhagen N. Aggelopoulos M. Kadohisa Hippocampal Neurophysiology H. Critchley J.-Z.Xiang Prefrontal cortex: top-down biased activation - a theory of attention Food intake control: sensory stimuli are modulated by satiety signals in the orbitofrontal cortex to produce food reward. Obesity: enhanced sensory-produced reward over-rides satiety signals; individual differences in reward value and cognitive control. Orbitofrontal cortex taste neuron Sensory factors: Taste Smell Texture Sight Effects of: Variety Sensory-specific satiety Palatability Food concentration Portion size Ready availability Cognitive factors: Conscious rational control Beliefs about the food Advertising Brain mechanisms: Sensory factors modulated by satiety signals produce reward value and appetite. Individual differences. Satiety / hunger signals: Adipose signals Gut hormones Gastric distension Eating Autonomic and Endocrine effects Rolls 2011 Int J Obesity Taste, olfactory, somatosensory and visual inputs to the orbitofrontal cortex Amount fed (ml) Sensoryspecific satiety in the macaque orbitofrontal cortex Rolls,E.T. et al (1989) European Journal of Neuroscience 1: 53-60. 1
Primary taste cortex: no effect of feeding to satiety with glucose on the neuronal response to glucose taste Sensory-specific satiety for the texture of fat Volume (ml) of glucose ingested Rolls, Scott, Sienkiewicz and Yaxley (1988) Journal of Physiology 397: 1-12. Rolls,E.T., Critchley,H.D., Browning,A.S., Hernadi,A. and Lenard,L. (1999) Responses to the sensory properties of fat of neurons in the primate orbitofrontal cortex. Journal of Neuroscience 19: 1532-1540. Each orbitofrontal cortex neuron responds to a different combination of taste, odor, texture and temperature stimuli: as a population they provide information about a rich variety of reward stimuli: and provide for behaviours such as sensory-specific satiety that are specific to combinations. A reward representation as a goal for action must be specific for a particular reward (cf dopamine). Verhagen, Rolls and Kadohisa (2003) J Neurophysiology 90 Firig rate (spikes/sec; mean+/-sem) 25 20 15 10 5 0 Some neurons encode the viscosity of oral stimuli, as shown by their responses to a viscosity series (CMC). Fat responses from these neurons can be predicted by viscosity e.g. neuron bk291c2. vegetable oil mineral oil 25 silicone oil 55 1 10 100 1000 10000 Viscosity (cp) 280 CMC series Rolls,E.T., Verhagen,J.V. and Kadohisa,M. (2003) Representations of the texture of food in the primate orbitofrontal cortex: neurons responding to viscosity, grittiness, and capsaicin. Journal of Neurophysiology 90: 3711-3724. Orbitofrontal cortex fat texture-responsive neurons 20 Fat responsive neurons respond independently of viscosity e.g. bk265 Firing rate (spikes/sec; mean+/-sem) 15 10 5 vegetable oil 55 safflower oil 50 280 mineral oil 25 coconut oil 40 CMC series silicone oil 0 1 10 100 1000 10000 Viscosity (cp) Rolls,E.T., Critchley,H.D., Browning,A.S., Hernadi,A. and Lenard,L. (1999) Responses to the sensory properties of fat of neurons in the primate orbitofrontal cortex. Journal of Neuroscience 19: 1532-1540. Verhagen,J.V., Rolls,E.T. and Kadohisa,M. (2003) Neurons in the primate orbitofrontal cortex respond to fat texture independently of viscosity. Journal of Neurophysiology 90: 1514-1525. 2
Orbitofrontal cortex neuronal responses Taste reward. Only respond if hungry. Implement sensory-specific satiety. All tastes are represented, including sweet, salt, bitter, sour and umami, and all are primary reinforcers. Olfactory reward. Hunger dependent. Implement olfactory sensory-specific satiety. 40% reflect olfactory-taste association learning. Texture. Reward - hunger dependent. Fat texture. Fat is coded by texture, not by unsaturated fatty acids (eg linoleic) Separate viscosity system. Separate astringency (tannic acid) system. Temperature. Visual Reward. One-trial visual-to-taste association learning. Hunger dependent. Implement visual sensory-specific satiety. Face-selective neurons Auditory, e.g vocalization Non-reward, error detection: negative reward prediction error neurons Activated from brain-stimulation reward sites High-dimensional representation of a very wide range of the sensory properties of both rewards and punishers, with secondary reinforcers linked to primary by stimulus-reinforcement association learning and reversal. A neuronal representation of stimulus value not of behavioral responses. Rolls and Grabenhorst 2008 Progress in Neurobiology; Grabenhorst and Rolls 2011 Trends in Cog Sci Representation of oral fat texture in the human brain CMC = Carboxymethyl cellulose; Fat = veg oil Responses to oral fat in the anterior cingulate cortex and ventral striatum are independent of viscosity Whole Food Sensory-specific Satiety: Correlation between the BOLD signal in the OFC and subjective, conscious, pleasantness ratings Fat & Sucrose: Convergence Correlation with pleasantness ratings Kringelbach,M.L., O Doherty,J., Rolls,E.T. andandrews,c. (2003) Activation of the human orbitofrontal cortex to a liquid food stimulus is correlated with its subjective pleasantness. Cerebral Cortex 13: 1064-1071. Responses to oral fat and sucrose converge in the most anterior part of the cingulate cortex De Araujo and Rolls (2004) Representation in the human brain of food texture and oral fat. J Neuroscience 24: 3086-3093. Whole Food Sensory-specific Satiety Conclusions A direct correlation between the subjective state of pleasure produced by food and the activation of the orbitofrontal cortex has been demonstrated. The human orbitofrontal cortex, as in non-human primates, plays an important role in representing the reward value of food stimuli, including chocolate, and a food rich in umami, tomato This is consistent with the hypothesis that the pleasantness of the flavour of food (with taste, olfactory and texture components) is represented in the human OFC. It is suggested that sensory-specific satiety is a general property of reward systems (Rolls, 2005, Emotion Explained. Oxford). Pleasantness of fat texture High fat vs low fat dairy drink Vanilla vs strawberry flavor 2x2 factorial design Activations in the pregenual cingulate cortex and orbitofrontal cortex were correlated with the pleasantness of fat texture Kringelbach,M.L., O Doherty,J., Rolls,E.T. and Andrews,C. (2003) Activation of the human orbitofrontal cortex to a liquid food stimulus is correlated with its subjective pleasantness. Cerebral Cortex 13: 1064-1071. Grabenhorst, Rolls et al 2010 Cerebral Cortex Conclusion 3
Effects of cognition on perception Rolls and De Araujo University of Oxford Isovaleric Acid Conclusions: A visually presented word label modulates representations of odors in olfactory areas in the orbitofrontal cortex, amygdala, and olfactory tubercle. Cognition can influence subjective, conscious, affective representations in the orbitofrontal and pregenual cingulate cortices. Isovaleric Acid Cheddar Cheese? Body odour? Cheddar Cheese Body odour (CH 3 ) 2 CHCH 2 COOH (CH 3 ) 2 CHCH 2 COOH De Araujo,I.E.T., Rolls,E.T., Velazco,M.I., Margot,C. and Cayeux,I. (2005) Cognitive modulation of olfactory processing. Neuron 46: 671-679. Cognitive modulation revealed by a correlation between the BOLD signal and the pleasantness ratings given to the Test odour. A, B: anterior cingulate; C: amygdala; D: olfactory tubercle Selective attention to affective value alters how the brain processes taste stimuli We delivered the identical taste (MSG) on every trial. Instructions were on different trials to remember and rate pleasantness or to remember and rate intensity Taste delivery, instruction Swallow Rinse delivery Swallow Delay Rating Pleasantness or intensity 0 9 11 18 20 25 30 s Grabenhorst and Rolls (2008) European Journal of Neuroscience Correlation analysis of BOLD signal with pleasantness ratings given to Clean Air. A, B: anterior cingulate; C: amygdala; D: olfactory tubercle Orbitofrontal and pregenual cingulate cortex Paying attention to pleasantness vs intensity produces greater activation to taste in the orbitofrontal and pregenual cingulate cortices. Activations were correlated with subjective pleasantness ratings. Grabenhorst and Rolls (2008) European Journal of Neuroscience 4
Anterior and mid insular cortex Paying attention to intensity vs pleasantness produced larger activations in the anterior and mid insula. Activations were correlated with subjective intensity ratings. Grabenhorst and Rolls (2008) European Journal of Neuroscience The effects of chocolate in the mouth: activation of the insular primary taste cortex. This was the same in cravers and noncravers. Selective attention to affective value alters how the brain processes taste stimuli Top-down selective attention allows processing in different brain areas to be emphasized for different types of decisionmaking Decisions about affective value recruit the orbitofrontal cortex and pregenual cingulate cortex. Decisions about the intensity of a stimulus recruit the primary sensory (insular taste) cortical area. In sensory testing, psychophysics and marketing, it is important to ensure that attention is being paid to pleasantness or to the physical properties: different brain systems are engaged by these two types of attention The flavour and sight of chocolate combination: more activation was found in the cravers than in the noncravers in the anterior and pregenual cingulate cortex, and in the ventral striatum. Grabenhorst and Rolls (2008) European Journal of Neuroscience Do responses of the orbitofrontal cortex and pregenual cingulate cortex enable individual differences in affective behaviour and decision-making to be predicted? Chocolate craving: a craving and a non-craving group Chocolate in the mouth flavour differences? Sight of chocolate conditioned cue differences? Sight of chocolate and chocolate in the mouth greater supralinearity? Cognitive biasing: dark chocolate word label vs white chocolate word label Condensed milk similar texture and sweetness to chocolate, but not craved. Tasteless control solution 8 cravers and 8 non-craver participants. SPM fmri group random effects analysis with full correction or svc. Rolls and McCabe (2007) European Journal of Neuroscience 26 a. More activation of mid and medial orbitofrontal cortex in cravers than non-cravers c. More activation in the ventral striatum in cravers than noncravers Sight of chocolate 5
fmri of chocolate craving: individual differences in brain activations predict craving and food intake There were no differences between chocolate cravers and non-cravers in responses to the flavour of chocolate in the primary taste cortex. Moreover the activations in the primary taste cortex were not correlated with the pleasantness or wanting ratings for chocolate. Thus it was not differences in the primary taste cortex, or physical sensitivity to taste and oral texture, that separated the cravers from the non-cravers. The flavour of chocolate produced more activation in cravers than non-cravers in the medial orbitofrontal cortex. The sight of chocolate produced more activation in chocolate cravers in the medial orbitofrontal cortex (OFC) and ventral striatum. A combination of the sight and flavour of chocolate produced more activation than the sum of the components in the medial orbitofrontal cortex, pregenual cingulate cortex, and striatum. This non-linearity was greater in the cravers than the non-cravers. The subjective pleasantness ratings of the chocolate and chocolate-related stimuli had higher positive correlations with the fmri BOLD signals in the pregenual cingulate cortex and medial OFC in the cravers than in the non-cravers. The amount of chocolate eaten on a regular basis was higher in the cravers (370 g / week) than the non-cravers (22 g / week). Understanding individual differences in brain responses to very pleasant foods helps to understand the mechanisms that drive the liking for foods, and thus food intake and decision-making. Individual differences and personality. Rolls and McCabe (2007) European Journal of Neuroscience 26: 1067-1076 Pleasuremap Yellow: Correlation with subjective pleasantness. White: Correlate with subjective unpleasantness. Taste:1: 2: (Grabenhorstand Rolls, 2008);odor: 3, 4: (Grabenhorst et al., 2007);5: (Grabenhorst and Rolls, 2009); 6, 7: (Rolls et al., 2003a);8, 9: (Anderson et al., 2003); 10: (de Araujoet al., 2005); flavor: 11, 12: (Grabenhorstet al., 2009b); 13: (Grabenhorst et al., 2008a); 14: (McCabeand Rolls, 2007); 15: (Kringelbachet al., 2003);16: (de Araujoet al., 2003b); oral texture:17, 18: (Grabenhorst et al., 2009b); chocolate: 19: (Rolls and McCabe, 2007);water: 20: (de Araujoet al., 2003a);wine: 21: (Plassmann et al., 2008); oral temperature:22, 23: (Guestet al., 2007); somatosensory temperature:24, 25: (Rolls et al., 2008a);sight of touch: 26, 27: (McCabeet al., 2008);facialattractiveness:28, 29: (O'Doherty et al., 2003);eroticpictures:30: (Walteret al., 2008); money: 31, 32: (O'Doherty et al., 2001); laser-inducedpain: 33: (Raij et al., 2005). Grabenhorst and Rolls (2011) TICS Personality and reward systems in the brain Each specific type of reward (taste, flavour, water, touch ) is represented by different neurons. Each specific type of gene-specified reward is subject to variation between individuals as part of evolution by natural selection. Therefore different individuals may have different sensitivity to different specific types of reward, nonreward, etc. This variation may contribute to personality, and to high dimensionality of the space. This may be reflected in differential sensitivity in different individuals in how brain reward systems respond to different reinforcers. What Reward Decision/Action Prefrontal cortex: top-down biased competition Orbitofrontal cortex neuroimaging Taste : Both pleasant and unpleasant tastes are represented Amygdala: pleasant tastes are as much represented as unpleasant tastes Olfactory reward. Olfactory sensory-specific satiety. Olfaction: pleasant odours activate a particular region of the OFC Anterior cingulate cortex is activated by pleasant and by unpleasant odors Primary olfactory cortical areas represent the identity and intensity of odours Cognitive inputs, word labels, modulate olfactory effects in the OFC and ant cingulate. Whole food: taste, odor and texture Reward reflects sensory-specific satiety. Correlation of OFC activation with the subjective pleasantness of the food Oral viscosity and fat texture: insula and orbitofrontal cortex Flavour: olfactory-taste convergence In the orbitofrontal cortex and adjoining agranular insula; MSG+savoury odor The primary taste cortex in the insula is unimodal Individual differences: chocolate craving: orbitofrontal cortex and pregenual cingulate Somatosensory pleasure and pain more than neutral Correlation of OFC activation with subjective pleasantness and pain Anterior cingulate cortex: anterior - pleasant touch; mid - pain Abstract (monetary) reward and punishment (loss) in a reversal task Separate representations of the magnitude of the gain (medial) and loss (lateral) Expected value in a probabilistic task: Activation = probability x reward value Face reversal cued by changing face expression: reward prediction error OFC activation related to an angry expression when it is used as a reversal cue Activation in the fusiform face area does not reflect the reversal Amphetamine activates the medial orbitofrontal cortex Cognitive affective modulation of taste, flavour, odour and touch is implemented in the orbitofrontal and pregenual cingulate cortex. Principles of the cortical processing of taste, texture and odor Taste reward and subjective pleasantness is represented in the orbitofrontal (secondary taste) cortex (OFC) Taste identity and intensity are represented in the primary taste cortex Olfaction and taste converge in the OFC onto the same neurons. In the OFC, odor is mapped by some neurons from molecular to taste space by associative learning. In the OFC, odor reward is represented. Visual to taste associative learning occurs in the OFC in 1 trial. Oral texture (viscosity, fat texture, capsaicin, roughness) and oral temperature are combined with taste in the primary and OFC taste cortex. Fat sensing is oral texture based, and is independent of fatty acids such as linoleic, and of viscosity. OFC oral texture channels reflect the crispness / freshness of a food. Fat texture pleasantness is represented in the OFC. Texture variety can increase appetite for a food: sensory-specific satiety for oral texture. Sensory-specific satiety is implemented by OFC combination-sensitive neurons. Cognition and attention modulate affective value representations in the OFC: a biased activation theory of selective attention. 6
Obesity: overstimulation of the food reward system in the brain 1) Endocrine and genetic factors - e.g. hyperinsulinemia or leptin deficiency. (Rare cannot accountfor the three-fold increase in humans with BMI > 30 since 1980) 2) Externality - Schachter: obese people may be more reactive to the sensory properties of food. Food craving, and its driving by the sight of food. Cognitive and attentional actors can directly influence palatability. 3) Palatability - enhancedpalatability in human diet, leading to imbalance between orosensory control signals, and gastrointestinaland post-absorptive satiety signals. 4) Variety - enhancedvariety in human diet, which leads to increased food intake becausesatiety is partly sensory-specific. 5) Meal pattern i) Inter-meal interval is normally regulated. ii) Obese tend to eat late in the day. iii) Ready availability of food. iv) Portion size v) Eating rate may be too fast to allow satiety signals to develop 6) Meal concentration / energy density - tends to be high in the obese, and only partial compensation is possible by e.g. gastric emptying rate. 7) Stress-induced eating. 8) Regulation to internal signals is poor - e.g. 2 weeks to adjust to altered caloric composition of the diet; intake is dominated by external sensory / reward signals. 9) Brown Adipose Tissue and energy output i) Important in thermo-regulation in rats. ii) Can be triggered by variety-induced eating in rats. iii) Little BAT in humans, although metabolism does normally partially compensatefor altered energy intake. 10) Exercise Rolls E T (2012) Proceedings of the Nutrition Society 71 Satiety signals from the digestive tract 1. Is gut feedbacknutrient-specific? (Some gut taste cells may have multiple types of taste receptor: Margolskee.) For example, does MSG in the gut reduce the pleasantnessof the flavor of protein? For example, does glucose in the gut reduce the pleasantnessof sweet taste? 2. Does MSG in the gut produce more satiety than other amino acids and proteins? Does MSG in the gut reduce dietary-induced obesity (K.Torii)? 3. Hormonal signals, e.g. leptin: may accountfor only small proportion of obesity; but useful in treatment in general? 4. Effects of food in the digestive tract: Unconditioned satiety Conditioned appetite Conditioned satiety Rolls E T (2011) International Journal of Obesity Rolls E T (2012) Proceedings of the Nutrition Society Dual routes to emotion Obesity, addiction and the mismatch hypothesis 1. Mismatch hypothesis: food palatability, availability, variety, and exposure by advertising have increased food reward in the last 30 years yet the satiety signals remain unchanged, and this contributes to overeating. Can the food industry develop very palatable foods, but with good nutrition, e.g. fat substitutes? 2. The control of eating is different from typical addictions in that eating is controlled by satiety signals from the periphery, and by sensory-specific satiety. 3. Typical drugs of addiction bypass the orbitofrontal cortex and amygdala where these satiety controls operate, and instead operate in regions to which they project, such as the ventral striatum. 4. Nevertheless, very pleasantfoods (chocolate, ice cream) may produce so much reward, in some individuals, that dopamine is released, and this may introduce some similarities of the behavior to addiction. 5. In addiction, the conditioned stimuli have a potent control on behavior. The same may be true in overeating and obesity. Limiting exposure to such potent conditionedstimuli (the sight and smell of food, advertising of high fat foods with poor nutrition) may help. Relation to impulsiveness. Rolls ET (2014) Emotion and Decision-Making Explained. Oxford University Press. 6. A largescale databaseof brain imaging results using a set of agreed measures (e.g. responsiveness to the sight and flavor of fattening vs healthy foods, BMI, questionnaires, genetic biomarkers) with a BMI follow-up might allow identification of which factors correlate with BMI, and a tendencyfor BMI to increase, and what subgroupsthere may be. Rolls E T (2011) International Journal of Obesity Rolls E T (2012) Proceedings of the Nutrition Society Obesity: overstimulation of the food reward system in the brain Future investigations: 1) Any of the factors described may promote over-eating and obesity: to prevent and treat obesity it may therefore be necessary to control all factors, and not to expectcontrol of one or several factors to be successful. 2) Are the brains of individuals who tend to become obese more sensitive to the reward properties of food? Individual differences in the sensitivity to different types of reward are expected, given how evolution selects for sensitivity to each of many different rewards. Investigation with fmri: this can reveal whether any differences are at the sensory or reward level; and may reveal effects about reward sensitivity that people may not know about, or acknowledge. 3) Does fmri reveal any differences between lean and high BMI individuals reward systems for the Sight, Taste, Flavour including texture, Cognitive modulation, Attentional modulation of food? Knowing which aspects of reward differ most may be important in guidance on how to control intake. 4) Are the brains of individuals who tend to become obese less sensitive to the satiating properties of food? This can now be studied with the fmri arterial spin labelling (ASL) method, which is sensitive to the level of satiety signals measured over a period of minutes, and which by providing an absolute value for the flow rate in different brain areas allows comparisons between subjects, and between treatments such as different food loads on different days. 5) Can we predict weight gain over one year from the sensitivity of the food reward system in the brain? 6) Decision-makingis probabilistic becauseof noise in the brain. 7) Decision-making: theexplicit reasoning vs the emotional system. 8) Cognitive / contextualmodulation can directly influence food reward value in the orbitofrontal cortex -- and also food intake intake? 9) Neuroeconomicsof food reward value, including effects of food cost, and of labels and descriptions. Rolls E T (2011) International Journal of Obesity Rolls E T (2012) Proceedings of the Nutrition Society Oxford University Press 2014 (Available now) Rolls E T (2012) Proceedings of the Nutrition Society 7