npg
© 2015 Nature America, Inc. All rights reserved.
news and views neurons. Using electron microscopy, they found not only a decrease in the total number of axo-spinous synapses, but also a tripling of axo-dendritic and axo-axonal synapses, which are usually much less frequent. The authors hypothesize that this spatial redistribution of synapses is responsible for the increase in the frequency and amplitude of excitatory events. These findings suggest that increased excitation onto the VTA-projecting prefrontal neurons drives increased excitation onto dopaminergic neurons, increasing dopaminergic neuron activity and striatal dopamine release. The authors tested this hypothesis by optogenetically stimulating prefrontal terminals in the VTA and measuring dopamine release in the ventral striatum. Stimulation of cortical terminals in the VTA indeed induced dopamine release, consistent with classical studies that measured dopamine release in the ventral striatum after electrical stimulation of the PFC1,2. Furthermore, the optogenetic stimulation also enhanced locomotion. As in conditional ArpC3 knockout mice, this hyperlocomotion was blocked by haloperidol, again suggesting that the stimulated dopamine release is responsible for the behavioral activation. Several issues remain outstanding, suggesting future experiments. For instance, it is unclear whether the increased excitatory input demonstrated in vitro results in increased PFC neuronal activity in vivo. Increased input could be offset by a decrease in excitability due to altered intrinsic membrane properties; such synaptic scaling is not unusual and has been observed in other circuits5. In vivo recordings are needed to address this question directly. Similarly, it would be important to confirm increases in
firing rates of dopaminergic neurons in ArpC3 knockout mice, as suggested by the current findings. In addition, it remains possible that increased PFC input onto dopamine neurons is not responsible for the increases in striatal dopamine and hyperlocomotion seen in the mutants; as an alternative, increased corticostriatal activity could modulate tonic extrasynaptic dopamine levels through decreased inhibition of VTA neurons by the ventral pallidum6. Differentiating between these two possibilities would require experiments inhibiting specific cortical projections. Finally, there is the issue of which dopamine cells are actually innervated by PFC inputs. In a classical electron microscopy study combining anterograde and retrograde tracers with immunohistochemical labeling for tyrosine hydroxylase, Carr and Sesack confirmed monosynaptic projections from the medial PFC to dopamine neurons3. However, in that study, dopamine neurons receiving synaptic inputs from the cortex were found to project back to the cortex. In contrast, GABAergic neurons receiving synaptic input from the cortex were found to project to the nucleus accumbens3. Stimulation of the PFC induces burst activity in midbrain dopamine neurons7,8, but whether these neurons project to the ventral striatum is unclear. Recording from retrogradely labeled, ventral striatum– projecting VTA neurons during optical activation of prefrontal terminals in VTA slices should be able to answer this question. The results of Kim et al.4 are highly intriguing in the context of schizophrenia. Post-mortem analyses have repeatedly found decreased spine density in PFC from patients with schizophrenia9.
Moreover, brain imaging–based measures of presynaptic dopaminergic function are enhanced in patients with schizophrenia10. Although it is unclear whether prefrontal cortical spine density and presynaptic dopamine disturbances coexist in the same patients, the results of Kim et al.4 suggest that decreased spine density could cause a hyperdopaminergic phenotype in the striatum (Fig. 1). In this context, it is striking that frontal activation during cognitive testing has been found to be related to striatal dopamine when measured in the same patients. Conflicting data on the directionality of that relationship—one study suggests a positive correlation, whereas another suggests the inverse11,12—cautions against a simplistic translation of the current findings to patient populations. Nonetheless, this curious tale of cortical spines and striatal dopamine is intriguing. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Taber, M.T., Das, S. & Fibiger, H.C. J. Neurochem. 65, 1407–1410 (1995). 2. Karreman, M. & Moghaddam, B. J. Neurochem. 66, 589–598 (1996). 3. Carr, D.B. & Sesack, S.R. J. Neurosci. 20, 3864–3873 (2000). 4. Kim, I.H. et al. Nat. Neurosci. 18, 883–891 (2015). 5. Turrigiano, G.G. Cell 135, 422–435 (2008). 6. Floresco, S.B., West, A.R., Ash, B., Moore, H. & Grace, A.A. Nat. Neurosci. 6, 968–973 (2003). 7. Murase, S., Grenhoff, J., Chouvet, G., Gonon, F.G. & Svensson, T.H. Neurosci. Lett. 157, 53–56 (1993). 8. Tong, Z.Y., Overton, P.G. & Clark, D. Synapse 22, 195–208 (1996). 9. Glausier, J.R. & Lewis, D.A. Neuroscience 251, 90–107 (2013). 10. Howes, O.D. et al. Arch. Gen. Psychiatry 69, 776–786 (2012). 11. Meyer-Lindenberg, A. et al. Nat. Neurosci. 5, 267–271 (2002). 12. Fusar-Poli, P. et al. Mol. Psychiatry 16, 67–75 (2011).
Hunger logic Richard Palmiter Activation of AgRP-expressing ‘hunger’ neurons promotes robust feeding. Recent studies reveal the valence, dynamics and neural circuits engaged by AgRP neurons. Dwindling energy reserves are a threat to homeostasis that, if not countered, leads to starvation. How does the brain detect and then coordinate appropriate responses to dwindling energy? As fasting progresses, hormonal changes, coupled with metabolic changes, regulate the activity of neurons in the brain, resulting Richard Palmiter is in the Department of Biochemistry, University of Washington, Seattle, Washington, USA. e-mail: [email protected]
in activation of an ‘alarm system’ whose output increases in intensity. Remarkably, a small cluster of neurons (about 0.01% of all brain neurons) in the arcuate region of the hypothalamus appears to be the alarm system that promotes foraging, feeding and energy conservation. These neurons, referred to as AgRP neurons because they alone make agoutirelated peptide, are inhibitory neurons that also use neuropeptide Y and GABA as neurotransmitters. Activation of AgRP neurons inhibits other neurons in the brain to promote
nature neuroscience volume 18 | number 6 | JUNE 2015
feeding and conserve energy1,2, whereas their ablation in adult mice results in starvation3. Delineating the neural circuitry by which AgRP neurons defend against starvation is akin to deciphering metabolic pathways. However, unlike metabolism, the logic of the wiring diagram of neurons connected to AgRP neurons is just beginning to be understood. In this issue of Nature Neuroscience, Garfield et al.4 enhance our understanding of this feeding circuitry by elucidating a branch that extends from AgRP neurons to MC4R-expressing 789
news and views a
PBN ??
PVH
ARC Excitatory MC4R neurons Inhibitory AgRP neurons
Fluorescence
b
Microscope
Food
ARC
npg
© 2015 Nature America, Inc. All rights reserved.
Minutes
AgRP neurons with GCaMP6
Figure 1 Logic of AgRP neuron signaling. (a) Activation of AgRP neurons in the arcuate (ARC, red) nucleus inhibits MC4R-expressing neurons in the PVH (green) that excite neurons in the PBN to elicit robust feeding. Activation of AgRP neurons has a negative valence (sad face), whereas activation of the PVH→PBN circuit has a positive valence (happy face). (b) The activity of AgRP neurons (calcium-induced fluorescence emitted by GCaMP6 protein introduced into the neurons with a virus) can be measured in real time with a microscope attached to the skull of the mouse. Inset, when a hungry mouse observes food, the activity of AgRP neurons rapidly declines.
neurons in the paraventricular hypothalamus (PVH) and from there to a newly identified group of neurons in the parabrachial nucleus (PBN). Meanwhile, Betley et al.5 report in Nature how the activity of AgRP neurons helps teach mice to avoid hunger. The logic of the neural circuitry by which AgRP neurons are connected to other neurons is critical for understanding how animals respond to the threat of starvation. It has been known for some time that AgRP neurons project their axons to many brain regions6. Recent experiments indicate that any given small group of AgRP neurons sends axons to only a single brain region, whereas other small groups send axons to different brain regions, rather than individual AgRP neurons sending axons to all regions via collaterals7. Nevertheless, most AgRP neurons appear to be activated together; thus, all brain regions that are responsive to AgRP neurons become inhibited as energy levels fall. To explore the function of specific neurons, genetic and viral transduction techniques can be combined to allow artificial activation or inhibition of specific neurons while observing the behavioral consequences. For example, activation of AgRP neurons by optical or chemical means 790
results in robust food consumption during the middle of the day, when mice are normally sleeping1,2. Mice stimulated in this manner will eat about as much as they would normally eat after an overnight fast. A nice feature of optical activation techniques is that one can selectively activate terminals in just one of the many brain regions to which AgRP neurons project. Using this strategy, Betley et al.7 found that the PVH was one of several spots in which activating AgRP axons elicited robust feeding by mice that were well fed, whereas activation of axon projections from AgRP neurons to other brain regions had no effect on feeding; those projections probably mediate other responses to starvation. AgRP acts on the melanocortin 4 receptor (MC4R) located on postsynaptic neurons8; thus, to extend the neural circuit beyond the PVH, Garfield et al.4 targeted MC4R-expressing neurons in the PVH to examine how artificially modulating their activity affects feeding behavior. They discovered an excitatory projection to the PBN that inhibited feeding when the MC4R cell bodies or their terminals in the PBN were activated, whereas inhibiting the MC4R neurons in the PVH stimulated feeding (Fig. 1a). Their results predict that direct inhibition of PBN neurons that receive innervation from the PVH should promote robust feeding. Because AgRP neurons are inhibitory, they are thought to elicit feeding by suppressing an excitatory satiety signal from the PVH to the PBN. The identity of the neurons in the PBN that are activated by the MC4R neuron projection is unknown, but they are not the calcitonin gene–related peptide (CGRP)-expressing neurons in the PBN that also promote satiety when activated9. Animals engage in rewarding activities while avoiding aversive ones. They can learn by performing activities that result in positive state (for example, eating) or by performing activities that reduce a negative state (for example, hunger). Betley et al.5 demonstrated that activation of AgRP neurons (without food) results in a negative state (or valence) and that mice learn to engage in activities that mitigate the negative feelings of hunger. Garfield et al.4 designed behavioral experiments to show that mice like to have their MC4R-expressing neurons activated (positive valence); that is, the mice prefer the feeling of satiety. Likewise, lateral hypothalamic neurons with positive valence have been identified that promote feeding by well-fed mice10; these neurons may be part of a hedonic system that promotes need-free eating (for example, consuming dessert when already full). The bad feeling of hunger and the good feeling of satiety help explain
why dieting is so difficult. Perhaps medicinal regulation of satiety or hedonic circuits can reduce the urge to eat without triggering the negative feelings associated with hunger. An unexpected result from recent research on AgRP neuron activity is that they respond rapidly to the sight of food, or even to cues that predict food5,11. These observations are derived from experiments in which a calciumsensitive fluorescent reporter (GCaMP6) is introduced into AgRP neurons. A fiber optic probe is then used to record changes in the fluorescence of the AgRP population11 or the fluorescence of individual AgRP neurons can be recorded with a lens attached to a miniature microscope mounted on the head of the mouse5 (Fig. 1b). On the level of individual neurons, the changes in fluorescence are a proxy for neuron activity—one observes the neurons flickering on and off. The return of AgRP neuron fluorescence to baseline immediately when food becomes available turns off the hunger alarm, but, if sufficient nutritive food is not consumed, the alarm resets within a few minutes5—resembling a snooze alarm. The rapid resetting of the AgRP alarm before a mouse even eats must be a neuronal signal that is learned, but the source of that signal is unknown. Activation of AgRP neurons appears to signal impending starvation; hence, a major function of these inhibitory neurons may be to suppress energy-demanding behaviors (for example, fighting) and physiological processes (metabolism, growth and reproduction) while promoting foraging into risky territory. Rapid changes in AgRP neuron activity may be important for adjusting behavior to consume food and modify social interactions with other animals competing for food. Metabolism, growth and reproduction can recover gradually as energy balance is restored. Despite the seeming importance of AgRP neurons, mice can survive without them3,12; thus, there must be compensatory mechanisms that can maintain adequate feeding, although the mechanisms involved are unknown. Understanding the logic of the wiring diagram of neurons connected to AgRP neurons is in its infancy. Determining the identity of the PBN neurons that are activated by the MC4R neurons and their postsynaptic targets, not to mention deciphering the roles of all the other AgRP neuron projections, are logical next steps. There will undoubtedly be numerous intersections of the AgRP neuronal signaling pathways with pathways that control reward, satiety and other aspects of physiology and behavior. Elucidating the neurocircuitry that underlies learning to avoid hunger is a particularly important endeavor. Analyzing the logic of AgRP neuron signaling continues
volume 18 | number 6 | JUNE 2015 nature neuroscience
news and views to be an attractive system for understanding how the brain responds to threats and activates defenses that help maintain homeostasis. COMPETING FINANCIAL INTERESTS The author declares no competing financial interests. 1. Krashes, M.J. et al. J. Clin. Invest. 121, 1424–1428 (2011).
2. Aponte, Y., Atasoy, D. & Sternson, S.M. Nat. Neurosci. 14, 351–355 (2011). 3. Luquet, S., Perez, F.A., Hnasko, T.S. & Palmiter, R.D. Science 310, 683–685 (2005). 4. Garfield, A.S. et al. Nat. Neurosci. 18, 863–871 (2015). 5. Betley, J.N. et al. Nature published online, doi:10.1038/nature14416 (27 April 2015). 6. Broberger, C., Johansson, J., Johansson, C., Schalling, M. & Hokfelt, T. Proc. Natl. Acad. Sci. USA 95, 15043–15048 (1998).
7. Betley, J.N., Cao, Z.F.H., Ritola, K.D. & Sternson, S.M. Cell 155, 1337–1350 (2013). 8. Cone, R.D. Nat. Neurosci. 8, 571–578 (2005). 9. Carter, M.E., Soden, M.E., Zweifel, L.S. & Palmiter, R.D. Nature 503, 111–114 (2013). 10. Jennings, J.H. et al. Cell 160, 516–527 (2015). 11. Chen, Y., Lin, Y.C., Kuo, T.W. & Knight, Z.A. Cell 160, 829–841 (2015). 12. Wu, Q., Boyle, M.P. & Palmiter, R.D. Cell 137, 1225–1234 (2009).
Social nudges: utility conferred from others David V Smith & Mauricio R Delgado
Other individuals can profoundly influence our decisions. From purchasing a new car to ordering lunch at a restaurant, our knowledge of what others chose in similar situations affects our own choices. These social nudges can shape the courses of our lives, leading us to make better or worse decisions. For example, when having lunch with friends, you may have a desire to order the bacon cheeseburger but you may conform to a healthier salad option if that is what your friends order. Although psychologists have long recognized the importance of conformity, it remains unclear why some individuals are more likely to conform. In this issue of Nature Neuroscience, Chung et al.1 investigated conformity in the context of decisions involving uncertainty. The decisions that have the most effect on our future involve elements of uncertainty: for example, questioning whether that new car is right for you. In these situations, social information often influences our decision. But it does so in an asymmetric manner, leading some people to take the risk and buy the car and other people to take the safe choice and wait for a better deal. Chung et al.1 hypothesized that the influence of a social nudge on decisions involving uncertainty critically depends on how others influence the perceived utility of the chosen option. To test this intriguing hypothesis, they developed a simple gambling task. On each trial, participants choose between two gambles that differ in objective uncertainty. In this task, a ‘safer’ gamble carries less payoff variance (for example, 40% chance of $33 versus 60% chance of $23), whereas a ‘riskier’ gamble carries more payoff variance (for example, 40% chance of $57 versus 60% chance of $2). David V. Smith and Mauricio R. Delgado are in the Department of Psychology, Rutgers University, Newark, New Jersey, USA. e-mail: [email protected]
When the probability of a high payoff is low, the difference in expected value favors the safer option. However, as the probability of a high payoff increases, participants switch to the riskier option. If you switch relatively early, you would be considered risk seeking; if you switch relatively late, you would be considered risk averse (Fig. 1). In a clever twist, Chung et al.1 created a powerful social context by informing participants that some decisions would be made publicly in front of peers who were concurrently performing the task. This elegant design feature allowed Chung et al.1 to focus on trials in which the participant made their choice after observing the choices of two peers. Exposure to this social information had dramatic influences on risktaking behavior. Observing peers choose risky options encouraged participants to choose the risky option. Likewise, observing peers choose safe options encouraged participants to choose the safe option. Critically, if these effects are a result of the social context, and not merely the addition of information, then similar conformity effects should not be observed when the player observes the choices of a nonsocial control (for example, a computer). In line with this expectation, Chung et al.1 found, in a separate behavioral experiment, that observing the choices of a computer had no influence on risk-taking behavior, thereby confirming the social nature of the observed effect. The distinction between social and nonsocial contexts has been observed in several studies2. For example, recent work has suggested that social decisions are uniquely tied to computations in the temporal-parietal junction (TPJ)3. Striatal responses to reward are also sensitive to social context, with increasing social closeness predicting heightened striatal responses to monetary rewards4. In addition, winning a video game against a human opponent relative to a computerized opponent selectively
nature neuroscience volume 18 | number 6 | JUNE 2015
increases responses in the striatum and the ventromedial prefrontal cortex (VMPFC)5. Building on these observations, other studies have demonstrated that the VMPFC responds to social feedback6,7 and computes the subjective value of social information8. Consistent with these reports, Chung et al.1 found that VMPFC responses encoded the subjective value of the chosen gamble. However, unlike in previous studies, they also developed a computational model to demonstrate that the VMPFC encodes the added utility conferred by others’ choices and predicts the likelihood of conforming to those choices. These results critically depended on participants’ attitudes toward risk. Specifically, a risk-averse participant was more likely to conform to a safe influence, whereas a risk-seeking participant was more likely to conform to a risky influence. Interestingly, when participants conformed contrary to their attitudes toward risk Probability of choosing risky option (%)
npg
© 2015 Nature America, Inc. All rights reserved.
Observing the choices of others adds utility to the chosen option. The additional utility conferred by others’ choices is encoded by the ventromedial prefrontal cortex and explains the idiosyncratic effects of social influence.
100 Point of subjective equivalence
50
More risk seeking
More risk averse
0 20
40
60
80
Probability of earning high payoff (%) Figure 1 Individuals vary in their willingness to choose risky options. The probability of choosing the risky option increases as the probability of earning the high payoff increases. The point at which a participant is indifferent between the risky and safe option indicates that person’s individual preference for risk.
791
© 2015 Nature America, Inc. All rights reserved.
news and views neurons. Using electron microscopy, they found not only a decrease in the total number of axo-spinous synapses, but also a tripling of axo-dendritic and axo-axonal synapses, which are usually much less frequent. The authors hypothesize that this spatial redistribution of synapses is responsible for the increase in the frequency and amplitude of excitatory events. These findings suggest that increased excitation onto the VTA-projecting prefrontal neurons drives increased excitation onto dopaminergic neurons, increasing dopaminergic neuron activity and striatal dopamine release. The authors tested this hypothesis by optogenetically stimulating prefrontal terminals in the VTA and measuring dopamine release in the ventral striatum. Stimulation of cortical terminals in the VTA indeed induced dopamine release, consistent with classical studies that measured dopamine release in the ventral striatum after electrical stimulation of the PFC1,2. Furthermore, the optogenetic stimulation also enhanced locomotion. As in conditional ArpC3 knockout mice, this hyperlocomotion was blocked by haloperidol, again suggesting that the stimulated dopamine release is responsible for the behavioral activation. Several issues remain outstanding, suggesting future experiments. For instance, it is unclear whether the increased excitatory input demonstrated in vitro results in increased PFC neuronal activity in vivo. Increased input could be offset by a decrease in excitability due to altered intrinsic membrane properties; such synaptic scaling is not unusual and has been observed in other circuits5. In vivo recordings are needed to address this question directly. Similarly, it would be important to confirm increases in
firing rates of dopaminergic neurons in ArpC3 knockout mice, as suggested by the current findings. In addition, it remains possible that increased PFC input onto dopamine neurons is not responsible for the increases in striatal dopamine and hyperlocomotion seen in the mutants; as an alternative, increased corticostriatal activity could modulate tonic extrasynaptic dopamine levels through decreased inhibition of VTA neurons by the ventral pallidum6. Differentiating between these two possibilities would require experiments inhibiting specific cortical projections. Finally, there is the issue of which dopamine cells are actually innervated by PFC inputs. In a classical electron microscopy study combining anterograde and retrograde tracers with immunohistochemical labeling for tyrosine hydroxylase, Carr and Sesack confirmed monosynaptic projections from the medial PFC to dopamine neurons3. However, in that study, dopamine neurons receiving synaptic inputs from the cortex were found to project back to the cortex. In contrast, GABAergic neurons receiving synaptic input from the cortex were found to project to the nucleus accumbens3. Stimulation of the PFC induces burst activity in midbrain dopamine neurons7,8, but whether these neurons project to the ventral striatum is unclear. Recording from retrogradely labeled, ventral striatum– projecting VTA neurons during optical activation of prefrontal terminals in VTA slices should be able to answer this question. The results of Kim et al.4 are highly intriguing in the context of schizophrenia. Post-mortem analyses have repeatedly found decreased spine density in PFC from patients with schizophrenia9.
Moreover, brain imaging–based measures of presynaptic dopaminergic function are enhanced in patients with schizophrenia10. Although it is unclear whether prefrontal cortical spine density and presynaptic dopamine disturbances coexist in the same patients, the results of Kim et al.4 suggest that decreased spine density could cause a hyperdopaminergic phenotype in the striatum (Fig. 1). In this context, it is striking that frontal activation during cognitive testing has been found to be related to striatal dopamine when measured in the same patients. Conflicting data on the directionality of that relationship—one study suggests a positive correlation, whereas another suggests the inverse11,12—cautions against a simplistic translation of the current findings to patient populations. Nonetheless, this curious tale of cortical spines and striatal dopamine is intriguing. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. 1. Taber, M.T., Das, S. & Fibiger, H.C. J. Neurochem. 65, 1407–1410 (1995). 2. Karreman, M. & Moghaddam, B. J. Neurochem. 66, 589–598 (1996). 3. Carr, D.B. & Sesack, S.R. J. Neurosci. 20, 3864–3873 (2000). 4. Kim, I.H. et al. Nat. Neurosci. 18, 883–891 (2015). 5. Turrigiano, G.G. Cell 135, 422–435 (2008). 6. Floresco, S.B., West, A.R., Ash, B., Moore, H. & Grace, A.A. Nat. Neurosci. 6, 968–973 (2003). 7. Murase, S., Grenhoff, J., Chouvet, G., Gonon, F.G. & Svensson, T.H. Neurosci. Lett. 157, 53–56 (1993). 8. Tong, Z.Y., Overton, P.G. & Clark, D. Synapse 22, 195–208 (1996). 9. Glausier, J.R. & Lewis, D.A. Neuroscience 251, 90–107 (2013). 10. Howes, O.D. et al. Arch. Gen. Psychiatry 69, 776–786 (2012). 11. Meyer-Lindenberg, A. et al. Nat. Neurosci. 5, 267–271 (2002). 12. Fusar-Poli, P. et al. Mol. Psychiatry 16, 67–75 (2011).
Hunger logic Richard Palmiter Activation of AgRP-expressing ‘hunger’ neurons promotes robust feeding. Recent studies reveal the valence, dynamics and neural circuits engaged by AgRP neurons. Dwindling energy reserves are a threat to homeostasis that, if not countered, leads to starvation. How does the brain detect and then coordinate appropriate responses to dwindling energy? As fasting progresses, hormonal changes, coupled with metabolic changes, regulate the activity of neurons in the brain, resulting Richard Palmiter is in the Department of Biochemistry, University of Washington, Seattle, Washington, USA. e-mail: [email protected]
in activation of an ‘alarm system’ whose output increases in intensity. Remarkably, a small cluster of neurons (about 0.01% of all brain neurons) in the arcuate region of the hypothalamus appears to be the alarm system that promotes foraging, feeding and energy conservation. These neurons, referred to as AgRP neurons because they alone make agoutirelated peptide, are inhibitory neurons that also use neuropeptide Y and GABA as neurotransmitters. Activation of AgRP neurons inhibits other neurons in the brain to promote
nature neuroscience volume 18 | number 6 | JUNE 2015
feeding and conserve energy1,2, whereas their ablation in adult mice results in starvation3. Delineating the neural circuitry by which AgRP neurons defend against starvation is akin to deciphering metabolic pathways. However, unlike metabolism, the logic of the wiring diagram of neurons connected to AgRP neurons is just beginning to be understood. In this issue of Nature Neuroscience, Garfield et al.4 enhance our understanding of this feeding circuitry by elucidating a branch that extends from AgRP neurons to MC4R-expressing 789
news and views a
PBN ??
PVH
ARC Excitatory MC4R neurons Inhibitory AgRP neurons
Fluorescence
b
Microscope
Food
ARC
npg
© 2015 Nature America, Inc. All rights reserved.
Minutes
AgRP neurons with GCaMP6
Figure 1 Logic of AgRP neuron signaling. (a) Activation of AgRP neurons in the arcuate (ARC, red) nucleus inhibits MC4R-expressing neurons in the PVH (green) that excite neurons in the PBN to elicit robust feeding. Activation of AgRP neurons has a negative valence (sad face), whereas activation of the PVH→PBN circuit has a positive valence (happy face). (b) The activity of AgRP neurons (calcium-induced fluorescence emitted by GCaMP6 protein introduced into the neurons with a virus) can be measured in real time with a microscope attached to the skull of the mouse. Inset, when a hungry mouse observes food, the activity of AgRP neurons rapidly declines.
neurons in the paraventricular hypothalamus (PVH) and from there to a newly identified group of neurons in the parabrachial nucleus (PBN). Meanwhile, Betley et al.5 report in Nature how the activity of AgRP neurons helps teach mice to avoid hunger. The logic of the neural circuitry by which AgRP neurons are connected to other neurons is critical for understanding how animals respond to the threat of starvation. It has been known for some time that AgRP neurons project their axons to many brain regions6. Recent experiments indicate that any given small group of AgRP neurons sends axons to only a single brain region, whereas other small groups send axons to different brain regions, rather than individual AgRP neurons sending axons to all regions via collaterals7. Nevertheless, most AgRP neurons appear to be activated together; thus, all brain regions that are responsive to AgRP neurons become inhibited as energy levels fall. To explore the function of specific neurons, genetic and viral transduction techniques can be combined to allow artificial activation or inhibition of specific neurons while observing the behavioral consequences. For example, activation of AgRP neurons by optical or chemical means 790
results in robust food consumption during the middle of the day, when mice are normally sleeping1,2. Mice stimulated in this manner will eat about as much as they would normally eat after an overnight fast. A nice feature of optical activation techniques is that one can selectively activate terminals in just one of the many brain regions to which AgRP neurons project. Using this strategy, Betley et al.7 found that the PVH was one of several spots in which activating AgRP axons elicited robust feeding by mice that were well fed, whereas activation of axon projections from AgRP neurons to other brain regions had no effect on feeding; those projections probably mediate other responses to starvation. AgRP acts on the melanocortin 4 receptor (MC4R) located on postsynaptic neurons8; thus, to extend the neural circuit beyond the PVH, Garfield et al.4 targeted MC4R-expressing neurons in the PVH to examine how artificially modulating their activity affects feeding behavior. They discovered an excitatory projection to the PBN that inhibited feeding when the MC4R cell bodies or their terminals in the PBN were activated, whereas inhibiting the MC4R neurons in the PVH stimulated feeding (Fig. 1a). Their results predict that direct inhibition of PBN neurons that receive innervation from the PVH should promote robust feeding. Because AgRP neurons are inhibitory, they are thought to elicit feeding by suppressing an excitatory satiety signal from the PVH to the PBN. The identity of the neurons in the PBN that are activated by the MC4R neuron projection is unknown, but they are not the calcitonin gene–related peptide (CGRP)-expressing neurons in the PBN that also promote satiety when activated9. Animals engage in rewarding activities while avoiding aversive ones. They can learn by performing activities that result in positive state (for example, eating) or by performing activities that reduce a negative state (for example, hunger). Betley et al.5 demonstrated that activation of AgRP neurons (without food) results in a negative state (or valence) and that mice learn to engage in activities that mitigate the negative feelings of hunger. Garfield et al.4 designed behavioral experiments to show that mice like to have their MC4R-expressing neurons activated (positive valence); that is, the mice prefer the feeling of satiety. Likewise, lateral hypothalamic neurons with positive valence have been identified that promote feeding by well-fed mice10; these neurons may be part of a hedonic system that promotes need-free eating (for example, consuming dessert when already full). The bad feeling of hunger and the good feeling of satiety help explain
why dieting is so difficult. Perhaps medicinal regulation of satiety or hedonic circuits can reduce the urge to eat without triggering the negative feelings associated with hunger. An unexpected result from recent research on AgRP neuron activity is that they respond rapidly to the sight of food, or even to cues that predict food5,11. These observations are derived from experiments in which a calciumsensitive fluorescent reporter (GCaMP6) is introduced into AgRP neurons. A fiber optic probe is then used to record changes in the fluorescence of the AgRP population11 or the fluorescence of individual AgRP neurons can be recorded with a lens attached to a miniature microscope mounted on the head of the mouse5 (Fig. 1b). On the level of individual neurons, the changes in fluorescence are a proxy for neuron activity—one observes the neurons flickering on and off. The return of AgRP neuron fluorescence to baseline immediately when food becomes available turns off the hunger alarm, but, if sufficient nutritive food is not consumed, the alarm resets within a few minutes5—resembling a snooze alarm. The rapid resetting of the AgRP alarm before a mouse even eats must be a neuronal signal that is learned, but the source of that signal is unknown. Activation of AgRP neurons appears to signal impending starvation; hence, a major function of these inhibitory neurons may be to suppress energy-demanding behaviors (for example, fighting) and physiological processes (metabolism, growth and reproduction) while promoting foraging into risky territory. Rapid changes in AgRP neuron activity may be important for adjusting behavior to consume food and modify social interactions with other animals competing for food. Metabolism, growth and reproduction can recover gradually as energy balance is restored. Despite the seeming importance of AgRP neurons, mice can survive without them3,12; thus, there must be compensatory mechanisms that can maintain adequate feeding, although the mechanisms involved are unknown. Understanding the logic of the wiring diagram of neurons connected to AgRP neurons is in its infancy. Determining the identity of the PBN neurons that are activated by the MC4R neurons and their postsynaptic targets, not to mention deciphering the roles of all the other AgRP neuron projections, are logical next steps. There will undoubtedly be numerous intersections of the AgRP neuronal signaling pathways with pathways that control reward, satiety and other aspects of physiology and behavior. Elucidating the neurocircuitry that underlies learning to avoid hunger is a particularly important endeavor. Analyzing the logic of AgRP neuron signaling continues
volume 18 | number 6 | JUNE 2015 nature neuroscience
news and views to be an attractive system for understanding how the brain responds to threats and activates defenses that help maintain homeostasis. COMPETING FINANCIAL INTERESTS The author declares no competing financial interests. 1. Krashes, M.J. et al. J. Clin. Invest. 121, 1424–1428 (2011).
2. Aponte, Y., Atasoy, D. & Sternson, S.M. Nat. Neurosci. 14, 351–355 (2011). 3. Luquet, S., Perez, F.A., Hnasko, T.S. & Palmiter, R.D. Science 310, 683–685 (2005). 4. Garfield, A.S. et al. Nat. Neurosci. 18, 863–871 (2015). 5. Betley, J.N. et al. Nature published online, doi:10.1038/nature14416 (27 April 2015). 6. Broberger, C., Johansson, J., Johansson, C., Schalling, M. & Hokfelt, T. Proc. Natl. Acad. Sci. USA 95, 15043–15048 (1998).
7. Betley, J.N., Cao, Z.F.H., Ritola, K.D. & Sternson, S.M. Cell 155, 1337–1350 (2013). 8. Cone, R.D. Nat. Neurosci. 8, 571–578 (2005). 9. Carter, M.E., Soden, M.E., Zweifel, L.S. & Palmiter, R.D. Nature 503, 111–114 (2013). 10. Jennings, J.H. et al. Cell 160, 516–527 (2015). 11. Chen, Y., Lin, Y.C., Kuo, T.W. & Knight, Z.A. Cell 160, 829–841 (2015). 12. Wu, Q., Boyle, M.P. & Palmiter, R.D. Cell 137, 1225–1234 (2009).
Social nudges: utility conferred from others David V Smith & Mauricio R Delgado
Other individuals can profoundly influence our decisions. From purchasing a new car to ordering lunch at a restaurant, our knowledge of what others chose in similar situations affects our own choices. These social nudges can shape the courses of our lives, leading us to make better or worse decisions. For example, when having lunch with friends, you may have a desire to order the bacon cheeseburger but you may conform to a healthier salad option if that is what your friends order. Although psychologists have long recognized the importance of conformity, it remains unclear why some individuals are more likely to conform. In this issue of Nature Neuroscience, Chung et al.1 investigated conformity in the context of decisions involving uncertainty. The decisions that have the most effect on our future involve elements of uncertainty: for example, questioning whether that new car is right for you. In these situations, social information often influences our decision. But it does so in an asymmetric manner, leading some people to take the risk and buy the car and other people to take the safe choice and wait for a better deal. Chung et al.1 hypothesized that the influence of a social nudge on decisions involving uncertainty critically depends on how others influence the perceived utility of the chosen option. To test this intriguing hypothesis, they developed a simple gambling task. On each trial, participants choose between two gambles that differ in objective uncertainty. In this task, a ‘safer’ gamble carries less payoff variance (for example, 40% chance of $33 versus 60% chance of $23), whereas a ‘riskier’ gamble carries more payoff variance (for example, 40% chance of $57 versus 60% chance of $2). David V. Smith and Mauricio R. Delgado are in the Department of Psychology, Rutgers University, Newark, New Jersey, USA. e-mail: [email protected]
When the probability of a high payoff is low, the difference in expected value favors the safer option. However, as the probability of a high payoff increases, participants switch to the riskier option. If you switch relatively early, you would be considered risk seeking; if you switch relatively late, you would be considered risk averse (Fig. 1). In a clever twist, Chung et al.1 created a powerful social context by informing participants that some decisions would be made publicly in front of peers who were concurrently performing the task. This elegant design feature allowed Chung et al.1 to focus on trials in which the participant made their choice after observing the choices of two peers. Exposure to this social information had dramatic influences on risktaking behavior. Observing peers choose risky options encouraged participants to choose the risky option. Likewise, observing peers choose safe options encouraged participants to choose the safe option. Critically, if these effects are a result of the social context, and not merely the addition of information, then similar conformity effects should not be observed when the player observes the choices of a nonsocial control (for example, a computer). In line with this expectation, Chung et al.1 found, in a separate behavioral experiment, that observing the choices of a computer had no influence on risk-taking behavior, thereby confirming the social nature of the observed effect. The distinction between social and nonsocial contexts has been observed in several studies2. For example, recent work has suggested that social decisions are uniquely tied to computations in the temporal-parietal junction (TPJ)3. Striatal responses to reward are also sensitive to social context, with increasing social closeness predicting heightened striatal responses to monetary rewards4. In addition, winning a video game against a human opponent relative to a computerized opponent selectively
nature neuroscience volume 18 | number 6 | JUNE 2015
increases responses in the striatum and the ventromedial prefrontal cortex (VMPFC)5. Building on these observations, other studies have demonstrated that the VMPFC responds to social feedback6,7 and computes the subjective value of social information8. Consistent with these reports, Chung et al.1 found that VMPFC responses encoded the subjective value of the chosen gamble. However, unlike in previous studies, they also developed a computational model to demonstrate that the VMPFC encodes the added utility conferred by others’ choices and predicts the likelihood of conforming to those choices. These results critically depended on participants’ attitudes toward risk. Specifically, a risk-averse participant was more likely to conform to a safe influence, whereas a risk-seeking participant was more likely to conform to a risky influence. Interestingly, when participants conformed contrary to their attitudes toward risk Probability of choosing risky option (%)
npg
© 2015 Nature America, Inc. All rights reserved.
Observing the choices of others adds utility to the chosen option. The additional utility conferred by others’ choices is encoded by the ventromedial prefrontal cortex and explains the idiosyncratic effects of social influence.
100 Point of subjective equivalence
50
More risk seeking
More risk averse
0 20
40
60
80
Probability of earning high payoff (%) Figure 1 Individuals vary in their willingness to choose risky options. The probability of choosing the risky option increases as the probability of earning the high payoff increases. The point at which a participant is indifferent between the risky and safe option indicates that person’s individual preference for risk.
791
